enhance the initcall_debug boot option:
- measure the time the initcall took to execute and report
it in units of milliseconds.
- show the return code of initcalls (useful to see failures and
to make sure that an initcall hung)
[[email protected]: fix printk warning]
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Remove clustered APIC mode. There's little point in the use of clustered APIC
mode, broadcasting is limited to within the cluster only, and chipsets have
bugs in this area as well. So default to physical APIC mode when the CPU
count is large, and default to logical APIC mode when the CPU count is 8 or
smaller.
(this patch only removes the use of genapic_cluster and cleans up the
resulting genapic.c file - removal of all remaining traces of clustered
mode will be done by another patch.)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
Cc: Suresh Siddha <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: "Li, Shaohua" <[email protected]>
Cc: "Eric W. Biederman" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Fix a couple of inconsistencies/problems I found while reviewing the x86_64
genapic code (when I was chasing mysterious eth0 timeouts that would only
trigger if CPU_HOTPLUG is enabled):
- AMD systems defaulted to the slower flat-physical mode instead
of the flat-logical mode. The only restriction on AMD systems
is that they should not use clustered APIC mode.
- removed the CPU hotplug hacks, switching the default for small
systems back from phys-flat to logical-flat. The switching to logical
flat mode on small systems fixed sporadic ethernet driver timeouts i
was getting on a dual-core Athlon64 system:
NETDEV WATCHDOG: eth0: transmit timed out
eth0: Transmit timeout, status 0c 0005 c07f media 80.
eth0: Tx queue start entry 32 dirty entry 28.
eth0: Tx descriptor 0 is 0008a04a. (queue head)
eth0: Tx descriptor 1 is 0008a04a.
eth0: Tx descriptor 2 is 0008a04a.
eth0: Tx descriptor 3 is 0008a04a.
eth0: link up, 100Mbps, full-duplex, lpa 0xC5E1
- The use of '<= 8' was a bug by itself (the valid APIC ids
for logical flat mode go from 0 to 7, not 0 to 8). The new logic
is to use logical flat mode on both AMD and Intel systems, and
to only switch to physical mode when logical mode cannot be used.
If CPU hotplug is racy wrt. APIC shutdown then CPU hotplug needs
fixing, not the whole IRQ system be made inconsistent and slowed
down.
- minor cleanups: simplified some code constructs
build & booted on a couple of AMD and Intel SMP systems.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
Cc: Suresh Siddha <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: "Li, Shaohua" <[email protected]>
Cc: "Eric W. Biederman" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
show_state() (SysRq-T) developed the buggy habbit of not showing
TASK_RUNNING tasks. This was due to the mistaken belief that state_filter
== -1 would be a pass-through filter - while in reality it did not let
TASK_RUNNING == 0 p->state values through.
Fix this by restoring the original '!state_filter means all tasks'
special-case i had in the original version. Test-built and test-booted on
i686, SysRq-T now works as intended.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
the p->parent PID printout gives us all the information about the
task tree that we need - the eldest_child()/older_sibling()/
younger_sibling() printouts are mostly historic and i do not
remember ever having used those fields. (IMO in fact they confuse
the SysRq-T output.) So remove them.
This code has sentimental value though, those fields and
printouts are one of the oldest ones still surviving from
Linux v0.95's kernel/sched.c:
if (p->p_ysptr || p->p_osptr)
printk(" Younger sib=%d, older sib=%d\n\r",
p->p_ysptr ? p->p_ysptr->pid : -1,
p->p_osptr ? p->p_osptr->pid : -1);
else
printk("\n\r");
written 15 years ago, in early 1992.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus 'snif' Torvalds <[email protected]>
Soeren Sonnenburg reported that upon resume he is getting
this backtrace:
[<c0119637>] smp_apic_timer_interrupt+0x57/0x90
[<c0142d30>] retrigger_next_event+0x0/0xb0
[<c0104d30>] apic_timer_interrupt+0x28/0x30
[<c0142d30>] retrigger_next_event+0x0/0xb0
[<c0140068>] __kfifo_put+0x8/0x90
[<c0130fe5>] on_each_cpu+0x35/0x60
[<c0143538>] clock_was_set+0x18/0x20
[<c0135cdc>] timekeeping_resume+0x7c/0xa0
[<c02aabe1>] __sysdev_resume+0x11/0x80
[<c02ab0c7>] sysdev_resume+0x47/0x80
[<c02b0b05>] device_power_up+0x5/0x10
it turns out that on resume we mistakenly re-enable interrupts too
early. Do the timer retrigger only on the current CPU.
Signed-off-by: Ingo Molnar <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Acked-by: Soeren Sonnenburg <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
hrtimer_start() incorrectly set the 'reprogram' flag to enqueue_hrtimer(),
which should only be 1 if the hrtimer is queued to the current CPU.
Doing otherwise could result in a reprogramming of the current CPU's
clockevents device, with a timer that is not queued to it - resulting in a
bogus next expiry value.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Michal Piotrowski <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
failed VM entry on VMX might still change %fs or %gs, thus make sure
that KVM always reloads the segment selectors. This is crutial on both
x86 and x86_64: x86 has __KERNEL_PDA in %fs on which things like
'current' depends and x86_64 has 0 there and needs MSR_GS_BASE to work.
Signed-off-by: Ingo Molnar <[email protected]>
latest -git triggers an irqtrace/lockdep warning of a leaked
irqs-off condition:
BUG: at kernel/fork.c:1033 copy_process()
after some debugging it turns out that commit ca1b940c accidentally left
interrupts disabled - which trickled down all the way to the first time
we fork a kernel thread and triggered the warning.
the fix is to re-enable interrupts in the 'else' branch of
setup_boot_APIC_clock()'s pmtimers calibration path.
