linux-2.6
15 years agoperf_counter: Move child perfcounter init to after scheduler init
Ingo Molnar [Tue, 19 May 2009 13:50:30 +0000 (15:50 +0200)] 
perf_counter: Move child perfcounter init to after scheduler init

Initialize a task's perfcounters (inherit from parent, etc.) after
the child task's scheduler fields have been initialized already.

[ Impact: cleanup ]

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf top: Reduce display overhead
Mike Galbraith [Mon, 25 May 2009 07:57:56 +0000 (09:57 +0200)] 
perf top: Reduce display overhead

Iterate over the symbol table once per display interval, and
copy/sort/tally/decay only those symbols which are active.

Before:

 top - 10:14:53 up  4:08, 17 users,  load average: 1.17, 1.53, 1.49
 Tasks: 273 total,   5 running, 268 sleeping,   0 stopped,   0 zombie
 Cpu(s):  6.9%us, 38.2%sy,  0.0%ni, 19.9%id,  0.0%wa,  0.0%hi, 35.0%si,  0.0%st

   PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  P COMMAND
 28504 root      20   0  1044  260  164 S   58  0.0   0:04.19 2 netserver
 28499 root      20   0  1040  412  316 R   51  0.0   0:04.15 0 netperf
 28500 root      20   0  1040  408  316 R   50  0.0   0:04.14 1 netperf
 28503 root      20   0  1044  260  164 S   50  0.0   0:04.01 1 netserver
 28501 root      20   0  1044  260  164 S   49  0.0   0:03.99 0 netserver
 28502 root      20   0  1040  412  316 S   43  0.0   0:03.96 2 netperf
 28468 root      20   0 1892m 325m  972 S   16 10.8   0:10.50 3 perf
 28467 root      20   0 1892m 325m  972 R    2 10.8   0:00.72 3 perf

After:

 top - 10:16:30 up  4:10, 17 users,  load average: 2.27, 1.88, 1.62
 Tasks: 273 total,   6 running, 267 sleeping,   0 stopped,   0 zombie
 Cpu(s):  2.5%us, 39.7%sy,  0.0%ni, 24.6%id,  0.0%wa,  0.0%hi, 33.3%si,  0.0%st

   PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  P COMMAND
 28590 root      20   0  1040  412  316 S   54  0.0   0:07.85 2 netperf
 28589 root      20   0  1044  260  164 R   54  0.0   0:07.84 0 netserver
 28588 root      20   0  1040  412  316 R   50  0.0   0:07.89 1 netperf
 28591 root      20   0  1044  256  164 S   50  0.0   0:07.82 1 netserver
 28587 root      20   0  1040  408  316 R   47  0.0   0:07.61 0 netperf
 28592 root      20   0  1044  260  164 R   47  0.0   0:07.85 2 netserver
 28378 root      20   0  8732 1300  860 R    2  0.0   0:01.81 3 top
 28577 root      20   0 1892m 165m  972 R    2  5.5   0:00.48 3 perf
 28578 root      20   0 1892m 165m  972 S    2  5.5   0:00.04 3 perf

[ Impact: optimization ]

Signed-off-by: Mike Galbraith <efault@gmx.de>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter tools: increase limits, fix
Ingo Molnar [Mon, 25 May 2009 07:59:50 +0000 (09:59 +0200)] 
perf_counter tools: increase limits, fix

NR_CPUS and NR_COUNTERS goes up quadratic ... 1024x4096 was far
too ambitious upper limit - go for 256x256 which is still plenty.

[ Impact: reduce perf tool memory consumption ]

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Increase mmap limit
Ingo Molnar [Sun, 24 May 2009 07:02:37 +0000 (09:02 +0200)] 
perf_counter: Increase mmap limit

In a default 'perf top' run the tool will create a counter for
each online CPU. With enough CPUs this will eventually exhaust
the default limit.

So scale it up with the number of online CPUs.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf top: fix segfault
Mike Galbraith [Sun, 24 May 2009 06:35:49 +0000 (08:35 +0200)] 
perf top: fix segfault

c6eb13 increased stack usage such that perf-top now croaks on startup.

Take event_array and mmap_array off the stack to prevent segfault on boxen
with smallish ulimit -s setting.

Signed-off-by: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Remove perf_counter_context::nr_enabled
Peter Zijlstra [Sat, 23 May 2009 16:29:01 +0000 (18:29 +0200)] 
perf_counter: Remove perf_counter_context::nr_enabled

now that pctrl() no longer disables other people's counters,
remove the PMU cache code that deals with that.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090523163013.032998331@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Change pctrl() behaviour
Peter Zijlstra [Sat, 23 May 2009 16:29:00 +0000 (18:29 +0200)] 
perf_counter: Change pctrl() behaviour

Instead of en/dis-abling all counters acting on a particular
task, en/dis- able all counters we created.

[ v2: fix crash on first counter enable ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090523163012.916937244@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Simplify context cleanup
Peter Zijlstra [Sat, 23 May 2009 16:28:59 +0000 (18:28 +0200)] 
perf_counter: Simplify context cleanup

Use perf_counter_remove_from_context() to remove counters from
the context.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090523163012.796275849@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Fix userspace build
Peter Zijlstra [Sat, 23 May 2009 16:28:58 +0000 (18:28 +0200)] 
perf_counter: Fix userspace build

recent userspace (F11) seems to already include the
linux/unistd.h bits which means we cannot include the version
in the kernel sources due to the header guards being the same.

Ensure we include the kernel version first.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090523163012.739756497@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Sanitize context locking
Peter Zijlstra [Sat, 23 May 2009 16:28:57 +0000 (18:28 +0200)] 
perf_counter: Sanitize context locking

Ensure we're consistent with the context locks.

 context->mutex
   context->lock
     list_{add,del}_counter();

so that either lock is sufficient to stabilize the context.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090523163012.618790733@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Sanitize counter->mutex
Peter Zijlstra [Sat, 23 May 2009 16:28:56 +0000 (18:28 +0200)] 
perf_counter: Sanitize counter->mutex

s/counter->mutex/counter->child_mutex/ and make sure its only
used to protect child_list.

The usage in __perf_counter_exit_task() doesn't appear to be
problematic since ctx->mutex also covers anything related to fd
tear-down.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090523163012.533186528@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Fix dynamic irq_period logging
Peter Zijlstra [Sat, 23 May 2009 16:28:55 +0000 (18:28 +0200)] 
perf_counter: Fix dynamic irq_period logging

We call perf_adjust_freq() from perf_counter_task_tick() which
is is called under the rq->lock causing lock recursion.
However, it's no longer required to be called under the
rq->lock, so remove it from under it.

Also, fix up some related comments.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090523163012.476197912@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter tools: increase limits
Ingo Molnar [Fri, 22 May 2009 16:18:28 +0000 (18:18 +0200)] 
perf_counter tools: increase limits

I tried to run with 300 active counters and the tools bailed out
because our limit was at 64. So increase the counter limit to 1024
and the CPU limit to 4096.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: fix !PERF_COUNTERS build failure
Ingo Molnar [Fri, 22 May 2009 10:32:15 +0000 (12:32 +0200)] 
perf_counter: fix !PERF_COUNTERS build failure

Update the !CONFIG_PERF_COUNTERS prototype too, for
perf_counter_task_sched_out().

[ Impact: build fix ]

Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <18966.10666.517218.332164@cargo.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Optimize context switch between identical inherited contexts
Paul Mackerras [Fri, 22 May 2009 04:27:22 +0000 (14:27 +1000)] 
perf_counter: Optimize context switch between identical inherited contexts

When monitoring a process and its descendants with a set of inherited
counters, we can often get the situation in a context switch where
both the old (outgoing) and new (incoming) process have the same set
of counters, and their values are ultimately going to be added together.
In that situation it doesn't matter which set of counters are used to
count the activity for the new process, so there is really no need to
go through the process of reading the hardware counters and updating
the old task's counters and then setting up the PMU for the new task.

This optimizes the context switch in this situation.  Instead of
scheduling out the perf_counter_context for the old task and
scheduling in the new context, we simply transfer the old context
to the new task and keep using it without interruption.  The new
context gets transferred to the old task.  This means that both
tasks still have a valid perf_counter_context, so no special case
is introduced when the old task gets scheduled in again, either on
this CPU or another CPU.

