GitHub/exynos8895/android_kernel_samsung_universal8895.git
12 years agoperf tools: Fix getrusage() related build failure on glibc trunk
Markus Trippelsdorf [Wed, 4 Apr 2012 08:45:27 +0000 (10:45 +0200)]
perf tools: Fix getrusage() related build failure on glibc trunk

On a system running glibc trunk perf doesn't build:

    CC builtin-sched.o
builtin-sched.c: In function ‘get_cpu_usage_nsec_parent’: builtin-sched.c:399:16: error: storage size of ‘ru’ isn’t known builtin-sched.c:403:2: error: implicit declaration of function ‘getrusage’ [-Werror=implicit-function-declaration]
    [...]

Fix it by including sys/resource.h.

Signed-off-by: Markus Trippelsdorf <markus@trippelsdorf.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120404084527.GA294@x4
Signed-off-by: Ingo Molnar <mingo@kernel.org>
12 years agoperf/x86/p4: Add format attributes
Peter Zijlstra [Tue, 27 Mar 2012 14:50:42 +0000 (16:50 +0200)]
perf/x86/p4: Add format attributes

Steven reported his P4 not booting properly, the missing format
attributes cause a NULL ptr deref. Cure this by adding the
missing format specification.

I took the format description out of the comment near
p4_config_pack*() and hope that comment is still relatively
accurate.

Reported-by: Steven Rostedt <rostedt@goodmis.org>
Reported-by: Bruno Prémont <bonbons@linux-vserver.org>
Tested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Lin Ming <ming.m.lin@intel.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1332859842.16159.227.camel@twins
Signed-off-by: Ingo Molnar <mingo@kernel.org>
12 years agotracing, sched, vfs: Fix 'old_pid' usage in trace_sched_process_exec()
Oleg Nesterov [Fri, 30 Mar 2012 16:26:36 +0000 (18:26 +0200)]
tracing, sched, vfs: Fix 'old_pid' usage in trace_sched_process_exec()

1. TRACE_EVENT(sched_process_exec) forgets to actually use the
   old pid argument, it sets ->old_pid = p->pid.

2. search_binary_handler() uses the wrong pid number. tracepoint
   needs the global pid_t from the root namespace, while old_pid
   is the virtual pid number as it seen by the tracer/parent.

With this patch we have two pid_t's in search_binary_handler(),
not really nice. Perhaps we should switch to "struct pid*", but
in this case it would be better to cleanup the current code
first and move the "depth == 0" code outside.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: David Smith <dsmith@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Link: http://lkml.kernel.org/r/20120330162636.GA4857@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
12 years agoMerge branch 'perf/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/acme...
Ingo Molnar [Sat, 31 Mar 2012 07:27:41 +0000 (09:27 +0200)]
Merge branch 'perf/urgent' of git://git./linux/kernel/git/acme/linux into perf/urgent

12 years agoperf tools: Remove auto-generated bison/flex files
Ingo Molnar [Thu, 29 Mar 2012 16:08:38 +0000 (13:08 -0300)]
perf tools: Remove auto-generated bison/flex files

These should not be in the Git history - they are auto-generated.

Extend the Makefile rules of the parser files to include the generation
run.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/20120327183335.GA27621@gmail.com
[ committer note: Fixed up O= handling ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agoperf annotate: Fix off by one symbol hist size allocation and hit accounting
Arnaldo Carvalho de Melo [Wed, 28 Mar 2012 15:49:35 +0000 (12:49 -0300)]
perf annotate: Fix off by one symbol hist size allocation and hit accounting

We were not noticing it because symbol__inc_addr_samples was erroneously
dropping samples that hit the last byte in a function.

Working on a fix for a problem reported by David Miller, Stephane
Eranian and Sorin Dumitru, where addresses < sym->start were causing
problems, I noticed this other problem.

Cc: David Ahern <dsahern@gmail.com>
Cc: David Miller <davem@davemloft.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sorin Dumitru <dumitru.sorin87@gmail.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/n/tip-pqjaq4cr1xs2xen73pjhbav4@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agoperf tools: Add missing ref-cycles event back to event parser
Namhyung Kim [Thu, 29 Mar 2012 10:21:57 +0000 (19:21 +0900)]
perf tools: Add missing ref-cycles event back to event parser

The commit 89812fc81f8d ("perf tools: Add parser generator for events
parsing") changed event parsing engine but missed the ref-cycles event.
Add it.

Signed-off-by: Namhyung Kim <namhyung.kim@lge.com>
Acked-by: Jiri Olsa <jolsa@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1333016517-10591-1-git-send-email-namhyung.kim@lge.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agoperf annotate: addr2line wants addresses in same format as objdump
David Miller [Sun, 25 Mar 2012 20:28:12 +0000 (16:28 -0400)]
perf annotate: addr2line wants addresses in same format as objdump

Therefore, in symbol__get_source_line(), use map__rip_2objdump
instead of calling map->unmap_ip() unconditionally.

Link: http://lkml.kernel.org/r/20120325.162812.59519424882536855.davem@davemloft.net
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agoperf probe: Finder fails to resolve function name to address
Prashanth Nageshappa [Mon, 26 Mar 2012 10:06:49 +0000 (15:36 +0530)]
perf probe: Finder fails to resolve function name to address

If DIE entries corresponding to declarations appear before definition
entry, probe finder returns error instead of continuing to look further
for a definition entry.

This patch ensures we reach to the DIE entry corresponding to the
definition and get the function address.

V2: A simpler solution based on Masami's suggestion.

Signed-off-by: Prashanth Nageshappa <prashanth@linux.vnet.ibm.com>
Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/4F703FB9.9020407@linux.vnet.ibm.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agotracing: Fix ent_size in trace output
Steven Rostedt [Tue, 27 Mar 2012 14:43:28 +0000 (10:43 -0400)]
tracing: Fix ent_size in trace output

When reading the trace file, the records of each of the per_cpu buffers
are examined to find the next event to print out. At the point of looking
at the event, the size of the event is recorded. But if the first event is
chosen, the other events in the other CPU buffers will reset the event size
that is stored in the iterator descriptor, causing the event size passed to
the output functions to be incorrect.

In most cases this is not a problem, but for the case of stack traces, it
is. With the change to the stack tracing to record a dynamic number of
back traces, the output depends on the size of the entry instead of the
fixed 8 back traces. When the entry size is not correct, the back traces
would not be fully printed.

Note, reading from the per-cpu trace files were not affected.

Reported-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
12 years agoperf symbols: Handle NULL dso in dso__name_len
David Miller [Sun, 25 Mar 2012 20:28:22 +0000 (16:28 -0400)]
perf symbols: Handle NULL dso in dso__name_len

We should use "[unknown]" in this case, in concert with the code in
_hist_entry__dso_snprintf().

Otherwise we'll crash when recomputing the histogram column lengths in
hists__calc_col_len().

Signed-off-by: David S. Miller <davem@davemloft.net>
Link: http://lkml.kernel.org/r/20120325.162822.2267799792062571623.davem@davemloft.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agoperf symbols: Do not include libgen.h
David Miller [Tue, 27 Mar 2012 04:03:01 +0000 (00:03 -0400)]
perf symbols: Do not include libgen.h

That causes us to end up using the XPG version of basename which can
modify it's argument.

Signed-off-by: David S. Miller <davem@davemloft.net>
Link: http://lkml.kernel.org/r/20120327.000301.1122788061724345175.davem@davemloft.net
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agoperf tools: Fix bug in raw sample parsing
Stephane Eranian [Sat, 17 Mar 2012 22:23:18 +0000 (23:23 +0100)]
perf tools: Fix bug in raw sample parsing

In perf_event__parse_sample(), the array variable was not incremented
by the amount of data used by the raw_data.

That was okay until we added PERF_SAMPLE_BRANCH_STACK which depends on
the array variable pointing to the beginning of the branch stack data.

But that was not the case if branch stack was combined with raw mode
sampling. That led to bogus branch stack addresses and count.

The bug would show up with:
$ perf record -R -b foo

This patch fixes the problem by correctly moving the array pointer
forward for RAW samples.

Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20120317222317.GA8803@quad
[ committer note: Fix also later submitted by Jiri Olsa ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agoperf tools: Fix display of first level of callchains
Frederic Weisbecker [Fri, 23 Mar 2012 18:06:50 +0000 (19:06 +0100)]
perf tools: Fix display of first level of callchains

The callchain stdio mode display was written using a sorted by symbol
report. In this mode we have only one callchain root per hist so we
forgot to handle cases where we have multiple callchain root, as in per
dso sorting for example.

Fix this by handling these roots like any other branch, with the hist as
the parent.

Before:

     1.97%  libpthread-2.12.1.so
            |
            --- __libc_write
                create_worker
                bench_sched_messaging
                cmd_bench
                run_builtin
                main
                __libc_start_main

            |
            --- __libc_read
                create_worker
                bench_sched_messaging
                cmd_bench
                run_builtin
                main
                __libc_start_main

After:

     1.97%  libpthread-2.12.1.so
            |
            |--36.97%-- __libc_write
            |          create_worker
            |          bench_sched_messaging
            |          cmd_bench
            |          run_builtin
            |          main
            |          __libc_start_main
            |
            |--31.47%-- __libc_read
            |          create_worker
            |          bench_sched_messaging
            |          cmd_bench
            |          run_builtin
            |          main
            |          __libc_start_main
           ...

Single roots keep their entry without percentage because they have
the same overhead than the hist they refer to. ie: 100% in fractal
mode and the percentage of the hist in graph mode:

     0.00%  [k] reschedule_interrupt
            |
            --- default_idle
                amd_e400_idle
                cpu_idle
                start_secondary

Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1332526010-15400-1-git-send-email-fweisbec@gmail.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
12 years agoperf tools: Switch module.h into export.h
Jiri Olsa [Mon, 26 Mar 2012 09:17:05 +0000 (11:17 +0200)]
perf tools: Switch module.h into export.h

When merged to Linus's latest tree the perf build is broken
due to following change in lib/rbtree.c object:

  lib: reduce the use of module.h wherever possible
  commit 8bc3bcc93a2b4e47d5d410146f6546bca6171663
  Author: Paul Gortmaker <paul.gortmaker@windriver.com>
  Date:   Wed Nov 16 21:29:17 2011 -0500

We need to move module.h header into export.h.

Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Cc: acme@redhat.com
Cc: a.p.zijlstra@chello.nl
Cc: paulus@samba.org
Cc: cjashfor@linux.vnet.ibm.com
Cc: fweisbec@gmail.com
Link: http://lkml.kernel.org/r/1332753425-3299-1-git-send-email-jolsa@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
12 years agoMerge branch 'linus' into perf/urgent
Ingo Molnar [Mon, 26 Mar 2012 15:18:44 +0000 (17:18 +0200)]
Merge branch 'linus' into perf/urgent

Merge reason: we need to fix a non-trivial merge conflict.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
12 years agoMerge tag 'stable/for-linus-3.4-tag-two' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 24 Mar 2012 19:20:25 +0000 (12:20 -0700)]
Merge tag 'stable/for-linus-3.4-tag-two' of git://git./linux/kernel/git/konrad/xen

Pull more xen updates from Konrad Rzeszutek Wilk:
 "One tiny feature that accidentally got lost in the initial git pull:
   * Add fast-EOI acking of interrupts (clear a bit instead of
     hypercall)
  And bug-fixes:
   * Fix CPU bring-up code missing a call to notify other subsystems.
   * Fix reading /sys/hypervisor even if PVonHVM drivers are not loaded.
   * In Xen ACPI processor driver: remove too verbose WARN messages, fix
     up the Kconfig dependency to be a module by default, and add
     dependency on CPU_FREQ.
   * Disable CPU frequency drivers from loading when booting under Xen
     (as we want the Xen ACPI processor to be used instead).
   * Cleanups in tmem code."

* tag 'stable/for-linus-3.4-tag-two' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad/xen:
  xen/acpi: Fix Kconfig dependency on CPU_FREQ
  xen: initialize platform-pci even if xen_emul_unplug=never
  xen/smp: Fix bringup bug in AP code.
  xen/acpi: Remove the WARN's as they just create noise.
  xen/tmem: cleanup
  xen: support pirq_eoi_map
  xen/acpi-processor: Do not depend on CPU frequency scaling drivers.
  xen/cpufreq: Disable the cpu frequency scaling drivers from loading.
  provide disable_cpufreq() function to disable the API.

12 years agoFix potential endless loop in kswapd when compaction is not enabled
Rik van Riel [Sat, 24 Mar 2012 14:26:21 +0000 (10:26 -0400)]
Fix potential endless loop in kswapd when compaction is not enabled

We should only test compaction_suitable if the kernel is built with
CONFIG_COMPACTION, otherwise the stub compaction_suitable function will
always return COMPACT_SKIPPED and send kswapd into an infinite loop.

Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge tag 'device-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg...
Linus Torvalds [Sat, 24 Mar 2012 17:41:37 +0000 (10:41 -0700)]
Merge tag 'device-for-3.4' of git://git./linux/kernel/git/paulg/linux

Pull <linux/device.h> avoidance patches from Paul Gortmaker:
 "Nearly every subsystem has some kind of header with a proto like:

void foo(struct device *dev);

  and yet there is no reason for most of these guys to care about the
  sub fields within the device struct.  This allows us to significantly
  reduce the scope of headers including headers.  For this instance, a
  reduction of about 40% is achieved by replacing the include with the
  simple fact that the device is some kind of a struct.

  Unlike the much larger module.h cleanup, this one is simply two
  commits.  One to fix the implicit <linux/device.h> users, and then one
  to delete the device.h includes from the linux/include/ dir wherever
  possible."

* tag 'device-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux:
  device.h: audit and cleanup users in main include dir
  device.h: cleanup users outside of linux/include (C files)

12 years agoMerge tag 'module-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg...
Linus Torvalds [Sat, 24 Mar 2012 17:24:31 +0000 (10:24 -0700)]
Merge tag 'module-for-3.4' of git://git./linux/kernel/git/paulg/linux

Pull cleanup of fs/ and lib/ users of module.h from Paul Gortmaker:
 "Fix up files in fs/ and lib/ dirs to only use module.h if they really
  need it.

  These are trivial in scope vs the work done previously.  We now have
  things where any few remaining cleanups can be farmed out to arch or
  subsystem maintainers, and I have done so when possible.  What is
  remaining here represents the bits that don't clearly lie within a
  single arch/subsystem boundary, like the fs dir and the lib dir.

  Some duplicate includes arising from overlapping fixes from
  independent subsystem maintainer submissions are also quashed."

Fix up trivial conflicts due to clashes with other include file cleanups
(including some due to the previous bug.h cleanup pull).

* tag 'module-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux:
  lib: reduce the use of module.h wherever possible
  fs: reduce the use of module.h wherever possible
  includecheck: delete any duplicate instances of module.h

12 years agoMerge tag 'bug-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux
Linus Torvalds [Sat, 24 Mar 2012 17:08:39 +0000 (10:08 -0700)]
Merge tag 'bug-for-3.4' of git://git./linux/kernel/git/paulg/linux

Pull <linux/bug.h> cleanup from Paul Gortmaker:
 "The changes shown here are to unify linux's BUG support under the one
  <linux/bug.h> file.  Due to historical reasons, we have some BUG code
  in bug.h and some in kernel.h -- i.e.  the support for BUILD_BUG in
  linux/kernel.h predates the addition of linux/bug.h, but old code in
  kernel.h wasn't moved to bug.h at that time.  As a band-aid, kernel.h
  was including <asm/bug.h> to pseudo link them.

  This has caused confusion[1] and general yuck/WTF[2] reactions.  Here
  is an example that violates the principle of least surprise:

      CC      lib/string.o
      lib/string.c: In function 'strlcat':
      lib/string.c:225:2: error: implicit declaration of function 'BUILD_BUG_ON'
      make[2]: *** [lib/string.o] Error 1
      $
      $ grep linux/bug.h lib/string.c
      #include <linux/bug.h>
      $

  We've included <linux/bug.h> for the BUG infrastructure and yet we
  still get a compile fail! [We've not kernel.h for BUILD_BUG_ON.] Ugh -
  very confusing for someone who is new to kernel development.

  With the above in mind, the goals of this changeset are:

  1) find and fix any include/*.h files that were relying on the
     implicit presence of BUG code.
  2) find and fix any C files that were consuming kernel.h and hence
     relying on implicitly getting some/all BUG code.
  3) Move the BUG related code living in kernel.h to <linux/bug.h>
  4) remove the asm/bug.h from kernel.h to finally break the chain.

  During development, the order was more like 3-4, build-test, 1-2.  But
  to ensure that git history for bisect doesn't get needless build
  failures introduced, the commits have been reorderd to fix the problem
  areas in advance.

[1]  https://lkml.org/lkml/2012/1/3/90
[2]  https://lkml.org/lkml/2012/1/17/414"

Fix up conflicts (new radeon file, reiserfs header cleanups) as per Paul
and linux-next.

* tag 'bug-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux:
  kernel.h: doesn't explicitly use bug.h, so don't include it.
  bug: consolidate BUILD_BUG_ON with other bug code
  BUG: headers with BUG/BUG_ON etc. need linux/bug.h
  bug.h: add include of it to various implicit C users
  lib: fix implicit users of kernel.h for TAINT_WARN
  spinlock: macroize assert_spin_locked to avoid bug.h dependency
  x86: relocate get/set debugreg fcns to include/asm/debugreg.

12 years agoxen/acpi: Fix Kconfig dependency on CPU_FREQ
Konrad Rzeszutek Wilk [Sat, 24 Mar 2012 13:18:57 +0000 (09:18 -0400)]
xen/acpi: Fix Kconfig dependency on CPU_FREQ

The functions: "acpi_processor_*" sound like they depend on CONFIG_ACPI_PROCESSOR
but in reality they are exposed when CONFIG_CPU_FREQ=[y|m]. As such
update the Kconfig to have this dependency and fix compile issues:

ERROR: "acpi_processor_unregister_performance" [drivers/xen/xen-acpi-processor.ko] undefined!
ERROR: "acpi_processor_notify_smm" [drivers/xen/xen-acpi-processor.ko] undefined!
ERROR: "acpi_processor_register_performance" [drivers/xen/xen-acpi-processor.ko] undefined!
ERROR: "acpi_processor_preregister_performance" [drivers/xen/xen-acpi-processor.ko] undefined!

Note: We still need the CONFIG_ACPI
Reported-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoperf: Move mmap page data_head offset assertion out of header
Jiri Olsa [Fri, 23 Mar 2012 14:41:20 +0000 (15:41 +0100)]
perf: Move mmap page data_head offset assertion out of header

Having the build time assertion in header is making the perf
build fail on x86 with:

  ../../include/linux/perf_event.h:411:32: error: variably modified \
‘__assert_mmap_data_head_offset’ at file scope [-Werror]

I'm moving the build time validation out of the header, because
I think it's better than to lessen the perf build warn/error
check.

Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Cc: acme@redhat.com
Cc: a.p.zijlstra@chello.nl
Cc: paulus@samba.org
Cc: cjashfor@linux.vnet.ibm.com
Cc: fweisbec@gmail.com
Link: http://lkml.kernel.org/r/1332513680-7870-1-git-send-email-jolsa@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
12 years agoMerge branch 'tip/perf/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/roste...
Ingo Molnar [Sat, 24 Mar 2012 07:19:09 +0000 (08:19 +0100)]
Merge branch 'tip/perf/urgent' of git://git./linux/kernel/git/rostedt/linux-trace into perf/urgent

12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/sysctl
Linus Torvalds [Sat, 24 Mar 2012 01:08:58 +0000 (18:08 -0700)]
Merge git://git./linux/kernel/git/ebiederm/sysctl

Pull sysctl updates from Eric Biederman:

 - Rewrite of sysctl for speed and clarity.

   Insert/remove/Lookup in sysctl are all now O(NlogN) operations, and
   are no longer bottlenecks in the process of adding and removing
   network devices.

   sysctl is now focused on being a filesystem instead of system call
   and the code can all be found in fs/proc/proc_sysctl.c.  Hopefully
   this means the code is now approachable.

   Much thanks is owed to Lucian Grinjincu for keeping at this until
   something was found that was usable.

 - The recent proc_sys_poll oops found by the fuzzer during hibernation
   is fixed.

* git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/sysctl: (36 commits)
  sysctl: protect poll() in entries that may go away
  sysctl: Don't call sysctl_follow_link unless we are a link.
  sysctl: Comments to make the code clearer.
  sysctl: Correct error return from get_subdir
  sysctl: An easier to read version of find_subdir
  sysctl: fix memset parameters in setup_sysctl_set()
  sysctl: remove an unused variable
  sysctl: Add register_sysctl for normal sysctl users
  sysctl: Index sysctl directories with rbtrees.
  sysctl: Make the header lists per directory.
  sysctl: Move sysctl_check_dups into insert_header
  sysctl: Modify __register_sysctl_paths to take a set instead of a root and an nsproxy
  sysctl: Replace root_list with links between sysctl_table_sets.
  sysctl: Add sysctl_print_dir and use it in get_subdir
  sysctl: Stop requiring explicit management of sysctl directories
  sysctl: Add a root pointer to ctl_table_set
  sysctl: Rewrite proc_sys_readdir in terms of first_entry and next_entry
  sysctl: Rewrite proc_sys_lookup introducing find_entry and lookup_entry.
  sysctl: Normalize the root_table data structure.
  sysctl: Factor out insert_header and erase_header
  ...

