linux/kernel/trace/trace_functions_graph.c

900 lines
21 KiB
C
Raw Normal View History

/*
*
* Function graph tracer.
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
* Copyright (c) 2008-2009 Frederic Weisbecker <fweisbec@gmail.com>
* Mostly borrowed from function tracer which
* is Copyright (c) Steven Rostedt <srostedt@redhat.com>
*
*/
#include <linux/debugfs.h>
#include <linux/uaccess.h>
#include <linux/ftrace.h>
#include <linux/fs.h>
#include "trace.h"
#include "trace_output.h"
struct fgraph_data {
pid_t last_pid;
int depth;
};
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
#define TRACE_GRAPH_INDENT 2
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
/* Flag options */
#define TRACE_GRAPH_PRINT_OVERRUN 0x1
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
#define TRACE_GRAPH_PRINT_CPU 0x2
#define TRACE_GRAPH_PRINT_OVERHEAD 0x4
#define TRACE_GRAPH_PRINT_PROC 0x8
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
#define TRACE_GRAPH_PRINT_DURATION 0x10
#define TRACE_GRAPH_PRINT_ABS_TIME 0X20
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
static struct tracer_opt trace_opts[] = {
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
/* Display overruns? (for self-debug purpose) */
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
{ TRACER_OPT(funcgraph-overrun, TRACE_GRAPH_PRINT_OVERRUN) },
/* Display CPU ? */
{ TRACER_OPT(funcgraph-cpu, TRACE_GRAPH_PRINT_CPU) },
/* Display Overhead ? */
{ TRACER_OPT(funcgraph-overhead, TRACE_GRAPH_PRINT_OVERHEAD) },
/* Display proc name/pid */
{ TRACER_OPT(funcgraph-proc, TRACE_GRAPH_PRINT_PROC) },
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
/* Display duration of execution */
{ TRACER_OPT(funcgraph-duration, TRACE_GRAPH_PRINT_DURATION) },
/* Display absolute time of an entry */
{ TRACER_OPT(funcgraph-abstime, TRACE_GRAPH_PRINT_ABS_TIME) },
{ } /* Empty entry */
};
static struct tracer_flags tracer_flags = {
/* Don't display overruns and proc by default */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
.val = TRACE_GRAPH_PRINT_CPU | TRACE_GRAPH_PRINT_OVERHEAD |
TRACE_GRAPH_PRINT_DURATION,
.opts = trace_opts
};
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
/* pid on the last trace processed */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
/* Add a function return address to the trace stack on thread info.*/
int
ftrace_push_return_trace(unsigned long ret, unsigned long func, int *depth)
{
unsigned long long calltime;
int index;
if (!current->ret_stack)
return -EBUSY;
/* The return trace stack is full */
if (current->curr_ret_stack == FTRACE_RETFUNC_DEPTH - 1) {
atomic_inc(&current->trace_overrun);
return -EBUSY;
}
calltime = trace_clock_local();
index = ++current->curr_ret_stack;
barrier();
current->ret_stack[index].ret = ret;
current->ret_stack[index].func = func;
current->ret_stack[index].calltime = calltime;
current->ret_stack[index].subtime = 0;
*depth = index;
return 0;
}
/* Retrieve a function return address to the trace stack on thread info.*/
static void
ftrace_pop_return_trace(struct ftrace_graph_ret *trace, unsigned long *ret)
{
int index;
index = current->curr_ret_stack;
if (unlikely(index < 0)) {
ftrace_graph_stop();
WARN_ON(1);
/* Might as well panic, otherwise we have no where to go */
*ret = (unsigned long)panic;
return;
}
*ret = current->ret_stack[index].ret;
trace->func = current->ret_stack[index].func;
trace->calltime = current->ret_stack[index].calltime;
trace->overrun = atomic_read(&current->trace_overrun);
trace->depth = index;
}
/*
* Send the trace to the ring-buffer.
* @return the original return address.
