2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Copyright (C) 1995 Linus Torvalds
|
|
|
|
*
|
|
|
|
* Pentium III FXSR, SSE support
|
|
|
|
* Gareth Hughes <gareth@valinux.com>, May 2000
|
2008-01-30 12:31:03 +00:00
|
|
|
*
|
2005-04-16 22:20:36 +00:00
|
|
|
* X86-64 port
|
|
|
|
* Andi Kleen.
|
2005-06-25 21:55:00 +00:00
|
|
|
*
|
|
|
|
* CPU hotplug support - ashok.raj@intel.com
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This file handles the architecture-dependent parts of process handling..
|
|
|
|
*/
|
|
|
|
|
2008-02-14 08:44:08 +00:00
|
|
|
#include <linux/stackprotector.h>
|
2005-06-25 21:55:00 +00:00
|
|
|
#include <linux/cpu.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/errno.h>
|
|
|
|
#include <linux/sched.h>
|
2008-01-30 12:31:03 +00:00
|
|
|
#include <linux/fs.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/elfcore.h>
|
|
|
|
#include <linux/smp.h>
|
|
|
|
#include <linux/slab.h>
|
|
|
|
#include <linux/user.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/delay.h>
|
2008-01-30 12:31:03 +00:00
|
|
|
#include <linux/module.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
#include <linux/ptrace.h>
|
2006-01-11 21:44:36 +00:00
|
|
|
#include <linux/notifier.h>
|
2006-03-26 09:38:20 +00:00
|
|
|
#include <linux/kprobes.h>
|
2007-05-08 07:27:03 +00:00
|
|
|
#include <linux/kdebug.h>
|
2007-10-12 21:04:07 +00:00
|
|
|
#include <linux/tick.h>
|
2008-04-13 22:24:18 +00:00
|
|
|
#include <linux/prctl.h>
|
2008-07-29 05:48:51 +00:00
|
|
|
#include <linux/uaccess.h>
|
|
|
|
#include <linux/io.h>
|
2008-12-06 02:40:00 +00:00
|
|
|
#include <linux/ftrace.h>
|
2011-04-01 23:34:59 +00:00
|
|
|
#include <linux/cpuidle.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
#include <asm/pgtable.h>
|
|
|
|
#include <asm/system.h>
|
|
|
|
#include <asm/processor.h>
|
|
|
|
#include <asm/i387.h>
|
|
|
|
#include <asm/mmu_context.h>
|
|
|
|
#include <asm/prctl.h>
|
|
|
|
#include <asm/desc.h>
|
|
|
|
#include <asm/proto.h>
|
|
|
|
#include <asm/ia32.h>
|
2006-01-11 21:44:36 +00:00
|
|
|
#include <asm/idle.h>
|
2008-07-21 16:04:13 +00:00
|
|
|
#include <asm/syscalls.h>
|
2009-06-01 18:14:55 +00:00
|
|
|
#include <asm/debugreg.h>
|
2011-09-30 19:06:22 +00:00
|
|
|
#include <asm/nmi.h>
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
asmlinkage extern void ret_from_fork(void);
|
|
|
|
|
2009-01-18 15:38:58 +00:00
|
|
|
DEFINE_PER_CPU(unsigned long, old_rsp);
|
2009-01-18 15:38:59 +00:00
|
|
|
static DEFINE_PER_CPU(unsigned char, is_idle);
|
2009-01-18 15:38:58 +00:00
|
|
|
|
[PATCH] Notifier chain update: API changes
The kernel's implementation of notifier chains is unsafe. There is no
protection against entries being added to or removed from a chain while the
chain is in use. The issues were discussed in this thread:
http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2
We noticed that notifier chains in the kernel fall into two basic usage
classes:
"Blocking" chains are always called from a process context
and the callout routines are allowed to sleep;
"Atomic" chains can be called from an atomic context and
the callout routines are not allowed to sleep.
We decided to codify this distinction and make it part of the API. Therefore
this set of patches introduces three new, parallel APIs: one for blocking
notifiers, one for atomic notifiers, and one for "raw" notifiers (which is
really just the old API under a new name). New kinds of data structures are
used for the heads of the chains, and new routines are defined for
registration, unregistration, and calling a chain. The three APIs are
explained in include/linux/notifier.h and their implementation is in
kernel/sys.c.
With atomic and blocking chains, the implementation guarantees that the chain
links will not be corrupted and that chain callers will not get messed up by
entries being added or removed. For raw chains the implementation provides no
guarantees at all; users of this API must provide their own protections. (The
idea was that situations may come up where the assumptions of the atomic and
blocking APIs are not appropriate, so it should be possible for users to
handle these things in their own way.)
There are some limitations, which should not be too hard to live with. For
atomic/blocking chains, registration and unregistration must always be done in
a process context since the chain is protected by a mutex/rwsem. Also, a
callout routine for a non-raw chain must not try to register or unregister
entries on its own chain. (This did happen in a couple of places and the code
had to be changed to avoid it.)
Since atomic chains may be called from within an NMI handler, they cannot use
spinlocks for synchronization. Instead we use RCU. The overhead falls almost
entirely in the unregister routine, which is okay since unregistration is much
less frequent that calling a chain.
Here is the list of chains that we adjusted and their classifications. None
of them use the raw API, so for the moment it is only a placeholder.
ATOMIC CHAINS
-------------
arch/i386/kernel/traps.c: i386die_chain
arch/ia64/kernel/traps.c: ia64die_chain
arch/powerpc/kernel/traps.c: powerpc_die_chain
arch/sparc64/kernel/traps.c: sparc64die_chain
arch/x86_64/kernel/traps.c: die_chain
drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list
kernel/panic.c: panic_notifier_list
kernel/profile.c: task_free_notifier
net/bluetooth/hci_core.c: hci_notifier
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain
net/ipv6/addrconf.c: inet6addr_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain
net/netlink/af_netlink.c: netlink_chain
BLOCKING CHAINS
---------------
arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain
arch/s390/kernel/process.c: idle_chain
arch/x86_64/kernel/process.c idle_notifier
drivers/base/memory.c: memory_chain
drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list
drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list
drivers/macintosh/adb.c: adb_client_list
drivers/macintosh/via-pmu.c sleep_notifier_list
drivers/macintosh/via-pmu68k.c sleep_notifier_list
drivers/macintosh/windfarm_core.c wf_client_list
drivers/usb/core/notify.c usb_notifier_list
drivers/video/fbmem.c fb_notifier_list
kernel/cpu.c cpu_chain
kernel/module.c module_notify_list
kernel/profile.c munmap_notifier
kernel/profile.c task_exit_notifier
kernel/sys.c reboot_notifier_list
net/core/dev.c netdev_chain
net/decnet/dn_dev.c: dnaddr_chain
net/ipv4/devinet.c: inetaddr_chain
It's possible that some of these classifications are wrong. If they are,
please let us know or submit a patch to fix them. Note that any chain that
gets called very frequently should be atomic, because the rwsem read-locking
used for blocking chains is very likely to incur cache misses on SMP systems.
(However, if the chain's callout routines may sleep then the chain cannot be
atomic.)
The patch set was written by Alan Stern and Chandra Seetharaman, incorporating
material written by Keith Owens and suggestions from Paul McKenney and Andrew
Morton.
[jes@sgi.com: restructure the notifier chain initialization macros]
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 09:16:30 +00:00
|
|
|
static ATOMIC_NOTIFIER_HEAD(idle_notifier);
|
2006-01-11 21:44:36 +00:00
|
|
|
|
|
|
|
void idle_notifier_register(struct notifier_block *n)
|
|
|
|
{
|
[PATCH] Notifier chain update: API changes
The kernel's implementation of notifier chains is unsafe. There is no
protection against entries being added to or removed from a chain while the
chain is in use. The issues were discussed in this thread:
http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2
We noticed that notifier chains in the kernel fall into two basic usage
classes:
"Blocking" chains are always called from a process context
and the callout routines are allowed to sleep;
"Atomic" chains can be called from an atomic context and
the callout routines are not allowed to sleep.
We decided to codify this distinction and make it part of the API. Therefore
this set of patches introduces three new, parallel APIs: one for blocking
notifiers, one for atomic notifiers, and one for "raw" notifiers (which is
really just the old API under a new name). New kinds of data structures are
used for the heads of the chains, and new routines are defined for
registration, unregistration, and calling a chain. The three APIs are
explained in include/linux/notifier.h and their implementation is in
kernel/sys.c.