The equivalence of contexts is detected by keeping a pointer in
each cloned context pointing to the context it was cloned from.
To cope with the situation where a context is changed by adding
or removing counters after it has been cloned, we also keep a
generation number on each context which is incremented every time
a context is changed.  When a context is cloned we take a copy
of the parent's generation number, and two cloned contexts are
equivalent only if they have the same parent and the same
generation number.  In order that the parent context pointer
remains valid (and is not reused), we increment the parent
context's reference count for each context cloned from it.

Since we don't have individual fds for the counters in a cloned
context, the only thing that can make two clones of a given parent
different after they have been cloned is enabling or disabling all
counters with prctl.  To account for this, we keep a count of the
number of enabled counters in each context.  Two contexts must have
the same number of enabled counters to be considered equivalent.

Here are some measurements of the context switch time as measured with
the lat_ctx benchmark from lmbench, comparing the times obtained with
and without this patch series:

-----Unmodified----- With this patch series
Counters: none 2 HW 4H+4S none 2 HW 4H+4S

2 processes:
Average 3.44 6.45 11.24 3.12 3.39 3.60
St dev 0.04 0.04 0.13 0.05 0.17 0.19

8 processes:
Average 6.45 8.79 14.00 5.57 6.23 7.57
St dev 1.27 1.04 0.88 1.42 1.46 1.42

32 processes:
Average 5.56 8.43 13.78 5.28 5.55 7.15
St dev 0.41 0.47 0.53 0.54 0.57 0.81

The numbers are the mean and standard deviation of 20 runs of
lat_ctx.  The "none" columns are lat_ctx run directly without any
counters.  The "2 HW" columns are with lat_ctx run under perfstat,
counting cycles and instructions.  The "4H+4S" columns are lat_ctx run
under perfstat with 4 hardware counters and 4 software counters
(cycles, instructions, cache references, cache misses, task
clock, context switch, cpu migrations, and page faults).

[ Impact: performance optimization of counter context-switches ]

Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <18966.10666.517218.332164@cargo.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Dynamically allocate tasks' perf_counter_context struct
Paul Mackerras [Fri, 22 May 2009 04:17:31 +0000 (14:17 +1000)] 
perf_counter: Dynamically allocate tasks' perf_counter_context struct

This replaces the struct perf_counter_context in the task_struct with
a pointer to a dynamically allocated perf_counter_context struct.  The
main reason for doing is this is to allow us to transfer a
perf_counter_context from one task to another when we do lazy PMU
switching in a later patch.

This has a few side-benefits: the task_struct becomes a little smaller,
we save some memory because only tasks that have perf_counters attached
get a perf_counter_context allocated for them, and we can remove the
inclusion of <linux/perf_counter.h> in sched.h, meaning that we don't
end up recompiling nearly everything whenever perf_counter.h changes.

The perf_counter_context structures are reference-counted and freed
when the last reference is dropped.  A context can have references
from its task and the counters on its task.  Counters can outlive the
task so it is possible that a context will be freed well after its
task has exited.

Contexts are allocated on fork if the parent had a context, or
otherwise the first time that a per-task counter is created on a task.
In the latter case, we set the context pointer in the task struct
locklessly using an atomic compare-and-exchange operation in case we
raced with some other task in creating a context for the subject task.

This also removes the task pointer from the perf_counter struct.  The
task pointer was not used anywhere and would make it harder to move a
context from one task to another.  Anything that needed to know which
task a counter was attached to was already using counter->ctx->task.

The __perf_counter_init_context function moves up in perf_counter.c
so that it can be called from find_get_context, and now initializes
the refcount, but is otherwise unchanged.

We were potentially calling list_del_counter twice: once from
__perf_counter_exit_task when the task exits and once from
__perf_counter_remove_from_context when the counter's fd gets closed.
This adds a check in list_del_counter so it doesn't do anything if
the counter has already been removed from the lists.

Since perf_counter_task_sched_in doesn't do anything if the task doesn't
have a context, and leaves cpuctx->task_ctx = NULL, this adds code to
__perf_install_in_context to set cpuctx->task_ctx if necessary, i.e. in
the case where the current task adds the first counter to itself and
thus creates a context for itself.

This also adds similar code to __perf_counter_enable to handle a
similar situation which can arise when the counters have been disabled
using prctl; that also leaves cpuctx->task_ctx = NULL.

[ Impact: refactor counter context management to prepare for new feature ]

Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <18966.10075.781053.231153@cargo.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Fix context removal deadlock
Ingo Molnar [Wed, 20 May 2009 18:13:28 +0000 (20:13 +0200)] 
perf_counter: Fix context removal deadlock

Disable the PMU globally before removing a counter from a
context. This fixes the following lockup:

[22081.741922] ------------[ cut here ]------------
[22081.746668] WARNING: at arch/x86/kernel/cpu/perf_counter.c:803 intel_pmu_handle_irq+0x9b/0x24e()
[22081.755624] Hardware name: X8DTN
[22081.758903] perfcounters: irq loop stuck!
[22081.762985] Modules linked in:
[22081.766136] Pid: 11082, comm: perf Not tainted 2.6.30-rc6-tip #226
[22081.772432] Call Trace:
[22081.774940]  <NMI>  [<ffffffff81019aed>] ? intel_pmu_handle_irq+0x9b/0x24e
[22081.781993]  [<ffffffff81019aed>] ? intel_pmu_handle_irq+0x9b/0x24e
[22081.788368]  [<ffffffff8104505c>] ? warn_slowpath_common+0x77/0xa3
[22081.794649]  [<ffffffff810450d3>] ? warn_slowpath_fmt+0x40/0x45
[22081.800696]  [<ffffffff81019aed>] ? intel_pmu_handle_irq+0x9b/0x24e
[22081.807080]  [<ffffffff814d1a72>] ? perf_counter_nmi_handler+0x3f/0x4a
[22081.813751]  [<ffffffff814d2d09>] ? notifier_call_chain+0x58/0x86
[22081.819951]  [<ffffffff8105b250>] ? notify_die+0x2d/0x32
[22081.825392]  [<ffffffff814d1414>] ? do_nmi+0x8e/0x242
[22081.830538]  [<ffffffff814d0f0a>] ? nmi+0x1a/0x20
[22081.835342]  [<ffffffff8117e102>] ? selinux_file_free_security+0x0/0x1a
[22081.842105]  [<ffffffff81018793>] ? x86_pmu_disable_counter+0x15/0x41
[22081.848673]  <<EOE>>  [<ffffffff81018f3d>] ? x86_pmu_disable+0x86/0x103
[22081.855512]  [<ffffffff8108fedd>] ? __perf_counter_remove_from_context+0x0/0xfe
[22081.862926]  [<ffffffff8108fcbc>] ? counter_sched_out+0x30/0xce
[22081.868909]  [<ffffffff8108ff36>] ? __perf_counter_remove_from_context+0x59/0xfe
[22081.876382]  [<ffffffff8106808a>] ? smp_call_function_single+0x6c/0xe6
[22081.882955]  [<ffffffff81091b96>] ? perf_release+0x86/0x14c
[22081.888600]  [<ffffffff810c4c84>] ? __fput+0xe7/0x195
[22081.893718]  [<ffffffff810c213e>] ? filp_close+0x5b/0x62
[22081.899107]  [<ffffffff81046a70>] ? put_files_struct+0x64/0xc2
[22081.905031]  [<ffffffff8104841a>] ? do_exit+0x1e2/0x6ef
[22081.910360]  [<ffffffff814d0a60>] ? _spin_lock_irqsave+0x9/0xe
[22081.916292]  [<ffffffff8104898e>] ? do_group_exit+0x67/0x93
[22081.921953]  [<ffffffff810489cc>] ? sys_exit_group+0x12/0x16
[22081.927759]  [<ffffffff8100baab>] ? system_call_fastpath+0x16/0x1b
[22081.934076] ---[ end trace 3a3936ce3e1b4505 ]---

And could potentially also fix the lockup reported by Marcelo Tosatti.

Also, print more debug info in case of a detected lockup.