*/
unsigned long ftrace_return_to_handler(void)
{
struct ftrace_graph_ret trace;
unsigned long ret;
ftrace_pop_return_trace(&trace, &ret);
trace.rettime = trace_clock_local();
ftrace_graph_return(&trace);
barrier();
current->curr_ret_stack--;
if (unlikely(!ret)) {
ftrace_graph_stop();
WARN_ON(1);
/* Might as well panic. What else to do? */
ret = (unsigned long)panic;
}
return ret;
}
static int graph_trace_init(struct trace_array *tr)
{
int ret = register_ftrace_graph(&trace_graph_return,
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
&trace_graph_entry);
if (ret)
return ret;
tracing_start_cmdline_record();
return 0;
}
static void graph_trace_reset(struct trace_array *tr)
{
tracing_stop_cmdline_record();
unregister_ftrace_graph();
}
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
static inline int log10_cpu(int nb)
{
if (nb / 100)
return 3;
if (nb / 10)
return 2;
return 1;
}
static enum print_line_t
print_graph_cpu(struct trace_seq *s, int cpu)
{
int i;
int ret;
int log10_this = log10_cpu(cpu);
int log10_all = log10_cpu(cpumask_weight(cpu_online_mask));
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
/*
* Start with a space character - to make it stand out
* to the right a bit when trace output is pasted into
* email:
*/
ret = trace_seq_printf(s, " ");
/*
* Tricky - we space the CPU field according to the max
* number of online CPUs. On a 2-cpu system it would take
* a maximum of 1 digit - on a 128 cpu system it would
* take up to 3 digits:
*/
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
for (i = 0; i < log10_all - log10_this; i++) {
ret = trace_seq_printf(s, " ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
ret = trace_seq_printf(s, "%d) ", cpu);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
return TRACE_TYPE_HANDLED;
}
#define TRACE_GRAPH_PROCINFO_LENGTH 14
static enum print_line_t
print_graph_proc(struct trace_seq *s, pid_t pid)
{
char comm[TASK_COMM_LEN];
/* sign + log10(MAX_INT) + '\0' */
char pid_str[11];
int spaces = 0;
int ret;
int len;
int i;
trace_find_cmdline(pid, comm);
comm[7] = '\0';
sprintf(pid_str, "%d", pid);
/* 1 stands for the "-" character */
len = strlen(comm) + strlen(pid_str) + 1;
if (len < TRACE_GRAPH_PROCINFO_LENGTH)
spaces = TRACE_GRAPH_PROCINFO_LENGTH - len;
/* First spaces to align center */
for (i = 0; i < spaces / 2; i++) {
ret = trace_seq_printf(s, " ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
ret = trace_seq_printf(s, "%s-%s", comm, pid_str);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
/* Last spaces to align center */
for (i = 0; i < spaces - (spaces / 2); i++) {
ret = trace_seq_printf(s, " ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
return TRACE_TYPE_HANDLED;
}
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
/* If the pid changed since the last trace, output this event */
static enum print_line_t
verif_pid(struct trace_seq *s, pid_t pid, int cpu, struct fgraph_data *data)
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
{
pid_t prev_pid;
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
pid_t *last_pid;
int ret;
if (!data)
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
return TRACE_TYPE_HANDLED;
last_pid = &(per_cpu_ptr(data, cpu)->last_pid);
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
if (*last_pid == pid)
return TRACE_TYPE_HANDLED;
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
prev_pid = *last_pid;
*last_pid = pid;
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
if (prev_pid == -1)
return TRACE_TYPE_HANDLED;
/*
* Context-switch trace line:
------------------------------------------
| 1) migration/0--1 => sshd-1755
------------------------------------------
*/
ret = trace_seq_printf(s,
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
" ------------------------------------------\n");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
ret = print_graph_cpu(s, cpu);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
ret = print_graph_proc(s, prev_pid);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
ret = trace_seq_printf(s, " => ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
ret = print_graph_proc(s, pid);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
ret = trace_seq_printf(s,
"\n ------------------------------------------\n\n");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
return TRACE_TYPE_HANDLED;
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
}
static struct ftrace_graph_ret_entry *
get_return_for_leaf(struct trace_iterator *iter,
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
struct ftrace_graph_ent_entry *curr)
{
struct ring_buffer_iter *ring_iter;
struct ring_buffer_event *event;
struct ftrace_graph_ret_entry *next;
ring_iter = iter->buffer_iter[iter->cpu];
/* First peek to compare current entry and the next one */
if (ring_iter)
event = ring_buffer_iter_peek(ring_iter, NULL);
else {
/* We need to consume the current entry to see the next one */
ring_buffer_consume(iter->tr->buffer, iter->cpu, NULL);
event = ring_buffer_peek(iter->tr->buffer, iter->cpu,
NULL);
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
if (!event)
return NULL;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
next = ring_buffer_event_data(event);
if (next->ent.type != TRACE_GRAPH_RET)
return NULL;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
if (curr->ent.pid != next->ent.pid ||
curr->graph_ent.func != next->ret.func)
return NULL;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
/* this is a leaf, now advance the iterator */
if (ring_iter)
ring_buffer_read(ring_iter, NULL);
return next;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