Reported-by: Michal Piotrowski <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
check_tsc_sync_source() depends on being called with irqs disabled (it
checks whether the TSC is coherent across two specific CPUs). This is
incidentally true during bootup, but not during cpu hotplug __cpu_up().
This got found via smp_processor_id() debugging.
disable irqs explicitly and remove the unconditional enabling of
interrupts. Add touch_nmi_watchdog() to the cpu_online_map busy loop.
this bug is present both on i386 and on x86_64.
Reported-by: Michal Piotrowski <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
CONFIG_PARAVIRT broke old glibc bootup: it silently turned off the
selectability of CONFIG_COMPAT_VDSO and thus rendered distro kernels
unbootable on old-style VDSO glibc setups.
the proper solution is to keep COMPAT_VDSO available - if a hypervisor
needs any modification of that concept then we'll judge those changes in
full context, once those changes are submitted.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
there's a new NMI watchdog related problem: KVM crashes on certain
bzImages because ... we enable the NMI watchdog by default (even if the
user does not ask for it) , and no other OS on this planet does that so
KVM doesnt have emulation for that yet. So KVM injects a #GP, which
crashes the Linux guest:
general protection fault: 0000 [#1]
PREEMPT SMP
Modules linked in:
CPU: 0
EIP: 0060:[<c011a8ae>] Not tainted VLI
EFLAGS: 00000246 (2.6.20-rc5-rt0 #3)
EIP is at setup_apic_nmi_watchdog+0x26d/0x3d3
and no, i did /not/ request an nmi_watchdog on the boot command line!
Solution: turn off that darn thing! It's a debug tool, not a 'make life
harder' tool!!
with this patch the KVM guest boots up just fine.
And with this my laptop (Lenovo T60) also stopped its sporadic hard
hanging (sometimes in acpi_init(), sometimes later during bootup,
sometimes much later during actual use) as well. It hung with both
nmi_watchdog=1 and nmi_watchdog=2, so it's generally the fact of NMI
injection that is causing problems, not the NMI watchdog variant, nor
any particular bootup code.
[ NMI breaks on some systems, esp in combination with SMM -Arjan ]
Signed-off-by: Ingo Molnar <[email protected]>
Acked-by: Arjan van de Ven <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
do not use default=y for CONFIG_VMI (we do not do that for any driver or
special-hardware feature): the overwhelming majority of Linux users does
not need it, and interested users and distributions can enable it
as-needed.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Clarify the description of the CONFIG_VMI option: describe the reality
that VMI is a VMWare-only interface for now. Once that changes and
another hypervisor adopts the VMI ABI we can change the text.
As can be seen from the Xen paravirtualization patches submitted to lkml
the Xen project has chosen its own, non-VMI interface between Xen and
the para-Linux - so remove Xen from the description.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Temove the mistaken turning on of NO_IDLE_HZ on x86+PARAVIRT kernels.
It's an obsolete, limited form of dynticks.
Signed-off-by: Ingo Molnar <[email protected]>
Acked-by: Thomas Gleixner <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
This adds a special MSR based hypercall API to KVM. This is to be
used by paravirtual kernels and virtual drivers.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Avi Kivity <[email protected]>
The SMT scheduler incorrectly skips kernel threads even if they are
runnable (but they are preempted by a higher-prio user-space task which got
SMT-delayed by an even higher-priority task running on a sibling CPU).
Fix this for now by only doing the SMT-nice optimization if the
to-be-delayed task is the only runnable task. (This should cover most of
the real-life cases anyway.)
This bug has been in the SMT scheduler since 2.6.17 or so, but has only
been noticed now by the active check in the dynticks code.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Michal Piotrowski <[email protected]>
Cc: Nick Piggin <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Jeremy Fitzhardinge suggested the use of -freg-struct-return, which does
structure-returns (such as when using pte_t) in registers instead of on
the stack.
that is indeed so, and this option reduced the kernel size a bit:
text data bss dec hex filename
4799506 543456 3760128 9103090 8ae6f2 vmlinux.before
4798117 543456 3760128 9101701 8ae185 vmlinux.after
the resulting kernel booted fine on my testbox. Lets go for it.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Thomas is the maintainer and primary author of the high-res timers,
clockevents and dynticks code.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
When a CPU is needed for RCU the tick has to continue even when it was
stopped before.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Now that disable_irq() defaults to delayed-disable semantics, the IRQ_DISABLED
flag is not needed anymore.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Never mask interrupts immediately upon request. Disabling interrupts in
high-performance codepaths is rare, and on the other hand this change could
recover lost edges (or even other types of lost interrupts) by conservatively
only masking interrupts after they happen. (NOTE: with this change the
highlevel irq-disable code still soft-disables this IRQ line - and if such an
interrupt happens then the IRQ flow handler keeps the IRQ masked.)
Mark i8529A controllers as 'never loses an edge'.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
make the TSC synchronization code more robust, and unify it between x86_64 and
i386.
The biggest change is the removal of the 'fix up TSCs' code on x86_64 and
i386, in some rare cases it was /causing/ time-warps on SMP systems.
The new code only checks for TSC asynchronity - and if it can prove a
time-warp (if it can observe the TSC going backwards when going from one CPU
to another within a critical section), then the TSC clock-source is turned
off.
The TSC synchronization-checking code also got moved into a separate file.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: john stultz <[email protected]>
Cc: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Prevent timeout overflow if timer ticks are behind jiffies (due to high
softirq load or due to dyntick), by limiting the valid timeout range to
MAX_LONG/2.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: john stultz <[email protected]>
Cc: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Fix multiple conversion bugs in msecs_to_jiffies().
