During kernel bootup, a new T60 laptop (CoreDuo, 32-bit) hangs about
10%-20% of the time in acpi_init():
Calling initcall 0xc055ce1a: topology_init+0x0/0x2f()
Calling initcall 0xc055d75e: mtrr_init_finialize+0x0/0x2c()
Calling initcall 0xc05664f3: param_sysfs_init+0x0/0x175()
Calling initcall 0xc014cb65: pm_sysrq_init+0x0/0x17()
Calling initcall 0xc0569f99: init_bio+0x0/0xf4()
Calling initcall 0xc056b865: genhd_device_init+0x0/0x50()
Calling initcall 0xc056c4bd: fbmem_init+0x0/0x87()
Calling initcall 0xc056dd74: acpi_init+0x0/0x1ee()
It's a hard hang that not even an NMI could punch through! Frustratingly,
adding printks or function tracing to the ACPI code made the hangs go away
...
After some time an additional detail emerged: disabling the NMI watchdog
made these occasional hangs go away.
So i spent the better part of today trying to debug this and trying out
various theories when i finally found the likely reason for the hang: if
acpi_ns_initialize_devices() executes an _INI AML method and an NMI
happens to hit that AML execution in the wrong moment, the machine would
hang. (my theory is that this must be some sort of chipset setup method
doing stores to chipset mmio registers?)
Unfortunately given the characteristics of the hang it was sheer
impossible to figure out which of the numerous AML methods is impacted
by this problem.
As a workaround i wrote an interface to disable chipset-based NMIs while
executing _INI sections - and indeed this fixed the hang. I did a
boot-loop of 100 separate reboots and none hung - while without the patch
it would hang every 5-10 attempts. Out of caution i did not touch the
nmi_watchdog=2 case (it's not related to the chipset anyway and didnt
hang).
I implemented this for both x86_64 and i686, tested the i686 laptop both
with nmi_watchdog=1 [which triggered the hangs] and nmi_watchdog=2, and
tested an Athlon64 box with the 64-bit kernel as well. Everything builds
and works with the patch applied.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@suse.de>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
P6 CPUs and Core/Core 2 CPUs which has 'architectural perf mon' feature,
only supports write of low 32 bits in Performance Monitoring Counters.
Bits 32..39 are sign extended based on bit 31 and bits 40..63 are reserved
and should be zero.
This patch:
Change x86_64 nmi handler to handle this case cleanly.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Commit f2802e7f57 and its x86 version
(b7471c6da9) adds nmi_known_cpu() check
while parsing boot options in x86_64 and i386.
With that, "nmi_watchdog=2" stops working for me on Intel Core 2 CPU
based system.
The problem is, setup_nmi_watchdog is called while parsing the boot
option and identify_cpu is not done yet. So, the return value of
nmi_known_cpu() is not valid at this point.
So revert that check. This should not have any adverse effect as the
nmi_known_cpu() check is done again later in enable_lapic_nmi_watchdog().
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2.6.19 stopped booting (or booted based on build/config) on our x86_64
systems due to a bug introduced in 2.6.19. check_nmi_watchdog schedules an
IPI on all cpus to busy wait on a flag, but fails to set the busywait
flag if NMI functionality is disabled. This causes the secondary cpus
to spin in an endless loop, causing the kernel bootup to hang.
Depending upon the build, the busywait flag got overwritten (stack variable)
and caused the kernel to bootup on certain builds. Following patch fixes
the bug by setting the busywait flag before returning from check_nmi_watchdog.
I guess using a stack variable is not good here as the calling function could
potentially return while the busy wait loop is still spinning on the flag.
AK: I redid the patch significantly to be cleaner
Signed-off-by: Ravikiran Thirumalai <kiran@scalex86.org>
Signed-off-by: Shai Fultheim <shai@scalex86.org>
Signed-off-by: Andi Kleen <ak@suse.de>
When a spinlock lockup occurs, arrange for the NMI code to emit an all-cpu
backtrace, so we get to see which CPU is holding the lock, and where.
Cc: Andi Kleen <ak@muc.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Andi Kleen <ak@suse.de>
AK: This redoes the changes I temporarily reverted.
Intel now has support for Architectural Performance Monitoring Counters
( Refer to IA-32 Intel Architecture Software Developer's Manual
http://www.intel.com/design/pentium4/manuals/253669.htm ). This
feature is present starting from Intel Core Duo and Intel Core Solo processors.
What this means is, the performance monitoring counters and some performance
monitoring events are now defined in an architectural way (using cpuid).
And there will be no need to check for family/model etc for these architectural
events.
Below is the patch to use this performance counters in nmi watchdog driver.
Patch handles both i386 and x86-64 kernels.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
When a unknown NMI happened the panic would claim a NMI watchdog timeout.
Also it would check the variable set by nmi_watchdog=panic and panic then.
Fix up the panic message to be generic
Unconditionally panic on unknown NMI when panic on unknown nmi is enabled.