}
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
/* Signal a overhead of time execution to the output */
static int
print_graph_overhead(unsigned long long duration, struct trace_seq *s)
{
/* If duration disappear, we don't need anything */
if (!(tracer_flags.val & TRACE_GRAPH_PRINT_DURATION))
return 1;
/* Non nested entry or return */
if (duration == -1)
return trace_seq_printf(s, " ");
if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERHEAD) {
/* Duration exceeded 100 msecs */
if (duration > 100000ULL)
return trace_seq_printf(s, "! ");
/* Duration exceeded 10 msecs */
if (duration > 10000ULL)
return trace_seq_printf(s, "+ ");
}
return trace_seq_printf(s, " ");
}
static int print_graph_abs_time(u64 t, struct trace_seq *s)
{
unsigned long usecs_rem;
usecs_rem = do_div(t, NSEC_PER_SEC);
usecs_rem /= 1000;
return trace_seq_printf(s, "%5lu.%06lu | ",
(unsigned long)t, usecs_rem);
}
static enum print_line_t
print_graph_irq(struct trace_iterator *iter, unsigned long addr,
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
enum trace_type type, int cpu, pid_t pid)
{
int ret;
struct trace_seq *s = &iter->seq;
if (addr < (unsigned long)__irqentry_text_start ||
addr >= (unsigned long)__irqentry_text_end)
return TRACE_TYPE_UNHANDLED;
/* Absolute time */
if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME) {
ret = print_graph_abs_time(iter->ts, s);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
/* Cpu */
if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU) {
ret = print_graph_cpu(s, cpu);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
}
/* Proc */
if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC) {
ret = print_graph_proc(s, pid);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
ret = trace_seq_printf(s, " | ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
/* No overhead */
ret = print_graph_overhead(-1, s);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
if (type == TRACE_GRAPH_ENT)
ret = trace_seq_printf(s, "==========>");
else
ret = trace_seq_printf(s, "<==========");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
/* Don't close the duration column if haven't one */
if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
trace_seq_printf(s, " |");
ret = trace_seq_printf(s, "\n");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
return TRACE_TYPE_HANDLED;
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
enum print_line_t
trace_print_graph_duration(unsigned long long duration, struct trace_seq *s)
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
{
unsigned long nsecs_rem = do_div(duration, 1000);
/* log10(ULONG_MAX) + '\0' */
char msecs_str[21];
char nsecs_str[5];
int ret, len;
int i;
sprintf(msecs_str, "%lu", (unsigned long) duration);
/* Print msecs */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
ret = trace_seq_printf(s, "%s", msecs_str);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
len = strlen(msecs_str);
/* Print nsecs (we don't want to exceed 7 numbers) */
if (len < 7) {
snprintf(nsecs_str, 8 - len, "%03lu", nsecs_rem);
ret = trace_seq_printf(s, ".%s", nsecs_str);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
len += strlen(nsecs_str);
}
ret = trace_seq_printf(s, " us ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
/* Print remaining spaces to fit the row's width */
for (i = len; i < 7; i++) {
ret = trace_seq_printf(s, " ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
return TRACE_TYPE_HANDLED;
}
static enum print_line_t
print_graph_duration(unsigned long long duration, struct trace_seq *s)
{
int ret;
ret = trace_print_graph_duration(duration, s);
if (ret != TRACE_TYPE_HANDLED)
return ret;
ret = trace_seq_printf(s, "| ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
return TRACE_TYPE_HANDLED;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
}
/* Case of a leaf function on its call entry */
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
static enum print_line_t
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
print_graph_entry_leaf(struct trace_iterator *iter,
struct ftrace_graph_ent_entry *entry,
struct ftrace_graph_ret_entry *ret_entry, struct trace_seq *s)
{
struct fgraph_data *data = iter->private;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
struct ftrace_graph_ret *graph_ret;
struct ftrace_graph_ent *call;
unsigned long long duration;
int ret;
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
int i;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
graph_ret = &ret_entry->ret;
call = &entry->graph_ent;
duration = graph_ret->rettime - graph_ret->calltime;
if (data) {
int cpu = iter->cpu;
int *depth = &(per_cpu_ptr(data, cpu)->depth);
/*
* Comments display at + 1 to depth. Since
* this is a leaf function, keep the comments
* equal to this depth.
*/
*depth = call->depth - 1;
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
/* Overhead */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
ret = print_graph_overhead(duration, s);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
/* Duration */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
ret = print_graph_duration(duration, s);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
}
