Create a new config option under the RCU menu that put
CPUs under RCU extended quiescent state (as in dynticks
idle mode) when they run in userspace. This require
some contribution from architectures to hook into kernel
and userspace boundaries.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Alessio Igor Bogani <abogani@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Avi Kivity <avi@redhat.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Geoff Levand <geoff@infradead.org>
Cc: Gilad Ben Yossef <gilad@benyossef.com>
Cc: Hakan Akkan <hakanakkan@gmail.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Josh Triplett <josh@joshtriplett.org>
Cc: Kevin Hilman <khilman@ti.com>
Cc: Max Krasnyansky <maxk@qualcomm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sven-Thorsten Dietrich <thebigcorporation@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The current implementation of RCU_FAST_NO_HZ tries reasonably hard to rid
the current CPU of RCU callbacks. This is appropriate when the CPU is
entering idle, where it doesn't have much useful to do anyway, but is most
definitely not what you want when transitioning to user-mode execution.
This commit therefore detects the adaptive-tick case, and refrains from
burning CPU time getting rid of RCU callbacks in that case.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
In some cases, it is necessary to enter or exit userspace-RCU-idle mode
from an interrupt handler, for example, if some other CPU sends this
CPU a resched IPI. In this case, the current CPU would enter the IPI
handler in userspace-RCU-idle mode, but would need to exit the IPI handler
after having exited that mode.
To allow this to work, this commit adds two new APIs to TREE_RCU:
- rcu_user_enter_after_irq(). This must be called from an interrupt between
rcu_irq_enter() and rcu_irq_exit(). After the irq calls rcu_irq_exit(),
the irq handler will return into an RCU extended quiescent state.
In theory, this interrupt is never a nested interrupt, but in practice
it might interrupt softirq, which looks to RCU like a nested interrupt.
- rcu_user_exit_after_irq(). This must be called from a non-nesting
interrupt, interrupting an RCU extended quiescent state, also
between rcu_irq_enter() and rcu_irq_exit(). After the irq calls
rcu_irq_exit(), the irq handler will return in an RCU non-quiescent
state.
[ Combined with "Allow calls to rcu_exit_user_irq from nesting irqs." ]
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
RCU currently insists that only idle tasks can enter RCU idle mode, which
prohibits an adaptive tickless kernel (AKA nohz cpusets), which in turn
would mean that usermode execution would always take scheduling-clock
interrupts, even when there is only one task runnable on the CPU in
question.
This commit therefore adds rcu_user_enter() and rcu_user_exit(), which
allow non-idle tasks to enter RCU idle mode. These are quite similar
to rcu_idle_enter() and rcu_idle_exit(), respectively, except that they
omit the idle-task checks.
[ Updated to use "user" flag rather than separate check functions. ]
[ paulmck: Updated to drop exports of new functions based on Josh's patch
getting rid of the need for them. ]
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Alessio Igor Bogani <abogani@kernel.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Avi Kivity <avi@redhat.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Geoff Levand <geoff@infradead.org>
Cc: Gilad Ben Yossef <gilad@benyossef.com>
Cc: Hakan Akkan <hakanakkan@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Kevin Hilman <khilman@ti.com>
Cc: Max Krasnyansky <maxk@qualcomm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Sven-Thorsten Dietrich <thebigcorporation@gmail.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The conflicts between kernel/rcutree.h and kernel/rcutree_plugin.h
were due to adjacent insertions and deletions, which were resolved
by simply accepting the changes on both branches.
bigrt.2012.09.23a contains additional commits to reduce scheduling latency
from RCU on huge systems (many hundrends or thousands of CPUs).
doctorture.2012.09.23a contains documentation changes and rcutorture fixes.
fixes.2012.09.23a contains miscellaneous fixes.
hotplug.2012.09.23a contains CPU-hotplug-related changes.
idle.2012.09.23a fixes architectures for which RCU no longer considered
the idle loop to be a quiescent state due to earlier
adaptive-dynticks changes. Affected architectures are alpha,
cris, frv, h8300, m32r, m68k, mn10300, parisc, score, xtensa,
and ia64.
Rabik and Paul reported two different issues related to the same few
lines of code.
