perf trace: Read_tracing_data should die() another day

It better propagate errors, also if we do a simple:

[root@doppio linux-2.6-tip]# perf record -R -a -f sleep 3s ;
perf trace [ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.182 MB perf.data (~7972 samples) ]
Fatal: not an trace data file
[root@doppio linux-2.6-tip]#

That is what is expected, right? I.e. as we didn't specify any
tracepoint event via -e, it should gracefully bail out and not
SEGFAULT.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258821086-11521-3-git-send-email-acme@infradead.org>
[ Fixed the error messages some more ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Old versions of elf.h don't have NT_GNU_BUILD_ID

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258821086-11521-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Fixup kernel_maps__fixup_end end map

We better call this routine after both the kernel and modules
are loaded, because as it was if there weren't modules it would not
be called, resulting in kernel_map->end remaining at zero, so no
map would be found and consequently the kernel symtab wouldn't
get loaded, i.e. no kernel symbols would be resolved.

Also this fixes another case, that is when we _have_ modules,
but the last map would have its ->end address not set before we
loaded its symbols, which would never happen because ->end was
not set.

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258821086-11521-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf_events: Fix default watermark calculation

This patch fixes the default watermark value for the sampling
buffer. With the existing calculation (watermark =
max(PAGE_SIZE, max_size / 2)), no notification was ever received
when the buffer was exactly 1 page. This was because you would
never cross the threshold (there is no partial samples).

In certain configuration, there was no possibilty detecting the
problem because there was not enough space left to store the
LOST record.In fact, there may be a more generic problem here.
The kernel should ensure that there is alaways enough space to
store one LOST record.

This patch sets the default watermark to half the buffer size.
With such limit, we are guaranteed to get a notification even
with a single page buffer assuming no sample is bigger than a
page.

Signed-off-by: Stephane Eranian <eranian@gmail.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212509.344964101@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
LKML-Reference: <1256302576-6169-1-git-send-email-eranian@gmail.com>

perf: Fix locking for PERF_FORMAT_GROUP

We should hold event->child_mutex when iterating the inherited
counters, we should hold ctx->mutex when iterating siblings.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212509.251030114@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Fix event scaling for inherited counters

Properly account the full hierarchy of counters for both the
count (we already did so) and the scale times (new).

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212509.153379276@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Fix time locking

Most sites updating ctx->time and event times do so under
ctx->lock, make sure they all do.

This was made possible by removing the __perf_event_read() call
from __perf_event_sync_stat(), which already had this lock
taken.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212509.102316434@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Simplify __perf_event_read

cpuctx is always active, task context is always active for
current

the previous condition verifies that if its a task context its
for current, hence we can assume ctx->is_active.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212509.000272254@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Simplify __perf_event_sync_stat

Removes constraints from __perf_event_read() by leaving it with
a single callsite; this callsite had ctx->lock held, the other
one does not.

Removes some superfluous code from __perf_event_sync_stat().

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.918544317@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Optimize __perf_event_read()

Both callers actually have IRQs disabled, no need doing so
again.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.863685796@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Optimize perf_event_task_sched_out

Remove an update_context_time() call from the
perf_event_task_sched_out() path and into the branch its needed.

The call was both superfluous, because __perf_event_sched_out()
already does it, and wrong, because it was done without holding
ctx->lock.

Place it in perf_event_sync_stat(), which is the only place it
is needed and which does already hold ctx->lock.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.779516394@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Fix PERF_FORMAT_GROUP scale info

As Corey reported, the total_enabled and total_running times
could occasionally be 0, even though there were events counted.

It turns out this is because we record the times before reading
the counter while the latter updates the times.

This patch corrects that.

While looking at this code I found that there is a lot of
locking iffyness around, the following patches correct most of
that.

Reported-by: Corey Ashford <cjashfor@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.685559857@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Optimize perf_event_mmap_ctx()

Remove a rcu_read_{,un}lock() pair and a few conditionals.

We can remove the rcu_read_lock() by increasing the scope of one
in the calling function.

We can do away with the system_state check if the machine still
boots after this patch (seems to be the case).

We can do away with the list_empty() check because the bare
list_for_each_entry_rcu() reduces to that now that we've removed
everything else.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.606459548@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Optimize perf_event_comm_ctx()

Remove a rcu_read_{,un}lock() pair and a few conditionals.

We can remove the rcu_read_lock() by increasing the scope of one
in the calling function.

We can do away with the system_state check if the machine still
boots after this patch (seems to be the case).

We can do away with the list_empty() check because the bare
list_for_each_entry_rcu() reduces to that now that we've removed
everything else.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.527608793@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Optimize perf_event_task_ctx()

Remove a rcu_read_{,un}lock() pair and a few conditionals.

We can remove the rcu_read_lock() by increasing the scope of one
in the calling function.

We can do away with the system_state check if the machine still
boots after this patch (seems to be the case).

We can do away with the list_empty() check because the bare
list_for_each_entry_rcu() reduces to that now that we've removed
everything else.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.452227115@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Optimize perf_swevent_ctx_event()

Remove a rcu_read_{,un}lock() pair and a few conditionals.

We can remove the rcu_read_lock() by increasing the scope of one
in the calling function.

We can do away with the system_state check if the machine still
boots after this patch (seems to be the case).