12 years agoMerge tag 'amd64-edac-updates-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 24 Mar 2012 00:59:47 +0000 (17:59 -0700)]
Merge tag 'amd64-edac-updates-for-3.4' of git://git./linux/kernel/git/bp/bp

Pull AMD64 EDAC fixes from Borislav Petkov:
 "A bunch of fixes/updates for the AMD side of EDAC including

   * MCE decoding updates
   * tree-wide EDAC sweep making pci_device_ids __devinitconst
   * Scrub rate API correction
   * two amd64_edac corrections for K8 boxes and sysfs csrow nodes"

* tag 'amd64-edac-updates-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/bp/bp:
  MCE, AMD: Constify error tables
  MCE, AMD: Correct bank 5 error signatures
  MCE, AMD: Rework NB MCE signatures
  MCE, AMD: Correct VB data error description
  MCE, AMD: Correct ucode patch buffer description
  MCE, AMD: Correct some MC0 error types
  EDAC: Make pci_device_id tables __devinitconst.
  EDAC: Correct scrub rate API
  amd64_edac: Fix K8 revD and later chip select sizes
  amd64_edac: Fix missing csrows sysfs nodes

12 years agoMerge branch 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq
Linus Torvalds [Sat, 24 Mar 2012 00:56:39 +0000 (17:56 -0700)]
Merge branch 'next' of git://git./linux/kernel/git/davej/cpufreq

Pull cpufreq updates for 3.4 from Dave Jones: new drivers and some fixes.

* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq:
  provide disable_cpufreq() function to disable the API.
  EXYNOS5250: Add support cpufreq for EXYNOS5250
  EXYNOS4X12: Add support cpufreq for EXYNOS4X12
  [CPUFREQ] CPUfreq ondemand: update sampling rate without waiting for next sampling
  [CPUFREQ] Add S3C2416/S3C2450 cpufreq driver
  [CPUFREQ] Fix exposure of ARM_EXYNOS4210_CPUFREQ
  [CPUFREQ] EXYNOS4210: update the name of EXYNOS clock register
  [CPUFREQ] EXYNOS: Initialize locking_frequency with initial frequency
  [CPUFREQ] s3c64xx: Fix mis-cherry pick of VDDINT

Fix up trivial conflicts in Kconfig and Makefile due to just changes
next to each other (OMAP2PLUS changes vs some new EXYNOS cpufreq
drivers).

12 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq
Linus Torvalds [Sat, 24 Mar 2012 00:51:50 +0000 (17:51 -0700)]
Merge branch 'fixes' of git://git./linux/kernel/git/davej/cpufreq

Pull cpufreq fixes from Dave Jones:
 "I meant to get some of these in for 3.3 final, but left things too
  late, so I've got two trees this time."

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/davej/cpufreq:
  cpufreq: OMAP: specify range for voltage scaling
  cpufreq: OMAP: scale voltage along with frequency
  cpufreq: OMAP driver depends CPUfreq tables

12 years agoMerge branch 'pcmcia' of git://git.linaro.org/people/rmk/linux-arm
Linus Torvalds [Sat, 24 Mar 2012 00:37:40 +0000 (17:37 -0700)]
Merge branch 'pcmcia' of git://git.linaro.org/people/rmk/linux-arm

Pull #3 ARM updates from Russell King:
 "This adds gpio support to soc_common, allowing an amount of code to be
  deleted from each PCMCIA socket driver for the PXA/SA11x0 SoCs."

* 'pcmcia' of git://git.linaro.org/people/rmk/linux-arm:
  PCMCIA: sa1111: rename sa1111 socket drivers to have sa1111_ prefix.
  PCMCIA: make lubbock socket driver part of sa1111_cs
  PCMCIA: add Kconfig control for building sa11xx_base.c
  PCMCIA: sa1111: jornada720: no need to disable IRQs around sa1111_set_io
  PCMCIA: sa1111: pass along sa1111_pcmcia_configure_socket() failure code
  PCMCIA: soc_common: remove explicit wrprot initialization in socket drivers
  PCMCIA: soc_common: remove soc_pcmcia_*_irqs functions
  PCMCIA: sa11x0: h3600: convert to use new irq/gpio management
  PCMCIA: sa11x0: simpad: convert to use new irq/gpio management
  PCMCIA: sa11x0: shannon: convert to use new irq/gpio management
  PCMCIA: sa11x0: nanoengine: convert reset handling to use GPIO subsystem
  PCMCIA: sa11x0: nanoengine: convert to use new irq/gpio management
  PCMCIA: sa11x0: cerf: convert reset handling to use GPIO subsystem
  PCMCIA: sa11x0: cerf: convert to use new irq/gpio management
  PCMCIA: sa11x0: assabet: convert to use new irq/gpio management
  PCMCIA: sa1111: use new per-socket irq/gpio infrastructure
  PCMCIA: pxa: convert PXA socket drivers to use new irq/gpio management
  PCMCIA: soc_common: add GPIO support for card status signals
  PCMCIA: soc_common: move common initialization into soc_common

12 years agoMerge branch 'amba' of git://git.linaro.org/people/rmk/linux-arm
Linus Torvalds [Sat, 24 Mar 2012 00:36:29 +0000 (17:36 -0700)]
Merge branch 'amba' of git://git.linaro.org/people/rmk/linux-arm

Pull #2 ARM updates from Russell King:
 "Further ARM AMBA primecell updates which aren't included directly in
  the previous commit.  I wanted to keep these separate as they're
  touching stuff outside arch/arm/."

* 'amba' of git://git.linaro.org/people/rmk/linux-arm:
  ARM: 7362/1: AMBA: Add module_amba_driver() helper macro for amba_driver
  ARM: 7335/1: mach-u300: do away with MMC config files
  ARM: 7280/1: mmc: mmci: Cache MMCICLOCK and MMCIPOWER register
  ARM: 7309/1: realview: fix unconnected interrupts on EB11MP
  ARM: 7230/1: mmc: mmci: Fix PIO read for small SDIO packets
  ARM: 7227/1: mmc: mmci: Prepare for SDIO before setting up DMA job
  ARM: 7223/1: mmc: mmci: Fixup use of runtime PM and use autosuspend
  ARM: 7221/1: mmc: mmci: Change from using legacy suspend
  ARM: 7219/1: mmc: mmci: Change vdd_handler to a generic ios_handler
  ARM: 7218/1: mmc: mmci: Provide option to configure bus signal direction
  ARM: 7217/1: mmc: mmci: Put power register deviations in variant data
  ARM: 7216/1: mmc: mmci: Do not release spinlock in request_end
  ARM: 7215/1: mmc: mmci: Increase max_segs from 16 to 128

12 years agoMerge branch 'for-armsoc' of git://git.linaro.org/people/rmk/linux-arm
Linus Torvalds [Sat, 24 Mar 2012 00:30:49 +0000 (17:30 -0700)]
Merge branch 'for-armsoc' of git://git.linaro.org/people/rmk/linux-arm

Pull #1 ARM updates from Russell King:
 "This one covers stuff which Arnd is waiting for me to push, as this is
  shared between both our trees and probably other trees elsewhere.

  Essentially, this contains:
   - AMBA primecell device initializer updates - mostly shrinking the
     size of the device declarations in platform code to something more
     reasonable.
   - Getting rid of the NO_IRQ crap from AMBA primecell stuff.
   - Nicolas' idle cleanups.  This in combination with the restart
     cleanups from the last merge window results in a great many
     mach/system.h files being deleted."

Yay: ~80 files, ~2000 lines deleted.

* 'for-armsoc' of git://git.linaro.org/people/rmk/linux-arm: (60 commits)
  ARM: remove disable_fiq and arch_ret_to_user macros
  ARM: make entry-macro.S depend on !MULTI_IRQ_HANDLER
  ARM: rpc: make default fiq handler run-time installed
  ARM: make arch_ret_to_user macro optional
  ARM: amba: samsung: use common amba device initializers
  ARM: amba: spear: use common amba device initializers
  ARM: amba: nomadik: use common amba device initializers
  ARM: amba: u300: use common amba device initializers
  ARM: amba: lpc32xx: use common amba device initializers
  ARM: amba: netx: use common amba device initializers
  ARM: amba: bcmring: use common amba device initializers
  ARM: amba: ep93xx: use common amba device initializers
  ARM: amba: omap2: use common amba device initializers
  ARM: amba: integrator: use common amba device initializers
  ARM: amba: realview: get rid of private platform amba_device initializer
  ARM: amba: versatile: get rid of private platform amba_device initializer
  ARM: amba: vexpress: get rid of private platform amba_device initializer
  ARM: amba: provide common initializers for static amba devices
  ARM: amba: make use of -1 IRQs warn
  ARM: amba: u300: get rid of NO_IRQ initializers
  ...

12 years agoMerge tag 'for-3.4' of git://openrisc.net/jonas/linux
Linus Torvalds [Sat, 24 Mar 2012 00:24:25 +0000 (17:24 -0700)]
Merge tag 'for-3.4' of git://openrisc.net/jonas/linux

Pull OpenRISC changes for 3.4 from Jonas Bonn:
 "This series for the OpenRISC architecture consists of mostly trivial
  fixups.  The most interesting bits of the series are:

  * A fix to the timer code whereby the shortest trigger period is set
    to 100 cycles; previously, it was possible to set this to 1 cycle,
    but by the time the register was written, that time had already
    passed and the timer interrupt would not go off until the cycle
    counter had gone a full cycle.

  * Allowing a device tree binary to be passed in to the kernel from
    u-boot.  The OpenRISC architecture has been recently merged into
    upstream u-boot, so this change gets OpenRISC Linux into sync with
    that project."

* tag 'for-3.4' of git://openrisc.net/jonas/linux:
  OpenRISC: Remove memory_start/end prototypes
  openrisc: remove semicolon from KSTK_ defs
  openrisc: sanitize use of orig_gpr11
  openrisc: fix virt_addr_valid
  OpenRISC: Export dump_stack()
  OpenRISC: Select GENERIC_ATOMIC64
  openrisc: Set shortest clock event to 100 ticks
  openrisc: included linux/thread_info.h twice
  OpenRISC: Use set_current_blocked() and block_sigmask()
  OpenRISC: Don't mask signals if we fail to setup signal stack
  OpenRISC: No need to reset handler if SA_ONESHOT
  OpenRISC: Don't reimplement force_sigsegv()
  openrisc: enable passing of flattened device tree pointer
  arch/openrisc/mm/init.c: trivial: use BUG_ON

12 years agoMerge tag 'ia64-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl...
Linus Torvalds [Sat, 24 Mar 2012 00:19:37 +0000 (17:19 -0700)]
Merge tag 'ia64-for-linus' of git://git./linux/kernel/git/aegl/linux

Pull miscellaneous Itanium patches from Tony Luck.