Rabik's issue is that the nr_uninterruptible migration code is wrong in
that he sees artifacts due to this (Rabik please do expand in more
detail).
Paul's issue is that this code as it stands relies on us using
stop_machine() for unplug, we all would like to remove this assumption
so that eventually we can remove this stop_machine() usage altogether.
The only reason we'd have to migrate nr_uninterruptible is so that we
could use for_each_online_cpu() loops in favour of
for_each_possible_cpu() loops, however since nr_uninterruptible() is the
only such loop and its using possible lets not bother at all.
The problem Rabik sees is (probably) caused by the fact that by
migrating nr_uninterruptible we screw rq->calc_load_active for both rqs
involved.
So don't bother with fancy migration schemes (meaning we now have to
keep using for_each_possible_cpu()) and instead fold any nr_active delta
after we migrate all tasks away to make sure we don't have any skewed
nr_active accounting.
[ paulmck: Move call to calc_load_migration to CPU_DEAD to avoid
miscounting noted by Rakib. ]
Reported-by: Rakib Mullick <rakib.mullick@gmail.com>
Reported-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Posting a callback after the CPU_DEAD notifier effectively leaks
that callback unless/until that CPU comes back online. Silence is
unhelpful when attempting to track down such leaks, so this commit emits
a WARN_ON_ONCE() and unconditionally leaks the callback when an offline
CPU attempts to register a callback. The rdp->nxttail[RCU_NEXT_TAIL] is
set to NULL in the CPU_DEAD notifier and restored in the CPU_UP_PREPARE
notifier, allowing _call_rcu() to determine exactly when posting callbacks
is illegal.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Currently, _rcu_barrier() relies on preempt_disable() to prevent
any CPU from going offline, which in turn depends on CPU hotplug's
use of __stop_machine().
This patch therefore makes _rcu_barrier() use get_online_cpus() to
block CPU-hotplug operations. This has the added benefit of removing
the need for _rcu_barrier() to adopt callbacks: Because CPU-hotplug
operations are excluded, there can be no callbacks to adopt. This
commit simplifies the code accordingly.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The print_cpu_stall_fast_no_hz() function attempts to print -1 when
the ->idle_gp_timer is not pending, but unsigned arithmetic causes it
to instead print ULONG_MAX, which is 4294967295 on 32-bit systems and
18446744073709551615 on 64-bit systems. Neither of these are the most
reader-friendly values, so this commit instead causes "timer not pending"
to be printed when ->idle_gp_timer is not pending.
Reported-by: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
TINY_RCU's rcu_idle_enter_common() invokes rcu_sched_qs() in order
to inform the RCU core of the quiescent state implied by idle entry.
Of course, idle is also an extended quiescent state, so that the call
to rcu_sched_qs() speeds up RCU's invoking of any callbacks that might
be queued. This speed-up is important when entering into dyntick-idle
mode -- if there are no further scheduling-clock interrupts, the callbacks
might never be invoked, which could result in a system hang.
However, processing callbacks does event tracing, which in turn
implies RCU read-side critical sections, which are illegal in extended
quiescent states. This patch therefore moves the call to rcu_sched_qs()
so that it precedes the point at which we inform lockdep that RCU has
entered an extended quiescent state.
Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The can_stop_idle_tick() function complains if a softirq vector is
raised too late in the idle-entry process, presumably in order to
prevent dangling softirq invocations from being delayed across the
full idle period, which might be indefinitely long -- and if softirq
was asserted any later than the call to this function, such a delay
might well happen.
However, RCU needs to be able to use softirq to stop idle entry in
order to be able to drain RCU callbacks from the current CPU, which in
turn enables faster entry into dyntick-idle mode, which in turn reduces
power consumption. Because RCU takes this action at a well-defined
point in the idle-entry path, it is safe for RCU to take this approach.
This commit therefore silences the error message that is sometimes
produced when the going-idle CPU suddenly finds that it has an RCU_SOFTIRQ
to process. The error message will continue to be issued for other
softirq vectors.