The main problem is that this condition:
if (m > jiffies_to_msecs(MAX_JIFFY_OFFSET))
overflows if HZ is smaller than 1000!
This change is user-visible: for HZ=250 SUS-compliant poll()-timeout
value of -20 is mistakenly converted to 'immediate timeout'.
(The new dyntick code also triggered this, that's how we noticed.)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: john stultz <[email protected]>
Cc: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
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 <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Len Brown <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Clean up sched_clock() on i686: it will use the TSC if available and falls
back to jiffies only if the user asked for it to be disabled via notsc or
the CPU calibration code didnt figure out the right cpu_khz.
This generally makes the scheduler timestamps more finegrained, on all
hardware. (the current scheduler is pretty resistant against asynchronous
sched_clock() values on different CPUs, it will allow at most up to a jiffy
of jitter.)
Also simplify sched_clock()'s check for TSC availability: propagate the
desire and ability to use the TSC into the tsc_disable flag, previously
this flag only indicated whether the notsc option was passed. This makes
the rare low-res sched_clock() codepath a single branch off a read-mostly
flag.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
vcpu_load() can return NULL and it sometimes does in failure paths (for
example when the userspace ABI version is too old) - causing a preemption
count underflow in the ->vcpu_free() later on. So check for NULL.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Avi Kivity <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Forms like "0(%rsp)" generate an instruction with an unnecessary one byte
displacement under certain circumstances. replace with the equivalent
"(%rsp)".
Signed-off-by: Avi Kivity <[email protected]>
Cc: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
close_files() can sometimes take long enough to trigger the soft lockup
detector.
Cc: Eric Dumazet <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Catch illegally nested kmap_atomic()s even if the page that is mapped by
the 'inner' instance is from lowmem.
This avoids spuriously zapped kmap-atomic ptes and turns hard to find
crashes into clear asserts at the bug site.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
gcc emits this warning:
kernel/auditfilter.c: In function 'audit_filter_user':
kernel/auditfilter.c:1611: warning: 'state' is used uninitialized in this function
I tend to agree with gcc - there are a couple of plausible exit paths from
audit_filter_user_rules() where it does not set 'state', keeping the
variable uninitialized. For example if a filter rule has an AUDIT_POSSIBLE
action. Initialize to 'wont audit'. Fix whitespace damage too.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Al Viro <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
kmem_cache_free() was missing the check for freeing held locks.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Recently cpufreq support on my laptop (Lenovo T60) broke completely: when
it's plugged into AC it would never go higher than 1 GHz - neither 1.3 GHz
nor 1.83 GHz is possible - no matter which governor (userspace, speed or
ondemand) is used.
After some cpufreq debugging i tracked the regression back to the following
(totally correct) bug-fix commit:
commit 0916bd3ebb
Author: Dave Jones <[email protected]>
Date: Wed Nov 22 20:42:01 2006 -0500
[PATCH] Correct bound checking from the value returned from _PPC method.
This bugfix, which makes other laptops work, made a previously hidden
(BIOS) bug visible on my laptop.
The bug is the following: if the _PPC (Performance Present Capabilities)
optional ACPI object is queried /after/ bootup then the BIOS reports an
incorrect value of '2'.
My laptop (Lenovo T60) has the following performance states supported:
0: 1833000
1: 1333000
2: 1000000
Per ACPI specification, a _PPC value of '0' means that all 3 performance
states are usable. A _PPC value of '1' means states 1 .. 2 are usable, a
value of '2' means only state '2' (slowest) is usable.
now, the _PPC object is optional, and it also comes with notification.
Furthermore, when a CPU object is initialized, the _PPC object is
initialized as well. So the following evaluation of the _PPC object is
superfluous:
[<c028ba5f>] acpi_processor_get_platform_limit+0xa1/0xaf
[<c028c040>] acpi_processor_register_performance+0x3b9/0x3ef
[<c0111a85>] acpi_cpufreq_cpu_init+0xb7/0x596
[<c03dab74>] cpufreq_add_dev+0x160/0x4a8
[<c02bed90>] sysdev_driver_register+0x5a/0xa0
[<c03d9c4c>] cpufreq_register_driver+0xb4/0x176
[<c068ac08>] acpi_cpufreq_init+0xe5/0xeb
[<c010056e>] init+0x14f/0x3dd
And this is the point where my laptop's BIOS returns the incorrect value of
'2'. Note that it has not sent any notification event, so the value is
probably not really intentional (possibly spurious), and Windows likely
doesnt query it after bootup either. Maybe the value is kept at '2'
normally, and is only set to the real value when a true asynchronous event
(such as AC plug event, battery switch, etc.) occurs.
So i /think/ this is a grey area of the ACPI spec: per the letter of the
spec the _PPC value only changes when notified, so there's no reason to
query it after the system has booted up. So in my opinion the best (and
most compatible) strategy would be to do the change below, and to not
evaluate the _PPC object in the acpi_processor_get_performance_info() call,
but only evaluate it if _PPC is present during CPU object init, or if it's
notified during an asynchronous event. This change is more permissive than
the previous logic, so it definitely shouldnt break any existing system.
This also happens to fix my laptop, which is merrily chugging along at
1.83 GHz now. Yay!
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Dave Jones <[email protected]>
Acked-by: Len Brown <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
while lock-profiling the -rt kernel i noticed weird contention during
mmap-intense workloads, and the tracer showed the following gem, in one
of our MM hotpaths:
threaded-2771 1.... 65us : sys_munmap (sysenter_do_call)
threaded-2771 1.... 66us : profile_munmap (sys_munmap)
threaded-2771 1.... 66us : blocking_notifier_call_chain (profile_munmap)
threaded-2771 1.... 66us : rt_down_read (blocking_notifier_call_chain)
ouch! a global rw-semaphore taken in one of the most performance-
sensitive codepaths of the kernel. And i dont even have oprofile
enabled! All distro kernels have CONFIG_PROFILING enabled, so this
scalability problem affects the majority of Linux users.