[ Impact: fix lockup ]

Reported-by: Marcelo Tosatti <mtosatti@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Optimize sched in/out of counters
Peter Zijlstra [Wed, 20 May 2009 10:21:22 +0000 (12:21 +0200)] 
perf_counter: Optimize sched in/out of counters

Avoid a function call for !group counters by directly calling the counter
function.

[ Impact: micro-optimize the code ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090520102553.511933670@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Optimize disable of time based sw counters
Peter Zijlstra [Wed, 20 May 2009 10:21:21 +0000 (12:21 +0200)] 
perf_counter: Optimize disable of time based sw counters

Currently we call hrtimer_cancel() unconditionally on disable of time based
software counters. Avoid when possible.

[ Impact: micro-optimize the code ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090520102553.388185031@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Log irq_period changes
Peter Zijlstra [Wed, 20 May 2009 10:21:20 +0000 (12:21 +0200)] 
perf_counter: Log irq_period changes

For the dynamic irq_period code, log whenever we change the period so that
analyzing code can normalize the event flow.

[ Impact: add new feature to allow more precise profiling ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090520102553.298769743@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Solve the rotate_ctx vs inherit race differently
Peter Zijlstra [Wed, 20 May 2009 10:21:19 +0000 (12:21 +0200)] 
perf_counter: Solve the rotate_ctx vs inherit race differently

Instead of disabling RR scheduling of the counters, use a different list
that does not get rotated to iterate the counters on inheritance.

[ Impact: cleanup, optimization ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
LKML-Reference: <20090520102553.237504544@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: fix counter inheritance race
Ingo Molnar [Sun, 17 May 2009 09:24:08 +0000 (11:24 +0200)] 
perf_counter: fix counter inheritance race

Context rotation should not occur when we are in the middle of
walking the counter list when inheriting counters ...

[ Impact: fix occasionally incorrect perf stat results ]

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: fix counter freeing logic
Ingo Molnar [Sun, 17 May 2009 09:08:41 +0000 (11:08 +0200)] 
perf_counter: fix counter freeing logic

Fix counter lifetime bugs which explain the crashes reported by
Marcelo Tosatti and Arnaldo Carvalho de Melo.

The new rule is: flushing + freeing is only done for a task's
own counters, never for other tasks.

[ Impact: fix crashes/lockups with inherited counters ]

Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Reported-by: Marcelo Tosatti <mtosatti@redhat.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter, x86: speed up the scheduling fast-path
Ingo Molnar [Sun, 17 May 2009 17:37:25 +0000 (19:37 +0200)] 
perf_counter, x86: speed up the scheduling fast-path

We have to set up the LVT entry only at counter init time, not at
every switch-in time.

There's friction between NMI and non-NMI use here - we'll probably
remove the per counter configurability of it - but until then, dont
slow down things ...

[ Impact: micro-optimization ]

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: powerpc: initialize cpuhw pointer before use
Paul Mackerras [Mon, 18 May 2009 04:02:12 +0000 (14:02 +1000)] 
perf_counter: powerpc: initialize cpuhw pointer before use

Commit 9e35ad38 ("perf_counter: Rework the perf counter
disable/enable") added code to the powerpc hw_perf_enable (renamed
from hw_perf_restore) to test cpuhw->disabled and return immediately
if it is not set (i.e. if the PMU is already enabled).

Unfortunately the test got added before cpuhw was initialized,
resulting in an oops the first time hw_perf_enable got called.
This fixes it by moving the initialization of cpuhw to before
cpuhw->disabled is tested.

[ Impact: fix oops-causing bug on powerpc ]

Signed-off-by: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <18960.56772.869734.304631@drongo.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge commit 'v2.6.30-rc6' into perfcounters/core
Ingo Molnar [Mon, 18 May 2009 05:37:44 +0000 (07:37 +0200)] 
Merge commit 'v2.6.30-rc6' into perfcounters/core

Merge reason: this branch was on an -rc4 base, merge it up to -rc6
              to get the latest upstream fixes.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter, x86: fix zero irq_period counters
Ingo Molnar [Sun, 17 May 2009 08:04:45 +0000 (10:04 +0200)] 
perf_counter, x86: fix zero irq_period counters

The quirk to irq_period unearthed an unrobustness we had in the
hw_counter initialization sequence: we left irq_period at 0, which
was then quirked up to 2 ... which then generated a _lot_ of
interrupts during 'perf stat' runs, slowed them down and skewed
the counter results in general.

Initialize irq_period to the maximum instead.

[ Impact: fix perf stat results ]

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: fix threaded task exit
Ingo Molnar [Sun, 17 May 2009 09:24:08 +0000 (11:24 +0200)] 
perf_counter: fix threaded task exit

Flushing counters in __exit_signal() with irqs disabled is not
a good idea as perf_counter_exit_task() acquires mutexes. So
flush it before acquiring the tasklist lock.

(Note, we still need a fix for when the PID has been unhashed.)

[ Impact: fix crash with inherited counters ]

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Fix counter inheritance
Peter Zijlstra [Fri, 15 May 2009 18:45:59 +0000 (20:45 +0200)] 
perf_counter: Fix counter inheritance

Srivatsa Vaddagiri reported that a Java workload triggers this
warning in kernel/exit.c:

   WARN_ON_ONCE(!list_empty(&tsk->perf_counter_ctx.counter_list));

Add the inherited counter propagation on self-detach, this could
cause counter leaks and incomplete stats in threaded code like
the below:

  #include <pthread.h>
  #include <unistd.h>

  void *thread(void *arg)
  {
          sleep(5);
          return NULL;
  }

  void main(void)
  {
          pthread_t thr;
          pthread_create(&thr, NULL, thread, NULL);
  }

Reported-by: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Fix inheritance cleanup code
Peter Zijlstra [Fri, 15 May 2009 18:45:59 +0000 (20:45 +0200)] 
perf_counter: Fix inheritance cleanup code

Clean up code that open-coded the list_{add,del}_counter() code in
__perf_counter_exit_task() which consequently diverged. This could
lead to software counter crashes.

Also, fold the ctx->nr_counter inc/dec into those functions and clean
up some of the related code.

[ Impact: fix potential sw counter crash, cleanup ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@in.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoLinux 2.6.30-rc6 v2.6.30-rc6
Linus Torvalds [Sat, 16 May 2009 04:12:57 +0000 (21:12 -0700)] 
Linux 2.6.30-rc6

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes...
Linus Torvalds [Fri, 15 May 2009 23:47:55 +0000 (16:47 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/jbarnes/pci-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
  PCI MSI: Fix MSI-X with NIU cards
  PCI: Fix pci-e port driver slot_reset bad default return value

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/holtmann/bluetooth-2.6
Linus Torvalds [Fri, 15 May 2009 21:29:53 +0000 (14:29 -0700)] 
Merge git://git./linux/kernel/git/holtmann/bluetooth-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/holtmann/bluetooth-2.6:
  Bluetooth: Don't trigger disconnect timeout for security mode 3 pairing
  Bluetooth: Don't use hci_acl_connect_cancel() for incoming connections
  Bluetooth: Fix wrong module refcount when connection setup fails

Another case of me handling the fallout from Davem's unfortunate
addiction to shuffleboard.

Won't anybody think of the children? Join the anti-shuffleboard league
today!

15 years agoMerge branch 'drm-intel-next' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt...
Linus Torvalds [Fri, 15 May 2009 20:22:11 +0000 (13:22 -0700)] 
Merge branch 'drm-intel-next' of git://git./linux/kernel/git/anholt/drm-intel

* 'drm-intel-next' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt/drm-intel:
  drm/i915: Add new GET_PIPE_FROM_CRTC_ID ioctl.
  drm/i915: Set HDMI hot plug interrupt enable for only the output in question.
  drm/i915: Include 965GME pci ID in IS_I965GM(dev) to match UMS.
  drm/i915: Use the GM45 VGA hotplug workaround on G45 as well.
  drm/i915: ignore LVDS on intel graphics systems that lie about having it
  drm/i915: sanity check IER at wait_request time
  drm/i915: workaround IGD i2c bus issue in kernel side (v2)
  drm/i915: Don't allow binding objects into the last page of the aperture.
  drm/i915: save/restore fence registers across suspend/resume
  drm/i915: x86 always has writeq. Add I915_READ64 for symmetry.