Reported-by: Sedat Dilek <sedat.dilek@gmail.com>
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The use of raw_local_irq_save() is unnecessary, given that local_irq_save()
really does disable interrupts. Also, it appears to interfere with lockdep.
Therefore, this commit moves to local_irq_save().
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Tested-by: Fengguang Wu <fengguang.wu@intel.com>
The first memory barrier in __call_rcu() is supposed to order any
updates done beforehand by the caller against the actual queuing
of the callback. However, the second memory barrier (which is intended
to order incrementing the queue lengths before queuing the callback)
is also between the caller's updates and the queuing of the callback.
The second memory barrier can therefore serve both purposes.
This commit therefore removes the first memory barrier.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
If a given CPU avoids the idle loop but also avoids starting a new
RCU grace period for a full minute, RCU can issue spurious RCU CPU
stall warnings. This commit fixes this issue by adding a check for
ongoing grace period to avoid these spurious stall warnings.
Reported-by: Becky Bruce <bgillbruce@gmail.com>
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The print_other_cpu_stall() function accesses a number of rcu_node
fields without protection from the ->lock. In theory, this is not
a problem because the fields accessed are all integers, but in
practice the compiler can get nasty. Therefore, the commit extends
the existing critical section to cover the entire loop body.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The rcu_print_detail_task_stall_rnp() function invokes
rcu_preempt_blocked_readers_cgp() to verify that there are some preempted
RCU readers blocking the current grace period outside of the protection
of the rcu_node structure's ->lock. This means that the last blocked
reader might exit its RCU read-side critical section and remove itself
from the ->blkd_tasks list before the ->lock is acquired, resulting in
a segmentation fault when the subsequent code attempts to dereference
the now-NULL gp_tasks pointer.
This commit therefore moves the test under the lock. This will not
have measurable effect on lock contention because this code is invoked
only when printing RCU CPU stall warnings, in other words, in the common
case, never.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The increment_cpu_stall_ticks() function listed each RCU flavor
explicitly, with an ifdef to handle preemptible RCU. This commit
therefore applies for_each_rcu_flavor() to save a line of code.
Because this commit switches from a code-based enumeration of the
flavors of RCU to an rcu_state-list-based enumeration, it is no longer
possible to apply __get_cpu_var() to the per-CPU rcu_data structures.
We instead use __this_cpu_var() on the rcu_state structure's ->rda field
that references the corresponding rcu_data structures.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Commit 1217ed1b (rcu: permit rcu_read_unlock() to be called while holding
runqueue locks) made rcu_initiate_boost() restore irq state when releasing
the rcu_node structure's ->lock, but failed to update the header comment
accordingly. This commit therefore brings the header comment up to date.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The rcu_implicit_offline_qs() function implicitly assumed that execution
would progress predictably when interrupts are disabled, which is of course
not guaranteed when running on a hypervisor. Furthermore, this function
is short, and is called from one place only in a short function.
This commit therefore ensures that the timing is checked before
checking the condition, which guarantees correct behavior even given
indefinite delays. It also inlines rcu_implicit_offline_qs() into
rcu_implicit_dynticks_qs().
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The rcu_preempt_offline_tasks() moves all tasks queued on a given leaf
rcu_node structure to the root rcu_node, which is done when the last CPU
corresponding the the leaf rcu_node structure goes offline. Now that
RCU-preempt's synchronize_rcu_expedited() implementation blocks CPU-hotplug
operations during the initialization of each rcu_node structure's
->boost_tasks pointer, rcu_preempt_offline_tasks() can do a better job
of setting the root rcu_node's ->boost_tasks pointer.
The key point is that rcu_preempt_offline_tasks() runs as part of the
CPU-hotplug process, so that a concurrent synchronize_rcu_expedited()
is guaranteed to either have not started on the one hand (in which case
there is no boosting on behalf of the expedited grace period) or to be
completely initialized on the other (in which case, in the absence of
other priority boosting, all ->boost_tasks pointers will be initialized).
Therefore, if rcu_preempt_offline_tasks() finds that the ->boost_tasks
pointer is equal to the ->exp_tasks pointer, it can be sure that it is
correctly placed.