The fix is to enhance blocking_notifier_call_chain() to only take the
lock if there appears to be work on the call-chain.
With this patch applied i get nicely saturated system, and much higher
munmap performance, on SMP systems.
And as a bonus this also fixes a similar scalability bottleneck in the
thread-exit codepath: profile_task_exit() ...
Signed-off-by: Ingo Molnar <[email protected]>
Acked-by: Peter Zijlstra <[email protected]>
Acked-by: Nick Piggin <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Prevent the guest's loading of a corrupt cr3 (pointing at no guest phsyical
page) from crashing the host.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Avi Kivity <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
KVM does kmalloc() in an atomic section while having preemption disabled via
vcpu_load(). Fix this by moving the ->*_msr setup from the vcpu_setup method
to the vcpu_create method.
(This is also a small speedup for setting up a vcpu, which can in theory be
more frequent than the vcpu_create method).
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Avi Kivity <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
fix an GFP_KERNEL allocation in atomic section: kvm_dev_ioctl_create_vcpu()
called kvm_mmu_init(), which calls alloc_pages(), while holding the vcpu.
The fix is to set up the MMU state in two phases: kvm_mmu_create() and
kvm_mmu_setup().
(NOTE: free_vcpus does an kvm_mmu_destroy() call so there's no need for any
extra teardown branch on allocation/init failure here.)
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Avi Kivity <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Remove the __resched_legal() check: it is conceptually broken. The biggest
problem it had is that it can mask buggy cond_resched() calls. A
cond_resched() call is only legal if we are not in an atomic context, with
two narrow exceptions:
- if the system is booting
- a reacquire_kernel_lock() down() done while PREEMPT_ACTIVE is set
But __resched_legal() hid this and just silently returned whenever
these primitives were called from invalid contexts. (Same goes for
cond_resched_locked() and cond_resched_softirq()).
Furthermore, the __legal_resched(0) call was buggy in that it caused
unnecessarily long softirq latencies via cond_resched_softirq(). (which is
only called from softirq-off sections, hence the code did nothing.)
The fix is to resurrect the efficiency of the might_sleep checks and to
only allow the narrow exceptions.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
WARN_ON() ever triggering is a kernel bug. Do not try to paper over this
fact by suggesting to the user that this is 'only' a warning, as the
following recent commit does:
commit 30e25b71e7
Author: Jeremy Fitzhardinge <[email protected]>
Date: Fri Dec 8 02:36:24 2006 -0800
[PATCH] Fix generic WARN_ON message
A warning is a warning, not a BUG.
( it might make sense to rename BUG() to CRASH() and BUG_ON() to
CRASH_ON(), but that does not change the fact that WARN_ON()
signals a kernel bug. )
i and others objected to this change during lkml review:
http://marc.theaimsgroup.com/?l=linux-kernel&m=116115160710533&w=2
still the change slipped upstream - grumble :)
Also, use the standard "BUG: " format to make it easier to grep logs and
to make it easier to google for kernel bugs.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Jeremy Fitzhardinge <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Clark Williams reported that suspend doesnt work on his laptop on
2.6.20-rc1-rt kernels. The bug was introduced by the following cleanup
commit:
commit 112cecb2cc
Author: Siddha, Suresh B <[email protected]>
Date: Wed Dec 6 20:34:31 2006 -0800
[PATCH] suspend: don't change cpus_allowed for task initiating the suspend
because with this change 'error' is not initialized to 0 anymore, if
there are no other online CPUs. (i.e. if the system is single-CPU).
the fix is the initialize it to 0. The really weird thing is that my
version of gcc does not warn about this non-initialized variable
situation ...
(also fix the kernel printk in the error branch, it was missing a
newline)
Reported-by: Clark Williams <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Fernando Lopez-Lezcano reported frequent scheduling latencies and audio
xruns starting at the 2.6.18-rt kernel, and those problems persisted all
until current -rt kernels. The latencies were serious and unjustified by
system load, often in the milliseconds range.
After a patient and heroic multi-month effort of Fernando, where he
tested dozens of kernels, tried various configs, boot options,
test-patches of mine and provided latency traces of those incidents, the
following 'smoking gun' trace was captured by him:
_------=> CPU#
/ _-----=> irqs-off
| / _----=> need-resched
|| / _---=> hardirq/softirq
||| / _--=> preempt-depth
|||| /
||||| delay
cmd pid ||||| time | caller
\ / ||||| \ | /
IRQ_19-1479 1D..1 0us : __trace_start_sched_wakeup (try_to_wake_up)
IRQ_19-1479 1D..1 0us : __trace_start_sched_wakeup <<...>-5856> (37 0)
IRQ_19-1479 1D..1 0us : __trace_start_sched_wakeup (c01262ba 0 0)
IRQ_19-1479 1D..1 0us : resched_task (try_to_wake_up)
IRQ_19-1479 1D..1 0us : __spin_unlock_irqrestore (try_to_wake_up)
...
<idle>-0 1...1 11us!: default_idle (cpu_idle)
...
<idle>-0 0Dn.1 602us : smp_apic_timer_interrupt (c0103baf 1 0)
...