15 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Fri, 15 May 2009 19:04:37 +0000 (12:04 -0700)] 
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of ssh://master.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  libata: Media rotation rate and form factor heuristics
  libata: Report disk alignment and physical block size
  sata_fsl: Fix the command description of FSL SATA controller
  sata_fsl: Fix compile warnings
  [libata] sata_sx4: fixup interrupt handling
  [libata] sata_sx4: convert to new exception handling methods

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6
Linus Torvalds [Fri, 15 May 2009 19:01:59 +0000 (12:01 -0700)] 
Merge git://git./linux/kernel/git/linville/wireless-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless-2.6:
  iwlwifi: fix device id registration for 6000 series 2x2 devices
  ath5k: update channel in sw state after stopping RX and TX
  rtl8187: use DMA-aware buffers with usb_control_msg
  mac80211: avoid NULL ptr deref when finding max_rates in PID and minstrel
  airo: airo_get_encode{,ext} potential buffer overflow

Pulled directly by Linus because Davem is off playing shuffle-board at
some Alaskan cruise, and the NULL ptr deref issue hits people and should
get merged sooner rather than later.

David - make us proud on the shuffle-board tournament!

15 years agolibata: Media rotation rate and form factor heuristics
Martin K. Petersen [Fri, 15 May 2009 04:40:35 +0000 (00:40 -0400)] 
libata: Media rotation rate and form factor heuristics

This patch provides new heuristics for parsing both the form factor and
media rotation rate ATA IDENFITY words.

The reported ATA version must be 7 or greater and the device must return
values defined as valid in the standard.  Only then are the
characteristics reported to SCSI via the VPD B1 page.

This seems like a reasonable compromise to me considering that we have
been shipping several kernel releases that key off the rotation rate bit
without any version checking whatsoever.  With no complaints so far.

Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agolibata: Report disk alignment and physical block size
Martin K. Petersen [Fri, 15 May 2009 04:40:34 +0000 (00:40 -0400)] 
libata: Report disk alignment and physical block size

For disks with 4KB sectors, report the correct block size and alignment
when filling out the READ CAPACITY(16) response.

This patch is based upon code from Matthew Wilcox' 4KB ATA tree.  I
fixed the bug I reported a while back caused by ATA and SCSI using
different approaches to describing the alignment.

Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agosata_fsl: Fix the command description of FSL SATA controller
Dave Liu [Thu, 14 May 2009 14:47:07 +0000 (09:47 -0500)] 
sata_fsl: Fix the command description of FSL SATA controller

The bit 11 of command description is reserved bit in Freescale
SATA controller and needs to be set to '1'.  This is needed to
make sure the last write from the controller to the buffer
descriptor is seen before an interrupt is raised.

Signed-off-by: Dave Liu <daveliu@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agosata_fsl: Fix compile warnings
Kumar Gala [Thu, 14 May 2009 03:10:50 +0000 (22:10 -0500)] 
sata_fsl: Fix compile warnings

We we build with dma_addr_t as a 64-bit quantity we get:

drivers/ata/sata_fsl.c: In function 'sata_fsl_fill_sg':
drivers/ata/sata_fsl.c:340: warning: format '%x' expects type 'unsigned int', but argument 4 has type 'dma_addr_t'

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years ago[libata] sata_sx4: fixup interrupt handling
David Milburn [Wed, 13 May 2009 23:02:21 +0000 (18:02 -0500)] 
[libata] sata_sx4: fixup interrupt handling

Issuing ATA_CMD_SET_FEATURES (0xef) times out because
pdc20621_interrupt ignores command completion since
ATA_TFLAG_POLLING flag is set.

This has already been fixed for sata_promise:

commit 51b94d2a5a90d4800e74d7348bcde098a28f4fb3
Author: Tejun Heo <htejun@gmail.com>
Date:   Fri Jun 8 13:46:55 2007 -0700

    sata_promise: use TF interface for polling NODATA commands

Also, this patch includes Mikael's original patches:

http://marc.info/?l=linux-ide&m=121135828227724&w=2
http://marc.info/?l=linux-ide&m=121144512109826&w=2

Signed-off-by: Mikael Pettersson <mikpe@it.uu.se>
Signed-off-by: David Milburn <dmilburn@redhat.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years ago[libata] sata_sx4: convert to new exception handling methods
Jeff Garzik [Wed, 8 Apr 2009 20:02:18 +0000 (16:02 -0400)] 
[libata] sata_sx4: convert to new exception handling methods

Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
15 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
Linus Torvalds [Fri, 15 May 2009 15:07:25 +0000 (08:07 -0700)] 
Merge branch 'for_linus' of git://git./linux/kernel/git/tytso/ext4

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: Fix race in ext4_inode_info.i_cached_extent
  ext4: Clear the unwritten buffer_head flag after the extent is initialized
  ext4: Use a fake block number for delayed new buffer_head
  ext4: Fix sub-block zeroing for writes into preallocated extents

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Fri, 15 May 2009 15:06:56 +0000 (08:06 -0700)] 
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ASoC: DaVinci EVM board support buildfixes
  ASoC: DaVinci I2S updates
  ASoC: davinci-pcm buildfixes
  ALSA: pcsp: fix printk format warning
  ALSA: riptide: postfix increment and off by one
  pxa2xx-ac97: fix reset gpio mode setting
  ASoC: soc-core: fix crash when removing not instantiated card

15 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jwessel...
Linus Torvalds [Fri, 15 May 2009 15:06:45 +0000 (08:06 -0700)] 
Merge branch 'for_linus' of git://git./linux/kernel/git/jwessel/linux-2.6-kgdb

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jwessel/linux-2.6-kgdb:
  kgdb: gdb documentation fix
  kgdb,i386: use address that SP register points to in the exception frame
  sysrq, intel_fb: fix sysrq g collision

15 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
Linus Torvalds [Fri, 15 May 2009 15:05:37 +0000 (08:05 -0700)] 
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  Revert "mm: add /proc controls for pdflush threads"
  viocd: needs to depend on BLOCK
  block: fix the bio_vec array index out-of-bounds test

15 years agoMerge branch 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc
Linus Torvalds [Fri, 15 May 2009 15:05:02 +0000 (08:05 -0700)] 
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/benh/powerpc:
  powerpc: Fix PCI ROM access
  powerpc/pseries: Really fix the oprofile CPU type on pseries
  serial/nwpserial: Fix wrong register read address and add interrupt acknowledge.
  powerpc/cell: Make ptcal more reliable
  powerpc: Allow mem=x cmdline to work with 4G+
  powerpc/mpic: Fix incorrect allocation of interrupt rev-map
  powerpc: Fix oprofile sampling of marked events on POWER7
  powerpc/iseries: Fix pci breakage due to bad dma_data initialization
  powerpc: Fix mktree build error on Mac OS X host
  powerpc/virtex: Fix duplicate level irq events.
  powerpc/virtex: Add uImage to the default images list
  powerpc/boot: add simpleImage.* to clean-files list
  powerpc/8xx: Update defconfigs
  powerpc/embedded6xx: Update defconfigs
  powerpc/86xx: Update defconfigs
  powerpc/85xx: Update defconfigs
  powerpc/83xx: Update defconfigs
  powerpc/fsl_soc: Remove mpc83xx_wdt_init, again

15 years agodevpts: correctly set default options
Sukadev Bhattiprolu [Fri, 15 May 2009 02:38:24 +0000 (19:38 -0700)] 
devpts: correctly set default options

devpts_get_sb() calls memset(0) to clear mount options and calls
parse_mount_options() if user specified any mount options.

The memset(0) is bogus since the 'mode' and 'ptmxmode' options are
non-zero by default.  parse_mount_options() restores options to default
anyway and can properly deal with NULL mount options.

So in devpts_get_sb() remove memset(0) and call parse_mount_options() even
for NULL mount options.

Bug reported by Eric Paris: http://lkml.org/lkml/2009/5/7/448.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Tested-by: Marc Dionne <marc.c.dionne@gmail.com>
Reported-by: Eric Paris <eparis@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Reviewed-by: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoperf_counter: powerpc: supply more precise information on counter overflow events
Paul Mackerras [Thu, 14 May 2009 03:31:48 +0000 (13:31 +1000)] 
perf_counter: powerpc: supply more precise information on counter overflow events

This uses values from the MMCRA, SIAR and SDAR registers on
powerpc to supply more precise information for overflow events,
including a data address when PERF_RECORD_ADDR is specified.