The conflicts in arch/ia64/hp/sim/simserial.c were due to patches to
simserial that had alredy been included (with lots of further cleanups)
in the serial tree.

* tag 'ia64-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux:
  Documentation/kernel-parameters: remove inttest parameter
  [IA64] Fix ISA IRQ trigger model and polarity setting
  [IA64] Fix a couple of warnings for EXPORT_SYMBOL
  [IA64] Check return from device_register() in cx_device_register()
  [IA64] Fix warning from machine_kexec.c
  [IA64] simserial, bail out when request_irq fails
  [IA64] hpsim, initialize chip for assigned irqs
  [IA64] simserial, include some headers
  [IA64] hpsim, fix SAL handling in fw-emu
  [IA64] genirq fixup for SGI/SN
  [IA64] disable interrupts when exiting from ia64_mca_cmc_int_handler()

12 years agoMerge branch 'mmci' into amba
Russell King [Sat, 24 Mar 2012 00:10:36 +0000 (00:10 +0000)]
Merge branch 'mmci' into amba

12 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 24 Mar 2012 00:10:09 +0000 (17:10 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull additional x86 fixes from Peter Anvin:
 - address a long-standing bug related to when a kernel-spawned process
   gets a signal on an i386 kernel compiled without CONFIG_VM86.

 - fix the newly introduced build warning in arch/x86/boot.

 - fix a typo in the i386 system call table which affects building some
   libcs.

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86-32: Fix endless loop when processing signals for kernel tasks
  x86, boot: Correct CFLAGS for hostprogs
  x86-32: Fix typo for mq_getsetattr in syscall table

12 years agoMerge branch 'akpm' (Andrew's patch-bomb)
Linus Torvalds [Fri, 23 Mar 2012 23:59:10 +0000 (16:59 -0700)]
Merge branch 'akpm' (Andrew's patch-bomb)

Merge second batch of patches from Andrew Morton:
 - various misc things
 - core kernel changes to prctl, exit, exec, init, etc.
 - kernel/watchdog.c updates
 - get_maintainer
 - MAINTAINERS
 - the backlight driver queue
 - core bitops code cleanups
 - the led driver queue
 - some core prio_tree work
 - checkpatch udpates
 - largeish crc32 update
 - a new poll() feature for the v4l guys
 - the rtc driver queue
 - fatfs
 - ptrace
 - signals
 - kmod/usermodehelper updates
 - coredump
 - procfs updates

* emailed from Andrew Morton <akpm@linux-foundation.org>: (141 commits)
  seq_file: add seq_set_overflow(), seq_overflow()
  proc-ns: use d_set_d_op() API to set dentry ops in proc_ns_instantiate().
  procfs: speed up /proc/pid/stat, statm
  procfs: add num_to_str() to speed up /proc/stat
  proc: speed up /proc/stat handling
  fs/proc/kcore.c: make get_sparsemem_vmemmap_info() static
  coredump: add VM_NODUMP, MADV_NODUMP, MADV_CLEAR_NODUMP
  coredump: remove VM_ALWAYSDUMP flag
  kmod: make __request_module() killable
  kmod: introduce call_modprobe() helper
  usermodehelper: ____call_usermodehelper() doesn't need do_exit()
  usermodehelper: kill umh_wait, renumber UMH_* constants
  usermodehelper: implement UMH_KILLABLE
  usermodehelper: introduce umh_complete(sub_info)
  usermodehelper: use UMH_WAIT_PROC consistently
  signal: zap_pid_ns_processes: s/SEND_SIG_NOINFO/SEND_SIG_FORCED/
  signal: oom_kill_task: use SEND_SIG_FORCED instead of force_sig()
  signal: cosmetic, s/from_ancestor_ns/force/ in prepare_signal() paths
  signal: give SEND_SIG_FORCED more power to beat SIGNAL_UNKILLABLE
  Hexagon: use set_current_blocked() and block_sigmask()
  ...

12 years agoseq_file: add seq_set_overflow(), seq_overflow()
KAMEZAWA Hiroyuki [Fri, 23 Mar 2012 22:02:55 +0000 (15:02 -0700)]
seq_file: add seq_set_overflow(), seq_overflow()

It is undocumented but a seq_file's overflow state is indicated by
m->count == m->size.  Add seq_set_overflow() and seq_overflow() to
set/check overflow status explicitly.

Based on an idea from Eric Dumazet.

[akpm@linux-foundation.org: tweak code comment]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoproc-ns: use d_set_d_op() API to set dentry ops in proc_ns_instantiate().
Pravin B Shelar [Fri, 23 Mar 2012 22:02:55 +0000 (15:02 -0700)]
proc-ns: use d_set_d_op() API to set dentry ops in proc_ns_instantiate().

The namespace cleanup path leaks a dentry which holds a reference count
on a network namespace.  Keeping that network namespace from being freed
when the last user goes away.  Leaving things like vlan devices in the
leaked network namespace.

If you use ip netns add for much real work this problem becomes apparent
pretty quickly.  It light testing the problem hides because frequently
you simply don't notice the leak.

Use d_set_d_op() so that DCACHE_OP_* flags are set correctly.

This issue exists back to 3.0.

Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>
Reported-by: Justin Pettit <jpettit@nicira.com>
Signed-off-by: Pravin B Shelar <pshelar@nicira.com>
Signed-off-by: Jesse Gross <jesse@nicira.com>
Cc: David Miller <davem@davemloft.net>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoprocfs: speed up /proc/pid/stat, statm
KAMEZAWA Hiroyuki [Fri, 23 Mar 2012 22:02:54 +0000 (15:02 -0700)]
procfs: speed up /proc/pid/stat, statm

Process accounting applications as top, ps visit some files under
/proc/<pid>.  With seq_put_decimal_ull(), we can optimize /proc/<pid>/stat
and /proc/<pid>/statm files.

This patch adds
  - seq_put_decimal_ll() for signed values.
  - allow delimiter == 0.
  - convert seq_printf() to seq_put_decimal_ull/ll in /proc/stat, statm.

Test result on a system with 2000+ procs.

Before patch:
  [kamezawa@bluextal test]$ top -b -n 1 | wc -l
  2223
  [kamezawa@bluextal test]$ time top -b -n 1 > /dev/null

  real    0m0.675s
  user    0m0.044s
  sys     0m0.121s

  [kamezawa@bluextal test]$ time ps -elf > /dev/null

  real    0m0.236s
  user    0m0.056s
  sys     0m0.176s

After patch:
  kamezawa@bluextal ~]$ time top -b -n 1 > /dev/null

  real    0m0.657s
  user    0m0.052s
  sys     0m0.100s

  [kamezawa@bluextal ~]$ time ps -elf > /dev/null

  real    0m0.198s
  user    0m0.050s
  sys     0m0.145s

Considering top, ps tend to scan /proc periodically, this will reduce cpu
consumption by top/ps to some extent.

[akpm@linux-foundation.org: checkpatch fixes]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoprocfs: add num_to_str() to speed up /proc/stat
KAMEZAWA Hiroyuki [Fri, 23 Mar 2012 22:02:54 +0000 (15:02 -0700)]
procfs: add num_to_str() to speed up /proc/stat

== stat_check.py
num = 0
with open("/proc/stat") as f:
        while num < 1000 :
                data = f.read()
                f.seek(0, 0)
                num = num + 1
==

perf shows

    20.39%  stat_check.py  [kernel.kallsyms]    [k] format_decode
    13.41%  stat_check.py  [kernel.kallsyms]    [k] number
    12.61%  stat_check.py  [kernel.kallsyms]    [k] vsnprintf
    10.85%  stat_check.py  [kernel.kallsyms]    [k] memcpy
     4.85%  stat_check.py  [kernel.kallsyms]    [k] radix_tree_lookup
     4.43%  stat_check.py  [kernel.kallsyms]    [k] seq_printf

This patch removes most of calls to vsnprintf() by adding num_to_str()
and seq_print_decimal_ull(), which prints decimal numbers without rich
functions provided by printf().

On my 8cpu box.
== Before patch ==
[root@bluextal test]# time ./stat_check.py

real    0m0.150s
user    0m0.026s
sys     0m0.121s

== After patch ==
[root@bluextal test]# time ./stat_check.py

real    0m0.055s
user    0m0.022s
sys     0m0.030s

[akpm@linux-foundation.org: remove incorrect comment, use less statck in num_to_str(), move comment from .h to .c, simplify seq_put_decimal_ull()]
[andrea@betterlinux.com: avoid breaking the ABI in /proc/stat]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrea Righi <andrea@betterlinux.com>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Turner <pjt@google.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoproc: speed up /proc/stat handling
Eric Dumazet [Fri, 23 Mar 2012 22:02:53 +0000 (15:02 -0700)]
proc: speed up /proc/stat handling

On a typical 16 cpus machine, "cat /proc/stat" gives more than 4096 bytes,
and is slow :

  # strace -T -o /tmp/STRACE cat /proc/stat | wc -c
  5826
  # grep "cpu " /tmp/STRACE
  read(0, "cpu  1949310 19 2144714 12117253"..., 32768) = 5826 <0.001504>

Thats partly because show_stat() must be called twice since initial
buffer size is too small (4096 bytes for less than 32 possible cpus)

Fix this by :

 1) Taking into account nr_irqs in the initial buffer sizing.

 2) Using ksize() to allow better filling of initial buffer.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Glauber Costa <glommer@parallels.com>
Cc: Russell King - ARM Linux <linux@arm.linux.org.uk>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Paul Turner <pjt@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agofs/proc/kcore.c: make get_sparsemem_vmemmap_info() static
Djalal Harouni [Fri, 23 Mar 2012 22:02:52 +0000 (15:02 -0700)]
fs/proc/kcore.c: make get_sparsemem_vmemmap_info() static

get_sparsemem_vmemmap_info() is only used inside fs/proc/kcore.c

Signed-off-by: Djalal Harouni <tixxdz@opendz.org>
Reviewed-by: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocoredump: add VM_NODUMP, MADV_NODUMP, MADV_CLEAR_NODUMP
Jason Baron [Fri, 23 Mar 2012 22:02:51 +0000 (15:02 -0700)]
coredump: add VM_NODUMP, MADV_NODUMP, MADV_CLEAR_NODUMP

Since we no longer need the VM_ALWAYSDUMP flag, let's use the freed bit
for 'VM_NODUMP' flag.  The idea is is to add a new madvise() flag:
MADV_DONTDUMP, which can be set by applications to specifically request
memory regions which should not dump core.

The specific application I have in mind is qemu: we can add a flag there
that wouldn't dump all of guest memory when qemu dumps core.  This flag
might also be useful for security sensitive apps that want to absolutely
make sure that parts of memory are not dumped.  To clear the flag use:
MADV_DODUMP.