ftrace: add cpu annotation for function graph tracer Impact: enhancement for function graph tracer When run on a SMP box, the function graph tracer is confusing because it shows the different CPUS as changes in the trace. This patch adds the annotation of 'CPU[###]' where ### is a three digit number. The output will look similar to this: CPU[001] dput() { CPU[000] } 726 CPU[001] } 487 CPU[000] do_softirq() { CPU[001] } 2221 CPU[000] __do_softirq() { CPU[000] __local_bh_disable() { CPU[001] unroll_tree_refs() { CPU[000] } 569 CPU[001] } 501 CPU[000] rcu_process_callbacks() { CPU[001] kfree() { What makes this nice is that now you can grep the file and produce readable format for a particular CPU. # cat /debug/tracing/trace > /tmp/trace # grep '^CPU\[000\]' /tmp/trace > /tmp/trace0 # grep '^CPU\[001\]' /tmp/trace > /tmp/trace1 Will give you: # head /tmp/trace0 CPU[000] ------------8<---------- thread sshd-3899 ------------8<---------- CPU[000] inotify_dentry_parent_queue_event() { CPU[000] } 2531 CPU[000] inotify_inode_queue_event() { CPU[000] } 505 CPU[000] } 69626 CPU[000] } 73089 CPU[000] audit_syscall_exit() { CPU[000] path_put() { CPU[000] dput() { # head /tmp/trace1 CPU[001] ------------8<---------- thread pcscd-3446 ------------8<---------- CPU[001] } 4186 CPU[001] dput() { CPU[001] } 543 CPU[001] vfs_permission() { CPU[001] inode_permission() { CPU[001] shmem_permission() { CPU[001] generic_permission() { CPU[001] } 501 CPU[001] } 2205 Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-26 05:16:27 +00:00
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
/* Function */
for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
ret = trace_seq_printf(s, " ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
ret = seq_print_ip_sym(s, call->func, 0);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
ret = trace_seq_printf(s, "();\n");
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
return TRACE_TYPE_HANDLED;
}
static enum print_line_t
print_graph_entry_nested(struct trace_iterator *iter,
struct ftrace_graph_ent_entry *entry,
struct trace_seq *s, int cpu)
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
{
struct ftrace_graph_ent *call = &entry->graph_ent;
struct fgraph_data *data = iter->private;
int ret;
int i;
if (data) {
int cpu = iter->cpu;
int *depth = &(per_cpu_ptr(data, cpu)->depth);
*depth = call->depth;
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
/* No overhead */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
ret = print_graph_overhead(-1, s);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
/* No time */
if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
ret = trace_seq_printf(s, " | ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
/* Function */
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
for (i = 0; i < call->depth * TRACE_GRAPH_INDENT; i++) {
ret = trace_seq_printf(s, " ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
}
ret = seq_print_ip_sym(s, call->func, 0);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
ret = trace_seq_printf(s, "() {\n");
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
/*
* we already consumed the current entry to check the next one
* and see if this is a leaf.
*/
return TRACE_TYPE_NO_CONSUME;
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
static enum print_line_t
print_graph_prologue(struct trace_iterator *iter, struct trace_seq *s,
int type, unsigned long addr)
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
{
struct fgraph_data *data = iter->private;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
struct trace_entry *ent = iter->ent;
int cpu = iter->cpu;
int ret;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
/* Pid */
if (verif_pid(s, ent->pid, cpu, data) == TRACE_TYPE_PARTIAL_LINE)
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
return TRACE_TYPE_PARTIAL_LINE;
if (type) {
/* Interrupt */
ret = print_graph_irq(iter, addr, type, cpu, ent->pid);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
/* Absolute time */
if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME) {
ret = print_graph_abs_time(iter->ts, s);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
/* Cpu */
if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU) {
ret = print_graph_cpu(s, cpu);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
}
/* Proc */
if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC) {
ret = print_graph_proc(s, ent->pid);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
ret = trace_seq_printf(s, " | ");
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
return 0;
}
static enum print_line_t
print_graph_entry(struct ftrace_graph_ent_entry *field, struct trace_seq *s,
struct trace_iterator *iter)
{
int cpu = iter->cpu;
struct ftrace_graph_ent *call = &field->graph_ent;
struct ftrace_graph_ret_entry *leaf_ret;
if (print_graph_prologue(iter, s, TRACE_GRAPH_ENT, call->func))
return TRACE_TYPE_PARTIAL_LINE;
leaf_ret = get_return_for_leaf(iter, field);
if (leaf_ret)
return print_graph_entry_leaf(iter, field, leaf_ret, s);
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
else
return print_graph_entry_nested(iter, field, s, cpu);
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
}
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
static enum print_line_t
print_graph_return(struct ftrace_graph_ret *trace, struct trace_seq *s,
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
struct trace_entry *ent, struct trace_iterator *iter)
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
{
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
unsigned long long duration = trace->rettime - trace->calltime;
struct fgraph_data *data = iter->private;
pid_t pid = ent->pid;
int cpu = iter->cpu;
int ret;
int i;
if (data) {
int cpu = iter->cpu;
int *depth = &(per_cpu_ptr(data, cpu)->depth);
/*
* Comments display at + 1 to depth. This is the
* return from a function, we now want the comments
* to display at the same level of the bracket.