Since POWER6 uses different bit positions in MMCRA from earlier
processors, this converts the struct power_pmu limited_pmc5_6
field, which only had 0/1 values, into a flags field and
defines bit values for its previous use (PPMU_LIMITED_PMC5_6)
and a new flag (PPMU_ALT_SIPR) to indicate that the processor
uses the POWER6 bit positions rather than the earlier
positions.  It also adds definitions in reg.h for the new and
old positions of the bit that indicates that the SIAR and SDAR
values come from the same instruction.

For the data address, the SDAR value is supplied if we are not
doing instruction sampling.  In that case there is no guarantee
that the address given in the PERF_RECORD_ADDR subrecord will
correspond to the instruction whose address is given in the
PERF_RECORD_IP subrecord.

If instruction sampling is enabled (e.g. because this counter
is counting a marked instruction event), then we only supply
the SDAR value for the PERF_RECORD_ADDR subrecord if it
corresponds to the instruction whose address is in the
PERF_RECORD_IP subrecord.  Otherwise we supply 0.

[ Impact: support more PMU hardware features on PowerPC ]

Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <18955.37028.48861.555309@drongo.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: allow arch to supply event misc flags and instruction pointer
Paul Mackerras [Thu, 14 May 2009 11:48:08 +0000 (21:48 +1000)] 
perf_counter: allow arch to supply event misc flags and instruction pointer

At present the values we put in overflow events for the misc
flags indicating processor mode and the instruction pointer are
obtained using the standard user_mode() and
instruction_pointer() functions. Those functions tell you where
the performance monitor interrupt was taken, which might not be
exactly where the counter overflow occurred, for example
because interrupts were disabled at the point where the
overflow occurred, or because the processor had many
instructions in flight and chose to complete some more
instructions beyond the one that caused the counter overflow.

Some architectures (e.g. powerpc) can supply more precise
information about where the counter overflow occurred and the
processor mode at that point.  This introduces new functions,
perf_misc_flags() and perf_instruction_pointer(), which arch
code can override to provide more precise information if
available.  They have default implementations which are
identical to the existing code.

This also adds a new misc flag value,
PERF_EVENT_MISC_HYPERVISOR, for the case where a counter
overflow occurred in the hypervisor.  We encode the processor
mode in the 2 bits previously used to indicate user or kernel
mode; the values for user and kernel mode are unchanged and
hypervisor mode is indicated by both bits being set.

[ Impact: generalize perfcounter core facilities ]

Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <18956.1272.818511.561835@cargo.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: powerpc: use u64 for event codes internally
Paul Mackerras [Thu, 14 May 2009 03:29:14 +0000 (13:29 +1000)] 
perf_counter: powerpc: use u64 for event codes internally

Although the perf_counter API allows 63-bit raw event codes,
internally in the powerpc back-end we had been using 32-bit
event codes.  This expands them to 64 bits so that we can add
bits for specifying threshold start/stop events and instruction
sampling modes later.

This also corrects the return value of can_go_on_limited_pmc;
we were returning an event code rather than just a 0/1 value in
some circumstances. That didn't particularly matter while event
codes were 32-bit, but now that event codes are 64-bit it
might, so this fixes it.

[ Impact: extend PowerPC perfcounter interfaces from u32 to u64 ]

Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <18955.36874.472452.353104@drongo.ozlabs.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: frequency based adaptive irq_period, 32-bit fix
Peter Zijlstra [Fri, 15 May 2009 13:37:47 +0000 (15:37 +0200)] 
perf_counter: frequency based adaptive irq_period, 32-bit fix

fix:

  kernel/built-in.o: In function `perf_counter_alloc':
  perf_counter.c:(.text+0x7ddc7): undefined reference to `__udivdi3'

[ Impact: build fix on 32-bit systems ]

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <1242394667.6642.1887.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoMerge branch 'fix/asoc' into for-linus
Takashi Iwai [Fri, 15 May 2009 13:38:26 +0000 (15:38 +0200)] 
Merge branch 'fix/asoc' into for-linus

* fix/asoc:
  ASoC: DaVinci EVM board support buildfixes
  ASoC: DaVinci I2S updates
  ASoC: davinci-pcm buildfixes
  pxa2xx-ac97: fix reset gpio mode setting
  ASoC: soc-core: fix crash when removing not instantiated card

15 years agoMerge branch 'fix/misc' into for-linus
Takashi Iwai [Fri, 15 May 2009 13:38:20 +0000 (15:38 +0200)] 
Merge branch 'fix/misc' into for-linus

* fix/misc:
  ALSA: pcsp: fix printk format warning
  ALSA: riptide: postfix increment and off by one

15 years agoperf top: update to use the new freq interface
Peter Zijlstra [Fri, 15 May 2009 13:19:29 +0000 (15:19 +0200)] 
perf top: update to use the new freq interface

Provide perf top -F as alternative to -c.

[ Impact: new 'perf top' feature ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <20090515132018.707922166@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: frequency based adaptive irq_period
Peter Zijlstra [Fri, 15 May 2009 13:19:28 +0000 (15:19 +0200)] 
perf_counter: frequency based adaptive irq_period

Instead of specifying the irq_period for a counter, provide a target interrupt
frequency and dynamically adapt the irq_period to match this frequency.

[ Impact: new perf-counter attribute/feature ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <20090515132018.646195868@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: per user mlock gift
Peter Zijlstra [Fri, 15 May 2009 13:19:27 +0000 (15:19 +0200)] 
perf_counter: per user mlock gift

Instead of a per-process mlock gift for perf-counters, use a
per-user gift so that there is less of a DoS potential.

[ Impact: allow less worst-case unprivileged memory consumption ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <20090515132018.496182835@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: remove perf_disable/enable exports
Peter Zijlstra [Fri, 15 May 2009 13:19:26 +0000 (15:19 +0200)] 
perf_counter: remove perf_disable/enable exports

Now that ACPI idle doesn't use it anymore, remove the exports.

[ Impact: remove dead code/data ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <20090515132018.429826617@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoext4: Fix race in ext4_inode_info.i_cached_extent
Theodore Ts'o [Fri, 15 May 2009 13:07:28 +0000 (09:07 -0400)] 
ext4: Fix race in ext4_inode_info.i_cached_extent

If two CPU's simultaneously call ext4_ext_get_blocks() at the same
time, there is nothing protecting the i_cached_extent structure from
being used and updated at the same time.  This could potentially cause
the wrong location on disk to be read or written to, including
potentially causing the corruption of the block group descriptors
and/or inode table.

This bug has been in the ext4 code since almost the very beginning of
ext4's development.  Fortunately once the data is stored in the page
cache cache, ext4_get_blocks() doesn't need to be called, so trying to
replicate this problem to the point where we could identify its root
cause was *extremely* difficult.  Many thanks to Kevin Shanahan for
working over several months to be able to reproduce this easily so we
could finally nail down the cause of the corruption.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
15 years agokgdb: gdb documentation fix
Frank Rowand [Fri, 15 May 2009 12:56:25 +0000 (07:56 -0500)] 
kgdb: gdb documentation fix

gdb command "set remote debug 1" is not valid, change to correct command.

Signed-off-by: Frank Rowand <frank.rowand@am.sony.com>
Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
15 years agokgdb,i386: use address that SP register points to in the exception frame
Jason Wessel [Thu, 12 Feb 2009 00:46:32 +0000 (18:46 -0600)] 
kgdb,i386: use address that SP register points to in the exception frame

The treatment of the SP register is different on x86_64 and i386.
This is a regression fix that lived outside the mainline kernel from
2.6.27 to now.  The regression was a result of the original merge
consolidation of the i386 and x86_64 archs to x86.

The incorrectly reported SP on i386 prevented stack tracebacks from
working correctly in gdb.

Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
15 years agosysrq, intel_fb: fix sysrq g collision
Jason Wessel [Thu, 14 May 2009 02:56:59 +0000 (21:56 -0500)] 
sysrq, intel_fb: fix sysrq g collision

Commit 79e539453b34e35f39299a899d263b0a1f1670bd introduced a
regression where you cannot use sysrq 'g' to enter kgdb.  The solution
is to move the intel fb sysrq over to V for video instead of G for
graphics.  The SMP VOYAGER code to register for the sysrq-v is not
anywhere to be found in the mainline kernel, so the comments in the
code were cleaned up as well.