Noticed by Jan Beulich
Cc: jbeulich@novell.com
Signed-off-by: Andi Kleen <ak@suse.de>
Making NMI suspend/resume work with SMP. We use CPU hotplug to offline
APs in SMP suspend/resume. Only BSP executes sysdev's .suspend/.resume
method. APs should follow CPU hotplug code path.
And:
+From: Don Zickus <dzickus@redhat.com>
Makes the start/stop paths of nmi watchdog more robust to handle the
suspend/resume cases more gracefully.
AK: I merged the two patches together
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Adds a new /proc/sys/kernel/nmi_watchdog call that will enable/disable the
nmi watchdog.
By entering a non-zero value here, a user can enable the nmi watchdog to
monitor the online cpus in the system. By entering a zero value here, a
user can disable the nmi watchdog and free up a performance counter which
could then be utilized by the oprofile subsystem, otherwise oprofile may be
short a counter when in use.
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Adds a new /proc/sys/kernel/nmi call that will enable/disable the nmi
watchdog.
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Removes the un/set_nmi_callback and reserve/release_lapic_nmi functions as
they are no longer needed. The various subsystems are modified to register
with the die_notifier instead.
Also includes compile fixes by Andrew Morton.
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
This patch cleans up the NMI interrupt path. Instead of being gated by if
the 'nmi callback' is set, the interrupt handler now calls everyone who is
registered on the die_chain and additionally checks the nmi watchdog,
reseting it if enabled. This allows more subsystems to hook into the NMI if
they need to (without being block by set_nmi_callback).
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
This patch includes the changes to make the nmi watchdog on x86_64 SMP
aware. A bunch of code was moved around to make it simpler to read. In
addition, it is now possible to determine if a particular NMI was the result
of the watchdog or not. This feature allows the kernel to filter out
unknown NMIs easier.
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Adds basic infrastructure to allow subsystems to reserve performance
counters on the x86 chips. Only UP kernels are supported in this patch to
make reviewing easier. The SMP portion makes a lot more changes.
Think of this as a locking mechanism where each bit represents a different
counter. In addition, each subsystem should also reserve an appropriate
event selection register that will correspond to the performance counter it
will be using (this is mainly neccessary for the Pentium 4 chips as they
break the 1:1 relationship to performance counters).
This will help prevent subsystems like oprofile from interfering with the
nmi watchdog.
Signed-off-by: Don Zickus <dzickus@redhat.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Make use of local_irq_enable_in_hardirq() API to annotate places that enable
hardirqs in hardirq context.
Has no effect on non-lockdep kernels.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Intel now has support for Architectural Performance Monitoring Counters
( Refer to IA-32 Intel Architecture Software Developer's Manual
http://www.intel.com/design/pentium4/manuals/253669.htm ). This
feature is present starting from Intel Core Duo and Intel Core Solo processors.
What this means is, the performance monitoring counters and some performance
monitoring events are now defined in an architectural way (using cpuid).
And there will be no need to check for family/model etc for these architectural
events.
Below is the patch to use this performance counters in nmi watchdog driver.
Patch handles both i386 and x86-64 kernels.
Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Machine checks can stall the machine for a long time and
it's not good to trigger the nmi watchdog during that.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
While the modular aspect of the respective i386 patch doesn't apply to
x86-64 (as the top level page directory entry is shared between modules
and the base kernel), handlers registered with register_die_notifier()
are still under similar constraints for touching ioremap()ed or
vmalloc()ed memory. The likelihood of this problem becoming visible is
of course significantly lower, as the assigned virtual addresses would
have to cross a 2**39 byte boundary. This is because the callback gets
invoked
(a) in the page fault path before the top level page table propagation
gets carried out (hence a fault to propagate the top level page table
entry/entries mapping to module's code/data would nest infinitly) and
(b) in the NMI path, where nested faults must absolutely not happen,
since otherwise the IRET from the nested fault re-enables NMIs,
potentially resulting in nested NMI occurences.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
When we stop allocating percpu memory for not-possible CPUs we must not touch
the percpu data for not-possible CPUs at all. The correct way of doing this
is to test cpu_possible() or to use for_each_cpu().
This patch is a kernel-wide sweep of all instances of NR_CPUS. I found very
few instances of this bug, if any. But the patch converts lots of open-coded
test to use the preferred helper macros.
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Acked-by: Kyle McMartin <kyle@parisc-linux.org>
Cc: Anton Blanchard <anton@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Andi Kleen <ak@muc.de>
Cc: Christian Zankel <chris@zankel.net>
Cc: Philippe Elie <phil.el@wanadoo.fr>
Cc: Nathan Scott <nathans@sgi.com>
Cc: Jens Axboe <axboe@suse.de>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Along with that, also suppress the memory touching altogether when the
watchdog is not running, to eliminate needless crosstalk. Plus ad a call
to it to make things consistent (one could also consider removing the call
in enable_timer_nmi_watchdog()).
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
A kprobe executes IRET early and that could cause NMI recursion
and stack corruption.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Fix
CC arch/x86_64/kernel/nmi.o
linux/arch/x86_64/kernel/nmi.c: In function ???check_nmi_watchdog???:
linux/arch/x86_64/kernel/nmi.c:155: warning: statement with no effect
on Uniprocessor builds.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Like previously done for i386, get the x86_64 watchdog tick calculation
into a state where it can also be used on CPUs with frequencies beyond
4GHz.