In the case where there was boosting ongoing at the time that the
synchronize_rcu_expedited() function started, different nodes might start
boosting the tasks blocking the expedited grace period at different times.
In this mixed case, the root node will either be boosting tasks for
the expedited grace period already, or it will start as soon as it gets
done boosting for the normal grace period -- but in this latter case,
the root node's tasks needed to be boosted in any case.
This commit therefore adds a check of the ->boost_tasks pointer against
the ->exp_tasks pointer to the list that prevents updating ->boost_tasks.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
There is a need to use RCU from interrupt context, but either before
rcu_irq_enter() is called or after rcu_irq_exit() is called. If the
interrupt occurs from idle, then lockdep-RCU will complain about such
uses, as they appear to be illegal uses of RCU from the idle loop.
In other environments, RCU_NONIDLE() could be used to properly protect
the use of RCU, but RCU_NONIDLE() currently cannot be invoked except
from process context.
This commit therefore modifies RCU_NONIDLE() to permit its use more
globally.
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
When rcu_preempt_offline_tasks() clears tasks from a leaf rcu_node
structure, it does not NULL out the structure's ->boost_tasks field.
This commit therefore fixes this issue.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Because TINY_RCU's idle detection keys directly off of the nesting
level, rather than from a separate variable as in TREE_RCU, the
TINY_RCU dyntick-idle tracing on transition to idle must happen
before the change to the nesting level. This commit therefore makes
this change by passing the desired new value (rather than the old value)
of the nesting level in to rcu_idle_enter_common().
[ paulmck: Add fix for wrong-variable bug spotted by
Michael Wang <wangyun@linux.vnet.ibm.com>. ]
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
There have been some recent bugs that were triggered only when
preemptible RCU's __rcu_read_unlock() was preempted just after setting
->rcu_read_lock_nesting to INT_MIN, which is a low-probability event.
Therefore, reproducing those bugs (to say nothing of gaining confidence
in alleged fixes) was quite difficult. This commit therefore creates
a new debug-only RCU kernel config option that forces a short delay
in __rcu_read_unlock() to increase the probability of those sorts of
bugs occurring.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
When you do something like "t = kthread_run(...)", it is possible that
the kthread will start running before the assignment to "t" happens.
If the child kthread expects to find a pointer to its task_struct in "t",
it will then be fatally disappointed. This commit therefore switches
such cases to kthread_create() followed by wake_up_process(), guaranteeing
that the assignment happens before the child kthread starts running.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Drop a few characters by switching kernel/rcutorture.c from
"printk(KERN_ALERT" to "pr_alert(".
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Many rcutorture runs include CPU-hotplug operations in their stress
testing. This commit accumulates statistics on the durations of these
operations in deference to the recent concern about the overhead and
latency of these operations.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
A number of new features have been added to rcutorture over the years, but
the defaults have not been updated to include them. This commit therefore
turns on a couple of them that have proven helpful and trustworthy, namely
periodic progress reports and testing of NO_HZ.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Currently, rcu_init_geometry() only reshapes RCU's combining trees
if the leaf fanout is changed at boot time. This means that by
default, kernels compiled with (say) NR_CPUS=4096 will keep oversized
data structures, even when running on systems with (say) four CPUs.
This commit therefore checks to see if the maximum number of CPUs on
the actual running system (nr_cpu_ids) differs from NR_CPUS, and if so
reshapes the combining trees accordingly.
Reported-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
If CONFIG_RCU_FANOUT_EXACT=y, if there are not enough CPUs (according
to nr_cpu_ids) to require more than a single rcu_node structure, but if
NR_CPUS is larger than would fit into a single rcu_node structure, then
the current rcu_init_levelspread() code is subject to integer overflow
in the eight-bit ->levelspread[] array in the rcu_state structure.
In this case, the solution is -not- to increase the size of the
elements in this array because the values in that array should be
constrained to the number of bits in an unsigned long. Instead, this
commit replaces NR_CPUS with nr_cpu_ids in the rcu_init_levelspread()
function's initialization of the cprv local variable. This results in
all of the arithmetic being consistently based off of the nr_cpu_ids
value, thus avoiding the overflow, which was caused by the mixing of
nr_cpu_ids and NR_CPUS.