This patch also cleans up the sysrq definitions for kgdb to make it
generic for the kernel debugger, such that the sysrq 'g' can be used
in the future to enter a gdbstub or another kernel debugger.

Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
15 years agoperf stat: handle Ctrl-C
Ingo Molnar [Fri, 15 May 2009 09:03:23 +0000 (11:03 +0200)] 
perf stat: handle Ctrl-C

Before this change, if a long-running perf stat workload was Ctrl-C-ed,
the utility exited without displaying statistics.

After the change, the Ctrl-C gets propagated into the workload (and
causes its early exit there), but perf stat itself will still continue
to run and will display counter results.

This is useful to run open-ended workloads, let them run for
a while, then Ctrl-C them to get the stats.

[ Impact: extend perf stat with new functionality ]

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoRevert "mm: add /proc controls for pdflush threads"
Jens Axboe [Fri, 15 May 2009 09:32:24 +0000 (11:32 +0200)] 
Revert "mm: add /proc controls for pdflush threads"

This reverts commit fafd688e4c0c34da0f3de909881117d374e4c7af.

Work is progressing to switch away from pdflush as the process backing
for flushing out dirty data. So it seems pointless to add more knobs
to control pdflush threads. The original author of the patch did not
have any specific use cases for adding the knobs, so we can easily
revert this before 2.6.30 to avoid having to maintain this API
forever.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
15 years agoASoC: DaVinci EVM board support buildfixes
David Brownell [Thu, 14 May 2009 20:01:59 +0000 (13:01 -0700)] 
ASoC: DaVinci EVM board support buildfixes

This is a build fix, resyncing the DaVinci EVM ASoC board code
with the version in the DaVinci tree.  That resync includes
support for the DM355 EVM, although that board isn't yet in
mainline.

(NOTE:  also includes a bugfix to the platform_add_resources
call, recently sent by Chaithrika U S <chaithrika@ti.com> but
not yet merged into the DaVinci tree.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
15 years agoASoC: DaVinci I2S updates
David Brownell [Thu, 14 May 2009 19:47:42 +0000 (12:47 -0700)] 
ASoC: DaVinci I2S updates

This resyncs the DaVinci I2S code with the version in the DaVinci
tree.  The behavioral change uses updated clock interfaces which
recently merged to mainline.  Two other changes include adding a
comment on the ASP/McBSP/McASP confusion, and dropping pdev->id in
order to support more boards than just the DM644x EVM.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
15 years agoASoC: davinci-pcm buildfixes
David Brownell [Thu, 14 May 2009 19:41:22 +0000 (12:41 -0700)] 
ASoC: davinci-pcm buildfixes

This is a buildfix for the DaVinci PCM code, resyncing it with
the version in the DaVinci tree.  The notable change is using
current EDMA interfaces, which recently merged to mainline.
(The older interfaces never made it into mainline.)

NOTE:  open issue, the DMA should be to/from SRAM; see chip
errata for more info.  The artifacts are extremely easy to
hear on DM355 hardware (not yet supported in mainline), but
don't seem as audible on DM6446 hardwaare (which does have
mainline support).

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
15 years agoperf_counter: Remove ACPI quirk
Ingo Molnar [Thu, 14 May 2009 03:16:59 +0000 (05:16 +0200)] 
perf_counter: Remove ACPI quirk

We had a disable/enable around acpi_idle_do_entry() due to an erratum
in an early prototype CPU i had access to. That erratum has been fixed
in the BIOS so remove the quirk.

The quirk also kept us from profiling interrupts that hit the ACPI idle
instruction - so this is an improvement as well, beyond a cleanup and
a micro-optimization.

[ Impact: improve profiling scope, cleanup, micro-optimization ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: x86: Protect against infinite loops in intel_pmu_handle_irq()
Ingo Molnar [Fri, 15 May 2009 06:26:20 +0000 (08:26 +0200)] 
perf_counter: x86: Protect against infinite loops in intel_pmu_handle_irq()

intel_pmu_handle_irq() can lock up in an infinite loop if the hardware
does not allow the acking of irqs. Alas, this happened in testing so
make this robust and emit a warning if it happens in the future.

Also, clean up the IRQ handlers a bit.

[ Impact: improve perfcounter irq/nmi handling robustness ]

Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: x86: Disallow interval of 1
Ingo Molnar [Fri, 15 May 2009 06:25:22 +0000 (08:25 +0200)] 
perf_counter: x86: Disallow interval of 1

On certain CPUs i have observed a stuck PMU if interval was set to
1 and NMIs were used. The PMU had PMC0 set in MSR_CORE_PERF_GLOBAL_STATUS,
but it was not possible to ack it via MSR_CORE_PERF_GLOBAL_OVF_CTRL,
and the NMI loop got stuck infinitely.

[ Impact: fix rare hangs during high perfcounter load ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: x86: Robustify interrupt handling
Peter Zijlstra [Thu, 14 May 2009 12:52:17 +0000 (14:52 +0200)] 
perf_counter: x86: Robustify interrupt handling

Two consecutive NMIs could daze and confuse the machine when the
first would handle the overflow of both counters.

[ Impact: fix false-positive syslog messages under multi-session profiling ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Rework the perf counter disable/enable
Peter Zijlstra [Wed, 13 May 2009 14:21:38 +0000 (16:21 +0200)] 
perf_counter: Rework the perf counter disable/enable

The current disable/enable mechanism is:

token = hw_perf_save_disable();
...
/* do bits */
...
hw_perf_restore(token);

This works well, provided that the use nests properly. Except we don't.

x86 NMI/INT throttling has non-nested use of this, breaking things. Therefore
provide a reference counter disable/enable interface, where the first disable
disables the hardware, and the last enable enables the hardware again.

[ Impact: refactor, simplify the PMU disable/enable logic ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: x86: Fix up the amd NMI/INT throttle
Peter Zijlstra [Wed, 13 May 2009 11:21:36 +0000 (13:21 +0200)] 
perf_counter: x86: Fix up the amd NMI/INT throttle

perf_counter_unthrottle() restores throttle_ctrl, buts its never set.
Also, we fail to disable all counters when throttling.

[ Impact: fix rare stuck perf-counters when they are throttled ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: Fix perf_output_copy() WARN to account for overflow
Peter Zijlstra [Wed, 13 May 2009 19:26:19 +0000 (21:26 +0200)] 
perf_counter: Fix perf_output_copy() WARN to account for overflow

The simple reservation test in perf_output_copy() failed to take
unsigned int overflow into account, fix this.

[ Impact: fix false positive warning with more than 4GB of profiling data ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: x86: Allow unpriviliged use of NMIs
Peter Zijlstra [Wed, 13 May 2009 08:02:57 +0000 (10:02 +0200)] 
perf_counter: x86: Allow unpriviliged use of NMIs

Apply sysctl_perf_counter_priv to NMIs. Also, fail the counter
creation instead of silently down-grading to regular interrupts.

[ Impact: allow wider perf-counter usage ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: x86: Fix throttling
Ingo Molnar [Wed, 13 May 2009 10:54:01 +0000 (12:54 +0200)] 
perf_counter: x86: Fix throttling

If counters are disabled globally when a perfcounter IRQ/NMI hits,
and if we throttle in that case, we'll promote the '0' value to
the next lapic IRQ and disable all perfcounters at that point,
permanently ...

Fix it.

[ Impact: fix hung perfcounters under load ]

Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf_counter: x86: More accurate counter update
Peter Zijlstra [Wed, 13 May 2009 07:45:19 +0000 (09:45 +0200)] 
perf_counter: x86: More accurate counter update

Take the counter width into account instead of assuming 32 bits.

In particular Nehalem has 44 bit wide counters, and all
arithmetics should happen on a 44-bit signed integer basis.

[ Impact: fix rare event imprecision, warning message on Nehalem ]

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Corey Ashford <cjashfor@linux.vnet.ibm.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agoperf record: Allow specifying a pid to record
Arnaldo Carvalho de Melo [Fri, 15 May 2009 01:50:46 +0000 (22:50 -0300)] 
perf record: Allow specifying a pid to record

Allow specifying a pid instead of always fork+exec'ing a command.