[akpm@linux-foundation.org: s/MADV_NODUMP/MADV_DONTDUMP/, s/MADV_CLEAR_NODUMP/MADV_DODUMP/, per Roland]
[akpm@linux-foundation.org: fix up the architectures which broke]
Signed-off-by: Jason Baron <jbaron@redhat.com>
Acked-by: Roland McGrath <roland@hack.frob.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Avi Kivity <avi@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Matt Turner <mattst88@gmail.com>
Cc: "James E.J. Bottomley" <jejb@parisc-linux.org>
Cc: Helge Deller <deller@gmx.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocoredump: remove VM_ALWAYSDUMP flag
Jason Baron [Fri, 23 Mar 2012 22:02:51 +0000 (15:02 -0700)]
coredump: remove VM_ALWAYSDUMP flag

The motivation for this patchset was that I was looking at a way for a
qemu-kvm process, to exclude the guest memory from its core dump, which
can be quite large.  There are already a number of filter flags in
/proc/<pid>/coredump_filter, however, these allow one to specify 'types'
of kernel memory, not specific address ranges (which is needed in this
case).

Since there are no more vma flags available, the first patch eliminates
the need for the 'VM_ALWAYSDUMP' flag.  The flag is used internally by
the kernel to mark vdso and vsyscall pages.  However, it is simple
enough to check if a vma covers a vdso or vsyscall page without the need
for this flag.

The second patch then replaces the 'VM_ALWAYSDUMP' flag with a new
'VM_NODUMP' flag, which can be set by userspace using new madvise flags:
'MADV_DONTDUMP', and unset via 'MADV_DODUMP'.  The core dump filters
continue to work the same as before unless 'MADV_DONTDUMP' is set on the
region.

The qemu code which implements this features is at:

  http://people.redhat.com/~jbaron/qemu-dump/qemu-dump.patch

In my testing the qemu core dump shrunk from 383MB -> 13MB with this
patch.

I also believe that the 'MADV_DONTDUMP' flag might be useful for
security sensitive apps, which might want to select which areas are
dumped.

This patch:

The VM_ALWAYSDUMP flag is currently used by the coredump code to
indicate that a vma is part of a vsyscall or vdso section.  However, we
can determine if a vma is in one these sections by checking it against
the gate_vma and checking for a non-NULL return value from
arch_vma_name().  Thus, freeing a valuable vma bit.

Signed-off-by: Jason Baron <jbaron@redhat.com>
Acked-by: Roland McGrath <roland@hack.frob.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Cc: Avi Kivity <avi@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agokmod: make __request_module() killable
Oleg Nesterov [Fri, 23 Mar 2012 22:02:50 +0000 (15:02 -0700)]
kmod: make __request_module() killable

As Tetsuo Handa pointed out, request_module() can stress the system
while the oom-killed caller sleeps in TASK_UNINTERRUPTIBLE.

The task T uses "almost all" memory, then it does something which
triggers request_module().  Say, it can simply call sys_socket().  This
in turn needs more memory and leads to OOM.  oom-killer correctly
chooses T and kills it, but this can't help because it sleeps in
TASK_UNINTERRUPTIBLE and after that oom-killer becomes "disabled" by the
TIF_MEMDIE task T.

Make __request_module() killable.  The only necessary change is that
call_modprobe() should kmalloc argv and module_name, they can't live in
the stack if we use UMH_KILLABLE.  This memory is freed via
call_usermodehelper_freeinfo()->cleanup.

Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agokmod: introduce call_modprobe() helper
Oleg Nesterov [Fri, 23 Mar 2012 22:02:49 +0000 (15:02 -0700)]
kmod: introduce call_modprobe() helper

No functional changes.  Move the call_usermodehelper code from
__request_module() into the new simple helper, call_modprobe().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agousermodehelper: ____call_usermodehelper() doesn't need do_exit()
Oleg Nesterov [Fri, 23 Mar 2012 22:02:49 +0000 (15:02 -0700)]
usermodehelper: ____call_usermodehelper() doesn't need do_exit()

Minor cleanup.  ____call_usermodehelper() can simply return, no need to
call do_exit() explicitely.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agousermodehelper: kill umh_wait, renumber UMH_* constants
Oleg Nesterov [Fri, 23 Mar 2012 22:02:48 +0000 (15:02 -0700)]
usermodehelper: kill umh_wait, renumber UMH_* constants

No functional changes.  It is not sane to use UMH_KILLABLE with enum
umh_wait, but obviously we do not want another argument in
call_usermodehelper_* helpers.  Kill this enum, use the plain int.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agousermodehelper: implement UMH_KILLABLE
Oleg Nesterov [Fri, 23 Mar 2012 22:02:47 +0000 (15:02 -0700)]
usermodehelper: implement UMH_KILLABLE

Implement UMH_KILLABLE, should be used along with UMH_WAIT_EXEC/PROC.
The caller must ensure that subprocess_info->path/etc can not go away
until call_usermodehelper_freeinfo().

call_usermodehelper_exec(UMH_KILLABLE) does
wait_for_completion_killable.  If it fails, it uses
xchg(&sub_info->complete, NULL) to serialize with umh_complete() which
does the same xhcg() to access sub_info->complete.

If call_usermodehelper_exec wins, it can safely return.  umh_complete()
should get NULL and call call_usermodehelper_freeinfo().

Otherwise we know that umh_complete() was already called, in this case
call_usermodehelper_exec() falls back to wait_for_completion() which
should succeed "very soon".

Note: UMH_NO_WAIT == -1 but it obviously should not be used with
UMH_KILLABLE.  We delay the neccessary cleanup to simplify the back
porting.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agousermodehelper: introduce umh_complete(sub_info)
Oleg Nesterov [Fri, 23 Mar 2012 22:02:47 +0000 (15:02 -0700)]
usermodehelper: introduce umh_complete(sub_info)

Preparation.  Add the new trivial helper, umh_complete().  Currently it
simply does complete(sub_info->complete).

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agousermodehelper: use UMH_WAIT_PROC consistently
Oleg Nesterov [Fri, 23 Mar 2012 22:02:46 +0000 (15:02 -0700)]
usermodehelper: use UMH_WAIT_PROC consistently

A few call_usermodehelper() callers use the hardcoded constant instead of
the proper UMH_WAIT_PROC, fix them.

Reported-by: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Lars Ellenberg <drbd-dev@lists.linbit.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Michal Januszewski <spock@gentoo.org>
Cc: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Kentaro Takeda <takedakn@nttdata.co.jp>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agosignal: zap_pid_ns_processes: s/SEND_SIG_NOINFO/SEND_SIG_FORCED/
Oleg Nesterov [Fri, 23 Mar 2012 22:02:46 +0000 (15:02 -0700)]
signal: zap_pid_ns_processes: s/SEND_SIG_NOINFO/SEND_SIG_FORCED/

Change zap_pid_ns_processes() to use SEND_SIG_FORCED, it looks more
clear compared to SEND_SIG_NOINFO which relies on from_ancestor_ns logic
send_signal().

It is also more efficient if we need to kill a lot of tasks because it
doesn't alloc sigqueue.

While at it, add the __fatal_signal_pending(task) check as a minor
optimization.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Anton Vorontsov <anton.vorontsov@linaro.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agosignal: oom_kill_task: use SEND_SIG_FORCED instead of force_sig()
Oleg Nesterov [Fri, 23 Mar 2012 22:02:45 +0000 (15:02 -0700)]
signal: oom_kill_task: use SEND_SIG_FORCED instead of force_sig()

Change oom_kill_task() to use do_send_sig_info(SEND_SIG_FORCED) instead
of force_sig(SIGKILL).  With the recent changes we do not need force_ to
kill the CLONE_NEWPID tasks.

And this is more correct.  force_sig() can race with the exiting thread
even if oom_kill_task() checks p->mm != NULL, while
do_send_sig_info(group => true) kille the whole process.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Anton Vorontsov <anton.vorontsov@linaro.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agosignal: cosmetic, s/from_ancestor_ns/force/ in prepare_signal() paths
Oleg Nesterov [Fri, 23 Mar 2012 22:02:45 +0000 (15:02 -0700)]
signal: cosmetic, s/from_ancestor_ns/force/ in prepare_signal() paths

Cosmetic, rename the from_ancestor_ns argument in prepare_signal()
paths.  After the previous change it doesn't match the reality.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Anton Vorontsov <anton.vorontsov@linaro.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agosignal: give SEND_SIG_FORCED more power to beat SIGNAL_UNKILLABLE
Oleg Nesterov [Fri, 23 Mar 2012 22:02:44 +0000 (15:02 -0700)]
signal: give SEND_SIG_FORCED more power to beat SIGNAL_UNKILLABLE

force_sig_info() and friends have the special semantics for synchronous
signals, this interface should not be used if the target is not current.
And it needs the fixes, in particular the clearing of SIGNAL_UNKILLABLE
is not exactly right.

However there are callers which have to use force_ exactly because it
clears SIGNAL_UNKILLABLE and thus it can kill the CLONE_NEWPID tasks,
although this is almost always is wrong by various reasons.

With this patch SEND_SIG_FORCED ignores SIGNAL_UNKILLABLE, like we do if
the signal comes from the ancestor namespace.

This makes the naming in prepare_signal() paths insane, fixed by the
next cleanup.

Note: this only affects SIGKILL/SIGSTOP, but this is enough for
force_sig() abusers.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Anton Vorontsov <anton.vorontsov@linaro.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoHexagon: use set_current_blocked() and block_sigmask()
Matt Fleming [Fri, 23 Mar 2012 22:02:43 +0000 (15:02 -0700)]
Hexagon: use set_current_blocked() and block_sigmask()

As described in e6fa16ab9c1e ("signal: sigprocmask() should do
retarget_shared_pending()") the modification of current->blocked is
incorrect as we need to check whether the signal we're about to block is
pending in the shared queue.

Also, use the new helper function introduced in commit 5e6292c0f28f
("signal: add block_sigmask() for adding sigmask to current->blocked")
which centralises the code for updating current->blocked after
successfully delivering a signal and reduces the amount of duplicate
code across architectures.  In the past some architectures got this code
wrong, so using this helper function should stop that from happening
again.

Acked-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Richard Kuo <rkuo@codeaurora.org>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoptrace: remove PTRACE_SEIZE_DEVEL bit
Denys Vlasenko [Fri, 23 Mar 2012 22:02:43 +0000 (15:02 -0700)]
ptrace: remove PTRACE_SEIZE_DEVEL bit

PTRACE_SEIZE code is tested and ready for production use, remove the
code which requires special bit in data argument to make PTRACE_SEIZE
work.