We can do away with the list_empty() check because the bare
list_for_each_entry_rcu() reduces to that now that we've removed
everything else.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.378188589@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Optimize some swcounter attr.sample_period==1 paths

Avoid the rather expensive perf_swevent_set_period() if we know
we have to sample every single event anyway.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.299508332@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Allow for custom overflow handlers

in-kernel perf users might wish to have custom actions on the
sample interrupt.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <20091120212508.222339539@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Change the kernel DSO name if it comes from kallsyms

So that the user have a clearer indication about the source of
the symbols, as we only state buildid mismatches in verbose
mode, because 'perf top' would overwrite such warning anyway.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258757489-5978-6-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Check vmlinux buildid

E.g.:

[root@doppio linux-2.6-tip]# perf top -v --vmlinux
../build/tip/vmlinux > /dev/null build_id in vmlinux is
e96699725a47413a50c231864a8e7a8ced40a31b while expected is
18e7cc53db62a7d35e9d6f6c9ddc23017d38ee9a, ignoring it

I.e. perf top was told to use a vmlinux file that is not the one
currently running on the machine, it ignores it and falls back
to using /proc/kallsyms.

This solves many, at first, mysterious results when people have
a stale vmlinux file while keeping the default of trying to use
the vmlinux file in the current directory in things like 'perf
annotate' where the DWARF info is required and thus we can't use
just /proc/kallsyms.

Modules buildids are already being checked as of the previous
changeset in this series, because we are using the default
dso__load routine, that will look at a series of places looking
for the best file with a matching buildid, starting in the
-debuginfo directories.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258757489-5978-5-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Do lazy symtab loading for the kernel & modules too

Just like we do with the other DSOs. This also simplifies the
kernel_maps setup process, now all that the tools need to do is
to call kernel_maps__init and the maps for the modules and
kernel will be created, then, later, when
kernel_maps__find_symbol() is used, it will also call
maps__find_symbol that already checks if the symtab was loaded,
loading it if needed.

Now if one does 'perf top --hide_kernel_symbols' we won't pay
the price of loading the (many) symbols in /proc/kallsyms or
vmlinux.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258757489-5978-4-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Introduce dso__build_id_equal

Will be used in more places.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258757489-5978-3-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Filename__read_build_id should look at .notes section too

In the kernel we have more than one notes section, so the linker
script combines all and puts them into a ".notes" combined
section. So we need to look at both sections and also traverse
them looking at multiple GElf_Nhdr entries till we find the one
we want, with the build_id.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258757489-5978-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Remove unrelated actions from dso__load_kernel_sym

It should just load kernel symbols, not load the list of
modules. There are more stuff to move to other routines, but
lets do it in several steps.

End goal is to be able to defer symbol table loading till we
find a hit for that map address range. So that the kernel &
modules are handled just like all the other DSOs in the system.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258757489-5978-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge branch 'tracing/hw-breakpoints' into perf/core

Conflicts:
arch/x86/kernel/kprobes.c
kernel/trace/Makefile

Merge reason: hw-breakpoints perf integration is looking
              good in testing and in reviews, plus conflicts
              are mounting up - so merge & resolve.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge branch 'perf/urgent' into perf/core

Conflicts:
tools/perf/util/symbol.c

Merge reason: this fix will get merged in .33, not .32, plus
              resolve the conflict.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf: Add 'perf kmem' tool

This tool is mostly a perf version of kmemtrace-user.

The following information is provided by this tool:

 - the total amount of memory allocated and fragmentation per
   call-site

 - the total amount of memory allocated and fragmentation per
   allocation

 - total memory allocated and fragmentation in the collected
   dataset - ...

Sample output:

 # ./perf kmem record
 ^C
 # ./perf kmem --stat caller --stat alloc -l 10

 ------------------------------------------------------------------------------
 Callsite          | Total_alloc/Per |  Total_req/Per  |  Hit   | Fragmentation
 ------------------------------------------------------------------------------
 0xc052f37a        |   790528/4096   |   790528/4096   |    193 |    0.000%
 0xc0541d70        |   524288/4096   |   524288/4096   |    128 |    0.000%
 0xc051cc68        |   481600/200    |   481600/200    |   2408 |    0.000%
 0xc0572623        |   297444/676    |   297440/676    |    440 |    0.001%
 0xc05399f1        |    73476/164    |    73472/164    |    448 |    0.005%
 0xc05243bf        |    51456/256    |    51456/256    |    201 |    0.000%
 0xc0730d0e        |    31844/497    |    31808/497    |     64 |    0.113%
 0xc0734c4e        |    17152/256    |    17152/256    |     67 |    0.000%
 0xc0541a6d        |    16384/128    |    16384/128    |    128 |    0.000%
 0xc059c217        |    13120/40     |    13120/40     |    328 |    0.000%
 0xc0501ee6        |    11264/88     |    11264/88     |    128 |    0.000%
 0xc04daef0        |     7504/682    |     7128/648    |     11 |    5.011%
 0xc04e14a3        |     4216/191    |     4216/191    |     22 |    0.000%
 0xc05041ca        |     3524/44     |     3520/44     |     80 |    0.114%
 0xc0734fa3        |     2104/701    |     1620/540    |      3 |   23.004%
 0xc05ec9f1        |     2024/289    |     2016/288    |      7 |    0.395%
 0xc06a1999        |     1792/256    |     1792/256    |      7 |    0.000%
 0xc0463b9a        |     1584/144    |     1584/144    |     11 |    0.000%
 0xc0541eb0        |     1024/16     |     1024/16     |     64 |    0.000%
 0xc06a19ac        |      896/128    |      896/128    |      7 |    0.000%
 0xc05721c0        |      772/12     |      768/12     |     64 |    0.518%
 0xc054d1e6        |      288/57     |      280/56     |      5 |    2.778%
 0xc04b562e        |      157/31     |      154/30     |      5 |    1.911%
 0xc04b536f        |       80/16     |       80/16     |      5 |    0.000%
 0xc05855a0        |       64/64     |       36/36     |      1 |   43.750%
 ------------------------------------------------------------------------------

