8 years agoperf tools: Robustify dynamic sample content fetch
Frederic Weisbecker [Sat, 21 May 2011 18:08:15 +0000]
perf tools: Robustify dynamic sample content fetch

Ensure the size of the dynamic fields such as callchains
or raw events don't overlap the whole event boundaries.

This prevents from dereferencing junk if the given size of
the callchain goes too eager.

Reported-by: Linus Torvalds <torvalds@linux-foundation.org>
Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>

8 years agoperf tools: Pre-check sample size before parsing
Frederic Weisbecker [Sat, 21 May 2011 17:33:04 +0000]
perf tools: Pre-check sample size before parsing

Check that the total size of the sample fields having a fixed
size do not exceed the one of the whole event. This robustifies
the sample parsing.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>

8 years agoperf tools: Move evlist sample helpers to evlist area
Frederic Weisbecker [Sat, 21 May 2011 15:49:00 +0000]
perf tools: Move evlist sample helpers to evlist area

These APIs should belong to evlist.c as they may not be
exclusively tied to the headers.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com

8 years agoperf tools: Remove junk code in mmap size handling
Frederic Weisbecker [Sat, 21 May 2011 15:07:24 +0000]
perf tools: Remove junk code in mmap size handling

size is overriden later and used only then. Those
lines are only junk, probably a leftover.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>

8 years agoperf tools: Check we are able to read the event size on mmap
Frederic Weisbecker [Sat, 21 May 2011 15:07:24 +0000]
perf tools: Check we are able to read the event size on mmap

Check we have enough mmaped space to read the current event
size from its headers, otherwise we may dereference some
hell there.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Stephane Eranian <eranian@google.com>

8 years agocore_kernel_data(): Fix architectures that do not define _sdata
Ingo Molnar [Thu, 19 May 2011 23:22:14 +0000]
core_kernel_data(): Fix architectures that do not define _sdata

Some architectures such as Alpha do not define _sdata but _data:

  kernel/built-in.o: In function `core_kernel_data':
  kernel/extable.c:77: undefined reference to `_sdata'

So expand the scope of the data range to the text addresses too,
this might be more correct anyway because this way we can
cover readonly variables as well.

Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-i878c8a0e0g0ep4v7i6vxnhz@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agoMerge branch 'tip/perf/core' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Ingo Molnar [Thu, 19 May 2011 17:48:03 +0000]
Merge branch 'tip/perf/core' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into perf/core

8 years agoMerge branch 'tip/perf/core-3' of git://git.kernel.org/pub/scm/linux/kernel/git/roste...
Ingo Molnar [Thu, 19 May 2011 17:25:55 +0000]
Merge branch 'tip/perf/core-3' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into perf/core

8 years agoperf stat: Add more cache-miss percentage printouts
Ingo Molnar [Thu, 19 May 2011 12:01:42 +0000]
perf stat: Add more cache-miss percentage printouts

Print out the cache-miss percentage as well if the cache refs were
collected, for all the generic cache event types.

Before:

   11,103,723,230 dTLB-loads                #  622.471 M/sec                    ( +-  0.30% )
       87,065,337 dTLB-load-misses          #    4.881 M/sec                    ( +-  0.90% )

After:

   11,353,713,242 dTLB-loads                #  626.020 M/sec                    ( +-  0.35% )
      113,393,472 dTLB-load-misses          #    1.00% of all dTLB cache hits   ( +-  0.49% )

Also ASCII color highlight too high percentages, them when it's executed on the console.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-lkhwxsevdbd9a8nymx0vxc3y@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agoperf stat: Add -d -d and -d -d -d options to show more CPU events
Ingo Molnar [Thu, 19 May 2011 11:30:56 +0000]
perf stat: Add -d -d and -d -d -d options to show more CPU events

Print even more detailed statistics if requested via perf stat -d:

       -d:          detailed events, L1 and LLC data cache
    -d -d:     more detailed events, dTLB and iTLB events
 -d -d -d:     very detailed events, adding prefetch events

Full output looks like this now:

 Performance counter stats for '/home/mingo/hackbench 10' (5 runs):

       1703.674707 task-clock                #    8.709 CPUs utilized            ( +-  4.19% )
            49,068 context-switches          #    0.029 M/sec                    ( +- 16.66% )
             8,303 CPU-migrations            #    0.005 M/sec                    ( +- 24.90% )
            17,397 page-faults               #    0.010 M/sec                    ( +-  0.46% )
     2,345,389,239 cycles                    #    1.377 GHz                      ( +-  4.61% ) [55.90%]
     1,884,503,527 stalled-cycles-frontend   #   80.35% frontend cycles idle     ( +-  5.67% ) [50.39%]
       743,919,737 stalled-cycles-backend    #   31.72% backend  cycles idle     ( +-  8.75% ) [49.91%]
     1,314,416,379 instructions              #    0.56  insns per cycle
                                             #    1.43  stalled cycles per insn  ( +-  2.53% ) [60.87%]
       272,592,567 branches                  #  160.003 M/sec                    ( +-  1.74% ) [56.56%]
         3,794,846 branch-misses             #    1.39% of all branches          ( +-  6.59% ) [58.50%]
       449,982,778 L1-dcache-loads           #  264.125 M/sec                    ( +-  2.47% ) [49.88%]
        22,404,961 L1-dcache-load-misses     #    4.98% of all L1-dcache hits    ( +-  6.08% ) [55.05%]
         6,204,750 LLC-loads                 #    3.642 M/sec                    ( +-  8.91% ) [43.75%]
         1,837,411 LLC-load-misses           #    1.078 M/sec                    ( +-  7.27% ) [12.07%]
       411,440,421 L1-icache-loads           #  241.502 M/sec                    ( +-  5.60% ) [36.52%]
        27,556,832 L1-icache-load-misses     #   16.175 M/sec                    ( +-  7.46% ) [46.72%]
       464,067,627 dTLB-loads                #  272.392 M/sec                    ( +-  4.46% ) [54.17%]
        10,765,648 dTLB-load-misses          #    6.319 M/sec                    ( +-  3.18% ) [48.68%]
     1,273,080,386 iTLB-loads                #  747.256 M/sec                    ( +-  3.38% ) [47.53%]
           117,481 iTLB-load-misses          #    0.069 M/sec                    ( +- 14.99% ) [47.01%]
         4,590,653 L1-dcache-prefetches      #    2.695 M/sec                    ( +-  4.49% ) [46.19%]
         1,712,660 L1-dcache-prefetch-misses #    1.005 M/sec                    ( +-  3.75% ) [44.82%]

        0.195622057  seconds time elapsed  ( +-  6.84% )

Also clean up the attribute construction code to be appending, and factor
it out into add_default_attributes().

Tweak the coverage percentage printout a bit, so that it's easier to view it
alongside the +- sttddev colum.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-to3kgu04449s64062val8b62@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agoftrace/kbuild: Add recordmcount files to force full build
Michal Marek [Tue, 17 May 2011 13:36:46 +0000]
ftrace/kbuild: Add recordmcount files to force full build

Modifications to recordmcount must be performed on all object
files to stay consistent with what the kernel code may expect.
Add the recordmcount files to the main dependencies to make sure
any change to them causes a full recompile.

Signed-off-by: Michal Marek <mmarek@suse.cz>
Link: http://lkml.kernel.org/r/20110517133646.GP13293@sepie.suse.cz
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Add self-tests for multiple function trace users
Steven Rostedt [Fri, 6 May 2011 04:08:51 +0000]
ftrace: Add self-tests for multiple function trace users

Add some basic sanity tests for multiple users of the function
tracer at startup.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Modify ftrace_set_filter/notrace to take ops
Steven Rostedt [Fri, 6 May 2011 02:54:01 +0000]
ftrace: Modify ftrace_set_filter/notrace to take ops

Since users of the function tracer can now pick and choose which
functions they want to trace agnostically from other users of the
function tracer, we need to pass the ops struct to the ftrace_set_filter()
functions.