<...>-5856 0D..2 618us : __switch_to (__schedule)
<...>-5856 0D..2 618us : __schedule <<idle>-0> (20 162)
<...>-5856 0D..2 619us : __spin_unlock_irq (__schedule)
<...>-5856 0...1 619us : trace_stop_sched_switched (__schedule)
<...>-5856 0D..1 619us : trace_stop_sched_switched <<...>-5856> (37 0)
what is visible in this trace is that CPU#1 ran try_to_wake_up() for
PID:5856, it placed PID:5856 on CPU#0's runqueue and ran resched_task()
for CPU#0. But it decided to not send an IPI that no CPU - due to
TS_POLLING. But CPU#0 never woke up after its NEED_RESCHED bit was set,
and only rescheduled to PID:5856 upon the next lapic timer IRQ. The
result was a 600+ usecs latency and a missed wakeup!
the bug turned out to be an idle-wakeup bug introduced into the mainline
kernel this summer via an optimization in the x86_64 tree:
commit 495ab9c045
Author: Andi Kleen <[email protected]>
Date: Mon Jun 26 13:59:11 2006 +0200
[PATCH] i386/x86-64/ia64: Move polling flag into thread_info_status
During some profiling I noticed that default_idle causes a lot of
memory traffic. I think that is caused by the atomic operations
to clear/set the polling flag in thread_info. There is actually
no reason to make this atomic - only the idle thread does it
to itself, other CPUs only read it. So I moved it into ti->status.
the problem is this type of change:
if (!hlt_counter && boot_cpu_data.hlt_works_ok) {
- clear_thread_flag(TIF_POLLING_NRFLAG);
+ current_thread_info()->status &= ~TS_POLLING;
smp_mb__after_clear_bit();
while (!need_resched()) {
local_irq_disable();
this changes clear_thread_flag() to an explicit clearing of TS_POLLING.
clear_thread_flag() is defined as:
clear_bit(flag, &ti->flags);
and clear_bit() is a LOCK-ed atomic instruction on all x86 platforms:
static inline void clear_bit(int nr, volatile unsigned long * addr)
{
__asm__ __volatile__( LOCK_PREFIX
"btrl %1,%0"
hence smp_mb__after_clear_bit() is defined as a simple compile barrier:
#define smp_mb__after_clear_bit() barrier()
but the explicit TS_POLLING clearing introduced by the patch:
+ current_thread_info()->status &= ~TS_POLLING;
is not an atomic op! So the clearing of the TS_POLLING bit is freely
reorderable with the reading of the NEED_RESCHED bit - and both now
reside in different memory addresses.
CPU idle wakeup very much depends on ordered memory ops, the clearing of
the TS_POLLING flag must always be done before we test need_resched()
and hit the idle instruction(s). [Symmetrically, the wakeup code needs
to set NEED_RESCHED before it tests the TS_POLLING flag, so memory
ordering is paramount.]
Fernando's dual-core Athlon64 system has a sufficiently advanced memory
ordering model so that it triggered this scenario very often.
( And it also turned out that the reason why these latencies never
triggered on my testsystems is that i routinely use idle=poll, which
was the only idle variant not affected by this bug. )
The fix is to change the smp_mb__after_clear_bit() to an smp_mb(), to
act as an absolute barrier between the TS_POLLING write and the
NEED_RESCHED read. This affects almost all idling methods (default,
ACPI, APM), on all 3 x86 architectures: i386, x86_64, ia64.
Signed-off-by: Ingo Molnar <[email protected]>
Tested-by: Fernando Lopez-Lezcano <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Matthew Wilcox noticed that the debug_locks_silent use should be inverted
in DEBUG_LOCKS_WARN_ON(). This bug was causing spurious stacktraces and
incorrect failures in the locking self-test on the parisc kernel.
Bug-found-by: Matthew Wilcox <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
fix the schedule_on_each_cpu() implementation: __queue_work() is now
stricter, hence set the work-pending bit before passing in the new work.
(found in the -rt tree, using Peter Zijlstra's files-lock scalability
patchset)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
if CONFIG_CALGARY_IOMMU is built into the kernel via
CONFIG_CALGARY_IOMMU_ENABLED_BY_DEFAULT, or is enabled via the
iommu=calgary boot option, then the detect_calgary() function runs to
detect the presence of a Calgary IOMMU.
detect_calgary() first searches the BIOS EBDA area for a "rio_table_hdr"
BIOS table. It has this parsing algorithm for the EBDA:
while (offset) {
...
/* The next offset is stored in the 1st word. 0 means no more */
offset = *((unsigned short *)(ptr + offset));
}
got that? Lets repeat it slowly: we've got a BIOS-supplied data
structure, plus Linux kernel code that will only break out of an
infinite parsing loop once the BIOS gives a zero offset. Ok?
Translation: what an excellent opportunity for BIOS writers to lock up
the Linux boot process in an utterly hard to debug place! Indeed the
BIOS jumped on that opportunity on my box, which has the following EBDA
chaining layout:
384, 65282, 65535, 65535, 65535, 65535, 65535, 65535 ...
see the pattern? So my, definitely non-Calgary system happily locks up
in detect_calgary()!
the patch below fixes the boot hang by trusting the BIOS-supplied data
structure a bit less: the parser always has to make forward progress,
and if it doesnt, we break out of the loop and i get the expected kernel
message:
Calgary: Unable to locate Rio Grande Table in EBDA - bailing!