With atomic and blocking chains, the implementation guarantees that the chain
links will not be corrupted and that chain callers will not get messed up by
entries being added or removed. For raw chains the implementation provides no
guarantees at all; users of this API must provide their own protections. (The
idea was that situations may come up where the assumptions of the atomic and
blocking APIs are not appropriate, so it should be possible for users to
handle these things in their own way.)
There are some limitations, which should not be too hard to live with. For
atomic/blocking chains, registration and unregistration must always be done in
a process context since the chain is protected by a mutex/rwsem. Also, a
callout routine for a non-raw chain must not try to register or unregister
entries on its own chain. (This did happen in a couple of places and the code
had to be changed to avoid it.)
Since atomic chains may be called from within an NMI handler, they cannot use
spinlocks for synchronization. Instead we use RCU. The overhead falls almost
entirely in the unregister routine, which is okay since unregistration is much
less frequent that calling a chain.
Here is the list of chains that we adjusted and their classifications. None
of them use the raw API, so for the moment it is only a placeholder.
ATOMIC CHAINS
-------------
arch/i386/kernel/traps.c: i386die_chain
arch/ia64/kernel/traps.c: ia64die_chain
arch/powerpc/kernel/traps.c: powerpc_die_chain
arch/sparc64/kernel/traps.c: sparc64die_chain
arch/x86_64/kernel/traps.c: die_chain
drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list
kernel/panic.c: panic_notifier_list
kernel/profile.c: task_free_notifier
net/bluetooth/hci_core.c: hci_notifier
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain
net/ipv6/addrconf.c: inet6addr_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain
net/netlink/af_netlink.c: netlink_chain
BLOCKING CHAINS
---------------
arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain
arch/s390/kernel/process.c: idle_chain
arch/x86_64/kernel/process.c idle_notifier
drivers/base/memory.c: memory_chain
drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list
drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list
drivers/macintosh/adb.c: adb_client_list
drivers/macintosh/via-pmu.c sleep_notifier_list
drivers/macintosh/via-pmu68k.c sleep_notifier_list
drivers/macintosh/windfarm_core.c wf_client_list
drivers/usb/core/notify.c usb_notifier_list
drivers/video/fbmem.c fb_notifier_list
kernel/cpu.c cpu_chain
kernel/module.c module_notify_list
kernel/profile.c munmap_notifier
kernel/profile.c task_exit_notifier
kernel/sys.c reboot_notifier_list
net/core/dev.c netdev_chain
net/decnet/dn_dev.c: dnaddr_chain
net/ipv4/devinet.c: inetaddr_chain
It's possible that some of these classifications are wrong. If they are,
please let us know or submit a patch to fix them. Note that any chain that
gets called very frequently should be atomic, because the rwsem read-locking
used for blocking chains is very likely to incur cache misses on SMP systems.
(However, if the chain's callout routines may sleep then the chain cannot be
atomic.)
The patch set was written by Alan Stern and Chandra Seetharaman, incorporating
material written by Keith Owens and suggestions from Paul McKenney and Andrew
Morton.
[jes@sgi.com: restructure the notifier chain initialization macros]
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 09:16:30 +00:00
|
|
|
atomic_notifier_chain_register(&idle_notifier, n);
|
2006-01-11 21:44:36 +00:00
|
|
|
}
|
2008-10-16 20:34:43 +00:00
|
|
|
EXPORT_SYMBOL_GPL(idle_notifier_register);
|
|
|
|
|
|
|
|
void idle_notifier_unregister(struct notifier_block *n)
|
|
|
|
{
|
|
|
|
atomic_notifier_chain_unregister(&idle_notifier, n);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(idle_notifier_unregister);
|
2006-01-11 21:44:36 +00:00
|
|
|
|
|
|
|
void enter_idle(void)
|
|
|
|
{
|
2009-01-18 15:38:59 +00:00
|
|
|
percpu_write(is_idle, 1);
|
[PATCH] Notifier chain update: API changes
The kernel's implementation of notifier chains is unsafe. There is no
protection against entries being added to or removed from a chain while the
chain is in use. The issues were discussed in this thread:
http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2
We noticed that notifier chains in the kernel fall into two basic usage
classes:
"Blocking" chains are always called from a process context
and the callout routines are allowed to sleep;
"Atomic" chains can be called from an atomic context and
the callout routines are not allowed to sleep.
We decided to codify this distinction and make it part of the API. Therefore
this set of patches introduces three new, parallel APIs: one for blocking
notifiers, one for atomic notifiers, and one for "raw" notifiers (which is
really just the old API under a new name). New kinds of data structures are
used for the heads of the chains, and new routines are defined for
registration, unregistration, and calling a chain. The three APIs are
explained in include/linux/notifier.h and their implementation is in
kernel/sys.c.
With atomic and blocking chains, the implementation guarantees that the chain
links will not be corrupted and that chain callers will not get messed up by
entries being added or removed. For raw chains the implementation provides no
guarantees at all; users of this API must provide their own protections. (The
idea was that situations may come up where the assumptions of the atomic and
blocking APIs are not appropriate, so it should be possible for users to
handle these things in their own way.)
There are some limitations, which should not be too hard to live with. For
atomic/blocking chains, registration and unregistration must always be done in
a process context since the chain is protected by a mutex/rwsem. Also, a
callout routine for a non-raw chain must not try to register or unregister
entries on its own chain. (This did happen in a couple of places and the code
had to be changed to avoid it.)
Since atomic chains may be called from within an NMI handler, they cannot use
spinlocks for synchronization. Instead we use RCU. The overhead falls almost
entirely in the unregister routine, which is okay since unregistration is much
less frequent that calling a chain.
Here is the list of chains that we adjusted and their classifications. None
of them use the raw API, so for the moment it is only a placeholder.
ATOMIC CHAINS
-------------
arch/i386/kernel/traps.c: i386die_chain
arch/ia64/kernel/traps.c: ia64die_chain
arch/powerpc/kernel/traps.c: powerpc_die_chain
arch/sparc64/kernel/traps.c: sparc64die_chain
arch/x86_64/kernel/traps.c: die_chain
drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list
kernel/panic.c: panic_notifier_list
kernel/profile.c: task_free_notifier
net/bluetooth/hci_core.c: hci_notifier
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain
net/ipv6/addrconf.c: inet6addr_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain
net/netlink/af_netlink.c: netlink_chain
BLOCKING CHAINS
---------------
arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain
arch/s390/kernel/process.c: idle_chain
arch/x86_64/kernel/process.c idle_notifier
drivers/base/memory.c: memory_chain
drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list
drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list
drivers/macintosh/adb.c: adb_client_list
drivers/macintosh/via-pmu.c sleep_notifier_list
drivers/macintosh/via-pmu68k.c sleep_notifier_list
drivers/macintosh/windfarm_core.c wf_client_list
drivers/usb/core/notify.c usb_notifier_list
drivers/video/fbmem.c fb_notifier_list
kernel/cpu.c cpu_chain
kernel/module.c module_notify_list
kernel/profile.c munmap_notifier
kernel/profile.c task_exit_notifier
kernel/sys.c reboot_notifier_list
net/core/dev.c netdev_chain
net/decnet/dn_dev.c: dnaddr_chain
net/ipv4/devinet.c: inetaddr_chain
It's possible that some of these classifications are wrong. If they are,
please let us know or submit a patch to fix them. Note that any chain that
gets called very frequently should be atomic, because the rwsem read-locking
used for blocking chains is very likely to incur cache misses on SMP systems.
(However, if the chain's callout routines may sleep then the chain cannot be
atomic.)
The patch set was written by Alan Stern and Chandra Seetharaman, incorporating
material written by Keith Owens and suggestions from Paul McKenney and Andrew
Morton.