Because the PERF_EVENT_COMM and PERF_EVENT_MMAP events happened before
we connected, we must synthesize them so that 'perf report' can get what
it needs.

[ Impact: add new command line option ]

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Clark Williams <williams@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <20090515015046.GA13664@ghostprotocols.net>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agopowerpc: Fix PCI ROM access
Benjamin Herrenschmidt [Thu, 14 May 2009 20:16:47 +0000 (20:16 +0000)] 
powerpc: Fix PCI ROM access

A couple of issues crept in since about 2.6.27 related to accessing PCI
device ROMs on various powerpc machines.

First, historically, we don't allocate the ROM resource in the resource
tree. I'm not entirely certain of why, I susepct they often contained
garbage on x86 but it's hard to tell. This causes the current generic
code to always call pci_assign_resource() when trying to access the said
ROM from sysfs, which will try to re-assign some new address regardless
of what the ROM BAR was already set to at boot time. This can be a
problem on hypervisor platforms like pSeries where we aren't supposed
to move PCI devices around (and in fact probably can't).

Second, our code that generates the PCI tree from the OF device-tree
(instead of doing config space probing) which we mostly use on pseries
at the moment, didn't set the (new) flag IORESOURCE_SIZEALIGN on any
resource. That means that any attempt at re-assigning such a resource
with pci_assign_resource() would fail due to resource_alignment()
returning 0.

This fixes this by doing these two things:

 - The code that calculates resource flags based on the OF device-node
is improved to set IORESOURCE_SIZEALIGN on any valid BAR, and while at
it also set IORESOURCE_READONLY for ROMs since we were lacking that too

 - We now allocate ROM resources as part of the resource tree. However
to limit the chances of nasty conflicts due to busted firmwares, we
only do it on the second pass of our two-passes allocation scheme,
so that all valid and enabled BARs get precedence.

This brings pSeries back the ability to access PCI ROMs via sysfs (and
thus initialize various video cards from X etc...).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/pseries: Really fix the oprofile CPU type on pseries
Benjamin Herrenschmidt [Thu, 14 May 2009 18:34:06 +0000 (18:34 +0000)] 
powerpc/pseries: Really fix the oprofile CPU type on pseries

My previous pach for fixing the oprofile CPU type got somewhat mismerged
(by my fault) when it collided with another related patch. This should
finally (fingers crossed) fix the whole thing.

We make sure we keep the -old- oprofile type and CPU type whenever
one of them was specified in the first pass through the function.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agoserial/nwpserial: Fix wrong register read address and add interrupt acknowledge.
Benjamin Krill [Wed, 13 May 2009 05:56:54 +0000 (05:56 +0000)] 
serial/nwpserial: Fix wrong register read address and add interrupt acknowledge.

The receive interrupt routine checks the wrong register if the
receive fifo is empty. Further an explicit interrupt acknowledge
write is introduced. In some circumstances another interrupt was
issued.

Signed-off-by: Benjamin Krill <ben@codiert.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/cell: Make ptcal more reliable
Gerhard Stenzel [Wed, 13 May 2009 05:50:46 +0000 (05:50 +0000)] 
powerpc/cell: Make ptcal more reliable

There have been a series of checkstops on QS21 related to
ptcal being set up incorrectly. On systems that only
have memory on a single node, ptcal fails when it gets
a pointer to memory on the remote node.

Moreover, agressive prefetching in memcpy and other
functions may accidentally touch the first cache line
of the page that we reserve for ptcal, which causes
an ECC checkstop.

We now allocate pages only from the specified node, moves the
ptcal area into the middle of the allocated page to avoid
potential prefetch problems and prints the address of the
ptcal area to facilitate diagnostics.

Signed-off-by: Gerhard Stenzel <gerhard.stenzel@de.ibm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Allow mem=x cmdline to work with 4G+
Becky Bruce [Fri, 8 May 2009 12:19:27 +0000 (12:19 +0000)] 
powerpc: Allow mem=x cmdline to work with 4G+

We're currently choking on mem=4g (and above) due to memory_limit
being specified as an unsigned long. Make memory_limit
phys_addr_t to fix this.

Signed-off-by: Becky Bruce <beckyb@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/mpic: Fix incorrect allocation of interrupt rev-map
Kumar Gala [Fri, 8 May 2009 12:08:20 +0000 (12:08 +0000)] 
powerpc/mpic: Fix incorrect allocation of interrupt rev-map

Before when we were setting up the irq host map for mpic we passed in
just isu_size for the size of the linear map.  However, for a number of
mpic implementations we have no isu (thus pass in 0) and will end up
with a no linear map (size = 0).  This causes us to always call
irq_find_mapping() from mpic_get_irq().

By moving the allocation of the host map to after we've determined the
number of sources we can actually benefit from having a linear map for
the non-isu users that covers all the interrupt sources.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Fix oprofile sampling of marked events on POWER7
Maynard Johnson [Thu, 7 May 2009 05:48:32 +0000 (05:48 +0000)] 
powerpc: Fix oprofile sampling of marked events on POWER7

Description
-----------
Change ppc64 oprofile kernel driver to use the SLOT bits (MMCRA[37:39]only on
older processors where those bits are defined.

Background
----------
The performance monitor unit of the 64-bit POWER processor family has the
ability to collect accurate instruction-level samples when profiling on marked
events (i.e., "PM_MRK_<event-name>").  In processors prior to POWER6, the MMCRA
register contained "slot information" that the oprofile kernel driver used to
adjust the value latched in the SIAR at the time of a PMU interrupt.  But as of
POWER6, these slot bits in MMCRA are no longer necessary for oprofile to use,
since the SIAR itself holds the accurate sampled instruction address.  With
POWER6, these MMCRA slot bits were zero'ed out by hardware so oprofile's use of
these slot bits was, in effect, a NOP.  But with POWER7, these bits are no
longer zero'ed out; however, they serve some other purpose rather than slot
information.  Thus, using these bits on POWER7 to adjust the SIAR value results
in samples being attributed to the wrong instructions.  The attached patch
changes the oprofile kernel driver to ignore these slot bits on all newer
processors starting with POWER6.

Signed-off-by: Maynard Johnson <maynardj@us.ibm.com>
Signed-off-by: Michael Wolf <mjw@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc/iseries: Fix pci breakage due to bad dma_data initialization
Stephen Rothwell x [Wed, 6 May 2009 14:07:52 +0000 (14:07 +0000)] 
powerpc/iseries: Fix pci breakage due to bad dma_data initialization

Commit 4fc665b88a79a45bae8bbf3a05563c27c7337c3d "powerpc: Merge 32 and
64-bit dma code" made changes to the PCI initialisation code that added
an assignment to archdata.dma_data but only for 32 bit code.  Commit
7eef440a545c7f812ed10b49d4a10a351df9cad6 "powerpc/pci: Cosmetic cleanups
of pci-common.c" removed the conditional compilation.  Unfortunately,
the iSeries code setup the archdata.dma_data before that assignment was
done - effectively overwriting the dma_data with NULL.

Fix this up by moving the iSeries setup of dma_data into a
pci_dma_dev_setup callback.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agopowerpc: Fix mktree build error on Mac OS X host
Timur Tabi [Thu, 30 Apr 2009 18:16:44 +0000 (18:16 +0000)] 
powerpc: Fix mktree build error on Mac OS X host

The mktree utility defines some variables as "uint", although this is not a
standard C type, and so cross-compiling on Mac OS X fails.  Change this to
"unsigned int".