Reported-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The current quiescent-state detection algorithm is needlessly
complex. It records the grace-period number corresponding to
the quiescent state at the time of the quiescent state, which
works, but it seems better to simply erase any record of previous
quiescent states at the time that the CPU notices the new grace
period. This has the further advantage of removing another piece
of RCU for which lockless reasoning is required.
Therefore, this commit makes this change.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The synchronize_rcu_expedited() function disables interrupts across a
scan of all leaf rcu_node structures, which is not good for real-time
scheduling latency on large systems (hundreds or especially thousands
of CPUs). This commit therefore holds off CPU-hotplug operations using
get_online_cpus(), and removes the prior acquisiion of the ->onofflock
(which required disabling interrupts).
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
In the C language, signed overflow is undefined. It is true that
twos-complement arithmetic normally comes to the rescue, but if the
compiler can subvert this any time it has any information about the values
being compared. For example, given "if (a - b > 0)", if the compiler
has enough information to realize that (for example) the value of "a"
is positive and that of "b" is negative, the compiler is within its
rights to optimize to a simple "if (1)", which might not be what you want.
This commit therefore converts synchronize_rcu_expedited()'s work-done
detection counter from signed to unsigned.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Now that the rcu_node structures' ->completed fields are unconditionally
assigned at grace-period cleanup time, they should already have the
correct value for the new grace period at grace-period initialization
time. This commit therefore inserts a WARN_ON_ONCE() to verify this
invariant.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Preemption greatly raised the probability of certain types of race
conditions, so this commit adds an anti-heisenbug to greatly increase
the collision cross section, also known as the probability of occurrence.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
The current approach to grace-period initialization is vulnerable to
extremely low-probability races. These races stem from the fact that
the old grace period is marked completed on the same traversal through
the rcu_node structure that is marking the start of the new grace period.
This means that some rcu_node structures will believe that the old grace
period is still in effect at the same time that other rcu_node structures
believe that the new grace period has already started.
These sorts of disagreements can result in too-short grace periods,
as shown in the following scenario:
1. CPU 0 completes a grace period, but needs an additional
grace period, so starts initializing one, initializing all
the non-leaf rcu_node structures and the first leaf rcu_node
structure. Because CPU 0 is both completing the old grace
period and starting a new one, it marks the completion of
the old grace period and the start of the new grace period
in a single traversal of the rcu_node structures.
Therefore, CPUs corresponding to the first rcu_node structure
can become aware that the prior grace period has completed, but
CPUs corresponding to the other rcu_node structures will see
this same prior grace period as still being in progress.
2. CPU 1 passes through a quiescent state, and therefore informs
the RCU core. Because its leaf rcu_node structure has already
been initialized, this CPU's quiescent state is applied to the
new (and only partially initialized) grace period.
3. CPU 1 enters an RCU read-side critical section and acquires
a reference to data item A. Note that this CPU believes that
its critical section started after the beginning of the new
grace period, and therefore will not block this new grace period.
4. CPU 16 exits dyntick-idle mode. Because it was in dyntick-idle
mode, other CPUs informed the RCU core of its extended quiescent
state for the past several grace periods. This means that CPU 16
is not yet aware that these past grace periods have ended. Assume
that CPU 16 corresponds to the second leaf rcu_node structure --
which has not yet been made aware of the new grace period.
5. CPU 16 removes data item A from its enclosing data structure
and passes it to call_rcu(), which queues a callback in the
RCU_NEXT_TAIL segment of the callback queue.
6. CPU 16 enters the RCU core, possibly because it has taken a
scheduling-clock interrupt, or alternatively because it has
more than 10,000 callbacks queued. It notes that the second
most recent grace period has completed (recall that because it
corresponds to the second as-yet-uninitialized rcu_node structure,
it cannot yet become aware that the most recent grace period has
completed), and therefore advances its callbacks. The callback
for data item A is therefore in the RCU_NEXT_READY_TAIL segment
of the callback queue.
7. CPU 0 completes initialization of the remaining leaf rcu_node
structures for the new grace period, including the structure
corresponding to CPU 16.