The functions ftrace_set_global_filter() and ftrace_set_global_notrace()
is added to keep the old filter functions which are used to modify
the generic function tracers.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Allow dynamically allocated function tracers
Steven Rostedt [Fri, 6 May 2011 01:14:55 +0000]
ftrace: Allow dynamically allocated function tracers

Now that functions may be selected individually, it only makes sense
that we should allow dynamically allocated trace structures to
be traced. This will allow perf to allocate a ftrace_ops structure
at runtime and use it to pick and choose which functions that
structure will trace.

Note, a dynamically allocated ftrace_ops will always be called
indirectly instead of being called directly from the mcount in
entry.S. This is because there's no safe way to prevent mcount
from being preempted before calling the function, unless we
modify every entry.S to do so (not likely). Thus, dynamically allocated
functions will now be called by the ftrace_ops_list_func() that
loops through the ops that are allocated if there are more than
one op allocated at a time. This loop is protected with a
preempt_disable.

To determine if an ftrace_ops structure is allocated or not, a new
util function was added to the kernel/extable.c called
core_kernel_data(), which returns 1 if the address is between
_sdata and _edata.

Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Implement separate user function filtering
Steven Rostedt [Wed, 4 May 2011 13:27:52 +0000]
ftrace: Implement separate user function filtering

ftrace_ops that are registered to trace functions can now be
agnostic to each other in respect to what functions they trace.
Each ops has their own hash of the functions they want to trace
and a hash to what they do not want to trace. A empty hash for
the functions they want to trace denotes all functions should
be traced that are not in the notrace hash.

Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Free hash with call_rcu_sched()
Steven Rostedt [Thu, 5 May 2011 22:03:47 +0000]
ftrace: Free hash with call_rcu_sched()

When a hash is modified and might be in use, we need to perform
a schedule RCU operation on it, as the hashes will soon be used
directly in the function tracer callback.

Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Have global_ops store the functions that are to be traced
Steven Rostedt [Wed, 4 May 2011 02:49:52 +0000]
ftrace: Have global_ops store the functions that are to be traced

This is a step towards each ops structure defining its own set
of functions to trace. As the current code with pid's and such
are specific to the global_ops, it is restructured to be used
with the global ops.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Add ops parameter to ftrace_startup/shutdown functions
Steven Rostedt [Wed, 4 May 2011 01:55:54 +0000]
ftrace: Add ops parameter to ftrace_startup/shutdown functions

In order to allow different ops to enable different functions,
the ftrace_startup() and ftrace_shutdown() functions need the
ops parameter passed to them.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Add enabled_functions file
Steven Rostedt [Tue, 3 May 2011 18:39:21 +0000]
ftrace: Add enabled_functions file

Add the enabled_functions file that is used to show all the
functions that have been enabled for tracing as well as their
ref counts. This helps seeing if any function has been registered
and what functions are being traced.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Use counters to enable functions to trace
Steven Rostedt [Tue, 3 May 2011 17:25:24 +0000]
ftrace: Use counters to enable functions to trace

Every function has its own record that stores the instruction
pointer and flags for the function to be traced. There are only
two flags: enabled and free. The enabled flag states that tracing
for the function has been enabled (actively traced), and the free
flag states that the record no longer points to a function and can
be used by new functions (loaded modules).

These flags are now moved to the MSB of the flags (actually just
the top 32bits). The rest of the bits (30 bits) are now used as
a ref counter. Everytime a tracer register functions to trace,
those functions will have its counter incremented.

When tracing is enabled, to determine if a function should be traced,
the counter is examined, and if it is non-zero it is set to trace.

When a ftrace_ops is registered to trace functions, its hashes
are examined. If the ftrace_ops filter_hash count is zero, then
all functions are set to be traced, otherwise only the functions
in the hash are to be traced. The exception to this is if a function
is also in the ftrace_ops notrace_hash. Then that function's counter
is not incremented for this ftrace_ops.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Separate hash allocation and assignment
Steven Rostedt [Mon, 2 May 2011 21:34:47 +0000]
ftrace: Separate hash allocation and assignment

When filtering, allocate a hash to insert the function records.
After the filtering is complete, assign it to the ftrace_ops structure.