Strace team prepares for a new release of strace, and we would like to
ship the code which uses PTRACE_SEIZE, preferably after this change goes
into released kernel.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoptrace: renumber PTRACE_EVENT_STOP so that future new options and events can match
Denys Vlasenko [Fri, 23 Mar 2012 22:02:42 +0000 (15:02 -0700)]
ptrace: renumber PTRACE_EVENT_STOP so that future new options and events can match

PTRACE_EVENT_foo and PTRACE_O_TRACEfoo used to match.

New PTRACE_EVENT_STOP is the first event which has no corresponding
PTRACE_O_TRACE option.  If we will ever want to add another such option,
its PTRACE_EVENT's value will collide with PTRACE_EVENT_STOP's value.

This patch changes PTRACE_EVENT_STOP value to prevent this.

While at it, added a comment - the one atop PTRACE_EVENT block, saying
"Wait extended result codes for the above trace options", is not true
for PTRACE_EVENT_STOP.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Tejun Heo <tj@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoptrace: make PTRACE_SEIZE set ptrace options specified in 'data' parameter
Denys Vlasenko [Fri, 23 Mar 2012 22:02:42 +0000 (15:02 -0700)]
ptrace: make PTRACE_SEIZE set ptrace options specified in 'data' parameter

This can be used to close a few corner cases in strace where we get
unwanted racy behavior after attach, but before we have a chance to set
options (the notorious post-execve SIGTRAP comes to mind), and removes
the need to track "did we set opts for this task" state in strace
internals.

While we are at it:

Make it possible to extend SEIZE in the future with more functionality
by passing non-zero 'addr' parameter.  To that end, error out if 'addr'
is non-zero.  PTRACE_ATTACH did not (and still does not) have such
check, and users (strace) do pass garbage there...  let's avoid
repeating this mistake with SEIZE.

Set all task->ptrace bits in one operation - before this change, we were
adding PT_SEIZED and PT_PTRACE_CAP with task->ptrace |= BIT ops.  This
was probably ok (not a bug), but let's be on a safer side.

Changes since v2: use (unsigned long) casts instead of (long) ones, move
PTRACE_SEIZE_DEVEL-related code to separate lines of code.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Pedro Alves <palves@redhat.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoptrace: simplify PTRACE_foo constants and PTRACE_SETOPTIONS code
Denys Vlasenko [Fri, 23 Mar 2012 22:02:41 +0000 (15:02 -0700)]
ptrace: simplify PTRACE_foo constants and PTRACE_SETOPTIONS code

Exchange PT_TRACESYSGOOD and PT_PTRACE_CAP bit positions, which makes
PT_option bits contiguous and therefore makes code in
ptrace_setoptions() much simpler.

Every PTRACE_O_TRACEevent is defined to (1 << PTRACE_EVENT_event)
instead of using explicit numeric constants, to ensure we don't mess up
relationship between bit positions and event ids.

PT_EVENT_FLAG_SHIFT was not particularly useful, PT_OPT_FLAG_SHIFT with
value of PT_EVENT_FLAG_SHIFT-1 is easier to use.

PT_TRACE_MASK constant is nuked, the only its use is replaced by
(PTRACE_O_MASK << PT_OPT_FLAG_SHIFT).

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoptrace: don't modify flags on PTRACE_SETOPTIONS failure
Denys Vlasenko [Fri, 23 Mar 2012 22:02:40 +0000 (15:02 -0700)]
ptrace: don't modify flags on PTRACE_SETOPTIONS failure

On ptrace(PTRACE_SETOPTIONS, pid, 0, <opts>), we used to set those
option bits which are known, and then fail with -EINVAL if there are
some unknown bits in <opts>.

This is inconsistent with typical error handling, which does not change
any state if input is invalid.

This patch changes PTRACE_SETOPTIONS behavior so that in this case, we
return -EINVAL and don't change any bits in task->ptrace.

It's very unlikely that there is userspace code in the wild which will
be affected by this change: it should have the form

    ptrace(PTRACE_SETOPTIONS, pid, 0, PTRACE_O_BOGUSOPT)

where PTRACE_O_BOGUSOPT is a constant unknown to the kernel.  But kernel
headers, naturally, don't contain any PTRACE_O_BOGUSOPTs, thus the only
way userspace can use one if it defines one itself.  I can't see why
anyone would do such a thing deliberately.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoptrace: don't send SIGTRAP on exec if SEIZED
Oleg Nesterov [Fri, 23 Mar 2012 22:02:40 +0000 (15:02 -0700)]
ptrace: don't send SIGTRAP on exec if SEIZED

ptrace_event(PTRACE_EVENT_EXEC) sends SIGTRAP if PT_TRACE_EXEC is not
set.  This is because this SIGTRAP predates PTRACE_O_TRACEEXEC option,
we do not need/want this with PT_SEIZED which can set the options during
attach.

Suggested-by: Pedro Alves <palves@redhat.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Chris Evans <scarybeasts@gmail.com>
Cc: Indan Zupancic <indan@nul.nu>
Cc: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoptrace: the killed tracee should not enter the syscall
Oleg Nesterov [Fri, 23 Mar 2012 22:02:39 +0000 (15:02 -0700)]
ptrace: the killed tracee should not enter the syscall

Another old/known problem.  If the tracee is killed after it reports
syscall_entry, it starts the syscall and debugger can't control this.
This confuses the users and this creates the security problems for
ptrace jailers.

Change tracehook_report_syscall_entry() to return non-zero if killed,
this instructs syscall_trace_enter() to abort the syscall.

Reported-by: Chris Evans <scarybeasts@gmail.com>
Tested-by: Indan Zupancic <indan@nul.nu>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Denys Vlasenko <vda.linux@googlemail.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Pedro Alves <palves@redhat.com>
Cc: Jan Kratochvil <jan.kratochvil@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agofat: fix bug in enforcing Long File Name length
Namjae Jeon [Fri, 23 Mar 2012 22:02:39 +0000 (15:02 -0700)]
fat: fix bug in enforcing Long File Name length

Since '*outlen' is initialized to zero, it is currently possible to
create a filename of length (FAT_LFN_LEN + 1) when utf8 is not enabled.
To enforce the FAT_LFN_LEN limit, we must perform one less iteration.

Signed-off-by: Namjae Jeon <linkinjeon@gmail.com>
Signed-off-by: Ravishankar N <cyberax82@gmail.com>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agofat: clean up xlate_to_uni()
Namjae Jeon [Fri, 23 Mar 2012 22:02:38 +0000 (15:02 -0700)]
fat: clean up xlate_to_uni()

xlate_to_uni() is called by vfat_build_slots() with sbi->nls_io as the
final argument.  nls_io can never be null at this point because the
check is already being done in fat_fill_super() wherein the mount fails
if it is null.

Signed-off-by: Namjae Jeon <linkinjeon@gmail.com>
Signed-off-by: Ravishankar N <cyberax82@gmail.com>
Acked-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc: ds1307: generalise ram size and offset
Austin Boyle [Fri, 23 Mar 2012 22:02:38 +0000 (15:02 -0700)]
rtc: ds1307: generalise ram size and offset

Generalise NVRAM to support RAM with other size and offset, such as the
64 bytes of SRAM on the mcp7941x.

[rdunlap@xenotime.net: fix printk format warning]
Signed-off-by: Austin Boyle <Austin.Boyle@aviatnet.com>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Cc: David Anders <danders.dev@gmail.com>
Cc: Alessandro Zummo <alessandro.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc: ds1307: comment and format cleanup
David Anders [Fri, 23 Mar 2012 22:02:37 +0000 (15:02 -0700)]
rtc: ds1307: comment and format cleanup

Do some cleanup of the comment sections as well as correct some
formatting issues reported by checkpatch.pl.

Signed-off-by: David Anders <x0132446@ti.com>
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Cc: Austin Boyle <Austin.Boyle@aviatnet.com>
Cc: Alessandro Zummo <alessandro.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc: ds1307: simplify irq setup code
Wolfram Sang [Fri, 23 Mar 2012 22:02:37 +0000 (15:02 -0700)]
rtc: ds1307: simplify irq setup code

No need to have two seperate if-blocks for setting up the irq.

Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Tested-by: David Anders <danders.dev@gmail.com>
Cc: Austin Boyle <Austin.Boyle@aviatnet.com>
Cc: Alessandro Zummo <alessandro.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc: ds1307: refactor chip_desc table
Wolfram Sang [Fri, 23 Mar 2012 22:02:36 +0000 (15:02 -0700)]
rtc: ds1307: refactor chip_desc table

The chip_desc table is suboptimal.  Currently it requires an entry for
every new chip type, even if it is empty.  This has already been
forgotten for the ds1388.  Refactor the code, so new entries are only
needed, when they chip type really needs a (non-empty) description.
Also make the table visually more appealing.

Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Cc: Austin Boyle <Austin.Boyle@aviatnet.com>
Cc: David Anders <danders.dev@gmail.com>
Cc: Alessandro Zummo <alessandro.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc: driver for DA9052/53 PMIC v1
Ashish Jangam [Fri, 23 Mar 2012 22:02:36 +0000 (15:02 -0700)]
rtc: driver for DA9052/53 PMIC v1

RTC Driver for Dialog Semiconductor DA9052/53 PMICs.

This patch is functionally tested on Samsung SMDKV6410.

[akpm@linux-foundation.org: clean up file header layout, remove unneeded initialisation of local arrays]
Signed-off-by: David Dajun Chen <dchen@diasemi.com>
Signed-off-by: Ashish Jangam <ashish.jangam@kpitcummins.com>
Cc: Paul Gortmaker <p_gortmaker@yahoo.com>
Cc: David Dajun Chen <dchen@diasemi.com>
Cc: Samuel Ortiz <sameo@linux.intel.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc/rtc-max8925.c: fix alarm->enabled mistake in max8925_rtc_read_alarm/max89...
Kevin Liu [Fri, 23 Mar 2012 22:02:36 +0000 (15:02 -0700)]
drivers/rtc/rtc-max8925.c: fix alarm->enabled mistake in max8925_rtc_read_alarm/max8925_rtc_set_alarm

max8925_rtc_read_alarm() should set alrm->enabled based on both
ALARM_IRQ_MASK and ALARM_CTRL setting.  max8925_rtc_set_alarm() should
enable/disable alarm according to ALARM_CTRL reg setting.