[jes@sgi.com: restructure the notifier chain initialization macros]
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 09:16:30 +00:00
|
|
|
atomic_notifier_call_chain(&idle_notifier, IDLE_START, NULL);
|
2006-01-11 21:44:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static void __exit_idle(void)
|
|
|
|
{
|
2009-01-18 15:38:59 +00:00
|
|
|
if (x86_test_and_clear_bit_percpu(0, is_idle) == 0)
|
2006-09-26 08:52:40 +00:00
|
|
|
return;
|
[PATCH] Notifier chain update: API changes
The kernel's implementation of notifier chains is unsafe. There is no
protection against entries being added to or removed from a chain while the
chain is in use. The issues were discussed in this thread:
http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2
We noticed that notifier chains in the kernel fall into two basic usage
classes:
"Blocking" chains are always called from a process context
and the callout routines are allowed to sleep;
"Atomic" chains can be called from an atomic context and
the callout routines are not allowed to sleep.
We decided to codify this distinction and make it part of the API. Therefore
this set of patches introduces three new, parallel APIs: one for blocking
notifiers, one for atomic notifiers, and one for "raw" notifiers (which is
really just the old API under a new name). New kinds of data structures are
used for the heads of the chains, and new routines are defined for
registration, unregistration, and calling a chain. The three APIs are
explained in include/linux/notifier.h and their implementation is in
kernel/sys.c.
With atomic and blocking chains, the implementation guarantees that the chain
links will not be corrupted and that chain callers will not get messed up by
entries being added or removed. For raw chains the implementation provides no
guarantees at all; users of this API must provide their own protections. (The
idea was that situations may come up where the assumptions of the atomic and
blocking APIs are not appropriate, so it should be possible for users to
handle these things in their own way.)
There are some limitations, which should not be too hard to live with. For
atomic/blocking chains, registration and unregistration must always be done in
a process context since the chain is protected by a mutex/rwsem. Also, a
callout routine for a non-raw chain must not try to register or unregister
entries on its own chain. (This did happen in a couple of places and the code
had to be changed to avoid it.)
Since atomic chains may be called from within an NMI handler, they cannot use
spinlocks for synchronization. Instead we use RCU. The overhead falls almost
entirely in the unregister routine, which is okay since unregistration is much
less frequent that calling a chain.
Here is the list of chains that we adjusted and their classifications. None
of them use the raw API, so for the moment it is only a placeholder.
ATOMIC CHAINS
-------------
arch/i386/kernel/traps.c: i386die_chain
arch/ia64/kernel/traps.c: ia64die_chain
arch/powerpc/kernel/traps.c: powerpc_die_chain
arch/sparc64/kernel/traps.c: sparc64die_chain
arch/x86_64/kernel/traps.c: die_chain
drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list
kernel/panic.c: panic_notifier_list
kernel/profile.c: task_free_notifier
net/bluetooth/hci_core.c: hci_notifier
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain
net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain
net/ipv6/addrconf.c: inet6addr_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_chain
net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain
net/netlink/af_netlink.c: netlink_chain
BLOCKING CHAINS
---------------
arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain
arch/s390/kernel/process.c: idle_chain
arch/x86_64/kernel/process.c idle_notifier
drivers/base/memory.c: memory_chain
drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list
drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list
drivers/macintosh/adb.c: adb_client_list
drivers/macintosh/via-pmu.c sleep_notifier_list
drivers/macintosh/via-pmu68k.c sleep_notifier_list
drivers/macintosh/windfarm_core.c wf_client_list
drivers/usb/core/notify.c usb_notifier_list
drivers/video/fbmem.c fb_notifier_list
kernel/cpu.c cpu_chain
kernel/module.c module_notify_list
kernel/profile.c munmap_notifier
kernel/profile.c task_exit_notifier
kernel/sys.c reboot_notifier_list
net/core/dev.c netdev_chain
net/decnet/dn_dev.c: dnaddr_chain
net/ipv4/devinet.c: inetaddr_chain
It's possible that some of these classifications are wrong. If they are,
please let us know or submit a patch to fix them. Note that any chain that
gets called very frequently should be atomic, because the rwsem read-locking
used for blocking chains is very likely to incur cache misses on SMP systems.
(However, if the chain's callout routines may sleep then the chain cannot be
atomic.)
The patch set was written by Alan Stern and Chandra Seetharaman, incorporating
material written by Keith Owens and suggestions from Paul McKenney and Andrew
Morton.
[jes@sgi.com: restructure the notifier chain initialization macros]
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 09:16:30 +00:00
|
|
|
atomic_notifier_call_chain(&idle_notifier, IDLE_END, NULL);
|
2006-01-11 21:44:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Called from interrupts to signify idle end */
|
|
|
|
void exit_idle(void)
|
|
|
|
{
|
2006-09-26 08:52:40 +00:00
|
|
|
/* idle loop has pid 0 */
|
|
|
|
if (current->pid)
|
2006-01-11 21:44:36 +00:00
|
|
|
return;
|
|
|
|
__exit_idle();
|
|
|
|
}
|
|
|
|
|
2008-09-03 13:30:23 +00:00
|
|
|
#ifndef CONFIG_SMP
|
2005-06-25 21:55:00 +00:00
|
|
|
static inline void play_dead(void)
|
|
|
|
{
|
|
|
|
BUG();
|
|
|
|
}
|
2008-09-03 13:30:23 +00:00
|
|
|
#endif
|
2005-06-25 21:55:00 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* The idle thread. There's no useful work to be
|
|
|
|
* done, so just try to conserve power and have a
|
|
|
|
* low exit latency (ie sit in a loop waiting for
|
|
|
|
* somebody to say that they'd like to reschedule)
|
|
|
|
*/
|
2008-01-30 12:30:00 +00:00
|
|
|
void cpu_idle(void)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2006-06-26 11:59:11 +00:00
|
|
|
current_thread_info()->status |= TS_POLLING;
|
2008-05-12 13:44:31 +00:00
|
|
|
|
|
|
|
/*
|
2009-02-11 07:31:00 +00:00
|
|
|
* If we're the non-boot CPU, nothing set the stack canary up
|
|
|
|
* for us. CPU0 already has it initialized but no harm in
|
|
|
|
* doing it again. This is a good place for updating it, as
|
|
|
|
* we wont ever return from this function (so the invalid
|
|
|
|
* canaries already on the stack wont ever trigger).
|
2008-05-12 13:44:31 +00:00
|
|
|
*/
|
2008-02-14 08:42:02 +00:00
|
|
|
boot_init_stack_canary();
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* endless idle loop with no priority at all */
|
|
|
|
while (1) {
|
2011-10-07 16:22:08 +00:00
|
|
|
tick_nohz_idle_enter();
|
2005-04-16 22:20:36 +00:00
|
|
|
while (!need_resched()) {
|
|
|
|
|
|
|
|
rmb();
|
2008-06-09 14:59:53 +00:00
|
|
|
|
2005-06-25 21:55:00 +00:00
|
|
|
if (cpu_is_offline(smp_processor_id()))
|
|
|
|
play_dead();
|
2006-12-07 01:14:13 +00:00
|
|
|
/*
|
|
|
|
* Idle routines should keep interrupts disabled
|
|
|
|
* from here on, until they go to idle.
|
|
|
|
* Otherwise, idle callbacks can misfire.