Signed-off-by: Ingo Molnar <[email protected]>
Acked-by: Muli Ben-Yehuda <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
one of my boxes didnt boot the 2.6.20-rc1-rt0 kernel rpm, it hung during
early bootup. After an hour or two of happy debugging i narrowed it down
to the CALGARY_IOMMU_ENABLED_BY_DEFAULT option, which was freshly added
to 2.6.20 via the x86_64 tree and /enabled by default/.
commit bff6547bb6 claims:
[PATCH] Calgary: allow compiling Calgary in but not using it by default
This patch makes it possible to compile Calgary in but not use it by
default. In this mode, use 'iommu=calgary' to activate it.
but the change does not actually practice it:
config CALGARY_IOMMU_ENABLED_BY_DEFAULT
bool "Should Calgary be enabled by default?"
default y
depends on CALGARY_IOMMU
help
Should Calgary be enabled by default? if you choose 'y', Calgary
will be used (if it exists). If you choose 'n', Calgary will not be
used even if it exists. If you choose 'n' and would like to use
Calgary anyway, pass 'iommu=calgary' on the kernel command line.
If unsure, say Y.
it's both 'default y', and says "If unsure, say Y". Clearly not a typo.
disabling this option makes my box boot again. The patch below fixes the
Kconfig entry. Grumble.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Jarek Poplawski noticed that lockdep global state could be accessed in a
racy way if one CPU did a lockdep assert (shutting lockdep down), while the
other CPU would try to do something that changes its global state.
This patch fixes those races and cleans up lockdep's internal locking by
adding a graph_lock()/graph_unlock()/debug_locks_off_graph_unlock helpers.
(Also note that as we all know the Linux kernel is, by definition, bug-free
and perfect, so this code never triggers, so these fixes are highly
theoretical. I wrote this patch for aesthetic reasons alone.)
[[email protected]: build fix]
[[email protected]: build fix's refix]
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Jarek Poplawski <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
When we print an assert due to scheduling-in-atomic bugs, and if lockdep
is enabled, then the IRQ tracing information of lockdep can be printed
to pinpoint the code location that disabled interrupts. This saved me
quite a bit of debugging time in cases where the backtrace did not
identify the irq-disabling site well enough.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
CONFIG_DEBUG_LOCKDEP is unacceptably slow because it does not utilize
the chain-hash. Turn the chain-hash back on in this case too.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Cleanup: the VERY_VERBOSE define was unnecessarily dependent on #ifdef VERBOSE
- while the VERBOSE switch is 0 or 1 (always defined).
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Clear all the chains during lockdep_reset(). This fixes some locking-selftest
false positives i saw on -rt. (never saw those on mainline though, but it
could happen.)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Make verbose lockdep messages (off by default) more informative by printing
out the hash chain key. (this patch was what helped me catch the earlier
lockdep hash-collision bug)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Fix typo in the class_filter() function. (filtering is not used by default so
this only affects lockdep-internal debugging cases)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Most distributions enable sysrq support but set it to 0 by default. Add a
sysrq_always_enabled boot option to always-enable sysrq keys. Useful for
debugging - without having to modify the disribution's config files (which
might not be possible if the kernel is on a live CD, etc.).
Also, while at it, clean up the sysrq interfaces.
[[email protected]: make sysrq_always_enabled_setup() static]
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Adrian Bunk <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
current -git doesnt boot on my laptop due to netpoll not unlocking the
tx lock in the else branch.
booted this up on my laptop with lockdep enabled and there are no
locking complaints and it works fine.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
seqlock_init() needs to use spin_lock_init() for dynamic locks, so that
lockdep is notified about the presence of a new lock.
(this is a fallout of the recent networking merge, which started using
the so-far unused seqlock_init() API.)
This fix solves the following lockdep-internal warning on current -git:
INFO: trying to register non-static key.
the code is fine but needs lockdep annotation.
turning off the locking correctness validator.
__lock_acquire+0x10c/0x9f9
lock_acquire+0x56/0x72
_spin_lock+0x35/0x42
neigh_destroy+0x9d/0x12e
neigh_periodic_timer+0x10a/0x15c
run_timer_softirq+0x126/0x18e
__do_softirq+0x6b/0xe6
do_softirq+0x64/0xd2
ksoftirqd+0x82/0x138
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Sometimes the kernel prints something interesting while userspace bootup
keeps messages turned off via loglevel. Enable the printing of /all/
kernel messages via the "ignore_loglevel" boot option. Off by default.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The hash_lock must only ever be taken with irqs disabled. This happens in
all the important places, except one codepath: register_lock_class(). The
race should trigger rarely because register_lock_class() is quite rare and
single-threaded (happens during init most of the time).
The fix is to disable irqs.
( bug found live in -rt: there preemption is alot more agressive and
preempting with the hash-lock held caused a lockup.)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Make the locking self-test failures (of 'FAILURE' type) easier to debug by
printing more information.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Some have reported a chain-table overflow - double its size.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
There was lots of #ifdef noise in the kernel due to hotcpu_notifier(fn,
prio) not correctly marking 'fn' as used in the !HOTPLUG_CPU case, and thus
generating compiler warnings of unused symbols, hence forcing people to add
#ifdefs.
the compiler can skip truly unused functions just fine:
text data bss dec hex filename
1624412 728710 3674856 6027978 5bfaca vmlinux.before
1624412 728710 3674856 6027978 5bfaca vmlinux.after
[[email protected]: topology.c fix]
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Implement prof=sleep profiling. TASK_UNINTERRUPTIBLE sleeps will be taken
as a profile hit, and every millisecond spent sleeping causes a profile-hit
for the call site that initiated the sleep.
Sample readprofile output on i386:
306 ps2_sendbyte 1.3973
432 call_usermodehelper_keys 1.9548
484 ps2_command 0.6453
790 __driver_attach 4.7879
1593 msleep 44.2500
3976 sync_buffer 64.1290
4076 do_lookup 12.4648
8587 sync_page 122.6714
20820 total 0.0067
(NOTE: architectures need to check whether get_wchan() can be called from
deep within the wakeup path.)
akpm: we need to mark more functions __sched. lock_sock(), msleep(), others..
akpm: the contention in do_lookup() is a surprise. Presumably doing disk
reads for directory contents while holding i_mutex.