This allows the ftrace_ops structure to have a much smaller array of
hash buckets instead of wasting a lot of memory.

A read only empty_hash is created to be the minimum size that any ftrace_ops
can point to.

When a new hash is created, it has the following steps:

o Allocate a default hash.
o Walk the function records assigning the filtered records to the hash
o Allocate a new hash with the appropriate size buckets
o Move the entries from the default hash to the new hash.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Create a global_ops to hold the filter and notrace hashes
Steven Rostedt [Mon, 2 May 2011 16:29:25 +0000]
ftrace: Create a global_ops to hold the filter and notrace hashes

Combine the filter and notrace hashes to be accessed by a single entity,
the global_ops. The global_ops is a ftrace_ops structure that is passed
to different functions that can read or modify the filtering of the
function tracer.

The ftrace_ops structure was modified to hold a filter and notrace
hashes so that later patches may allow each ftrace_ops to have its own
set of rules to what functions may be filtered.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Use hash instead for FTRACE_FL_FILTER
Steven Rostedt [Sat, 30 Apr 2011 00:59:51 +0000]
ftrace: Use hash instead for FTRACE_FL_FILTER

When multiple users are allowed to have their own set of functions
to trace, having the FTRACE_FL_FILTER flag will not be enough to
handle the accounting of those users. Each user will need their own
set of functions.

Replace the FTRACE_FL_FILTER with a filter_hash instead. This is
temporary until the rest of the function filtering accounting
gets in.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Replace FTRACE_FL_NOTRACE flag with a hash of ignored functions
Steven Rostedt [Fri, 29 Apr 2011 19:12:32 +0000]
ftrace: Replace FTRACE_FL_NOTRACE flag with a hash of ignored functions

To prepare for the accounting system that will allow multiple users of
the function tracer, having the FTRACE_FL_NOTRACE as a flag in the
dyn_trace record does not make sense.

All ftrace_ops will soon have a hash of functions they should trace
and not trace. By making a global hash of functions not to trace makes
this easier for the transition.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoperf bench, x86: Add alternatives-asm.h wrapper
Ingo Molnar [Wed, 18 May 2011 19:00:44 +0000]
perf bench, x86: Add alternatives-asm.h wrapper

perf bench needs this to build the kernel's memcpy routine:

In file included from bench/mem-memcpy-x86-64-asm.S:2:0:
bench/../../../arch/x86/lib/memcpy_64.S:7:33: fatal error: asm/alternative-asm.h: No such file or directory

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/n/tip-c5d41xibgullk8h2280q4gv0@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agoMerge branch 'x86/mem' into perf/core
Ingo Molnar [Wed, 18 May 2011 18:59:27 +0000]
Merge branch 'x86/mem' into perf/core

Merge reason: memcpy_64.S changes an assumption perf bench has, so merge this
              here so we can fix it.

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

8 years agoMerge branch 'tip/perf/core-2' of git://git.kernel.org/pub/scm/linux/kernel/git/roste...
Ingo Molnar [Wed, 18 May 2011 17:46:10 +0000]
Merge branch 'tip/perf/core-2' of git://git./linux/kernel/git/rostedt/linux-2.6-trace into perf/core

8 years agox86, 64-bit: Fix copy_[to/from]_user() checks for the userspace address limit
Jiri Olsa [Thu, 12 May 2011 14:30:30 +0000]
x86, 64-bit: Fix copy_[to/from]_user() checks for the userspace address limit

As reported in BZ #30352:

  https://bugzilla.kernel.org/show_bug.cgi?id=30352

there's a kernel bug related to reading the last allowed page on x86_64.

The _copy_to_user() and _copy_from_user() functions use the following
check for address limit:

  if (buf + size >= limit)
fail();

while it should be more permissive:

  if (buf + size > limit)
fail();

That's because the size represents the number of bytes being
read/write from/to buf address AND including the buf address.
So the copy function will actually never touch the limit
address even if "buf + size == limit".