|
|
|
|
*/
|
2011-09-30 19:06:22 +00:00
|
|
|
local_touch_nmi();
|
2006-12-07 01:14:13 +00:00
|
|
|
local_irq_disable();
|
2006-01-11 21:44:36 +00:00
|
|
|
enter_idle();
|
2008-05-12 19:20:42 +00:00
|
|
|
/* Don't trace irqs off for idle */
|
|
|
|
stop_critical_timings();
|
2011-10-07 16:22:08 +00:00
|
|
|
|
|
|
|
/* enter_idle() needs rcu for notifiers */
|
|
|
|
rcu_idle_enter();
|
|
|
|
|
2011-04-01 23:34:59 +00:00
|
|
|
if (cpuidle_idle_call())
|
|
|
|
pm_idle();
|
2011-10-07 16:22:08 +00:00
|
|
|
|
|
|
|
rcu_idle_exit();
|
2008-05-12 19:20:42 +00:00
|
|
|
start_critical_timings();
|
2010-06-14 11:37:20 +00:00
|
|
|
|
2006-09-26 08:52:40 +00:00
|
|
|
/* In many cases the interrupt that ended idle
|
|
|
|
has already called exit_idle. But some idle
|
|
|
|
loops can be woken up without interrupt. */
|
2006-01-11 21:44:36 +00:00
|
|
|
__exit_idle();
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
2011-10-07 16:22:08 +00:00
|
|
|
tick_nohz_idle_exit();
|
2005-11-09 05:39:01 +00:00
|
|
|
preempt_enable_no_resched();
|
2005-04-16 22:20:36 +00:00
|
|
|
schedule();
|
2005-11-09 05:39:01 +00:00
|
|
|
preempt_disable();
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-01-30 12:31:03 +00:00
|
|
|
/* Prints also some state that isn't saved in the pt_regs */
|
2008-04-03 13:40:48 +00:00
|
|
|
void __show_regs(struct pt_regs *regs, int all)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs;
|
2007-07-21 15:10:42 +00:00
|
|
|
unsigned long d0, d1, d2, d3, d6, d7;
|
2008-01-30 12:31:03 +00:00
|
|
|
unsigned int fsindex, gsindex;
|
|
|
|
unsigned int ds, cs, es;
|
2009-12-08 08:29:42 +00:00
|
|
|
|
|
|
|
show_regs_common();
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip);
|
2008-01-30 12:33:08 +00:00
|
|
|
printk_address(regs->ip, 1);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "RSP: %04lx:%016lx EFLAGS: %08lx\n", regs->ss,
|
2008-07-29 05:48:52 +00:00
|
|
|
regs->sp, regs->flags);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "RAX: %016lx RBX: %016lx RCX: %016lx\n",
|
2008-01-30 12:30:56 +00:00
|
|
|
regs->ax, regs->bx, regs->cx);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "RDX: %016lx RSI: %016lx RDI: %016lx\n",
|
2008-01-30 12:30:56 +00:00
|
|
|
regs->dx, regs->si, regs->di);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "RBP: %016lx R08: %016lx R09: %016lx\n",
|
2008-01-30 12:30:56 +00:00
|
|
|
regs->bp, regs->r8, regs->r9);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "R10: %016lx R11: %016lx R12: %016lx\n",
|
2008-07-29 05:48:51 +00:00
|
|
|
regs->r10, regs->r11, regs->r12);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "R13: %016lx R14: %016lx R15: %016lx\n",
|
2008-07-29 05:48:51 +00:00
|
|
|
regs->r13, regs->r14, regs->r15);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-07-29 05:48:51 +00:00
|
|
|
asm("movl %%ds,%0" : "=r" (ds));
|
|
|
|
asm("movl %%cs,%0" : "=r" (cs));
|
|
|
|
asm("movl %%es,%0" : "=r" (es));
|
2005-04-16 22:20:36 +00:00
|
|
|
asm("movl %%fs,%0" : "=r" (fsindex));
|
|
|
|
asm("movl %%gs,%0" : "=r" (gsindex));
|
|
|
|
|
|
|
|
rdmsrl(MSR_FS_BASE, fs);
|
2008-07-29 05:48:51 +00:00
|
|
|
rdmsrl(MSR_GS_BASE, gs);
|
|
|
|
rdmsrl(MSR_KERNEL_GS_BASE, shadowgs);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-04-03 13:40:48 +00:00
|
|
|
if (!all)
|
|
|
|
return;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2007-07-22 09:12:29 +00:00
|
|
|
cr0 = read_cr0();
|
|
|
|
cr2 = read_cr2();
|
|
|
|
cr3 = read_cr3();
|
|
|
|
cr4 = read_cr4();
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "FS: %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n",
|
2008-07-29 05:48:51 +00:00
|
|
|
fs, fsindex, gs, gsindex, shadowgs);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "CS: %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds,
|
2008-07-29 05:48:52 +00:00
|
|
|
es, cr0);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3,
|
2008-07-29 05:48:52 +00:00
|
|
|
cr4);
|
2007-07-21 15:10:42 +00:00
|
|
|
|
|
|
|
get_debugreg(d0, 0);
|
|
|
|
get_debugreg(d1, 1);
|
|
|
|
get_debugreg(d2, 2);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2);
|
2007-07-21 15:10:42 +00:00
|
|
|
get_debugreg(d3, 3);
|
|
|
|
get_debugreg(d6, 6);
|
|
|
|
get_debugreg(d7, 7);
|
2009-12-28 08:26:59 +00:00
|
|
|
printk(KERN_DEFAULT "DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
void release_thread(struct task_struct *dead_task)
|
|
|
|
{
|
|
|
|
if (dead_task->mm) {
|
|
|
|
if (dead_task->mm->context.size) {
|
|
|
|
printk("WARNING: dead process %8s still has LDT? <%p/%d>\n",
|
|
|
|
dead_task->comm,
|
|
|
|
dead_task->mm->context.ldt,
|
|
|
|
dead_task->mm->context.size);
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void set_32bit_tls(struct task_struct *t, int tls, u32 addr)
|
|
|
|
{
|
2008-01-30 12:31:03 +00:00
|
|
|
struct user_desc ud = {
|
2005-04-16 22:20:36 +00:00
|
|
|
.base_addr = addr,
|
|
|
|
.limit = 0xfffff,
|
|
|
|
.seg_32bit = 1,
|
|
|
|
.limit_in_pages = 1,
|
|
|
|
.useable = 1,
|
|
|
|
};
|
2008-01-30 12:33:23 +00:00
|
|
|
struct desc_struct *desc = t->thread.tls_array;
|
2005-04-16 22:20:36 +00:00
|
|
|
desc += tls;
|
2008-01-30 12:31:13 +00:00
|
|
|
fill_ldt(desc, &ud);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline u32 read_32bit_tls(struct task_struct *t, int tls)
|
|
|
|
{
|
2008-01-30 12:30:45 +00:00
|
|
|
return get_desc_base(&t->thread.tls_array[tls]);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This gets called before we allocate a new thread and copy
|
|
|
|
* the current task into it.
|
|
|
|
*/
|
|
|
|
void prepare_to_copy(struct task_struct *tsk)
|
|
|
|
{
|
|
|
|
unlazy_fpu(tsk);
|
|
|
|
}
|
|
|
|
|
2009-04-02 23:56:59 +00:00
|
|
|
int copy_thread(unsigned long clone_flags, unsigned long sp,
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long unused,
|
2008-07-29 05:48:51 +00:00
|
|
|
struct task_struct *p, struct pt_regs *regs)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
|
|
|
int err;
|
2008-07-29 05:48:51 +00:00
|
|
|
struct pt_regs *childregs;
|
2005-04-16 22:20:36 +00:00
|
|
|
struct task_struct *me = current;
|
|
|
|
|
2005-11-05 16:25:54 +00:00
|
|
|
childregs = ((struct pt_regs *)
|
2006-01-12 09:05:39 +00:00
|
|
|
(THREAD_SIZE + task_stack_page(p))) - 1;
|
2005-04-16 22:20:36 +00:00
|
|
|
*childregs = *regs;
|
|
|
|
|
2008-01-30 12:30:56 +00:00
|
|
|
childregs->ax = 0;
|
2009-12-09 17:34:41 +00:00
|
|
|
if (user_mode(regs))
|
|
|
|
childregs->sp = sp;
|
|
|
|
else
|
2008-01-30 12:30:56 +00:00
|
|
|
childregs->sp = (unsigned long)childregs;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-01-30 12:31:02 +00:00
|
|
|
p->thread.sp = (unsigned long) childregs;
|
|
|
|
p->thread.sp0 = (unsigned long) (childregs+1);
|
|
|
|
p->thread.usersp = me->thread.usersp;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2006-01-12 09:05:38 +00:00
|
|
|
set_tsk_thread_flag(p, TIF_FORK);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2009-06-01 18:14:55 +00:00
|
|
|
p->thread.io_bitmap_ptr = NULL;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-06-25 04:19:00 +00:00
|
|
|
savesegment(gs, p->thread.gsindex);
|
2010-04-23 23:17:40 +00:00
|
|
|
p->thread.gs = p->thread.gsindex ? 0 : me->thread.gs;
|
2008-06-25 04:19:00 +00:00
|
|
|
savesegment(fs, p->thread.fsindex);
|
2010-04-23 23:17:40 +00:00
|
|
|
p->thread.fs = p->thread.fsindex ? 0 : me->thread.fs;
|
2008-06-25 04:19:00 +00:00
|
|
|
savesegment(es, p->thread.es);
|
|
|
|
savesegment(ds, p->thread.ds);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2009-06-01 18:14:55 +00:00
|
|
|
err = -ENOMEM;
|
2009-09-09 17:22:48 +00:00
|
|
|
memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));
|
2009-06-01 18:14:55 +00:00
|
|
|
|
2006-09-26 08:52:28 +00:00
|
|
|
if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) {
|
2011-11-17 22:43:40 +00:00
|
|
|
p->thread.io_bitmap_ptr = kmemdup(me->thread.io_bitmap_ptr,
|
|
|
|
IO_BITMAP_BYTES, GFP_KERNEL);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (!p->thread.io_bitmap_ptr) {
|
|
|
|
p->thread.io_bitmap_max = 0;
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2006-09-26 08:52:28 +00:00
|
|
|
set_tsk_thread_flag(p, TIF_IO_BITMAP);
|
2008-01-30 12:31:03 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Set a new TLS for the child thread?