*/
*depth = trace->depth - 1;
}
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
if (print_graph_prologue(iter, s, 0, 0))
ftrace: add cpu annotation for function graph tracer Impact: enhancement for function graph tracer When run on a SMP box, the function graph tracer is confusing because it shows the different CPUS as changes in the trace. This patch adds the annotation of 'CPU[###]' where ### is a three digit number. The output will look similar to this: CPU[001] dput() { CPU[000] } 726 CPU[001] } 487 CPU[000] do_softirq() { CPU[001] } 2221 CPU[000] __do_softirq() { CPU[000] __local_bh_disable() { CPU[001] unroll_tree_refs() { CPU[000] } 569 CPU[001] } 501 CPU[000] rcu_process_callbacks() { CPU[001] kfree() { What makes this nice is that now you can grep the file and produce readable format for a particular CPU. # cat /debug/tracing/trace > /tmp/trace # grep '^CPU\[000\]' /tmp/trace > /tmp/trace0 # grep '^CPU\[001\]' /tmp/trace > /tmp/trace1 Will give you: # head /tmp/trace0 CPU[000] ------------8<---------- thread sshd-3899 ------------8<---------- CPU[000] inotify_dentry_parent_queue_event() { CPU[000] } 2531 CPU[000] inotify_inode_queue_event() { CPU[000] } 505 CPU[000] } 69626 CPU[000] } 73089 CPU[000] audit_syscall_exit() { CPU[000] path_put() { CPU[000] dput() { # head /tmp/trace1 CPU[001] ------------8<---------- thread pcscd-3446 ------------8<---------- CPU[001] } 4186 CPU[001] dput() { CPU[001] } 543 CPU[001] vfs_permission() { CPU[001] inode_permission() { CPU[001] shmem_permission() { CPU[001] generic_permission() { CPU[001] } 501 CPU[001] } 2205 Signed-off-by: Steven Rostedt <srostedt@redhat.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-26 05:16:27 +00:00
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
/* Overhead */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
ret = print_graph_overhead(duration, s);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
/* Duration */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
ret = print_graph_duration(duration, s);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
}
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
/* Closing brace */
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
for (i = 0; i < trace->depth * TRACE_GRAPH_INDENT; i++) {
ret = trace_seq_printf(s, " ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
}
tracing/function-graph-tracer: adjustments of the trace informations Impact: increase the visual qualities of the call-graph-tracer output This patch applies various trace output formatting changes: - CPU is now a decimal number, followed by a parenthesis. - Overhead is now on the second column (gives a good visibility) - Cost is now on the third column, can't exceed 9999.99 us. It is followed by a virtual line based on a "|" character. - Functions calls are now the last column on the right. This way, we haven't dynamic column (which flow is harder to follow) on its right. - CPU and Overhead have their own option flag. They are default-on but you can disable them easily: echo nofuncgraph-cpu > trace_options echo nofuncgraph-overhead > trace_options TODO: _ Refactoring of the thread switch output. _ Give a default-off option to output the thread and its pid on each row. _ Provide headers _ .... Here is an example of the new trace style: 0) | mutex_unlock() { 0) 0.639 us | __mutex_unlock_slowpath(); 0) 1.607 us | } 0) | remove_wait_queue() { 0) 0.616 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.779 us | } 0) 0.495 us | n_tty_set_room(); 0) ! 9999.999 us | } 0) | tty_ldisc_deref() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.616 us | _spin_unlock_irqrestore(); 0) 2.793 us | } 0) | current_fs_time() { 0) 0.488 us | current_kernel_time(); 0) 0.495 us | timespec_trunc(); 0) 2.486 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) ! 9999.999 us | } 0) | sys_read() { 0) 0.796 us | fget_light(); 0) | vfs_read() { 0) | rw_verify_area() { 0) | security_file_permission() { 0) 0.488 us | cap_file_permission(); 0) 1.720 us | } 0) 3. 4 us | } 0) | tty_read() { 0) 0.488 us | tty_paranoia_check(); 0) | tty_ldisc_ref_wait() { 0) | tty_ldisc_try() { 0) 0.615 us | _spin_lock_irqsave(); 0) 0.615 us | _spin_unlock_irqrestore(); 0) 5.436 us | } 0) 6.427 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 23:42:46 +00:00
ret = trace_seq_printf(s, "}\n");
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: enhancements for the trace output Impact: enhance the output of the graph-tracer This patch applies some ideas of Ingo Molnar and Steven Rostedt. * Output leaf functions in one line with parenthesis, semicolon and duration output. * Add a second column (after cpu) for an overhead sign. if duration > 100 us, "!" if duration > 10 us, "+" else " " * Print output in us with remaining nanosec: u.n * Print duration on the right end, following the indentation of the functions. Use also visual clues: "-" on entry call (no duration to output) and "+" on return (duration output). The name of the tracer has been fixed as well: function-branch becomes function_branch. Here is an example of the new output: CPU[000] dequeue_entity() { - CPU[000] update_curr() { - CPU[000] update_min_vruntime(); + 0.512 us CPU[000] } + 1.504 us CPU[000] clear_buddies(); + 0.481 us