Following program fails to use the last page as buffer
due to the wrong limit check:

 #include <sys/mman.h>
 #include <sys/socket.h>
 #include <assert.h>

 #define PAGE_SIZE       (4096)
 #define LAST_PAGE       ((void*)(0x7fffffffe000))

 int main()
 {
        int fds[2], err;
        void * ptr = mmap(LAST_PAGE, PAGE_SIZE, PROT_READ | PROT_WRITE,
                          MAP_ANONYMOUS | MAP_PRIVATE | MAP_FIXED, -1, 0);
        assert(ptr == LAST_PAGE);
        err = socketpair(AF_LOCAL, SOCK_STREAM, 0, fds);
        assert(err == 0);
        err = send(fds[0], ptr, PAGE_SIZE, 0);
        perror("send");
        assert(err == PAGE_SIZE);
        err = recv(fds[1], ptr, PAGE_SIZE, MSG_WAITALL);
        perror("recv");
        assert(err == PAGE_SIZE);
        return 0;
 }

The other place checking the addr limit is the access_ok() function,
which is working properly. There's just a misleading comment
for the __range_not_ok() macro - which this patch fixes as well.

The last page of the user-space address range is a guard page and
Brian Gerst observed that the guard page itself due to an erratum on K8 cpus
(#121 Sequential Execution Across Non-Canonical Boundary Causes Processor
Hang).

However, the test code is using the last valid page before the guard page.
The bug is that the last byte before the guard page can't be read
because of the off-by-one error. The guard page is left in place.

This bug would normally not show up because the last page is
part of the process stack and never accessed via syscalls.

Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Brian Gerst <brgerst@gmail.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: <stable@kernel.org>
Link: http://lkml.kernel.org/r/1305210630-7136-1-git-send-email-jolsa@redhat.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agox86, mem: memset_64.S: Optimize memset by enhanced REP MOVSB/STOSB
Fenghua Yu [Tue, 17 May 2011 22:29:18 +0000]
x86, mem: memset_64.S: Optimize memset by enhanced REP MOVSB/STOSB

Support memset() with enhanced rep stosb. On processors supporting enhanced
REP MOVSB/STOSB, the alternative memset_c_e function using enhanced rep stosb
overrides the fast string alternative memset_c and the original function.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-10-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, mem: memmove_64.S: Optimize memmove by enhanced REP MOVSB/STOSB
Fenghua Yu [Tue, 17 May 2011 22:29:17 +0000]
x86, mem: memmove_64.S: Optimize memmove by enhanced REP MOVSB/STOSB

Support memmove() by enhanced rep movsb. On processors supporting enhanced
REP MOVSB/STOSB, the alternative memmove() function using enhanced rep movsb
overrides the original function.

The patch doesn't change the backward memmove case to use enhanced rep
movsb.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-9-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, mem: memcpy_64.S: Optimize memcpy by enhanced REP MOVSB/STOSB
Fenghua Yu [Tue, 17 May 2011 22:29:16 +0000]
x86, mem: memcpy_64.S: Optimize memcpy by enhanced REP MOVSB/STOSB

Support memcpy() with enhanced rep movsb. On processors supporting enhanced
rep movsb, the alternative memcpy() function using enhanced rep movsb overrides the original function and the fast string
function.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-8-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, mem: copy_user_64.S: Support copy_to/from_user by enhanced REP MOVSB/STOSB
Fenghua Yu [Tue, 17 May 2011 22:29:15 +0000]
x86, mem: copy_user_64.S: Support copy_to/from_user by enhanced REP MOVSB/STOSB

Support copy_to_user/copy_from_user() by enhanced REP MOVSB/STOSB.
On processors supporting enhanced REP MOVSB/STOSB, the alternative
copy_user_enhanced_fast_string function using enhanced rep movsb overrides the
original function and the fast string function.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-7-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, mem: clear_page_64.S: Support clear_page() with enhanced REP MOVSB/STOSB
Fenghua Yu [Tue, 17 May 2011 22:29:14 +0000]
x86, mem: clear_page_64.S: Support clear_page() with enhanced REP MOVSB/STOSB

Intel processors are adding enhancements to REP MOVSB/STOSB and the use of
REP MOVSB/STOSB for optimal memcpy/memset or similar functions is recommended.
Enhancement availability is indicated by CPUID.7.0.EBX[9] (Enhanced REP MOVSB/
STOSB).