|
|
|
|
*/
|
|
|
|
if (clone_flags & CLONE_SETTLS) {
|
|
|
|
#ifdef CONFIG_IA32_EMULATION
|
|
|
|
if (test_thread_flag(TIF_IA32))
|
2008-01-30 12:30:46 +00:00
|
|
|
err = do_set_thread_area(p, -1,
|
2008-01-30 12:30:56 +00:00
|
|
|
(struct user_desc __user *)childregs->si, 0);
|
2008-07-29 05:48:51 +00:00
|
|
|
else
|
|
|
|
#endif
|
|
|
|
err = do_arch_prctl(p, ARCH_SET_FS, childregs->r8);
|
|
|
|
if (err)
|
2005-04-16 22:20:36 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
err = 0;
|
|
|
|
out:
|
|
|
|
if (err && p->thread.io_bitmap_ptr) {
|
|
|
|
kfree(p->thread.io_bitmap_ptr);
|
|
|
|
p->thread.io_bitmap_max = 0;
|
|
|
|
}
|
2009-06-01 18:14:55 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2009-10-09 22:56:53 +00:00
|
|
|
static void
|
|
|
|
start_thread_common(struct pt_regs *regs, unsigned long new_ip,
|
|
|
|
unsigned long new_sp,
|
|
|
|
unsigned int _cs, unsigned int _ss, unsigned int _ds)
|
2008-02-21 04:18:40 +00:00
|
|
|
{
|
2008-06-25 04:19:00 +00:00
|
|
|
loadsegment(fs, 0);
|
2009-10-09 22:56:53 +00:00
|
|
|
loadsegment(es, _ds);
|
|
|
|
loadsegment(ds, _ds);
|
2008-02-21 04:18:40 +00:00
|
|
|
load_gs_index(0);
|
|
|
|
regs->ip = new_ip;
|
|
|
|
regs->sp = new_sp;
|
2009-01-18 15:38:58 +00:00
|
|
|
percpu_write(old_rsp, new_sp);
|
2009-10-09 22:56:53 +00:00
|
|
|
regs->cs = _cs;
|
|
|
|
regs->ss = _ss;
|
2009-10-09 01:02:54 +00:00
|
|
|
regs->flags = X86_EFLAGS_IF;
|
2008-03-10 22:28:05 +00:00
|
|
|
/*
|
|
|
|
* Free the old FP and other extended state
|
|
|
|
*/
|
|
|
|
free_thread_xstate(current);
|
2008-02-21 04:18:40 +00:00
|
|
|
}
|
2009-10-09 22:56:53 +00:00
|
|
|
|
|
|
|
void
|
|
|
|
start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
|
|
|
|
{
|
|
|
|
start_thread_common(regs, new_ip, new_sp,
|
|
|
|
__USER_CS, __USER_DS, 0);
|
|
|
|
}
|
2008-02-21 04:18:40 +00:00
|
|
|
|
2009-10-09 01:02:54 +00:00
|
|
|
#ifdef CONFIG_IA32_EMULATION
|
|
|
|
void start_thread_ia32(struct pt_regs *regs, u32 new_ip, u32 new_sp)
|
|
|
|
{
|
2009-10-09 22:56:53 +00:00
|
|
|
start_thread_common(regs, new_ip, new_sp,
|
|
|
|
__USER32_CS, __USER32_DS, __USER32_DS);
|
2009-10-09 01:02:54 +00:00
|
|
|
}
|
|
|
|
#endif
|
2008-02-21 04:18:40 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* switch_to(x,y) should switch tasks from x to y.
|
|
|
|
*
|
2008-01-30 12:31:03 +00:00
|
|
|
* This could still be optimized:
|
2005-04-16 22:20:36 +00:00
|
|
|
* - fold all the options into a flag word and test it with a single test.
|
|
|
|
* - could test fs/gs bitsliced
|
2006-02-03 20:51:38 +00:00
|
|
|
*
|
|
|
|
* Kprobes not supported here. Set the probe on schedule instead.
|
2008-12-06 02:40:00 +00:00
|
|
|
* Function graph tracer not supported too.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2008-12-06 02:40:00 +00:00
|
|
|
__notrace_funcgraph struct task_struct *
|
2005-11-05 16:25:54 +00:00
|
|
|
__switch_to(struct task_struct *prev_p, struct task_struct *next_p)
|
2005-04-16 22:20:36 +00:00
|
|
|
{
|
2008-07-08 22:06:26 +00:00
|
|
|
struct thread_struct *prev = &prev_p->thread;
|
|
|
|
struct thread_struct *next = &next_p->thread;
|
2008-01-30 12:31:03 +00:00
|
|
|
int cpu = smp_processor_id();
|
2005-04-16 22:20:36 +00:00
|
|
|
struct tss_struct *tss = &per_cpu(init_tss, cpu);
|
2008-06-25 04:19:24 +00:00
|
|
|
unsigned fsindex, gsindex;
|
i387: re-introduce FPU state preloading at context switch time
After all the FPU state cleanups and finally finding the problem that
caused all our FPU save/restore problems, this re-introduces the
preloading of FPU state that was removed in commit b3b0870ef3ff ("i387:
do not preload FPU state at task switch time").
However, instead of simply reverting the removal, this reimplements
preloading with several fixes, most notably
- properly abstracted as a true FPU state switch, rather than as
open-coded save and restore with various hacks.
In particular, implementing it as a proper FPU state switch allows us
to optimize the CR0.TS flag accesses: there is no reason to set the
TS bit only to then almost immediately clear it again. CR0 accesses
are quite slow and expensive, don't flip the bit back and forth for
no good reason.
- Make sure that the same model works for both x86-32 and x86-64, so
that there are no gratuitous differences between the two due to the
way they save and restore segment state differently due to
architectural differences that really don't matter to the FPU state.
- Avoid exposing the "preload" state to the context switch routines,
and in particular allow the concept of lazy state restore: if nothing
else has used the FPU in the meantime, and the process is still on
the same CPU, we can avoid restoring state from memory entirely, just
re-expose the state that is still in the FPU unit.
That optimized lazy restore isn't actually implemented here, but the
infrastructure is set up for it. Of course, older CPU's that use
'fnsave' to save the state cannot take advantage of this, since the
state saving also trashes the state.
In other words, there is now an actual _design_ to the FPU state saving,
rather than just random historical baggage. Hopefully it's easier to
follow as a result.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-02-18 20:56:35 +00:00
|
|
|
fpu_switch_t fpu;
|
2006-09-26 08:52:36 +00:00
|
|
|
|
i387: re-introduce FPU state preloading at context switch time
After all the FPU state cleanups and finally finding the problem that
caused all our FPU save/restore problems, this re-introduces the
preloading of FPU state that was removed in commit b3b0870ef3ff ("i387:
do not preload FPU state at task switch time").
However, instead of simply reverting the removal, this reimplements
preloading with several fixes, most notably
- properly abstracted as a true FPU state switch, rather than as
open-coded save and restore with various hacks.
In particular, implementing it as a proper FPU state switch allows us
to optimize the CR0.TS flag accesses: there is no reason to set the
TS bit only to then almost immediately clear it again. CR0 accesses
are quite slow and expensive, don't flip the bit back and forth for
no good reason.
- Make sure that the same model works for both x86-32 and x86-64, so
that there are no gratuitous differences between the two due to the
way they save and restore segment state differently due to
architectural differences that really don't matter to the FPU state.