CPU[000] update_min_vruntime(); + 0.504 us CPU[000] } + 4.557 us CPU[000] hrtick_update() { - CPU[000] hrtick_start_fair(); + 0.489 us CPU[000] } + 1.443 us CPU[000] + } + 14.655 us CPU[000] + } + 15.678 us CPU[000] + } + 16.686 us CPU[000] msecs_to_jiffies(); + 0.481 us CPU[000] put_prev_task_fair(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.482 us CPU[000] pick_next_task_rt(); + 0.504 us CPU[000] pick_next_task_fair(); + 0.481 us CPU[000] pick_next_task_idle(); + 0.489 us CPU[000] _spin_trylock(); + 0.655 us CPU[000] _spin_unlock(); + 0.609 us CPU[000] ------------8<---------- thread bash-2794 ------------8<---------- CPU[000] finish_task_switch() { - CPU[000] _spin_unlock_irq(); + 0.722 us CPU[000] } + 2.369 us CPU[000] ! } + 501972.605 us CPU[000] ! } + 501973.763 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.670 us CPU[000] _spin_unlock_irqrestore(); + 0.699 us CPU[000] copy_to_user() { - CPU[000] might_fault() { - CPU[000] __might_sleep(); + 0.503 us CPU[000] } + 1.632 us CPU[000] __copy_to_user_ll(); + 0.542 us CPU[000] } + 3.858 us CPU[000] tty_audit_add_data() { - CPU[000] _spin_lock_irq(); + 0.609 us CPU[000] _spin_unlock_irq(); + 0.624 us CPU[000] } + 3.196 us CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.625 us CPU[000] + } + 13.611 us CPU[000] copy_from_read_buf() { - CPU[000] _spin_lock_irqsave(); + 0.624 us CPU[000] _spin_unlock_irqrestore(); + 0.616 us CPU[000] } + 2.820 us CPU[000] Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-27 00:46:33 +00:00
/* Overrun */
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
if (tracer_flags.val & TRACE_GRAPH_PRINT_OVERRUN) {
ret = trace_seq_printf(s, " (Overruns: %lu)\n",
trace->overrun);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
}
ret = print_graph_irq(iter, trace->func, TRACE_GRAPH_RET, cpu, pid);
if (ret == TRACE_TYPE_PARTIAL_LINE)
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
return TRACE_TYPE_HANDLED;
}
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
static enum print_line_t
print_graph_comment(struct trace_seq *s, struct trace_entry *ent,
struct trace_iterator *iter)
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
{
unsigned long sym_flags = (trace_flags & TRACE_ITER_SYM_MASK);
struct fgraph_data *data = iter->private;
struct trace_event *event;
int depth = 0;
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
int ret;
int i;
if (data)
depth = per_cpu_ptr(data, iter->cpu)->depth;
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
if (print_graph_prologue(iter, s, 0, 0))
return TRACE_TYPE_PARTIAL_LINE;
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
/* No overhead */
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
ret = print_graph_overhead(-1, s);
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
/* No time */
if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION) {
ret = trace_seq_printf(s, " | ");
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
/* Indentation */
if (depth > 0)
for (i = 0; i < (depth + 1) * TRACE_GRAPH_INDENT; i++) {
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
ret = trace_seq_printf(s, " ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
}
/* The comment */
tracing/core: drop the old trace_printk() implementation in favour of trace_bprintk() Impact: faster and lighter tracing Now that we have trace_bprintk() which is faster and consume lesser memory than trace_printk() and has the same purpose, we can now drop the old implementation in favour of the binary one from trace_bprintk(), which means we move all the implementation of trace_bprintk() to trace_printk(), so the Api doesn't change except that we must now use trace_seq_bprintk() to print the TRACE_PRINT entries. Some changes result of this: - Previously, trace_bprintk depended of a single tracer and couldn't work without. This tracer has been dropped and the whole implementation of trace_printk() (like the module formats management) is now integrated in the tracing core (comes with CONFIG_TRACING), though we keep the file trace_printk (previously trace_bprintk.c) where we can find the module management. Thus we don't overflow trace.c - changes some parts to use trace_seq_bprintk() to print TRACE_PRINT entries. - change a bit trace_printk/trace_vprintk macros to support non-builtin formats constants, and fix 'const' qualifiers warnings. But this is all transparent for developers. - etc... V2: - Rebase against last changes - Fix mispell on the changelog V3: - Rebase against last changes (moving trace_printk() to kernel.h) Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> LKML-Reference: <1236356510-8381-5-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-03-06 16:21:49 +00:00
ret = trace_seq_printf(s, "/* ");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
switch (iter->ent->type) {
case TRACE_BPRINT:
ret = trace_print_bprintk_msg_only(iter);
if (ret != TRACE_TYPE_HANDLED)
return ret;
break;
case TRACE_PRINT:
ret = trace_print_printk_msg_only(iter);
if (ret != TRACE_TYPE_HANDLED)
return ret;
break;
default:
event = ftrace_find_event(ent->type);
if (!event)
return TRACE_TYPE_UNHANDLED;
ret = event->trace(iter, sym_flags);
if (ret != TRACE_TYPE_HANDLED)
return ret;
}
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
/* Strip ending newline */