Support clear_page() with rep stosb for processor supporting enhanced REP MOVSB
/STOSB. On processors supporting enhanced REP MOVSB/STOSB, the alternative
clear_page_c_e function using enhanced REP STOSB overrides the original function
and the fast string function.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-6-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, alternative: Add altinstruction_entry macro
Fenghua Yu [Tue, 17 May 2011 22:29:13 +0000]
x86, alternative: Add altinstruction_entry macro

Add altinstruction_entry macro to generate .altinstructions section
entries from assembly code.  This should be less failure-prone than
open-coding.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-5-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, alternative, doc: Add comment for applying alternatives order
Fenghua Yu [Tue, 17 May 2011 22:29:12 +0000]
x86, alternative, doc: Add comment for applying alternatives order

Some string operation functions may be patched twice, e.g. on enhanced REP MOVSB
/STOSB processors, memcpy is patched first by fast string alternative function,
then it is patched by enhanced REP MOVSB/STOSB alternative function.

Add comment for applying alternatives order to warn people who may change the
applying alternatives order for any reason.

[ Documentation-only patch ]

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-4-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, mem, intel: Initialize Enhanced REP MOVSB/STOSB
Fenghua Yu [Tue, 17 May 2011 22:29:11 +0000]
x86, mem, intel: Initialize Enhanced REP MOVSB/STOSB

If kernel intends to use enhanced REP MOVSB/STOSB, it must ensure
IA32_MISC_ENABLE.Fast_String_Enable (bit 0) is set and CPUID.(EAX=07H, ECX=0H):
EBX[bit 9] also reports 1.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-3-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, cpufeature: Add CPU feature bit for enhanced REP MOVSB/STOSB
Fenghua Yu [Tue, 17 May 2011 22:29:10 +0000]
x86, cpufeature: Add CPU feature bit for enhanced REP MOVSB/STOSB

Intel processors are adding enhancements to REP MOVSB/STOSB and the use of
REP MOVSB/STOSB for optimal memcpy/memset or similar functions is recommended.
Enhancement availability is indicated by CPUID.7.0.EBX[9] (Enhanced REP MOVSB/
STOSB).

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305671358-14478-2-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>

8 years agox86, cpufeature: Fix cpuid leaf 7 feature detection
Fenghua Yu [Tue, 17 May 2011 19:33:26 +0000]
x86, cpufeature: Fix cpuid leaf 7 feature detection

CPUID leaf 7, subleaf 0 returns the maximum subleaf in EAX, not the
number of subleaves.  Since so far only subleaf 0 is defined (and only
the EBX bitfield) we do not need to qualify the test.

Signed-off-by: Fenghua Yu <fenghua.yu@intel.com>
Link: http://lkml.kernel.org/r/1305660806-17519-1-git-send-email-fenghua.yu@intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
Cc: <stable@kernel.org> 2.6.36..39

8 years agoperf: Fix multi-event parsing bug
Stephane Eranian [Tue, 17 May 2011 13:36:19 +0000]
perf: Fix multi-event parsing bug

This patch fixes an issue with event parsing.
The following commit appears to have broken the
ability to specify a comma separated list of events:

   commit ceb53fbf6dbb1df26d38379a262c6981fe73dd36
   Author: Ingo Molnar <mingo@elte.hu>
   Date:   Wed Apr 27 04:06:33 2011 +0200

       perf stat: Fail more clearly when an invalid modifier is specified

This patch fixes this while preserving the desired effect:

$ perf stat -e instructions:u,instructions:k ls /dev/null /dev/null

 Performance counter stats for 'ls /dev/null':