- Avoid exposing the "preload" state to the context switch routines,
and in particular allow the concept of lazy state restore: if nothing
else has used the FPU in the meantime, and the process is still on
the same CPU, we can avoid restoring state from memory entirely, just
re-expose the state that is still in the FPU unit.
That optimized lazy restore isn't actually implemented here, but the
infrastructure is set up for it. Of course, older CPU's that use
'fnsave' to save the state cannot take advantage of this, since the
state saving also trashes the state.
In other words, there is now an actual _design_ to the FPU state saving,
rather than just random historical baggage. Hopefully it's easier to
follow as a result.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-02-18 20:56:35 +00:00
|
|
|
fpu = switch_fpu_prepare(prev_p, next_p);
|
2012-02-17 03:11:15 +00:00
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/*
|
|
|
|
* Reload esp0, LDT and the page table pointer:
|
|
|
|
*/
|
2008-01-30 12:31:31 +00:00
|
|
|
load_sp0(tss, next);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-07-29 05:48:51 +00:00
|
|
|
/*
|
2005-04-16 22:20:36 +00:00
|
|
|
* Switch DS and ES.
|
|
|
|
* This won't pick up thread selector changes, but I guess that is ok.
|
|
|
|
*/
|
2008-06-25 04:19:00 +00:00
|
|
|
savesegment(es, prev->es);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (unlikely(next->es | prev->es))
|
2008-07-29 05:48:51 +00:00
|
|
|
loadsegment(es, next->es);
|
2008-06-25 04:19:00 +00:00
|
|
|
|
|
|
|
savesegment(ds, prev->ds);
|
2005-04-16 22:20:36 +00:00
|
|
|
if (unlikely(next->ds | prev->ds))
|
|
|
|
loadsegment(ds, next->ds);
|
|
|
|
|
2008-06-25 04:19:24 +00:00
|
|
|
|
|
|
|
/* We must save %fs and %gs before load_TLS() because
|
|
|
|
* %fs and %gs may be cleared by load_TLS().
|
|
|
|
*
|
|
|
|
* (e.g. xen_load_tls())
|
|
|
|
*/
|
|
|
|
savesegment(fs, fsindex);
|
|
|
|
savesegment(gs, gsindex);
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
load_TLS(next, cpu);
|
|
|
|
|
2008-06-25 04:19:23 +00:00
|
|
|
/*
|
|
|
|
* Leave lazy mode, flushing any hypercalls made here.
|
|
|
|
* This must be done before restoring TLS segments so
|
|
|
|
* the GDT and LDT are properly updated, and must be
|
|
|
|
* done before math_state_restore, so the TS bit is up
|
|
|
|
* to date.
|
|
|
|
*/
|
2009-02-18 19:18:57 +00:00
|
|
|
arch_end_context_switch(next_p);
|
2008-06-25 04:19:23 +00:00
|
|
|
|
2008-07-29 05:48:51 +00:00
|
|
|
/*
|
2005-04-16 22:20:36 +00:00
|
|
|
* Switch FS and GS.
|
2008-07-08 22:06:26 +00:00
|
|
|
*
|
|
|
|
* Segment register != 0 always requires a reload. Also
|
|
|
|
* reload when it has changed. When prev process used 64bit
|
|
|
|
* base always reload to avoid an information leak.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2008-07-08 22:06:26 +00:00
|
|
|
if (unlikely(fsindex | next->fsindex | prev->fs)) {
|
|
|
|
loadsegment(fs, next->fsindex);
|
2008-07-29 05:48:51 +00:00
|
|
|
/*
|
2008-07-08 22:06:26 +00:00
|
|
|
* Check if the user used a selector != 0; if yes
|
|
|
|
* clear 64bit base, since overloaded base is always
|
|
|
|
* mapped to the Null selector
|
|
|
|
*/
|
|
|
|
if (fsindex)
|
2008-07-29 05:48:51 +00:00
|
|
|
prev->fs = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2008-07-08 22:06:26 +00:00
|
|
|
/* when next process has a 64bit base use it */
|
|
|
|
if (next->fs)
|
|
|
|
wrmsrl(MSR_FS_BASE, next->fs);
|
|
|
|
prev->fsindex = fsindex;
|
|
|
|
|
|
|
|
if (unlikely(gsindex | next->gsindex | prev->gs)) {
|
|
|
|
load_gs_index(next->gsindex);
|
|
|
|
if (gsindex)
|
2008-07-29 05:48:51 +00:00
|
|
|
prev->gs = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2008-07-08 22:06:26 +00:00
|
|
|
if (next->gs)
|
|
|
|
wrmsrl(MSR_KERNEL_GS_BASE, next->gs);
|
|
|
|
prev->gsindex = gsindex;
|
2005-04-16 22:20:36 +00:00
|
|
|
|
i387: re-introduce FPU state preloading at context switch time
After all the FPU state cleanups and finally finding the problem that
caused all our FPU save/restore problems, this re-introduces the
preloading of FPU state that was removed in commit b3b0870ef3ff ("i387:
do not preload FPU state at task switch time").
However, instead of simply reverting the removal, this reimplements
preloading with several fixes, most notably
- properly abstracted as a true FPU state switch, rather than as
open-coded save and restore with various hacks.
In particular, implementing it as a proper FPU state switch allows us
to optimize the CR0.TS flag accesses: there is no reason to set the
TS bit only to then almost immediately clear it again. CR0 accesses
are quite slow and expensive, don't flip the bit back and forth for
no good reason.
- Make sure that the same model works for both x86-32 and x86-64, so
that there are no gratuitous differences between the two due to the
way they save and restore segment state differently due to
architectural differences that really don't matter to the FPU state.
- Avoid exposing the "preload" state to the context switch routines,
and in particular allow the concept of lazy state restore: if nothing
else has used the FPU in the meantime, and the process is still on
the same CPU, we can avoid restoring state from memory entirely, just
re-expose the state that is still in the FPU unit.
That optimized lazy restore isn't actually implemented here, but the
infrastructure is set up for it. Of course, older CPU's that use
'fnsave' to save the state cannot take advantage of this, since the
state saving also trashes the state.
In other words, there is now an actual _design_ to the FPU state saving,
rather than just random historical baggage. Hopefully it's easier to
follow as a result.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-02-18 20:56:35 +00:00
|
|
|
switch_fpu_finish(next_p, fpu);
|
|
|
|
|
2008-07-29 05:48:51 +00:00
|
|
|
/*
|
2006-03-25 15:29:25 +00:00
|
|
|
* Switch the PDA and FPU contexts.
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2009-01-18 15:38:58 +00:00
|
|
|
prev->usersp = percpu_read(old_rsp);
|
|
|
|
percpu_write(old_rsp, next->usersp);
|
2009-01-18 15:38:58 +00:00
|
|
|
percpu_write(current_task, next_p);
|
2006-04-20 00:36:45 +00:00
|
|
|
|
2009-01-18 15:38:58 +00:00
|
|
|
percpu_write(kernel_stack,
|
2008-07-08 22:06:26 +00:00
|
|
|
(unsigned long)task_stack_page(next_p) +
|
2009-01-18 15:38:58 +00:00
|
|
|
THREAD_SIZE - KERNEL_STACK_OFFSET);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
/*
|
2006-09-26 08:52:28 +00:00
|
|
|
* Now maybe reload the debug registers and handle I/O bitmaps
|
2005-04-16 22:20:36 +00:00
|
|
|
*/
|
2008-01-30 12:31:09 +00:00
|
|
|
if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT ||
|
|
|
|
task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV))
|
2006-09-26 08:52:28 +00:00
|
|
|
__switch_to_xtra(prev_p, next_p, tss);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
return prev_p;
|
|
|
|
}
|
|
|
|
|
|
|
|
void set_personality_64bit(void)
|
|
|
|
{
|
|
|
|
/* inherit personality from parent */
|
|
|
|
|
|
|
|
/* Make sure to be in 64bit mode */
|
2008-01-30 12:31:03 +00:00
|
|
|
clear_thread_flag(TIF_IA32);
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2011-03-13 19:49:14 +00:00
|
|
|
/* Ensure the corresponding mm is not marked. */
|
|
|
|
if (current->mm)
|
|
|
|
current->mm->context.ia32_compat = 0;
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
/* TBD: overwrites user setup. Should have two bits.