[[email protected]: various fixes]
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Commit 2c8c0e6b8d ("[PATCH] Convert x86-64
to early param") broke the earlyprintk=...,keep feature.
This restores that functionality. Tested on x86_64. Must-have for
v2.6.19, no risk.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
lockdep got confused by certain locks in modules:
INFO: trying to register non-static key.
the code is fine but needs lockdep annotation.
turning off the locking correctness validator.
Call Trace:
[<ffffffff8026f40d>] dump_trace+0xaa/0x3f2
[<ffffffff8026f78f>] show_trace+0x3a/0x60
[<ffffffff8026f9d1>] dump_stack+0x15/0x17
[<ffffffff802abfe8>] __lock_acquire+0x724/0x9bb
[<ffffffff802ac52b>] lock_acquire+0x4d/0x67
[<ffffffff80267139>] rt_spin_lock+0x3d/0x41
[<ffffffff8839ed3f>] :ip_conntrack:__ip_ct_refresh_acct+0x131/0x174
[<ffffffff883a1334>] :ip_conntrack:udp_packet+0xbf/0xcf
[<ffffffff8839f9af>] :ip_conntrack:ip_conntrack_in+0x394/0x4a7
[<ffffffff8023551f>] nf_iterate+0x41/0x7f
[<ffffffff8025946a>] nf_hook_slow+0x64/0xd5
[<ffffffff802369a2>] ip_rcv+0x24e/0x506
[...]
Steven Rostedt found the bug: static_obj() check did not take
PERCPU_ENOUGH_ROOM into account, so in-module DEFINE_PER_CPU-area locks
were triggering this message.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Steven Rostedt <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
The scheduler on Andreas Friedrich's hyperthreading system stopped
working properly: the scheduler would never move tasks to another CPU!
The lask known working kernel was 2.6.8.
After a couple of attempts to corner the bug, the following smoking gun
was found:
BIOS reported wrong ACPI idfor the processor
CPU#1: set_cpus_allowed(), swapper:1, 3 -> 2
[<c0103bbe>] show_trace_log_lvl+0x34/0x4a
[<c0103ceb>] show_trace+0x2c/0x2e
[<c01045f8>] dump_stack+0x2b/0x2d
[<c0116a77>] set_cpus_allowed+0x52/0xec
[<c0101d86>] cpu_idle_wait+0x2e/0x100
[<c0259c57>] acpi_processor_power_exit+0x45/0x58
[<c0259752>] acpi_processor_remove+0x46/0xea
[<c025c6fb>] acpi_start_single_object+0x47/0x54
[<c025cee5>] acpi_bus_register_driver+0xa4/0xd3
[<c04ab2d7>] acpi_processor_init+0x57/0x77
[<c01004d7>] init+0x146/0x2fd
[<c0103a87>] kernel_thread_helper+0x7/0x10
a quick look at cpu_idle_wait() shows how broken that code is
on i386: it changes the init task's affinity map but never
restores it ...
and because all userspace tasks get forked by init, they all
inherited that single-CPU affinity mask. x86_64 cloned this
bug too.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Andreas Friedrich <[email protected]>
Cc: Wolfgang Erig <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Adrian Bunk <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
the new dwarf2 unwinder crashes while trying to dump the stack:
Leftover inexact backtrace:
Unable to handle kernel paging request at ffffffff82800000 RIP:
[<ffffffff8026cf26>] dump_trace+0x35b/0x3d2
PGD 203027 PUD 205027 PMD 0
Oops: 0000 [2] PREEMPT SMP
CPU 0
Modules linked in:
Pid: 30, comm: khelper Not tainted 2.6.19-rc6-rt1 #11
RIP: 0010:[<ffffffff8026cf26>] [<ffffffff8026cf26>] dump_trace+0x35b/0x3d2
RSP: 0000:ffff81003fb9d848 EFLAGS: 00010006
RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffffffff805b3520 RDI: 0000000000000000
RBP: ffffffff827ffff9 R08: ffffffff80aad000 R09: 0000000000000005
R10: ffffffff80aae000 R11: ffffffff8037961b R12: ffff81003fb9d858
R13: 0000000000000000 R14: ffffffff80598460 R15: ffffffff80ab1fc0
FS: 0000000000000000(0000) GS:ffffffff806c4200(0000) knlGS:0000000000000000
CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b
CR2: ffffffff82800000 CR3: 0000000000201000 CR4: 00000000000006e0
this crash happened because it did not sanitize the dwarf2 data it
got, and got an unaligned stack pointer - which happily walked past
the process stack (and eventually reached the end of kernel memory
and pagefaulted there) due to this naive iteration condition:
HANDLE_STACK (((long) stack & (THREAD_SIZE-1)) != 0);
note that i386 is alot more conservative when it comes to trusting
stack pointers:
static inline int valid_stack_ptr(struct thread_info *tinfo, void *p)
{
return p > (void *)tinfo &&
p < (void *)tinfo + THREAD_SIZE - 3;
}
but the x86_64 code did not take this bit of i386 code.
The fix is to align the stack pointer.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Andi Kleen <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
on x86_64, the CONFIG_CC_STACKPROTECTOR build fails if used in a
distcc setup that has "CC" defined to "distcc gcc":
gcc: gcc: linker input file unused because linking not done
gcc: gcc: linker input file unused because linking not done
gcc: gcc: linker input file unused because linking not done
this is because the gcc-x86_64-has-stack-protector.sh script
has a 2-parameters assumption. Fix this by passing $(CC) as
a single parameter.