            365956 instructions:u           #    0.00  insns per cycle
            731806 instructions:k           #    0.00  insns per cycle

        0.001108862  seconds time elapsed

$ perf stat -e task-clock-msecs true
invalid event modifier: '-msecs'
Run 'perf list' for a list of valid events and modifiers

Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: acme@redhat.com
Cc: peterz@infradead.org
Cc: fweisbec@gmail.com
Link: http://lkml.kernel.org/r/20110517133619.GA6999@quad
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agoftrace/s390: mcount offset calculation
Martin Schwidefsky [Tue, 10 May 2011 08:10:43 +0000]
ftrace/s390: mcount offset calculation

Do the mcount offset adjustment in the recordmcount.pl/recordmcount.[ch]
at compile time and not in ftrace_call_adjust at run time.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/x86: mcount offset calculation
Martin Schwidefsky [Tue, 10 May 2011 08:10:41 +0000]
ftrace/x86: mcount offset calculation

Do the mcount offset adjustment in the recordmcount.pl/recordmcount.[ch]
at compile time and not in ftrace_call_adjust at run time.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/recordmcount: mcount address adjustment
Martin Schwidefsky [Tue, 10 May 2011 08:10:40 +0000]
ftrace/recordmcount: mcount address adjustment

Introduce mcount_adjust{,_32,_64} to the C implementation of
recordmcount analog to $mcount_adjust in the perl script.
The adjustment is added to the address of the relocations
against the mcount symbol. If this adjustment is done by
recordmcount at compile time the ftrace_call_adjust function
can be turned into a nop.

Cc: John Reiser <jreiser@bitwagon.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/recordmcount: Add helper function get_sym_str_and_relp()
Steven Rostedt [Thu, 21 Apr 2011 01:13:06 +0000]
ftrace/recordmcount: Add helper function get_sym_str_and_relp()

The code to get the symbol, string, and relp pointers in the two functions
sift_rel_mcount() and nop_mcount() are identical and also non-trivial.
Moving this duplicate code into a single helper function makes the code
easier to read and more maintainable.

Cc: John Reiser <jreiser@bitwagon.com>
Link: http://lkml.kernel.org/r/20110421023739.723658553@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/recordmcount: Remove duplicate code to find mcount symbol
Steven Rostedt [Thu, 21 Apr 2011 00:47:34 +0000]
ftrace/recordmcount: Remove duplicate code to find mcount symbol

The code in sift_rel_mcount() and nop_mcount() to get the mcount symbol
number is identical. Replace the two locations with a call to a function
that does the work.

Cc: John Reiser <jreiser@bitwagon.com>
Link: http://lkml.kernel.org/r/20110421023739.488093407@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/x86: Do not trace .discard.text section
Steven Rostedt [Wed, 13 Apr 2011 20:43:29 +0000]
ftrace/x86: Do not trace .discard.text section

The section called .discard.text has tracing attached to it and is
currently ignored by ftrace. But it does include a call to the mcount
stub. Adding a notrace to the code keeps gcc from adding the useless
mcount caller to it.

Link: http://lkml.kernel.org/r/20110421023739.243651696@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Avoid recording mcount on .init sections directly
Steven Rostedt [Tue, 12 Apr 2011 23:06:39 +0000]
ftrace: Avoid recording mcount on .init sections directly

The init and exit sections should not be traced and adding a call to
mcount to them is a waste of text and instruction cache. Have the
macro section attributes include notrace to ignore these functions
for tracing from the build.

Link: http://lkml.kernel.org/r/20110421023738.953028219@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agokbuild/recordmcount: Add RECORDMCOUNT_WARN to warn about mcount callers
Steven Rostedt [Tue, 12 Apr 2011 22:59:10 +0000]
kbuild/recordmcount: Add RECORDMCOUNT_WARN to warn about mcount callers

When mcount is called in a section that ftrace will not modify it into
a nop, we want to warn about this. But not warn about this always. Now
if the user builds the kernel with the option RECORDMCOUNT_WARN=1 then
the build will warn about mcount callers that are ignored and will just
waste execution time.