|
|
|
|
But 64bit processes have always behaved this way,
|
|
|
|
so it's not too bad. The main problem is just that
|
2008-01-30 12:31:03 +00:00
|
|
|
32bit childs are affected again. */
|
2005-04-16 22:20:36 +00:00
|
|
|
current->personality &= ~READ_IMPLIES_EXEC;
|
|
|
|
}
|
|
|
|
|
2010-01-29 06:14:43 +00:00
|
|
|
void set_personality_ia32(void)
|
|
|
|
{
|
|
|
|
/* inherit personality from parent */
|
|
|
|
|
|
|
|
/* Make sure to be in 32bit mode */
|
|
|
|
set_thread_flag(TIF_IA32);
|
2010-02-16 14:02:13 +00:00
|
|
|
current->personality |= force_personality32;
|
2010-01-29 06:14:43 +00:00
|
|
|
|
2011-03-13 19:49:14 +00:00
|
|
|
/* Mark the associated mm as containing 32-bit tasks. */
|
|
|
|
if (current->mm)
|
|
|
|
current->mm->context.ia32_compat = 1;
|
|
|
|
|
2010-01-29 06:14:43 +00:00
|
|
|
/* Prepare the first "return" to user space */
|
|
|
|
current_thread_info()->status |= TS_COMPAT;
|
|
|
|
}
|
|
|
|
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long get_wchan(struct task_struct *p)
|
|
|
|
{
|
|
|
|
unsigned long stack;
|
2008-07-29 05:48:51 +00:00
|
|
|
u64 fp, ip;
|
2005-04-16 22:20:36 +00:00
|
|
|
int count = 0;
|
|
|
|
|
2008-07-29 05:48:51 +00:00
|
|
|
if (!p || p == current || p->state == TASK_RUNNING)
|
|
|
|
return 0;
|
2006-01-12 09:05:39 +00:00
|
|
|
stack = (unsigned long)task_stack_page(p);
|
2008-10-07 21:15:11 +00:00
|
|
|
if (p->thread.sp < stack || p->thread.sp >= stack+THREAD_SIZE)
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
2008-01-30 12:31:02 +00:00
|
|
|
fp = *(u64 *)(p->thread.sp);
|
2008-07-29 05:48:51 +00:00
|
|
|
do {
|
2005-11-05 16:25:54 +00:00
|
|
|
if (fp < (unsigned long)stack ||
|
2008-10-07 21:15:11 +00:00
|
|
|
fp >= (unsigned long)stack+THREAD_SIZE)
|
2008-07-29 05:48:51 +00:00
|
|
|
return 0;
|
2008-01-30 12:30:56 +00:00
|
|
|
ip = *(u64 *)(fp+8);
|
|
|
|
if (!in_sched_functions(ip))
|
|
|
|
return ip;
|
2008-07-29 05:48:51 +00:00
|
|
|
fp = *(u64 *)fp;
|
|
|
|
} while (count++ < 16);
|
2005-04-16 22:20:36 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
long do_arch_prctl(struct task_struct *task, int code, unsigned long addr)
|
2008-07-29 05:48:51 +00:00
|
|
|
{
|
|
|
|
int ret = 0;
|
2005-04-16 22:20:36 +00:00
|
|
|
int doit = task == current;
|
|
|
|
int cpu;
|
|
|
|
|
2008-07-29 05:48:51 +00:00
|
|
|
switch (code) {
|
2005-04-16 22:20:36 +00:00
|
|
|
case ARCH_SET_GS:
|
[PATCH] x86_64: TASK_SIZE fixes for compatibility mode processes
Appended patch will setup compatibility mode TASK_SIZE properly. This will
fix atleast three known bugs that can be encountered while running
compatibility mode apps.
a) A malicious 32bit app can have an elf section at 0xffffe000. During
exec of this app, we will have a memory leak as insert_vm_struct() is
not checking for return value in syscall32_setup_pages() and thus not
freeing the vma allocated for the vsyscall page. And instead of exec
failing (as it has addresses > TASK_SIZE), we were allowing it to
succeed previously.
b) With a 32bit app, hugetlb_get_unmapped_area/arch_get_unmapped_area
may return addresses beyond 32bits, ultimately causing corruption
because of wrap-around and resulting in SEGFAULT, instead of returning
ENOMEM.
c) 32bit app doing this below mmap will now fail.
mmap((void *)(0xFFFFE000UL), 0x10000UL, PROT_READ|PROT_WRITE,
MAP_FIXED|MAP_PRIVATE|MAP_ANON, 0, 0);
Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:32 +00:00
|
|
|
if (addr >= TASK_SIZE_OF(task))
|
2008-07-29 05:48:51 +00:00
|
|
|
return -EPERM;
|
2005-04-16 22:20:36 +00:00
|
|
|
cpu = get_cpu();
|
2008-07-29 05:48:51 +00:00
|
|
|
/* handle small bases via the GDT because that's faster to
|
2005-04-16 22:20:36 +00:00
|
|
|
switch. */
|
2008-07-29 05:48:51 +00:00
|
|
|
if (addr <= 0xffffffff) {
|
|
|
|
set_32bit_tls(task, GS_TLS, addr);
|
|
|
|
if (doit) {
|
2005-04-16 22:20:36 +00:00
|
|
|
load_TLS(&task->thread, cpu);
|
2008-07-29 05:48:51 +00:00
|
|
|
load_gs_index(GS_TLS_SEL);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2008-07-29 05:48:51 +00:00
|
|
|
task->thread.gsindex = GS_TLS_SEL;
|
2005-04-16 22:20:36 +00:00
|
|
|
task->thread.gs = 0;
|
2008-07-29 05:48:51 +00:00
|
|
|
} else {
|
2005-04-16 22:20:36 +00:00
|
|
|
task->thread.gsindex = 0;
|
|
|
|
task->thread.gs = addr;
|
|
|
|
if (doit) {
|
2005-11-05 16:25:54 +00:00
|
|
|
load_gs_index(0);
|
|
|
|
ret = checking_wrmsrl(MSR_KERNEL_GS_BASE, addr);
|
2008-07-29 05:48:51 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
put_cpu();
|
|
|
|
break;
|
|
|
|
case ARCH_SET_FS:
|
|
|
|
/* Not strictly needed for fs, but do it for symmetry
|
|
|
|
with gs */
|
[PATCH] x86_64: TASK_SIZE fixes for compatibility mode processes
Appended patch will setup compatibility mode TASK_SIZE properly. This will
fix atleast three known bugs that can be encountered while running
compatibility mode apps.
a) A malicious 32bit app can have an elf section at 0xffffe000. During
exec of this app, we will have a memory leak as insert_vm_struct() is
not checking for return value in syscall32_setup_pages() and thus not
freeing the vma allocated for the vsyscall page. And instead of exec
failing (as it has addresses > TASK_SIZE), we were allowing it to
succeed previously.
b) With a 32bit app, hugetlb_get_unmapped_area/arch_get_unmapped_area
may return addresses beyond 32bits, ultimately causing corruption
because of wrap-around and resulting in SEGFAULT, instead of returning
ENOMEM.
c) 32bit app doing this below mmap will now fail.