8. CPU 16 again enters the RCU core, again, possibly because it has
taken a scheduling-clock interrupt, or alternatively because
it now has more than 10,000 callbacks queued. It notes that
the most recent grace period has ended, and therefore advances
its callbacks. The callback for data item A is therefore in
the RCU_DONE_TAIL segment of the callback queue.
9. All CPUs other than CPU 1 pass through quiescent states. Because
CPU 1 already passed through its quiescent state, the new grace
period completes. Note that CPU 1 is still in its RCU read-side
critical section, still referencing data item A.
10. Suppose that CPU 2 wais the last CPU to pass through a quiescent
state for the new grace period, and suppose further that CPU 2
did not have any callbacks queued, therefore not needing an
additional grace period. CPU 2 therefore traverses all of the
rcu_node structures, marking the new grace period as completed,
but does not initialize a new grace period.
11. CPU 16 yet again enters the RCU core, yet again possibly because
it has taken a scheduling-clock interrupt, or alternatively
because it now has more than 10,000 callbacks queued. It notes
that the new grace period has ended, and therefore advances
its callbacks. The callback for data item A is therefore in
the RCU_DONE_TAIL segment of the callback queue. This means
that this callback is now considered ready to be invoked.
12. CPU 16 invokes the callback, freeing data item A while CPU 1
is still referencing it.
This scenario represents a day-zero bug for TREE_RCU. This commit
therefore ensures that the old grace period is marked completed in
all leaf rcu_node structures before a new grace period is marked
started in any of them.
That said, it would have been insanely difficult to force this race to
happen before the grace-period initialization process was preemptible.
Therefore, this commit is not a candidate for -stable.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Conflicts:
kernel/rcutree.c
The module parameters blimit, qhimark, and qlomark (and more
recently, rcu_fanout_leaf) have permission masks of zero, so
that their values are not visible from sysfs. This is unnecessary
and inconvenient to administrators who might like an easy way to
see what these values are on a running system. This commit therefore
sets their permission masks to 0444, allowing them to be read but
not written.
Reported-by: Rusty Russell <rusty@ozlabs.org>
Reported-by: Josh Triplett <josh@joshtriplett.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Although almost everyone is well-served by the defaults, some uses of RCU
benefit from shorter grace periods, while others benefit more from the
greater efficiency provided by longer grace periods. Situations requiring
a large number of grace periods to elapse (and wireshark startup has
been called out as an example of this) are helped by lower-latency
grace periods. Furthermore, in some embedded applications, people are
willing to accept a small degradation in update efficiency (due to there
being more of the shorter grace-period operations) in order to gain the
lower latency.
In contrast, those few systems with thousands of CPUs need longer grace
periods because the CPU overhead of a grace period rises roughly
linearly with the number of CPUs. Such systems normally do not make
much use of facilities that require large numbers of grace periods to
elapse, so this is a good tradeoff.
Therefore, this commit allows the durations to be controlled from sysfs.
There are two sysfs parameters, one named "jiffies_till_first_fqs" that
specifies the delay in jiffies from the end of grace-period initialization
until the first attempt to force quiescent states, and the other named
"jiffies_till_next_fqs" that specifies the delay (again in jiffies)
between subsequent attempts to force quiescent states. They both default
to three jiffies, which is compatible with the old hard-coded behavior.
At some future time, it may be possible to automatically increase the
grace-period length with the number of CPUs, but we do not yet have
sufficient data to do a good job. Preliminary data indicates that we
should add an addiitonal jiffy to each of the delays for every 200 CPUs
in the system, but more experimentation is needed. For now, the number
of systems with more than 1,000 CPUs is small enough that this can be
relegated to boot-time hand tuning.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Large systems running RCU_FAST_NO_HZ kernels see extreme memory
contention on the rcu_state structure's ->fqslock field. This
can be avoided by disabling RCU_FAST_NO_HZ, either at compile time
or at boot time (via the nohz kernel boot parameter), but large
systems will no doubt become sensitive to energy consumption.
This commit therefore uses a combining-tree approach to spread the
memory contention across new cache lines in the leaf rcu_node structures.
This can be thought of as a tournament lock that has only a try-lock
acquisition primitive.