Signed-off-by: Timur Tabi <timur@freescale.com>
Acked-by: Josh Boyer <jwboyer@linux.vnet.ibm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Fri, 15 May 2009 02:20:04 +0000 (19:20 -0700)] 
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  cifs: fix error handling in parse_DFS_referrals

15 years agoMerge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
Linus Torvalds [Fri, 15 May 2009 02:19:43 +0000 (19:19 -0700)] 
Merge branch 'upstream' of git://ftp.linux-mips.org/upstream-linus

* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus: (38 commits)
  MIPS: Sibyte: Fix locking in set_irq_affinity
  MIPS: Use force_sig when handling address errors.
  MIPS: Cavium: Add struct clocksource * argument to octeon_cvmcount_read()
  MIPS: Rewrite <asm/div64.h> to work with gcc 4.4.0.
  MIPS: Fix highmem.
  MIPS: Fix sign-extension bug in 32-bit kernel on 32-bit hardware.
  MIPS: MSP71xx: Remove the RAMROOT functions
  MIPS: Use -mno-check-zero-division
  MIPS: Set compiler options only after the compiler prefix has ben set.
  MIPS: IP27: Get rid of #ident.  Gcc 4.4.0 doesn't like it.
  MIPS: uaccess: Switch lock annotations to might_fault().
  MIPS: MSP71xx: Resolve use of non-existent GPIO routines in msp71xx reset
  MIPS: MSP71xx: Resolve multiple definition of plat_timer_setup
  MIPS: Make uaccess.h slightly more sparse friendly.
  MIPS: Make access_ok() sideeffect proof.
  MIPS: IP27: Fix clash with NMI_OFFSET from hardirq.h
  MIPS: Alchemy: Timer build fix
  MIPS: Kconfig: Delete duplicate definition of RWSEM_GENERIC_SPINLOCK.
  MIPS: Cavium: Add support for 8k and 32k page sizes.
  MIPS: TXx9: Fix possible overflow in clock calculations
  ...

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable
Linus Torvalds [Fri, 15 May 2009 02:18:44 +0000 (19:18 -0700)] 
Merge git://git./linux/kernel/git/mason/btrfs-unstable

* git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable:
  Btrfs: Spelling fix in btrfs_lookup_first_block_group comments
  Btrfs: make show_options result match actual option names
  Btrfs: remove outdated comment in btrfs_ioctl_resize()
  Btrfs: remove some WARN_ONs in the IO failure path
  Btrfs: Don't loop forever on metadata IO failures
  Btrfs: init inode ordered_data_close flag properly

15 years agodrm/i915: Add new GET_PIPE_FROM_CRTC_ID ioctl.
Carl Worth [Wed, 29 Apr 2009 21:43:54 +0000 (14:43 -0700)] 
drm/i915: Add new GET_PIPE_FROM_CRTC_ID ioctl.

This allows userlevel code to discover the pipe number corresponding
to a given CRTC ID. This is necessary for doing pipe-specific
operations such as waiting for vblank on a given CRTC.  Failure to use
the right pipe mapping can result in GPU hangs, or at least failure
to actually sync to vblank.

Signed-off-by: Carl Worth <cworth@cworth.org>
[anholt: Style touchups from review]
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Set HDMI hot plug interrupt enable for only the output in question.
Ma Ling [Mon, 11 May 2009 03:33:22 +0000 (11:33 +0800)] 
drm/i915: Set HDMI hot plug interrupt enable for only the output in question.

We detect HDMI output connection status by writing to HOT Plug Interrupt
Detect Enable bit in PORT_HOTPLUG_EN.  The behavior will generate a specified
interrupt, which is caught by audio driver, but during one detection driver
set all Detect Enable bits of HDMIB, HDMIC HDMID, and generate wrong
interrupt signals for current output, according to the signals audio driver
misunderstand device status. The patch intends to handle corresponding
output precisely.

It fixed freedesktop.org bug #21371

Signed-off-by: Ma Ling <ling.ma@intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Include 965GME pci ID in IS_I965GM(dev) to match UMS.
Ma Ling [Wed, 13 May 2009 07:08:27 +0000 (15:08 +0800)] 
drm/i915: Include 965GME pci ID in IS_I965GM(dev) to match UMS.

It fixed bug #21659

Signed-off-by: Ma Ling <ling.ma@intel.com>
[anholt: hand-applied because git-am is too picky]
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Use the GM45 VGA hotplug workaround on G45 as well.
Ma Ling [Wed, 13 May 2009 06:46:12 +0000 (14:46 +0800)] 
drm/i915: Use the GM45 VGA hotplug workaround on G45 as well.

Although spec say CRT_HOTPLUG_ACTIVATION_PERIOD_64 is only useful for
mobile platform, it is also required to detect vga on G4x desktops correctly.
Tested on G45/G43/Q45 platforms with no regressions.

It fixed freedesktop.org bug #21120 and part of bug #21210

Signed-off-by: Ma Ling <ling.ma@intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: ignore LVDS on intel graphics systems that lie about having it
Jarod Wilson [Tue, 5 May 2009 14:00:25 +0000 (10:00 -0400)] 
drm/i915: ignore LVDS on intel graphics systems that lie about having it

There are a number of small form factor desktop systems with Intel mobile
graphics chips that lie and say they have an LVDS. With kernel mode-setting,
this becomes a problem, and makes native resolution boot go haywire -- for
example, my Dell Studio Hybrid, hooked to a 1920x1080 display claims to
have a 1024x768 LVDS, and the resulting graphical boot on the 1920x1080
display uses only the top left 1024x768, and auto-configured X will end
up only 1024x768 as well. With this change, graphical boot and X
both do 1920x1080 as expected.

Note that we're simply embracing and extending the early bail-out code
in place for the Mac Mini here. The xorg intel driver uses pci subsystem
device and vendor id for matching, while we're using dmi lookups here.
The MSI addition is courtesy of and tested by Bill Nottingham.

Signed-off-by: Jarod Wilson <jarod@redhat.com>
Tested-by: Bill Nottingham <notting@redhat.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: sanity check IER at wait_request time
Jesse Barnes [Tue, 5 May 2009 23:03:48 +0000 (16:03 -0700)] 
drm/i915: sanity check IER at wait_request time

We might sleep here anyway so I hope an extra uncached read is ok to
add.

In #20896 we found that vbetool clobbers the IER.  In KMS mode this is
particularly bad since we don't set the interrupt regs late (in
EnterVT), so we'd fail to get *any* interrupts at all after X started
(since some distros have scripts that call vbetool at X startup
apparently).

So this patch checks IER at wait_request time, and re-enables
interrupts if it's been clobbered.  In a proper config this check
should never be triggered.

This is really a distro issue, but having a sanity check is nice, as
long as it doesn't have a real performance hit.

Tested-by: Mateusz Kaduk <mateusz.kaduk@gmail.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
[anholt: Moved the check inside of the sleeping case to avoid perf cost]
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: workaround IGD i2c bus issue in kernel side (v2)
Shaohua Li [Tue, 7 Apr 2009 03:02:28 +0000 (11:02 +0800)] 
drm/i915: workaround IGD i2c bus issue in kernel side (v2)

In IGD, DPCUNIT_CLOCK_GATE_DISABLE bit should be set, otherwise i2c
access will be wrong.

v2: Disable CLOCK_GATE_DISABLE bit after bit bashing as suggested by Eric.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agodrm/i915: Don't allow binding objects into the last page of the aperture.
Eric Anholt [Tue, 12 May 2009 22:27:36 +0000 (15:27 -0700)] 
drm/i915: Don't allow binding objects into the last page of the aperture.

This should avoid a class of bugs where the hardware prefetches past the
end of the object, and walks into unallocated memory when the object is
bound to the last page of the aperture.

fd.o bug #21488

Signed-off-by: Eric Anholt <eric@anholt.net>
15 years agoext4: Clear the unwritten buffer_head flag after the extent is initialized
Aneesh Kumar K.V [Thu, 14 May 2009 21:05:39 +0000 (17:05 -0400)] 
ext4: Clear the unwritten buffer_head flag after the extent is initialized

The BH_Unwritten flag indicates that the buffer is allocated on disk
but has not been written; that is, the disk was part of a persistent
preallocation area.  That flag should only be set when a get_blocks()
function is looking up a inode's logical to physical block mapping.

When ext4_get_blocks_wrap() is called with create=1, the uninitialized
extent is converted into an initialized one, so the BH_Unwritten flag
is no longer appropriate.  Hence, we need to make sure the
BH_Unwritten is not left set, since the combination of BH_Mapped and
BH_Unwritten is not allowed; among other things, it will result ext4's
get_block() to be called over and over again during the write_begin
phase of write(2).

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoBtrfs: Spelling fix in btrfs_lookup_first_block_group comments
Sankar P [Thu, 14 May 2009 17:52:22 +0000 (13:52 -0400)] 
Btrfs: Spelling fix in btrfs_lookup_first_block_group comments

Signed-off-by: Sankar P <sankar.curiosity@gmail.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>