mmap((void *)(0xFFFFE000UL), 0x10000UL, PROT_READ|PROT_WRITE,
MAP_FIXED|MAP_PRIVATE|MAP_ANON, 0, 0);
Signed-off-by: Zou Nan hai <nanhai.zou@intel.com>
Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:32 +00:00
|
|
|
if (addr >= TASK_SIZE_OF(task))
|
2008-01-30 12:31:03 +00:00
|
|
|
return -EPERM;
|
2005-04-16 22:20:36 +00:00
|
|
|
cpu = get_cpu();
|
2008-01-30 12:31:03 +00:00
|
|
|
/* handle small bases via the GDT because that's faster to
|
2005-04-16 22:20:36 +00:00
|
|
|
switch. */
|
2008-01-30 12:31:03 +00:00
|
|
|
if (addr <= 0xffffffff) {
|
2005-04-16 22:20:36 +00:00
|
|
|
set_32bit_tls(task, FS_TLS, addr);
|
2008-01-30 12:31:03 +00:00
|
|
|
if (doit) {
|
|
|
|
load_TLS(&task->thread, cpu);
|
2008-06-25 04:19:00 +00:00
|
|
|
loadsegment(fs, FS_TLS_SEL);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
task->thread.fsindex = FS_TLS_SEL;
|
|
|
|
task->thread.fs = 0;
|
2008-01-30 12:31:03 +00:00
|
|
|
} else {
|
2005-04-16 22:20:36 +00:00
|
|
|
task->thread.fsindex = 0;
|
|
|
|
task->thread.fs = addr;
|
|
|
|
if (doit) {
|
|
|
|
/* set the selector to 0 to not confuse
|
|
|
|
__switch_to */
|
2008-06-25 04:19:00 +00:00
|
|
|
loadsegment(fs, 0);
|
2005-11-05 16:25:54 +00:00
|
|
|
ret = checking_wrmsrl(MSR_FS_BASE, addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
put_cpu();
|
|
|
|
break;
|
2008-01-30 12:31:03 +00:00
|
|
|
case ARCH_GET_FS: {
|
|
|
|
unsigned long base;
|
2005-04-16 22:20:36 +00:00
|
|
|
if (task->thread.fsindex == FS_TLS_SEL)
|
|
|
|
base = read_32bit_tls(task, FS_TLS);
|
2005-11-05 16:25:54 +00:00
|
|
|
else if (doit)
|
2005-04-16 22:20:36 +00:00
|
|
|
rdmsrl(MSR_FS_BASE, base);
|
2005-11-05 16:25:54 +00:00
|
|
|
else
|
2005-04-16 22:20:36 +00:00
|
|
|
base = task->thread.fs;
|
2008-01-30 12:31:03 +00:00
|
|
|
ret = put_user(base, (unsigned long __user *)addr);
|
|
|
|
break;
|
2005-04-16 22:20:36 +00:00
|
|
|
}
|
2008-01-30 12:31:03 +00:00
|
|
|
case ARCH_GET_GS: {
|
2005-04-16 22:20:36 +00:00
|
|
|
unsigned long base;
|
[PATCH] x86_64: Plug GS leak in arch_prctl()
In linux-2.6.16, we have noticed a problem where the gs base value
returned from an arch_prtcl(ARCH_GET_GS, ...) call will be incorrect if:
- the current/calling task has NOT set its own gs base yet to a
non-zero value,
- some other task that ran on the same processor previously set their
own gs base to a non-zero value.
In this situation, the ARCH_GET_GS code will read and return the
MSR_KERNEL_GS_BASE msr register.
However, since the __switch_to() code does NOT load/zero the
MSR_KERNEL_GS_BASE register when the task that is switched IN has a zero
next->gs value, the caller of arch_prctl(ARCH_GET_GS, ...) will get back
the value of some previous tasks's gs base value instead of 0.
Change the arch_prctl() ARCH_GET_GS code to only read and return
the MSR_KERNEL_GS_BASE msr register if the 'gs' register of the calling
task is non-zero.
Side note: Since in addition to using arch_prctl(ARCH_SET_GS, ...),
a task can also setup a gs base value by using modify_ldt() and write
an index value into 'gs' from user space, the patch below reads
'gs' instead of using thread.gs, since in the modify_ldt() case,
the thread.gs value will be 0, and incorrect value would be returned
(the task->thread.gs value).
When the user has not set its own gs base value and the 'gs'
register is zero, then the MSR_KERNEL_GS_BASE register will not be
read and a value of zero will be returned by reading and returning
'task->thread.gs'.
The first patch shown below is an attempt at implementing this
approach.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-07 17:50:25 +00:00
|
|
|
unsigned gsindex;
|
2005-04-16 22:20:36 +00:00
|
|
|
if (task->thread.gsindex == GS_TLS_SEL)
|
|
|
|
base = read_32bit_tls(task, GS_TLS);
|
[PATCH] x86_64: Plug GS leak in arch_prctl()
In linux-2.6.16, we have noticed a problem where the gs base value
returned from an arch_prtcl(ARCH_GET_GS, ...) call will be incorrect if:
- the current/calling task has NOT set its own gs base yet to a
non-zero value,
- some other task that ran on the same processor previously set their
own gs base to a non-zero value.
In this situation, the ARCH_GET_GS code will read and return the
MSR_KERNEL_GS_BASE msr register.
However, since the __switch_to() code does NOT load/zero the
MSR_KERNEL_GS_BASE register when the task that is switched IN has a zero
next->gs value, the caller of arch_prctl(ARCH_GET_GS, ...) will get back
the value of some previous tasks's gs base value instead of 0.
Change the arch_prctl() ARCH_GET_GS code to only read and return
the MSR_KERNEL_GS_BASE msr register if the 'gs' register of the calling
task is non-zero.
Side note: Since in addition to using arch_prctl(ARCH_SET_GS, ...),
a task can also setup a gs base value by using modify_ldt() and write
an index value into 'gs' from user space, the patch below reads
'gs' instead of using thread.gs, since in the modify_ldt() case,
the thread.gs value will be 0, and incorrect value would be returned
(the task->thread.gs value).
When the user has not set its own gs base value and the 'gs'
register is zero, then the MSR_KERNEL_GS_BASE register will not be
read and a value of zero will be returned by reading and returning
'task->thread.gs'.
The first patch shown below is an attempt at implementing this
approach.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-07 17:50:25 +00:00
|
|
|
else if (doit) {
|
2008-06-25 04:19:00 +00:00
|
|
|
savesegment(gs, gsindex);
|
[PATCH] x86_64: Plug GS leak in arch_prctl()
In linux-2.6.16, we have noticed a problem where the gs base value
returned from an arch_prtcl(ARCH_GET_GS, ...) call will be incorrect if:
- the current/calling task has NOT set its own gs base yet to a
non-zero value,
- some other task that ran on the same processor previously set their
own gs base to a non-zero value.
In this situation, the ARCH_GET_GS code will read and return the
MSR_KERNEL_GS_BASE msr register.
However, since the __switch_to() code does NOT load/zero the
MSR_KERNEL_GS_BASE register when the task that is switched IN has a zero
next->gs value, the caller of arch_prctl(ARCH_GET_GS, ...) will get back
the value of some previous tasks's gs base value instead of 0.
Change the arch_prctl() ARCH_GET_GS code to only read and return
the MSR_KERNEL_GS_BASE msr register if the 'gs' register of the calling
task is non-zero.
Side note: Since in addition to using arch_prctl(ARCH_SET_GS, ...),
a task can also setup a gs base value by using modify_ldt() and write
an index value into 'gs' from user space, the patch below reads
'gs' instead of using thread.gs, since in the modify_ldt() case,
the thread.gs value will be 0, and incorrect value would be returned
(the task->thread.gs value).
When the user has not set its own gs base value and the 'gs'
register is zero, then the MSR_KERNEL_GS_BASE register will not be
read and a value of zero will be returned by reading and returning
'task->thread.gs'.
The first patch shown below is an attempt at implementing this
approach.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-04-07 17:50:25 +00:00
|
|
|
if (gsindex)
|
|
|
|
rdmsrl(MSR_KERNEL_GS_BASE, base);
|
|
|
|
else
|
|
|
|
base = task->thread.gs;
|
2008-07-29 05:48:51 +00:00
|
|
|
} else
|
2005-04-16 22:20:36 +00:00
|
|
|
base = task->thread.gs;
|
2008-01-30 12:31:03 +00:00
|
|
|
ret = put_user(base, (unsigned long __user *)addr);
|
2005-04-16 22:20:36 +00:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
default:
|
|
|
|
ret = -EINVAL;
|
|
|
|
break;
|
2008-01-30 12:31:03 +00:00
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
2008-01-30 12:31:03 +00:00
|
|
|
return ret;
|
|
|
|
}
|
2005-04-16 22:20:36 +00:00
|
|
|
|
|
|
|
long sys_arch_prctl(int code, unsigned long addr)
|
|
|
|
{
|
|
|
|
return do_arch_prctl(current, code, addr);
|
|
|
|
}
|
|
|
|
|
2009-11-03 09:22:40 +00:00
|
|
|
unsigned long KSTK_ESP(struct task_struct *task)
|
|
|
|
{
|
|
|
|
return (test_tsk_thread_flag(task, TIF_IA32)) ?
|
|
|
|
(task_pt_regs(task)->sp) : ((task)->thread.usersp);
|
|
|
|
}
|