 ------------------------------------------------------------------------------
 Alloc Ptr         | Total_alloc/Per |  Total_req/Per  |  Hit   | Fragmentation
 ------------------------------------------------------------------------------
 0xda884000        |  1052672/4096   |  1052672/4096   |    257 |    0.000%
 0xda886000        |   262144/4096   |   262144/4096   |     64 |    0.000%
 0xf60c7c00        |    16512/128    |    16512/128    |    129 |    0.000%
 0xf59a4118        |    13120/40     |    13120/40     |    328 |    0.000%
 0xdfd4b2c0        |    11264/88     |    11264/88     |    128 |    0.000%
 0xf5274600        |     7680/256    |     7680/256    |     30 |    0.000%
 0xe8395000        |     5948/594    |     5464/546    |     10 |    8.137%
 0xe59c3c00        |     5748/479    |     5712/476    |     12 |    0.626%
 0xf4cd1a80        |     3524/44     |     3520/44     |     80 |    0.114%
 0xe5bd1600        |     2892/482    |     2856/476    |      6 |    1.245%
 ...               | ...             | ...             | ...    | ...
 ------------------------------------------------------------------------------

SUMMARY
=======
Total bytes requested: 2333626
Total bytes allocated: 2353712
Total bytes wasted on internal fragmentation: 20086
Internal fragmentation: 0.853375%

TODO:
- show sym+offset in 'callsite' column
- show cross node allocation stats
- collect more useful stats?
- ...

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Eduard - Gabriel Munteanu <eduard.munteanu@linux360.ro>
Cc: linux-mm@kvack.org <linux-mm@kvack.org>
LKML-Reference: <4B064AF5.9060208@cn.fujitsu.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: Instruction decoder test should generate build warning

Since some instructions are not decoded correctly by older
versions of objdump, it may cause false positive error in insn
decoder posttest.

This changes build error of insn decoder test to build warning.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: systemtap <systemtap@sources.redhat.com>
Cc: DLE <dle-develop@lists.sourceforge.net>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Randy Dunlap <rdunlap@xenotime.net>
Cc: Jim Keniston <jkenisto@us.ibm.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
LKML-Reference: <20091116230631.5250.41579.stgit@harusame>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Don't die() in mmap_dispatch_perf_file

Propagate the error, that, interestingly, are already handled by
all callers :-)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258649757-17554-3-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Eliminate some more die() uses in library functions

This time in perf_header__adds_write, propagating the do_write
error returns.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258649757-17554-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: perf_header__read() shouldn't die()

And also don't call the constructor in it, this way it adheres
to the model the other methods follow.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258649757-17554-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Capture the running kernel buildid too

[root@doppio linux-2.6-tip]# perf record -a -f sleep 3s ; perf
buildid-list | grep vmlinux
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.171 MB perf.data (~7489
samples) ] 18e7cc53db62a7d35e9d6f6c9ddc23017d38ee9a vmlinux
[root@doppio linux-2.6-tip]#

Several refactorings were needed so that we can have symmetry
between dsos__load_modules() and dsos__load_kernel(), i.e. those
functions will respectively create and add to the dsos list the
loaded modules and kernel, with its buildids, but not load its
symbols. That is something the subcomands that need will have to
call dso__load_kernel_sym(), just like we do with modules with
dsos__load_module_sym()/dso__load_module_sym().

Next csets will actually use this info to stop producing bogus
results using mismatched vmlinux and .ko files.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258582853-8579-4-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Record the build_ids of kernel modules too

[root@doppio linux-2.6-tip]# perf record -a sleep 2s;perf
buildid-list|tail [ perf record: Woken up 1 times to write data
] [ perf record: Captured and wrote 0.162 MB perf.data (~7078
samples) ] 881588fa57b3c1696bc91e5e804a11304f093535 [cfg80211]
4d47ce1da9d16bad00c962c072451b7c681e82df [snd_page_alloc]
5146377e89a7caac617f9782f1a02e46263d3a31 [rfkill]
2153b937bff0d345fea83b63a2e1d3138569f83d [i915]
4e6fb1bb97362e3ee4d306988b9ad6912d5fb9ae [drm_kms_helper]
f56ef2bf853e3a798f0d8d51f797622e5dc4420e [drm]
b0d157a3b5c4e017329ffc07c64623cd6ad65e95 [i2c_algo_bit]
8125374b905ef9fa8b65d98e166b008ad952f198 [i2c_core]
fc875c6e5a90e7b915e9d445d0efc859e1b2678c [video]
4b43c5006589f977e9762fdfc7ac1a92b72fca52 [output]
[root@doppio linux-2.6-tip]#

elfutils libdwfl/linux-kernel-modules.c was used as reference,
as suggested by Roland McGrath.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258582853-8579-3-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Kill struct build_id_list and die() another day

No need for this struct and its allocations, we can just use the
->build_id member we already have in struct dso, then ask for it
to be read, and later traverse the dsos list, writing the
buildid table to the perf.data file.