if (s->buffer[s->len - 1] == '\n') {
s->buffer[s->len - 1] = '\0';
s->len--;
}
tracing/function-graph-tracer: handle ftrace_printk entries Handle the TRACE_PRINT entries from the function grapg tracer and output them as a C comment just below the function that called it, as if it was a comment inside this function. Example with an ftrace_printk inside might_sleep() function: void __might_sleep(char *file, int line) { static unsigned long prev_jiffy; /* ratelimiting */ ftrace_printk("Hi I'm a comment in might_sleep() :-)"); A chunk of a resulting trace: 0) | _reiserfs_free_block() { 0) | reiserfs_read_bitmap_block() { 0) | __bread() { 0) | __getblk() { 0) | __find_get_block() { 0) 0.698 us | mark_page_accessed(); 0) 2.267 us | } 0) | __might_sleep() { 0) | /* Hi I'm a comment in might_sleep() :-) */ 0) 1.321 us | } 0) 5.872 us | } 0) 7.313 us | } 0) 8.718 us | } And this patch brings two minor fixes: - The newline after a switch-out task has disappeared - The "|" sign just before the cpu number on task-switch has been deleted. 0) 0.616 us | pick_next_task_rt(); 0) 1.457 us | _spin_trylock(); 0) 0.653 us | _spin_unlock(); 0) 0.728 us | _spin_trylock(); 0) 0.631 us | _spin_unlock(); 0) 0.729 us | native_load_sp0(); 0) 0.593 us | native_load_tls(); ------------------------------------------ 0) cat-2834 => migrati-3 ------------------------------------------ 0) | finish_task_switch() { 0) 0.841 us | _spin_unlock_irq(); 0) 0.616 us | post_schedule_rt(); 0) 3.882 us | } Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-12-03 22:45:11 +00:00
ret = trace_seq_printf(s, " */\n");
if (!ret)
return TRACE_TYPE_PARTIAL_LINE;
return TRACE_TYPE_HANDLED;
}
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
enum print_line_t
print_graph_function(struct trace_iterator *iter)
{
struct trace_entry *entry = iter->ent;
struct trace_seq *s = &iter->seq;
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
switch (entry->type) {
case TRACE_GRAPH_ENT: {
struct ftrace_graph_ent_entry *field;
trace_assign_type(field, entry);
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
return print_graph_entry(field, s, iter);
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
}
case TRACE_GRAPH_RET: {
struct ftrace_graph_ret_entry *field;
trace_assign_type(field, entry);
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
return print_graph_return(&field->ret, s, entry, iter);
tracing/function-return-tracer: set a more human readable output Impact: feature This patch sets a C-like output for the function graph tracing. For this aim, we now call two handler for each function: one on the entry and one other on return. This way we can draw a well-ordered call stack. The pid of the previous trace is loosely stored to be compared against the one of the current trace to see if there were a context switch. Without this little feature, the call tree would seem broken at some locations. We could use the sched_tracer to capture these sched_events but this way of processing is much more simpler. 2 spaces have been chosen for indentation to fit the screen while deep calls. The time of execution in nanosecs is printed just after closed braces, it seems more easy this way to find the corresponding function. If the time was printed as a first column, it would be not so easy to find the corresponding function if it is called on a deep depth. I plan to output the return value but on 32 bits CPU, the return value can be 32 or 64, and its difficult to guess on which case we are. I don't know what would be the better solution on X86-32: only print eax (low-part) or even edx (high-part). Actually it's thee same problem when a function return a 8 bits value, the high part of eax could contain junk values... Here is an example of trace: sys_read() { fget_light() { } 526 vfs_read() { rw_verify_area() { security_file_permission() { cap_file_permission() { } 519 } 1564 } 2640 do_sync_read() { pipe_read() { __might_sleep() { } 511 pipe_wait() { prepare_to_wait() { } 760 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { update_min_vruntime() { } 504 } 1587 clear_buddies() { } 512 add_cfs_task_weight() { } 519 update_min_vruntime() { } 511 } 5602 dequeue_entity() { update_curr() { update_min_vruntime() { } 496 } 1631 clear_buddies() { } 496 update_min_vruntime() { } 527 } 4580 hrtick_update() { hrtick_start_fair() { } 488 } 1489 } 13700 } 14949 } 16016 msecs_to_jiffies() { } 496 put_prev_task_fair() { } 504 pick_next_task_fair() { } 489 pick_next_task_rt() { } 496 pick_next_task_fair() { } 489 pick_next_task_idle() { } 489 ------------8<---------- thread 4 ------------8<---------- finish_task_switch() { } 1203 do_softirq() { __do_softirq() { __local_bh_disable() { } 669 rcu_process_callbacks() { __rcu_process_callbacks() { cpu_quiet() { rcu_start_batch() { } 503 } 1647 } 3128 __rcu_process_callbacks() { } 542 } 5362 _local_bh_enable() { } 587 } 8880 } 9986 kthread_should_stop() { } 669 deactivate_task() { dequeue_task() { dequeue_task_fair() { dequeue_entity() { update_curr() { calc_delta_mine() { } 511 update_min_vruntime() { } 511 } 2813 Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-11-25 23:57:25 +00:00
}
default:
return print_graph_comment(s, entry, iter);
}
return TRACE_TYPE_HANDLED;
}