Signed-off-by: Kevin Liu <kliu5@marvell.com>
Signed-off-by: Haojian Zhuang <haojian.zhuang@marvell.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc/rtc-max8925.c: fix max8925_rtc_read_alarm() return value error
Kevin Liu [Fri, 23 Mar 2012 22:02:35 +0000 (15:02 -0700)]
drivers/rtc/rtc-max8925.c: fix max8925_rtc_read_alarm() return value error

max8925_rtc_read_alarm should always return 0 with success

Signed-off-by: Kevin Liu <kliu5@marvell.com>
Signed-off-by: Haojian Zhuang <haojian.zhuang@marvell.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc/rtc-pm8xxx.c: make pm8xxx_rtc_pm_ops static
Navin P [Fri, 23 Mar 2012 22:02:34 +0000 (15:02 -0700)]
drivers/rtc/rtc-pm8xxx.c: make pm8xxx_rtc_pm_ops static

Signed-off-by: Navin P <zicrim@gmail.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc: remove IRQF_DISABLED
Yong Zhang [Fri, 23 Mar 2012 22:02:34 +0000 (15:02 -0700)]
drivers/rtc: remove IRQF_DISABLED

Since commit e58aa3d2d0cc ("genirq: run irq handlers with interrupts
disabled") we run all interrupt handlers with interrupts disabled and we
even check and yell when an interrupt handler returns with interrupts
enabled - see commit b738a50a2026 ("genirq: warn when handler enables
interrupts").

So now this flag is a NOOP and can be removed.

Signed-off-by: Yong Zhang <yong.zhang0@gmail.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Acked-by: Wan ZongShun <mcuos.com@gmail.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc/rtc-twl.c: return correct RTC event from ISR
Venu Byravarasu [Fri, 23 Mar 2012 22:02:34 +0000 (15:02 -0700)]
drivers/rtc/rtc-twl.c: return correct RTC event from ISR

Following changes are made as part of this change:

1. As TWL RTC supports periodic interrupt, the correct event should be
   RTC_PF instead of RTC_UF.

2. No need to initialize variable "events" to 0 & then OR it with the
   event values.  Hence fixing it.

Signed-off-by: Venu Byravarasu <vbyravarasu@nvidia.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc/rtc-twl.c: simplify RTC interrupt clearing
Venu Byravarasu [Fri, 23 Mar 2012 22:02:33 +0000 (15:02 -0700)]
drivers/rtc/rtc-twl.c: simplify RTC interrupt clearing

For clearing RTC interrupt, programming ALARM bit only is sufficient, as
all other bits are any way not affected by writing 0 to them.

Hence removed unwanted OR operation.

Signed-off-by: Venu Byravarasu <vbyravarasu@nvidia.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc/rtc-twl.c: enable RTC irrespective of its prior state
Venu Byravarasu [Fri, 23 Mar 2012 22:02:33 +0000 (15:02 -0700)]
drivers/rtc/rtc-twl.c: enable RTC irrespective of its prior state

As part of probe, before enabling RTC, RTC_CTRL register is read to check
if it is already running.  If RTC is used by kernel alone, then this read
is not required.  Even if RTC was enabled already by boot loader, setting
STOP_RTC bit again should not harm.  Hence removed unwanted read
operation.

Signed-off-by: Venu Byravarasu <vbyravarasu@nvidia.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/rtc/rtc-twl.c: optimize IRQ bit access
Venu Byravarasu [Fri, 23 Mar 2012 22:02:32 +0000 (15:02 -0700)]
drivers/rtc/rtc-twl.c: optimize IRQ bit access

As the TWL RTC driver has a cached copy of enabled RTC interrupt bits in
variable rtc_irq_bits, that can be checked before really setting or
masking any of the interrupt bits.

Signed-off-by: Venu Byravarasu <vbyravarasu@nvidia.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMIPS: add RTC support for loongson1B
zhao zhang [Fri, 23 Mar 2012 22:02:32 +0000 (15:02 -0700)]
MIPS: add RTC support for loongson1B

Add RTC support(TOY counter0) for loongson1B SOC

Signed-off-by: zhao zhang <zhzhl555@gmail.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc: convert rtc i2c drivers to module_i2c_driver
Axel Lin [Fri, 23 Mar 2012 22:02:31 +0000 (15:02 -0700)]
rtc: convert rtc i2c drivers to module_i2c_driver

Factor out some boilerplate code for i2c driver registration into
module_i2c_driver.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Cc: Piotr Ziecik <kosmo@semihalf.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Scott Wood <scottwood@freescale.com>
Cc: Srikanth Srinivasan <srikanth.srinivasan@freescale.com>
Cc: Mike Rapoport <mike@compulab.co.il>
Cc: Sergey Lapin <slapin@ossfans.org>
Cc: Roman Fietze <roman.fietze@telemotive.de>
Cc: Herbert Valerio Riedel <hvr@gnu.org>
Cc: Alexander Bigga <ab@mycable.de>
Cc: Dale Farnsworth <dale@farnsworth.org>
Cc: Gregory Hermant <gregory.hermant@calao-systems.com>
Cc: Wolfgang Grandegger <wg@grandegger.com>
Cc: Martyn Welch <martyn.welch@ge.com>
Cc: Byron Bradley <byron.bbradley@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc: convert rtc spi drivers to module_spi_driver
Axel Lin [Fri, 23 Mar 2012 22:02:30 +0000 (15:02 -0700)]
rtc: convert rtc spi drivers to module_spi_driver

Factor out some boilerplate code for spi driver registration into
module_spi_driver.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Cc: Mark Jackson <mpfj@mimc.co.uk>
Cc: Dennis Aberilla <denzzzhome@yahoo.com>
Cc: Nikolaus Voss <n.voss@weinmann.de>
Cc: "Kim B. Heino" <Kim.Heino@bluegiga.com>
Cc: Raphael Assenat <raph@raphnet.net>
Cc: Chris Verges <chrisv@cyberswitching.com>
Cc: Magnus Damm <damm@opensource.se>
Cc: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc/rtc-spear: call platform_set_drvdata() before registering rtc device
Viresh Kumar [Fri, 23 Mar 2012 22:02:30 +0000 (15:02 -0700)]
rtc/rtc-spear: call platform_set_drvdata() before registering rtc device

rtc_device_register() calls rtc-spear routines internally.  These
routines call dev_get_drvdata() to get struct spear_rtc_config.
Currently, platform_set_drvdata is called after rtc device is
registered.  This causes system to crash, as dev_get_drvdata returns
NULL.

For this we need to call platform_set_drvdata() before registering rtc
device.  This requires further cleanup, that leads to removal of
dev_set_drvdata on rtc->dev, which was just not required at all.

Also, we change the parameter to request_irq and pass pointer to config
instead of pointer to rtc struct.

This patch brings all above changes.

Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Cc: Shiraz Hashim <shiraz.hashim@st.com>
Cc: Deepak Sikri <deepak.sikri@st.com>
Acked-by: Rajeev Kumar <rajeev-dlh.kumar@st.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc/spear: fix for RTC_AIE_ON and RTC_AIE_OFF ioctl errors
Shiraz Hashim [Fri, 23 Mar 2012 22:02:29 +0000 (15:02 -0700)]
rtc/spear: fix for RTC_AIE_ON and RTC_AIE_OFF ioctl errors

Define API for '.alarm_irq_enable' to enable and disable alarm irq. This
is required by the framework else RTC_AIE_ON and RTC_AIE_OFF ioctls
return errors.

Signed-off-by: Shiraz Hashim <shiraz.hashim@st.com>
Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Cc: Deepak Sikri <deepak.sikri@st.com>
Acked-by: Rajeev Kumar <rajeev-dlh.kumar@st.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agortc-spear: fix for balancing the enable_irq_wake in Power Mgmt
Deepak Sikri [Fri, 23 Mar 2012 22:02:29 +0000 (15:02 -0700)]
rtc-spear: fix for balancing the enable_irq_wake in Power Mgmt

Handle the fix for unbalanced irq for the cases when enable_irq_wake
fails, and a warning related to same is displayed on the console.  The
workaround is handled at the driver level.

Signed-off-by: Deepak Sikri <deepak.sikri@st.com>
Signed-off-by: Viresh Kumar <viresh.kumar@st.com>
Acked-by: Rajeev Kumar <rajeev-dlh.kumar@st.com>
Cc: Shiraz Hashim <shiraz.hashim@st.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoinit/do_mounts.c: print error code on mount failure
Bernhard Walle [Fri, 23 Mar 2012 22:02:28 +0000 (15:02 -0700)]
init/do_mounts.c: print error code on mount failure

Printing the error code makes it easier to debug the cause of a mount
failure.  For example I had the problem that the root file system could
not be mounted read-writeable because my SD card was write-protected.
Without an error code it looks like the SD card was not detected at all.

Signed-off-by: Bernhard Walle <bernhard@bwalle.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoinit: check printed flag to skip printing message
Diwakar Tundlam [Fri, 23 Mar 2012 22:02:28 +0000 (15:02 -0700)]
init: check printed flag to skip printing message

Otherwise the 'Calibration skipped' message gets printed everytime a CPU
is hotplugged in, cluttering console for systems that frequently hotplug
CPUs.

Signed-off-by: Diwakar Tundlam <dtundlam@nvidia.com>
Cc: Phil Carmody <ext-phil.2.carmody@nokia.com>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Greg KH <greg@kroah.com>
Cc: Sameer Nanda <snanda@chromium.org>
Cc: Peter De Schrijver <pdeschrijver@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoepoll: remove unneeded variable in reverse_path_check()
Dan Carpenter [Fri, 23 Mar 2012 22:02:28 +0000 (15:02 -0700)]
epoll: remove unneeded variable in reverse_path_check()

We never use the length variable.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Jason Baron <jbaron@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoepoll: comment the funky #ifdef
Steven Rostedt [Fri, 23 Mar 2012 22:02:27 +0000 (15:02 -0700)]
epoll: comment the funky #ifdef

Looking for a bug in -rt, I stumbled across this code here from: commit
2dfa4eeab0fc ("epoll keyed wakeups: teach epoll about hints coming with
the wakeup key"), specifically:

  #ifdef CONFIG_DEBUG_LOCK_ALLOC
  static inline void ep_wake_up_nested(wait_queue_head_t *wqueue,
                                      unsigned long events, int subclass)
  {
         unsigned long flags;

         spin_lock_irqsave_nested(&wqueue->lock, flags, subclass);
         wake_up_locked_poll(wqueue, events);
         spin_unlock_irqrestore(&wqueue->lock, flags);
  }
  #else
  static inline void ep_wake_up_nested(wait_queue_head_t *wqueue,
                                      unsigned long events, int subclass)
  {
         wake_up_poll(wqueue, events);
  }
  #endif

You change the function of ep_wake_up_nested() depending on whether
CONFIG_DEBUG_LOCK_ALLOC is set or not.  This looks awfully suspicious,
and there's no comment to explain why.  I initially thought that this
was trying to fool lockdep, and hiding a real bug.

Investigating it, I found the creation of wake_up_nested() (which no
longer exists) but was created for the sole purpose of epoll and its
strange wake ups, as explained in commit 0ccf831cbee9 ("lockdep:
annotate epoll")

Although the commit message says "annotate epoll" the change log is much
better at explaining what is happening than what is in the actual code.
Thus a comment is really necessary here.  And to save the time of other
developers from having to go trudging through the git logs trying to
figure out why this code exists.