As a bonus, one more die() function got killed.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258582853-8579-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Fix comparision of build_ids

When we read the build_id from the DSO name to then index into
/usr/lib/debug/.buildid/DSO_BUILD_ID[0:2]/DSO_BUILD_ID[2:], we
were jumping directly to the comparision with the buildid we
already have in dso->build_id (that came from the perf.data
build_id section, collected at perf record time)
unconditionally, even if we didn't had recorded it, and
furthermore, comparing a formatted buildid with a rawbuildid, yikes.

Fix it by deleting the dso__read_build_id() function, that was
really misdesigned anyway, and do the necessary checks and
correct comparison of raw buildids.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258582853-8579-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf top: Don't allocate the source parsing members upfront

Defer to parse_source() time allocating it.

Now we use about this much memory:

 1724 root      20   0 42104  10m  940 S  0.0  0.4   0:00.23 perf

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258490282-1821-3-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf top: Allocate space only for the number of counters used

Reducing memory consumption on a typical desktop machine:

From:

32710 root      20   0  172m 142m 1056 S  0.0  4.7   0:00.37
perf

To:

  420 root      20   0 47528  16m 1056 R  0.3  0.5   0:00.24 perf

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258490282-1821-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf top: Introduce helper function to access symbol from sym_entry

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258490282-1821-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf top: Suppress DSO column if only one is present

E.g.

[root@doppio ~]# perf top -U
---------------------------------------------------------------------------
   PerfTop:     482 irqs/sec  kernel:100.0% [1000Hz cycles],  (all, 2 CPUs)
---------------------------------------------------------------------------
DSO: vmlinux
             samples  pcnt function
             _______ _____ _________________________

              471.00 47.9% read_hpet
               57.00  5.8% acpi_os_read_port
               30.00  3.1% hpet_next_event
               30.00  3.1% find_busiest_group
               22.00  2.2% schedule
               18.00  1.8% sched_clock_local
               14.00  1.4% _spin_lock_irqsave
               14.00  1.4% native_read_tsc
               13.00  1.3% trace_hardirqs_off
                9.00  0.9% fget_light
                9.00  0.9% ioread8
                8.00  0.8% do_sys_poll

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258479655-28662-3-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf top: Auto adjust symbol and dso widths

We pre-calculate the symbol name length, then after we sort the
entries to print, calculate the biggest one and use that for the
symbol name width justification, then use the
dso->long_name->len to justificate the DSO name, deciding whether
using the short or long name depending on how much space we have
on the terminal.

IOW give as much info to the user as the terminal width allows.

Suggested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258479655-28662-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Add a long_name_len member to struct dso

Using a two bytes hole we already had and since we also need to
calculate this strlen for fetching the buildids. We'll use it in
'perf top' to auto-adjust the output based on the terminal
width.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frédéric Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <1258479655-28662-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Add ia64 support for tools/perf/

Compiler on ia64 rejects the "-m64" option.
Add arch specific pieces to perf.h

Signed-off-by: Tony Luck <tony.luck@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
LKML-Reference: <4b02d7f43514327a@agluck-desktop.sc.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf annotate: Allocate history size correctly

Symbol offset history table size does not get updated properly
when it is being resized. This leads to garbage results in
perf annotate.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Support static build

This makes it possible to build perf statically, by
performing:

  make LDFLAGS=-static

Since static libraries are only searched in the order they are
specified, move library list from LDFLAGS to EXTLIBS, so that
they are put at the end of linker command line.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <20091029152002.GA5406@redhat.com>
[ v2: resolved conflicts ]
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge branch 'perf/core' into perf/probes

Resolved merge conflict in tools/perf/Makefile

Merge reason: we want to queue up a dependent patch.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Don't die() in do_write()

Propagate the errors instead, the users are the ones to decide
what to do if a library call fails.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258427892-16312-4-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Don't die() in perf_header__new()

Propagate the errors instead, the users are the ones to decide
what to do if a library call fails.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258427892-16312-3-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Don't die() in perf_header_attr__add_id()

Propagate the errors instead, the users are the ones to decide
what to do if a library call fails.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258427892-16312-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Don't die() in perf_header__add_attr()

Propagate the errors instead, the users are the ones to decide
what to do if a library call fails.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258427892-16312-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf buildid-list: Always show the DSO name

Porcelain can ignore it, humans can make more sense of it.

Suggested-by: Frederic Weisbecker <fweisbec@gmail.com>
Suggested-by: Ingo Molnar <mingo@elte.hu>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258415125-15019-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf top: Introduce --hide_{user,kernel}_symbols

Default continues to be showing all symbols. 'K' and 'U' can be
used to toggle showing kernel and user symbols.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258415125-15019-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf top: Use all the lines in the screen

By querying the current number of rows, if the user specifies
the number of entries, use that instead. If the user uses the
'e' command to change the number of lines 0 will mean do it
automatically, any other number disables the auto resizing.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258407027-384-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Don't die in perf_header_attr__new()

We really should propagate such kinds of errors so that users of
these library functions decide what to do in such cases instead
of exiting in random places like now.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258407027-384-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: Show symbol name if insn decoder test failed

Show symbol name if insn decoder test find a difference.
This will help us to find out where the issue is.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: systemtap <systemtap@sources.redhat.com>
Cc: DLE <dle-develop@lists.sourceforge.net>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Randy Dunlap <rdunlap@xenotime.net>
Cc: Jim Keniston <jkenisto@us.ibm.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
LKML-Reference: <20091116230624.5250.49813.stgit@harusame>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: Add verbose option to insn decoder test

Add verbose option to insn decoder test. This dumps decoded
instruction when building kernel with V=1.

Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
Cc: systemtap <systemtap@sources.redhat.com>
Cc: DLE <dle-develop@lists.sourceforge.net>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Randy Dunlap <rdunlap@xenotime.net>
Cc: Jim Keniston <jkenisto@us.ibm.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
LKML-Reference: <20091116230618.5250.18762.stgit@harusame>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf buildid-list: New plumbing command

With this we can list the buildids in a perf.data file so that
we can pipe them to other, distro specific tools that from the
buildids can figure out separate packages (foo-debuginfo) where
we can find the matching symtabs so that perf report can do its
job.

E.g:

[acme@doppio linux-2.6-tip]$ perf buildid-list | head -5
8e08b117e5458ad3f85da16d42d0fc5cd21c5869
520c2387a587cc5acfcf881e27dba1caaeab4b1f
ec8dd400904ddfcac8b1c343263a790f977159dc
7caedbca5a6d8ab39a7fe44bd28c07d3e14a3f3f
379bb828fd08859dbea73279f04abefabc95a6a3
[acme@doppio linux-2.6-tip]$ perf buildid-list -v | head -5
8e08b117e5458ad3f85da16d42d0fc5cd21c5869 /sbin/init
520c2387a587cc5acfcf881e27dba1caaeab4b1f /lib64/ld-2.10.1.so
ec8dd400904ddfcac8b1c343263a790f977159dc /lib64/libc-2.10.1.so
7caedbca5a6d8ab39a7fe44bd28c07d3e14a3f3f /sbin/udevd
379bb828fd08859dbea73279f04abefabc95a6a3 /lib64/libdl-2.10.1.so
[acme@doppio linux-2.6-tip]$

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258396365-29217-5-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Introduce dsos__fprintf_buildid

To print the buildids in the list of dsos. Will be used by 'perf
buildid-list'

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258396365-29217-4-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Generalize perf_header__adds_read()

Renaming it to perf_header__process_sections() and passing a
callback to handle each feature.

The next changesets will introduce 'perf buildid-list' that will
handle just the HEADER_BUILD_ID table, ignoring all the other
features.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Acked-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258396365-29217-3-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Debug.h needs to include event.h for event_t

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258396365-29217-2-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Pass the offset to perf_header__read_build_ids()

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258396365-29217-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf symbols: Call the symbol filter in dso__synthesize_plt_symbols()

We need to pass the symbol to the filter so that, for instance,
'perf top' can do filtering and also set the private area it
manages, setting the ->map pointer, etc.

I found this while running 'perf top' on a machine where hits
happened on PLT symbols, where ->map wasn't being set up and
segfaults thus happened.

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
LKML-Reference: <1258386491-20278-1-git-send-email-acme@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf_event: Optimize perf_output_lock()

The purpose of perf_output_{un,}lock() is to:

 1) avoid publishing incomplete data
    [ possible when publishing a head that is ahead of an entry
      that is still being written ]

 2) guarantee fwd progress
    [ a simple refcount on pending writers doesn't need to drop to
      0, making it so would end up implementing something like forced
      quiecent states of RCU ]

To satisfy the above without undue complexity it serializes
between CPUs, this means that a pending writer can only be the
same cpu in a nested context, and since (under normal operation)
a cpu always makes progress we're good -- if the head is only
published when the bottom  most writer completes.

Now we don't need to disable IRQs in order to serialize between
CPUs, disabling preemption ought to be sufficient, esp since we
already deal with nesting due to NMIs.

This avoids potentially expensive (and needless) local IRQ
disable/enable ops.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1258373161.26714.254.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf stat: Do not print ratio when task-clock event is not counted

The ratio between the number of events and the time elapsed makes
sense only if task-clock event is counted. Otherwise it will be
simply a (confusing)

#      0.000 M/sec

This patch outputs the ratio only if task-clock event is counted.
Some test examples of before and after:

Before:

 [lucas@skywalker linux.trees.git]$ sudo perf stat -e branch-misses -a -- sleep 1

 Performance counter stats for 'sleep 1':

1367818  branch-misses            #      0.000 M/sec

    1.001494325  seconds time elapsed

After (without task-clock):

 [lucas@skywalker perf]$ sudo ./perf stat -e branch-misses -a -- sleep 1

 Performance counter stats for 'sleep 1':

1135044  branch-misses

    1.001370775  seconds time elapsed

After (with task-clock):

 [lucas@skywalker perf]$ sudo ./perf stat -e branch-misses -e task-clock -a -- sleep 1

 Performance counter stats for 'sleep 1':

1070111  branch-misses            #      0.534 M/sec
    2002.730893  task-clock-msecs         #      1.999 CPUs

    1.001640292  seconds time elapsed

Signed-off-by: Lucas De Marchi <lucas.de.marchi@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
LKML-Reference: <20091115140507.GB21561@skywalker.lan>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf tools: Add new perf_atoll() function to parse string representing size in bytes

This patch modifies util/string.[ch] to add new function:
perf_atoll() to parse string representing size in bytes.