The effect on small systems is minimal, because such systems have
an rcu_node "tree" consisting of a single node. In addition, this
functionality is not used on fastpaths.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Moving quiescent-state forcing into a kthread dispenses with the need
for the ->n_rp_need_fqs field, so this commit removes it.
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
RCU quiescent-state forcing is currently carried out without preemption
points, which can result in excessive latency spikes on large systems
(many hundreds or thousands of CPUs). This patch therefore inserts
a voluntary preemption point into force_qs_rnp(), which should greatly
reduce the magnitude of these spikes.
Reported-by: Mike Galbraith <mgalbraith@suse.de>
Reported-by: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
As the first step towards allowing quiescent-state forcing to be
preemptible, this commit moves RCU quiescent-state forcing into the
same kthread that is now used to initialize and clean up after grace
periods. This is yet another step towards keeping scheduling
latency down to a dull roar.
Updated to change from raw_spin_lock_irqsave() to raw_spin_lock_irq()
and to remove the now-unused rcu_state structure fields as suggested by
Peter Zijlstra.
Reported-by: Mike Galbraith <mgalbraith@suse.de>
Reported-by: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
The fields in the rcu_state structure that are protected by the
root rcu_node structure's ->lock can share a cache line with the
fields protected by ->onofflock. This can result in excessive
memory contention on large systems, so this commit applies
____cacheline_internodealigned_in_smp to the ->onofflock field in
order to segregate them.
Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Tested-by: Dimitri Sivanich <sivanich@sgi.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
In kernels built with CONFIG_RCU_FAST_NO_HZ=y, CPUs can accumulate a
large number of lazy callbacks, which as the name implies will be slow
to be invoked. This can be a problem on small-memory systems, where the
default 6-second sleep for CPUs having only lazy RCU callbacks could well
be fatal. This commit therefore installs an OOM hander that ensures that
every CPU with lazy callbacks has at least one non-lazy callback, in turn
ensuring timely advancement for these callbacks.
Updated to fix bug that disabled OOM killing, noted by Lai Jiangshan.
Updated to push the for_each_rcu_flavor() loop into rcu_oom_notify_cpu(),
thus reducing the number of IPIs, as suggested by Steven Rostedt. Also
to make the for_each_online_cpu() loop be preemptible. (Later, it might
be good to use smp_call_function(), as suggested by Peter Zijlstra.)
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Tested-by: Sasha Levin <levinsasha928@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Earlier versions of RCU invoked the RCU core from the CPU_DYING notifier
in order to note a quiescent state for the outgoing CPU. Because the
CPU is marked "offline" during the execution of the CPU_DYING notifiers,
the RCU core had to tolerate being invoked from an offline CPU. However,
commit b1420f1c (Make rcu_barrier() less disruptive) left only tracing
code in the CPU_DYING notifier, so the RCU core need no longer execute
on offline CPUs. This commit therefore enforces this restriction.
Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Then rcu_gp_kthread() function is too large and furthermore needs to
have the force_quiescent_state() code pulled in. This commit therefore
breaks up rcu_gp_kthread() into rcu_gp_init() and rcu_gp_cleanup().
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
RCU grace-period cleanup is currently carried out with interrupts
disabled, which can result in excessive latency spikes on large systems
(many hundreds or thousands of CPUs). This patch therefore makes the
RCU grace-period cleanup be preemptible, including voluntary preemption
points, which should eliminate those latency spikes. Similar spikes from
forcing of quiescent states will be dealt with similarly by later patches.
Updated to replace uses of spin_lock_irqsave() with spin_lock_irq(), as
suggested by Peter Zijlstra.
Reported-by: Mike Galbraith <mgalbraith@suse.de>
Reported-by: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
As a first step towards allowing grace-period cleanup to be preemptible,
this commit moves the RCU grace-period cleanup into the same kthread
that is now used to initialize grace periods. This is needed to keep
scheduling latency down to a dull roar.
[ paulmck: Get rid of stray spin_lock_irqsave() calls. ]
Reported-by: Mike Galbraith <mgalbraith@suse.de>
Reported-by: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Frederic Weisbecker