Signed-off-by: Ingo Molnar <[email protected]>
Please-Use-Me-More: make randconfig
Signed-off-by: Linus Torvalds <[email protected]>
In general, lockdep warnings are intended to be non-fatal, so I have put in
various practical limits on internal data structure failure modes. We haven't
had a /single/ lockdep-internal crash ever since lockdep went upstream [the
unwinder crashes are outside of lockdep], and that's largely due to the good
internal checks it does.
Recursion within the dependency graph is currently limited to 20, that's
probably not enough on some many-CPU boxes - this patch doubles it to 40. I
have written the lockdep functions to have as small stackframes as possible,
so 40 should be OK too. (The practical recursion limit should be somewhere
between 100 and 200 entries. If we hit that then I'll change the algorithm to
be iteration-based. Graph walking logic is so easy to program via recursion,
so i'd like to keep recursion as long as possible.)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Bug reported and fixed by Tilman Schmidt <[email protected]>: if lockdep is
enabled then log messages make it to /var/log/messages belatedly. The
reason is a missed wakeup of klogd.
Initially there was only a lockdep_internal() protection against lockdep
recursion within vprintk() - it grew the 'outer' lockdep_off()/on()
protection only later on. But that lockdep_off() made the
release_console_sem() within vprintk() always happen under the
lockdep_internal() condition, causing the bug.
The right solution to remove the inner protection against recursion here -
the outer one is enough.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Tilman Schmidt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
CONFIG_M386 turns on spinlock-based generic rwsems - which surprises the
semaphore.S rwsem stubs. Tested both with and without CONFIG_M386.
Reported-by: Klaus Knopper <[email protected]>
Triaged-by: Adrian Bunk <[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
This patch converts all the i386 PIC controllers (except VisWS and Voyager,
which I could not test - but which should still work as old-style IRQ layers)
to the new and simpler irq-chip interrupt handling layer.
[[email protected]: build fix]
[[email protected]: enable fasteoi handler for i386 level-triggered IO-APIC irqs]
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Cc: Roland Dreier <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
With CONFIG_DEBUG_LOCK_ALLOC turned off i was getting sporadic failures in
the locking self-test:
------------>
| Locking API testsuite:
----------------------------------------------------------------------------
| spin |wlock |rlock |mutex | wsem | rsem |
--------------------------------------------------------------------------
A-A deadlock: ok | ok | ok | ok | ok | ok |
A-B-B-A deadlock: ok | ok | ok | ok | ok | ok |
A-B-B-C-C-A deadlock: ok | ok | ok | ok | ok | ok |
A-B-C-A-B-C deadlock: ok | ok | ok | ok | ok | ok |
A-B-B-C-C-D-D-A deadlock: ok |FAILED| ok | ok | ok | ok |
A-B-C-D-B-D-D-A deadlock: ok | ok | ok | ok | ok | ok |
A-B-C-D-B-C-D-A deadlock: ok | ok | ok | ok | ok |FAILED|
after much debugging it turned out to be caused by accidental chain-hash
key collisions. The current hash is:
#define iterate_chain_key(key1, key2) \
(((key1) << MAX_LOCKDEP_KEYS_BITS/2) ^ \
((key1) >> (64-MAX_LOCKDEP_KEYS_BITS/2)) ^ \
(key2))
where MAX_LOCKDEP_KEYS_BITS is 11. This hash is pretty good as it will
shift by 5 bits in every iteration, where every new ID 'mixed' into the
hash would have up to 11 bits. But because there was a 6 bits overlap
between subsequent IDs and their high bits tended to be similar, there was
a chance for accidental chain-hash collision for a low number of locks
held.
the solution is to shift by 11 bits:
#define iterate_chain_key(key1, key2) \
(((key1) << MAX_LOCKDEP_KEYS_BITS) ^ \
((key1) >> (64-MAX_LOCKDEP_KEYS_BITS)) ^ \
(key2))
This keeps the hash perfect up to 5 locks held, but even above that the
hash is still good because 11 bits is a relative prime to the total 64
bits, so a complete match will only occur after 64 held locks (which doesnt
happen in Linux). Even after 5 locks held, entropy of the 5 IDs mixed into
the hash is already good enough so that overlap doesnt generate a colliding
hash ID.
with this change the false positives went away.
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
while porting the -rt tree to 2.6.18-rc7 i noticed the following
screaming-IRQ scenario on an SMP system:
2274 0Dn.:1 0.001ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.010ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.020ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.029ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.039ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.048ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.058ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.068ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.077ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.087ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
2274 0Dn.:1 0.097ms: do_IRQ+0xc/0x103 <= (ret_from_intr+0x0/0xf)
as it turns out, the bug is caused by handle_level_irq(), which if it
races with another CPU already handling this IRQ, it _unmasks_ the IRQ
line on the way out. This is not how 2.6.17 works, and we introduced
this bug in one of the early genirq cleanups right before it went into
-mm. (the bug was not in the genirq patchset for a long time, and we
didnt notice the bug due to the lack of -rt rebase to the new genirq
code. -rt, and hardirq-preemption in particular opens up such races much
wider than anything else.)
Signed-off-by: Ingo Molnar <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Acked-by: Benjamin Herrenschmidt <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Miles Lane reported the "BUG: MAX_STACK_TRACE_ENTRIES too low!" message,
which means that during normal use his system produced enough lockdep
events so that the 128-thousand entries stack-trace array got exhausted.
Double the size of the array.
Signed-off-by: Ingo Molnar <[email protected]>
Cc: Miles Lane <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
Ingo Molnar