This function parses (\d+)(b|B|kb|KB|mb|MB|gb|GB) (e.g. "256MB")
and returns its numeric value. (e.g. 268435456)

Signed-off-by: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <1258285013-4759-1-git-send-email-mitake@dcl.info.waseda.ac.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge branches 'perf/powerpc' and 'perf/bench' into perf/core

Merge reason: Both 'perf bench' and the pending PowerPC changes
              are now ready for the next merge window.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge commit 'v2.6.32-rc7' into perf/core

Merge reason: pick up perf fixlets

Signed-off-by: Ingo Molnar <mingo@elte.hu>

hw-breakpoints, x86: Fix modular KVM build

This build error:

arch/x86/kvm/x86.c:3655: error: implicit declaration of function 'hw_breakpoint_restore'

Happens because in the CONFIG_KVM=m case there's no 'CONFIG_KVM' define
in the kernel - it's CONFIG_KVM_MODULE in that case.

Make the prototype available unconditionally.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Prasad <prasad@linux.vnet.ibm.com>
LKML-Reference: <1258114575-32655-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

perf_event: Fix invalid type in ioctl definition

u64 is invalid in userspace headers, including ioctl
definitions; use __u64 instead

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Cc: <stable@kernel.org>
LKML-Reference: <20091113214733.7cd76be9@infradead.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

hw-breakpoints: Fix build on !perf architectures

the arch/alpha build fails with:

 In file included from tip/kernel/exit.c:52:
 tip/include/linux/hw_breakpoint.h: In function 'hw_breakpoint_addr':
 tip/include/linux/hw_breakpoint.h:21: error: 'struct perf_event' has no member named 'attr'
 [...]

Move these helper inlines inside the CONFIG_HAVE_HW_BREAKPOINT ifdef.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Prasad <prasad@linux.vnet.ibm.com>
LKML-Reference: <1258114575-32655-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

hw-breakpoints: Provide an off-case for counter_arch_bp()

If an arch doesn't support the hw breakpoints, counter_arch_bp()
has no off case to cover the missing breakpoint info structure
from the perf event. The result is a build error in non-x86
configs.

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Prasad <prasad@linux.vnet.ibm.com>
LKML-Reference: <1258114575-32655-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Prasad <prasad@linux.vnet.ibm.com>

tracing: Rename 'lockdep' event subsystem into 'lock'

Lockdep events subsystem gathers various locking related events
such as a request, release, contention or acquisition of a lock.

The name of this event subsystem is a bit of a misnomer since
these events are not quite related to lockdep but more generally
to locking, ie: these events are not reporting lock dependencies
or possible deadlock scenario but pure locking events.

Hence this rename.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
LKML-Reference: <1258103194-843-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Linux 2.6.32-rc7

Merge branch 'omap-fixes-for-linus' of git://git./linux/kernel/git/tmlind/linux-omap-2.6

* 'omap-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tmlind/linux-omap-2.6:
  omap3: Decrease cpufreq transition latency
  omap3: update Pandora defconfig
  omap3: 3430sdp: Enable Linux Regulator framework
  omap3: beagle: Fix USB host port power control
  omap3: pandora: Fix keypad keymap
  omap1: Amstrad Delta defconfig fixes
  omap: Fix omapfb/lcdc on OMAP1510 broken when PM set
  omap: Use resource_size
  omap: Fix race condition in omap dma driver

__generic_block_fiemap(): fix for files bigger than 4GB

Because of an integer overflow on start_blk, various kind of wrong results
would be returned by the generic_block_fiemap() handler, such as no
extents when there is a 4GB+ hole at the beginning of the file, or wrong
fe_logical when an extent starts after the first 4GB.

Signed-off-by: Mike Hommey <mh@glandium.org>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Eric Sandeen <sandeen@sgi.com>
Cc: Josef Bacik <jbacik@redhat.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pps: events reporting fix up

PPS events must be recorded according to PPS's mode settings.

If a process asks for (i.e.) capture-assert events only, when the PPS
client calls the pps_event() function to save the current PPS event, we
should verify the event type and then discard unwanted ones.

Also, without this patch userland processes waiting for a specific PPS
event (assert or clear but not both) may be awakened at wrong time.

Signed-off-by: Rodolfo Giometti <giometti@linux.it>
Tested-by: William S. Brasher <billb958@door.net>
Tested-by: Reg Clemens <clemens@dwf.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pps: locking scheme fix up for PPS_GETPARAMS

Userland programs may read/write PPS parameters at same time and these
operations may corrupt PPS data.

Signed-off-by: Rodolfo Giometti <giometti@linux.it>
Tested-by: Reg Clemens <clemens@dwf.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/video/msm: update to new kernel

TASK_INTERRUPTIBLE and friends are now only available after including
<linux/sched.h>, so include it when needed.

bus_id is no longer available/necessary, so remove that.

Android pmem driver is not available in mainline, so remove its hooks
from drivers/video.

Signed-off-by: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gpiolib: fix device_create() result check

In case of failure, device_create() returns not NULL but the error code.
The current code checks for non-NULL though which causes kernel oops in
sysfs_create_group() when device_create() fails.  Check for error using
IS_ERR() and propagate the error value using PTR_ERR() instead of fixed
-ENODEV code returned now...

Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com>
Cc: David Brownell <david-b@pacbell.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rtc: v3020: fix v3020_mmio_read_bit()

v3020_mmio_read_bit() always returns 0 when left_shift > 7.

v3020_mmio_read_bit()'s return type is (unsigned char).  The code returns
a value masked by (1 << left_shift) that is casted to the return type.  If
left_shift is larger than 7, the cast will always result in a 0 return
value.  The problem was discovered with left_shift = 16, and the included
patch corrects the problem.

The bug was introduced in the last (Apr 3 2009) commit of the file, kernel
versions 2.6.30 and later.

Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Paul Gortmaker <p_gortmaker@yahoo.com>
Cc: Raphael Assenat <raph@8d.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rtc-vr41xx: fix do_div() warning

drivers/rtc/rtc-vr41xx.c: In function 'vr41xx_rtc_irq_set_freq':
drivers/rtc/rtc-vr41xx.c:217: warning: comparison of distinct pointer types lacks a cast
drivers/rtc/rtc-vr41xx.c:217: warning: right shift count >= width of type
drivers/rtc/rtc-vr41xx.c:217: warning: passing argument 1 of '__div64_32' from incompatible pointer type
include/asm-generic/div64.h:35: note: expected 'uint64_t *' but argument is of type 'long unsigned int *'

Signed-off-by: Yoichi Yuasa <yuasa@linux-mips.org>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Cc: Paul Gortmaker <p_gortmaker@yahoo.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

exec: setup_arg_pages() fails to return errors

In setup_arg_pages we work hard to assign a value to ret, but on exit we
always return 0.

Also remove a now duplicated exit path and branch to out_unlock instead.

Signed-off-by: Anton Blanchard <anton@samba.org>
Acked-by: Serge Hallyn <serue@us.ibm.com>
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>

fb: put framebuffer drivers back under the "frame buffer devices" menu

The config FB_PRE_INIT_FB entry in drivers/video/Kconfig pushes all entries
below it out of the menuconfig selection. Fix this.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MAINTAINERS: correct SECURITY SUBSYSTEM git entry

Use git.kernel.org not www.kernel.org

Signed-off-by: Joe Perches <joe@perches.com>
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MAINTAINERS: correct NETFILTER git entry format

Signed-off-by: Joe Perches <joe@perches.com>
Cc: Patrick McHardy <kaber@trash.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MAINTAINERS: correct 9P FILE SYSTEM git entry

Signed-off-by: Joe Perches <joe@perches.com>
Cc: Latchesar Ionkov <lucho@ionkov.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mmci-omap: free irq resource

Free IRQ on remove.

Signed-off-by: Ladislav Michl <ladis@linux-mips.org>
Acked-by: Tony Lindgren <tony@atomide.com>
Cc: <linux-mmc@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

serial: add support for the Lava Quattro PCI quad-port 16550A card

This seems to be a different model (with a different PCI ID) than the
"Quatro" card that is also in the list.

Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

serial: fix printk format specifiers for struct uart_port::iobase

struct uart_port::iobase is unsigned long, so use %lx as printk format
specifier.

Signed-off-by: Lennert Buytenhek <buytenh@wantstofly.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rtc: pcf50633: consider alrm->enable in pcf50633_rtc_set_alarm

According to Documentation/rtc.txt, RTC_WKALM_SET sets the alarm time and
enables/disables the alarm.  We implement RTC_WKALM_SET through
pcf50633_rtc_set_alarm.  The enabling/disabling part was missing.

Signed-off-by: Werner Almesberger <werner@openmoko.org>
Reported-by: Michael 'Mickey' Lauer <mickey@openmoko.org>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Cc: Paul Gortmaker <p_gortmaker@yahoo.com>
Cc: Balaji Rao <balajirrao@openmoko.org>
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>

rtc: pcf50633: fix month off-by-one error

The PCF50633 stores a month value of 1-12, but the kernel wants 0-11.

Signed-off-by: Rask Ingemann Lambertsen <rask@sygehus.dk>
Signed-off-by: Paul Fertser <fercerpav@gmail.com>
Cc: Paul Gortmaker <p_gortmaker@yahoo.com>
Cc: Balaji Rao <balajirrao@openmoko.org>
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>

atmel_lcdfb: new alternate pixel clock formula

at91sam9g45 non ES lots have an alternate pixel clock calculation formula.
Introduce this one with condition on the cpu_is_xxxxx() macros.

Newer 9g45 SOC will not have good pixel clock calculation without this
fix.

Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Cc: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs: add missing compat_ptr handling for FS_IOC_RESVSP ioctl

For FS_IOC_RESVSP and FS_IOC_RESVSP64 compat_sys_ioctl() uses its
arg argument as a pointer to userspace. However it is missing a
a call to compat_ptr() which will do a proper pointer conversion.

This was introduced with 3e63cbb1 "fs: Add new pre-allocation ioctls
to vfs for compatibility with legacy xfs ioctls".

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Ankit Jain <me@ankitjain.org>
Acked-by: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Arnd Bergmann <arndbergmann@googlemail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: <stable@kernel.org> [2.6.31.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pidns: fix a leak in /proc dentries and inodes with pid namespaces.

Daniel Lezcano reported a leak in 'struct pid' and 'struct pid_namespace'
that is discussed in:

http://lkml.org/lkml/2009/10/2/159.

To summarize the thread, when container-init is terminated, it sets the
PF_EXITING flag, zaps other processes in the container and waits to reap
them.  As a part of reaping, the container-init should flush any /proc
dentries associated with the processes.  But because the container-init is
itself exiting and the following PF_EXITING check, the dentries are not
flushed, resulting in leak in /proc inodes and dentries.

This fix reverts the commit 7766755a2f249e7e0 ("Fix /proc dcache deadlock
in do_exit") which introduced the check for PF_EXITING.  At the time of
the commit, shrink_dcache_parent() flushed dentries from other filesystems
also and could have caused a deadlock which the commit fixed.  But as
pointed out by Eric Biederman, after commit 0feae5c47aabdde59,
shrink_dcache_parent() no longer affects other filesystems.  So reverting
the commit is now safe.