static void print_graph_headers(struct seq_file *s)
{
/* 1st line */
seq_printf(s, "# ");
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME)
seq_printf(s, " TIME ");
if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU)
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
seq_printf(s, "CPU");
if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC)
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
seq_printf(s, " TASK/PID ");
if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
seq_printf(s, " DURATION ");
seq_printf(s, " FUNCTION CALLS\n");
/* 2nd line */
seq_printf(s, "# ");
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
if (tracer_flags.val & TRACE_GRAPH_PRINT_ABS_TIME)
seq_printf(s, " | ");
if (tracer_flags.val & TRACE_GRAPH_PRINT_CPU)
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
seq_printf(s, "| ");
if (tracer_flags.val & TRACE_GRAPH_PRINT_PROC)
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
seq_printf(s, " | | ");
if (tracer_flags.val & TRACE_GRAPH_PRINT_DURATION)
seq_printf(s, " | | ");
seq_printf(s, " | | | |\n");
}
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
static void graph_trace_open(struct trace_iterator *iter)
{
/* pid and depth on the last trace processed */
struct fgraph_data *data = alloc_percpu(struct fgraph_data);
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
int cpu;
if (!data)
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
pr_warning("function graph tracer: not enough memory\n");
else
for_each_possible_cpu(cpu) {
pid_t *pid = &(per_cpu_ptr(data, cpu)->last_pid);
int *depth = &(per_cpu_ptr(data, cpu)->depth);
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
*pid = -1;
*depth = 0;
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
}
iter->private = data;
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
}
static void graph_trace_close(struct trace_iterator *iter)
{
free_percpu(iter->private);
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
}
static struct tracer graph_trace __read_mostly = {
.name = "function_graph",
tracing/function-graph-tracer: various fixes and features This patch brings various bugfixes: - Drop the first irrelevant task switch on the very beginning of a trace. - Drop the OVERHEAD word from the headers, the DURATION word is sufficient and will not overlap other columns. - Make the headers fit well their respective columns whatever the selected options. Ie, default options: # tracer: function_graph # # CPU DURATION FUNCTION CALLS # | | | | | | | 1) 0.646 us | } 1) | mem_cgroup_del_lru_list() { 1) 0.624 us | lookup_page_cgroup(); 1) 1.970 us | } echo funcgraph-proc > trace_options # tracer: function_graph # # CPU TASK/PID DURATION FUNCTION CALLS # | | | | | | | | | 0) bash-2937 | 0.895 us | } 0) bash-2937 | 0.888 us | __rcu_read_unlock(); 0) bash-2937 | 0.864 us | conv_uni_to_pc(); 0) bash-2937 | 1.015 us | __rcu_read_lock(); echo nofuncgraph-cpu > trace_options echo nofuncgraph-proc > trace_options # tracer: function_graph # # DURATION FUNCTION CALLS # | | | | | | 3.752 us | native_pud_val(); 0.616 us | native_pud_val(); 0.624 us | native_pmd_val(); About features, one can now disable the duration (this will hide the overhead too for convenient reasons and because on doesn't need overhead if it hasn't the duration): echo nofuncgraph-duration > trace_options # tracer: function_graph # # FUNCTION CALLS # | | | | cap_vm_enough_memory() { __vm_enough_memory() { vm_acct_memory(); } } } And at last, an option to print the absolute time: //Restart from default options echo funcgraph-abstime > trace_options # tracer: function_graph # # TIME CPU DURATION FUNCTION CALLS # | | | | | | | | 261.339774 | 1) + 42.823 us | } 261.339775 | 1) 1.045 us | _spin_lock_irq(); 261.339777 | 1) 0.940 us | _spin_lock_irqsave(); 261.339778 | 1) 0.752 us | _spin_unlock_irqrestore(); 261.339780 | 1) 0.857 us | _spin_unlock_irq(); 261.339782 | 1) | flush_to_ldisc() { 261.339783 | 1) | tty_ldisc_ref() { 261.339783 | 1) | tty_ldisc_try() { 261.339784 | 1) 1.075 us | _spin_lock_irqsave(); 261.339786 | 1) 0.842 us | _spin_unlock_irqrestore(); 261.339788 | 1) 4.211 us | } 261.339788 | 1) 5.662 us | } The format is seconds.usecs. I guess no one needs the nanosec precision here, the main goal is to have an overview about the general timings of events, and to see the place when the trace switches from one cpu to another. ie: 274.874760 | 1) 0.676 us | _spin_unlock(); 274.874762 | 1) 0.609 us | native_load_sp0(); 274.874763 | 1) 0.602 us | native_load_tls(); 274.878739 | 0) 0.722 us | } 274.878740 | 0) 0.714 us | native_pmd_val(); 274.878741 | 0) 0.730 us | native_pmd_val(); Here there is a 4000 usecs difference when we switch the cpu. Changes in V2: - Completely fix the first pointless task switch. Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-01-23 01:04:53 +00:00
.open = graph_trace_open,
.close = graph_trace_close,
.wait_pipe = poll_wait_pipe,
.init = graph_trace_init,
.reset = graph_trace_reset,
.print_line = print_graph_function,
.print_header = print_graph_headers,
.flags = &tracer_flags,
#ifdef CONFIG_FTRACE_SELFTEST
.selftest = trace_selftest_startup_function_graph,
#endif
};
static __init int init_graph_trace(void)
{
return register_tracer(&graph_trace);
}
device_initcall(init_graph_trace);