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
... and with that all instances in arch/x86_64 are gone.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The x86_64 nmi code is missing a newline in one of its messages.
I added a space before the CPU id for readability and killed the trailing
space on the previous line as well.
Signed-off-by: Chuck Ebbert <76306.1226@compuserve.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Uses of RCU for dynamically changeable NMI handlers need to use the new
rcu_dereference() and rcu_assign_pointer() facilities. This change makes
it clear that these uses are safe from a memory-barrier viewpoint, but the
main purpose is to document exactly what operations are being protected by
RCU. This has been tested on x86 and x86-64, which are the only
architectures affected by this change.
Signed-off-by: <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch adds a new kernel debug feature: CONFIG_DETECT_SOFTLOCKUP.
When enabled then per-CPU watchdog threads are started, which try to run
once per second. If they get delayed for more than 10 seconds then a
callback from the timer interrupt detects this condition and prints out a
warning message and a stack dump (once per lockup incident). The feature
is otherwise non-intrusive, it doesnt try to unlock the box in any way, it
only gets the debug info out, automatically, and on all CPUs affected by
the lockup.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-Off-By: Matthias Urlichs <smurf@smurf.noris.de>
Signed-off-by: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Fix remaining bits of u32 vs. pm_message confusion. Should not break
anything.
Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch adds __cpuinit and __cpuinitdata sections that need to exist past
boot to support cpu hotplug.
Caveat: This is done *only* for EM64T CPU Hotplug support, on request from
Andi Kleen. Much of the generic hotplug code in kernel, and none of the other
archs that support CPU hotplug today, i386, ia64, ppc64, s390 and parisc dont
mark sections with __cpuinit, but only mark them as __devinit, and
__devinitdata.
If someone is motivated to change generic code, we need to make sure all
existing hotplug code does not break, on other arch's that dont use __cpuinit,
and __cpudevinit.
Signed-off-by: Ashok Raj <ashok.raj@intel.com>
Acked-by: Andi Kleen <ak@muc.de>
Acked-by: Zwane Mwaikambo <zwane@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Collected NMI watchdog fixes.
- Fix call of check_nmi_watchdog
- Remove earlier move of check_nmi_watchdog to later. It does not fix the
race it was supposed to fix fully.
- Remove unused P6 definitions
- Add support for performance counter based watchdog on P4 systems.
This allows to run it only once per second, which saves some CPU time.
Previously it would run at 1000Hz, which was too much.
Code ported from i386
Make this the default on Intel systems.
- Use check_nmi_watchdog with local APIC based nmi
- Fix race in touch_nmi_watchdog
- Fix bug that caused incorrect performance counters to be programmed in a
few cases on K8.
- Remove useless check for local APIC
- Use local_t and per_cpu variables for per CPU data.
- Keep other CPUs busy during check_nmi_watchdog to make sure they really
tick when in lapic mode.
- Only check CPUs that are actually online.
- Various other fixes.
- Fix fallback path when MSRs are unimplemented
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
NR_CPUs can be quite big these days. kmalloc the per CPU array instead of
putting it onto the stack
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
A bug against an xSeries system showed up recently noting that the
check_nmi_watchdog() test was failing.
I have been investigating it and discovered in both i386 and x86_64 the
recent change to the routine to use the cpu_callin_map has uncovered a
problem. Prior to that change, on an SMP box, the test was trivally
passing because all cpu's were found to not yet be online, but now with the
callin_map they are discovered, it goes on to test the counter and they
have not yet begun to increment, so it announces a CPU is stuck and bails
out.
On all the systems I have access to test, the announcement of failure is
also bougs... by the time you can login and check /proc/interrupts, the
NMI count is happily incrementing on all CPUs. Its just that the test is
being done too early.
I have tried moving the call to the test around a bit, and it was always
too early. I finally hit on this proposed solution, it delays the routine
via a late_initcall(), seems like the right solution to me.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I thought I'm done with fixing u32 vs. pm_message_t ... unfortunately that
turned out not to be the case... Here are fixes x86-64.
Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This will allow hotplug CPU in the future and in general cleans up a lot of
crufty code. It also should plug some races that the old hackish way
introduces. Remove one old race workaround in NMI watchdog setup that is not
needed anymore.
I removed the old total sum of bogomips reporting code. The brag value of
BogoMips has been greatly devalued in the last years on the open market.
Real CPU hotplug will need some more work, but the infrastructure for it is
there now.
One drawback: the new TSC sync algorithm is less accurate than before. The
old way of zeroing TSCs is too intrusive to do later. Instead the TSC of the
BP is duplicated now, which is less accurate.
akpm:
- sync_tsc_bp_init seems to have the sense of `init' inverted.
- SPIN_LOCK_UNLOCKED is deprecated - use DEFINE_SPINLOCK.
Cc: <rusty@rustcorp.com.au>
Cc: <mingo@elte.hu>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The NMI watchdog code did this incorrectly
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