As pointed out by Jan Kara, the leak is not as critical since the
unclaimed space will be reclaimed under memory pressure or by:

echo 3 > /proc/sys/vm/drop_caches

But since this check is no longer required, its best to remove it.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Reported-by: Daniel Lezcano <dlezcano@fr.ibm.com>
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Jan Kara <jack@ucw.cz>
Cc: Andrea Arcangeli <andrea@cpushare.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

alpha: move THREAD_SIZE definition outside #ifndef ASSEMBLY

arch/alpha/kernel/vmlinux.lds.S uses it:

arch/alpha/kernel/vmlinux.lds:241: undefined symbol `THREAD_SIZE' referenced in expression

Seems to have been caused by

commit 9d93f00580243cc059510d9d6ac4d2f5e97e5b83
Author:     Geoffrey Thomas <geofft@ksplice.com>
AuthorDate: Thu Sep 24 10:36:26 2009 -0400
Commit:     Linus Torvalds <torvalds@linux-foundation.org>
CommitDate: Thu Sep 24 17:16:22 2009 -0700

    alpha: Clean up linker script using new linker script macros.

    Note that .data.page_aligned and .data.cacheline_aligned are now after
    _data; it was probably a bug that they were before it.

    Also, some explicit ALIGN(8)'s between various initcall sections were
    removed; this should be harmless as the implicit alignment of
    initcall_t was already 8.

Cc: Geoffrey Thomas <geofft@ksplice.com>
Cc: Tim Abbott <tabbott@ksplice.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Richard Henderson <rth@twiddle.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

alpha: fix F_SETOWN_EX and F_GETLK64 conflict

Fix a bug in

    commit ba0a6c9f6fceed11c6a99e8326f0477fe383e6b5
    Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
    AuthorDate: Wed Sep 23 15:57:03 2009 -0700
    Commit:     Linus Torvalds <torvalds@linux-foundation.org>
    CommitDate: Thu Sep 24 07:21:01 2009 -0700

        fcntl: add F_[SG]ETOWN_EX

In asm-generic/fcntl.h, F_SETOWN_EX and F_GETLK64 both have value 12, and
F_GETOWN_EX and F_SETLK64 both have value 13.

Reported-by: "Joseph S. Myers" <joseph@codesourcery.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Andreas Schwab <schwab@redhat.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: fix wrong pointer initialization at page migration when memcg is disabled.

Lee Schermerhorn reported that he saw bad pointer dereference in
mem_cgroup_end_migration() when he disabled memcg by boot option.

memcg's page migration logic works as

mem_cgroup_prepare_migration(page, &ptr);
do page migration
mem_cgroup_end_migration(page, ptr);

Now, ptr is not initialized in prepare_migration when memcg is disabled
by boot option. This causes panic in end_migration. This patch fixes it.

Reported-by: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

savagefb: fix blanking mode on CRT display

Fix wrong bit mask for blanking register.  Due to the error a CRT monitor
blanks off due to wrong frequency (out of range) instead of PM signal
(vertical and horizontal frequencies cut off).

Just compare the mask with bits set in the switch(blank) clause below the
changed line.

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fb: remove fb_save_state() and fb_restore_state operations

Remove fb_save_state() and fb_restore_state operations from frame buffer layer.
They are used only in two drivers:
1. savagefb  - and cause bug #11248
2. uvesafb

Usage of these operations is misunderstood in both drivers so kill these
operations, fix the bug #11248 and avoid confusion in the future.

Tested on Savage 3D/MV card and the patch fixes the bug #11248.

The frame buffer layer uses these funtions during switch between graphics
and text mode of the console, but these drivers saves state before
switching of the frame buffer (in the fb_open) and after releasing it (in
the fb_release).  This defeats the purpose of these operations.

Addresses http://bugzilla.kernel.org/show_bug.cgi?id=11248

Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Reported-by: Jochen Hein <jochen@jochen.org>
Tested-by: Jochen Hein <jochen@jochen.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Michal Januszewski <spock@gentoo.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

page allocator: Do not allow interrupts to use ALLOC_HARDER

Commit 341ce06f69abfafa31b9468410a13dbd60e2b237 ("page allocator:
calculate the alloc_flags for allocation only once") altered watermark
logic slightly by allowing rt_tasks that are handling an interrupt to set
ALLOC_HARDER.  This patch brings the watermark logic more in line with
2.6.30.

This change results in a reduction of the number high-order GFP_ATOMIC
allocation failures reported.  See
http://www.gossamer-threads.com/lists/linux/kernel/1144153

[rientjes@google.com: Spotted the problem]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Reviewed-by: Rik van Riel <riel@redhat.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

page allocator: always wake kswapd when restarting an allocation attempt after direct reclaim failed

If a direct reclaim makes no forward progress, it considers whether it
should go OOM or not.  Whether OOM is triggered or not, it may retry the
allocation afterwards.  In times past, this would always wake kswapd as
well but currently, kswapd is not woken up after direct reclaim fails.
For order-0 allocations, this makes little difference but if there is a
heavy mix of higher-order allocations that direct reclaim is failing for,
it might mean that kswapd is not rewoken for higher orders as much as it
did previously.

This patch wakes up kswapd when an allocation is being retried after a
direct reclaim failure.  It would be expected that kswapd is already
awake, but this has the effect of telling kswapd to reclaim at the higher
order as well.

Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>