I took parts of the change log and placed it into a comment above the
affected code.  This will make the description of what is happening more
visible to new developers that have to look at this code for the first
time.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Davide Libenzi <davidel@xmailserver.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agopoll: add poll_requested_events() and poll_does_not_wait() functions
Hans Verkuil [Fri, 23 Mar 2012 22:02:27 +0000 (15:02 -0700)]
poll: add poll_requested_events() and poll_does_not_wait() functions

In some cases the poll() implementation in a driver has to do different
things depending on the events the caller wants to poll for.  An example
is when a driver needs to start a DMA engine if the caller polls for
POLLIN, but doesn't want to do that if POLLIN is not requested but instead
only POLLOUT or POLLPRI is requested.  This is something that can happen
in the video4linux subsystem among others.

Unfortunately, the current epoll/poll/select implementation doesn't
provide that information reliably.  The poll_table_struct does have it: it
has a key field with the event mask.  But once a poll() call matches one
or more bits of that mask any following poll() calls are passed a NULL
poll_table pointer.

Also, the eventpoll implementation always left the key field at ~0 instead
of using the requested events mask.

This was changed in eventpoll.c so the key field now contains the actual
events that should be polled for as set by the caller.

The solution to the NULL poll_table pointer is to set the qproc field to
NULL in poll_table once poll() matches the events, not the poll_table
pointer itself.  That way drivers can obtain the mask through a new
poll_requested_events inline.

The poll_table_struct can still be NULL since some kernel code calls it
internally (netfs_state_poll() in ./drivers/staging/pohmelfs/netfs.h).  In
that case poll_requested_events() returns ~0 (i.e.  all events).

Very rarely drivers might want to know whether poll_wait will actually
wait.  If another earlier file descriptor in the set already matched the
events the caller wanted to wait for, then the kernel will return from the
select() call without waiting.  This might be useful information in order
to avoid doing expensive work.

A new helper function poll_does_not_wait() is added that drivers can use
to detect this situation.  This is now used in sock_poll_wait() in
include/net/sock.h.  This was the only place in the kernel that needed
this information.

Drivers should no longer access any of the poll_table internals, but use
the poll_requested_events() and poll_does_not_wait() access functions
instead.  In order to enforce that the poll_table fields are now prepended
with an underscore and a comment was added warning against using them
directly.

This required a change in unix_dgram_poll() in unix/af_unix.c which used
the key field to get the requested events.  It's been replaced by a call
to poll_requested_events().

For qproc it was especially important to change its name since the
behavior of that field changes with this patch since this function pointer
can now be NULL when that wasn't possible in the past.

Any driver accessing the qproc or key fields directly will now fail to compile.

Some notes regarding the correctness of this patch: the driver's poll()
function is called with a 'struct poll_table_struct *wait' argument.  This
pointer may or may not be NULL, drivers can never rely on it being one or
the other as that depends on whether or not an earlier file descriptor in
the select()'s fdset matched the requested events.

There are only three things a driver can do with the wait argument:

1) obtain the key field:

events = wait ? wait->key : ~0;

   This will still work although it should be replaced with the new
   poll_requested_events() function (which does exactly the same).
   This will now even work better, since wait is no longer set to NULL
   unnecessarily.

2) use the qproc callback. This could be deadly since qproc can now be
   NULL. Renaming qproc should prevent this from happening. There are no
   kernel drivers that actually access this callback directly, BTW.

3) test whether wait == NULL to determine whether poll would return without
   waiting. This is no longer sufficient as the correct test is now
   wait == NULL || wait->_qproc == NULL.

   However, the worst that can happen here is a slight performance hit in
   the case where wait != NULL and wait->_qproc == NULL. In that case the
   driver will assume that poll_wait() will actually add the fd to the set
   of waiting file descriptors. Of course, poll_wait() will not do that
   since it tests for wait->_qproc. This will not break anything, though.

   There is only one place in the whole kernel where this happens
   (sock_poll_wait() in include/net/sock.h) and that code will be replaced
   by a call to poll_does_not_wait() in the next patch.

   Note that even if wait->_qproc != NULL drivers cannot rely on poll_wait()
   actually waiting. The next file descriptor from the set might match the
   event mask and thus any possible waits will never happen.

Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Reviewed-by: Jonathan Corbet <corbet@lwn.net>
Reviewed-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Davide Libenzi <davidel@xmailserver.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Cc: Mauro Carvalho Chehab <mchehab@infradead.org>
Cc: David Miller <davem@davemloft.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: select an algorithm via Kconfig
Darrick J. Wong [Fri, 23 Mar 2012 22:02:26 +0000 (15:02 -0700)]
crc32: select an algorithm via Kconfig

Allow the kernel builder to choose a crc32* algorithm for the kernel.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Cc: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: add self-test code for crc32c
Darrick J. Wong [Fri, 23 Mar 2012 22:02:26 +0000 (15:02 -0700)]
crc32: add self-test code for crc32c

Add self-test code for crc32c.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Cc: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrypto: crc32c should use library implementation
Darrick J. Wong [Fri, 23 Mar 2012 22:02:25 +0000 (15:02 -0700)]
crypto: crc32c should use library implementation

Since lib/crc32.c now provides crc32c, remove the software implementation
here and call the library function instead.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: bolt on crc32c
Darrick J. Wong [Fri, 23 Mar 2012 22:02:25 +0000 (15:02 -0700)]
crc32: bolt on crc32c

Reuse the existing crc32 code to stamp out a crc32c implementation.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Bob Pearson <rpearson@systemfabricworks.com>
Cc: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: add note about this patchset to crc32.c
Bob Pearson [Fri, 23 Mar 2012 22:02:24 +0000 (15:02 -0700)]
crc32: add note about this patchset to crc32.c

Add a comment at the top of crc32.c

[djwong@us.ibm.com: Minor changelog tweaks]
Signed-off-by: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: optimize loop counter for x86
Bob Pearson [Fri, 23 Mar 2012 22:02:24 +0000 (15:02 -0700)]
crc32: optimize loop counter for x86

Add two changes that improve the performance of x86 systems

1. replace main loop with incrementing counter this change improves
   the performance of the selftest by about 5-6% on Nehalem CPUs.  The
   apparent reason is that the compiler can use the loop index to perform
   an indexed memory access.  This is reported to make the performance of
   PowerPC CPUs to get worse.

2. replace the rem_len loop with incrementing counter this change
   improves the performance of the selftest, which has more than the usual
   number of occurances, by about 1-2% on x86 CPUs.  In actual work loads
   the length is most often a multiple of 4 bytes and this code does not
   get executed as often if at all.  Again this change is reported to make
   the performance of PowerPC get worse.

[djwong@us.ibm.com: Minor changelog tweaks]
Signed-off-by: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: add slice-by-8 algorithm to existing code
Bob Pearson [Fri, 23 Mar 2012 22:02:24 +0000 (15:02 -0700)]
crc32: add slice-by-8 algorithm to existing code

Add slicing-by-8 algorithm to the existing slicing-by-4 algorithm.  This
consists of:

- extend largest BITS size from 32 to 64
- extend tables from tab[4][256] to up to tab[8][256]
- Add code for inner loop.

[djwong@us.ibm.com: Minor changelog tweaks]
Signed-off-by: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: make CRC_*_BITS definition correspond to actual bit counts
Bob Pearson [Fri, 23 Mar 2012 22:02:23 +0000 (15:02 -0700)]
crc32: make CRC_*_BITS definition correspond to actual bit counts

crc32.c provides a choice of one of several algorithms for computing the
LSB and LSB versions of the CRC32 checksum based on the parameters
CRC_LE_BITS and CRC_BE_BITS.

In the original version the values 1, 2, 4 and 8 respectively selected
versions of the alrogithm that computed the crc 1, 2, 4 and 32 bits as a
time.

This patch series adds a new version that computes the CRC 64 bits at a
time.  To make things easier to understand the parameter has been
reinterpreted to actually stand for the number of bits processed in each
step of the algorithm so that the old value 8 has been replaced with the
value 32.

This also allows us to add in a widely used crc algorithm that computes
the crc 8 bits at a time called the Sarwate algorithm.

[djwong@us.ibm.com: Minor changelog tweaks]
Signed-off-by: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: fix mixing of endian-specific types
Bob Pearson [Fri, 23 Mar 2012 22:02:23 +0000 (15:02 -0700)]
crc32: fix mixing of endian-specific types

crc32.c in its original version freely mixed u32, __le32 and __be32 types
which caused warnings from sparse with __CHECK_ENDIAN__.  This patch fixes
these by forcing the types to u32.

[djwong@us.ibm.com: Minor changelog tweaks]
Signed-off-by: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: miscellaneous cleanups
Bob Pearson [Fri, 23 Mar 2012 22:02:22 +0000 (15:02 -0700)]
crc32: miscellaneous cleanups

Misc cleanup of lib/crc32.c and related files.

- remove unnecessary header files.

- straighten out some convoluted ifdef's

- rewrite some references to 2 dimensional arrays as 1 dimensional
  arrays to make them correct.  I.e.  replace tab[i] with tab[0][i].

- a few trivial whitespace changes

- fix a warning in gen_crc32tables.c caused by a mismatch in the type of
  the pointer passed to output table.  Since the table is only used at
  kernel compile time, it is simpler to make the table big enough to hold
  the largest column size used.  One cannot make the column size smaller
  in output_table because it has to be used by both the le and be tables
  and they can have different column sizes.

[djwong@us.ibm.com: Minor changelog tweaks]
Signed-off-by: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocrc32: simplify unit test code
Bob Pearson [Fri, 23 Mar 2012 22:02:22 +0000 (15:02 -0700)]
crc32: simplify unit test code

Replace the unit test provided in crc32.c, which doesn't have a makefile
and doesn't compile with current headers, with a simpler self test
routine that also gives a measure of performance and runs at module init
time.  The self test option can be enabled through a configuration
option CONFIG_CRC32_SELFTEST.

The test stresses the pre and post loops and is thus not very realistic
since actual uses will likely have addresses and lengths that are at
least 4 byte aligned.  However, the main loop is long enough so that the
performance is dominated by that loop.

The expected values for crc32_le and crc32_be were generated with the
original version of crc32.c using CRC_BITS_LE = 8 and CRC_BITS_BE = 8.
These values were then used to check all the values of the BITS
parameters in both the original and new versions.

The performance results show some variability from run to run in spite
of attempts to both warm the cache and reduce the amount of OS noise by
limiting interrutps during the test.  To get comparable results and to
analyse options wrt performance the best time reported over a small
sample of runs has been taken.

[djwong@us.ibm.com: Minor changelog tweaks]
Signed-off-by: Bob Pearson <rpearson@systemfabricworks.com>
Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>