Acked-by: Michal Marek <mmarek@suse.cz>
Cc: linux-kbuild@vger.kernel.org
Link: http://lkml.kernel.org/r/20110421023738.714956282@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/recordmcount: Add warning logic to warn on mcount not recorded
Steven Rostedt [Tue, 12 Apr 2011 22:53:25 +0000]
ftrace/recordmcount: Add warning logic to warn on mcount not recorded

There's some sections that should not have mcount recorded and should not have
modifications to the that code. But currently they waste some time by calling
mcount anyway (which simply returns). As the real answer should be to
either whitelist the section or have gcc ignore it fully.

This change adds a option to recordmcount to warn when it finds a section
that is ignored by ftrace but still contains mcount callers. This is not on
by default as developers may not know if the section should be completely
ignored or added to the whitelist.

Cc: John Reiser <jreiser@bitwagon.com>
Link: http://lkml.kernel.org/r/20110421023738.476989377@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/recordmcount: Make ignored mcount calls into nops at compile time
Steven Rostedt [Fri, 8 Apr 2011 07:58:48 +0000]
ftrace/recordmcount: Make ignored mcount calls into nops at compile time

There are sections that are ignored by ftrace for the function tracing because
the text is in a section that can be removed without notice. The mcount calls
in these sections are ignored and ftrace never sees them. The downside of this
is that the functions in these sections still call mcount. Although the mcount
function is defined in assembly simply as a return, this added overhead is
unnecessary.

The solution is to convert these callers into nops at compile time.
A better solution is to add 'notrace' to the section markers, but as new sections
come up all the time, it would be nice that they are delt with when they
are created.

Later patches will deal with finding these sections and doing the proper solution.

Thanks to H. Peter Anvin for giving me the right nops to use for x86.

Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: John Reiser <jreiser@bitwagon.com>
Link: http://lkml.kernel.org/r/20110421023738.237101176@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/recordmcount: Modify only executable sections
Steven Rostedt [Wed, 13 Apr 2011 17:31:08 +0000]
ftrace/recordmcount: Modify only executable sections

PROGBITS is not enough to determine if the section should be modified
or not. Only process sections that are marked as executable.

Cc: John Reiser <jreiser@bitwagon.com>
Link: http://lkml.kernel.org/r/20110421023737.991485123@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace: Add .kprobe.text section to whitelist for recordmcount.c
Steven Rostedt [Wed, 6 Apr 2011 18:10:22 +0000]
ftrace: Add .kprobe.text section to whitelist for recordmcount.c

The .kprobe.text section is safe to modify mcount to nop and tracing.
Add it to the whitelist in recordmcount.c and recordmcount.pl.

Cc: John Reiser <jreiser@bitwagon.com>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Link: http://lkml.kernel.org/r/20110421023737.743350547@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/trivial: Clean up record mcount to use Linux switch style
Steven Rostedt [Wed, 6 Apr 2011 17:32:24 +0000]
ftrace/trivial: Clean up record mcount to use Linux switch style

The Linux style for switch statements is:

switch (var) {
case x:
[...]
break;
}

Not:
switch (var) {
case x: {
[...]
} break;

Cc: John Reiser <jreiser@bitwagon.com>
Link: http://lkml.kernel.org/r/20110421023737.523968644@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agoftrace/trivial: Clean up recordmcount.c to use Linux style comparisons
Steven Rostedt [Wed, 6 Apr 2011 17:21:17 +0000]
ftrace/trivial: Clean up recordmcount.c to use Linux style comparisons

The Linux ftrace subsystem style for comparing is:

  var == 1
  var > 0

and not:

  1 == var
  0 < var

It is considered that Linux developers are smart enough not to do the

  if (var = 1)

mistake.

Cc: John Reiser <jreiser@bitwagon.com>
Link: http://lkml.kernel.org/r/20110421023737.290712238@goodmis.org
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

8 years agox86: Remove warning and warning_symbol from struct stacktrace_ops
Richard Weinberger [Thu, 12 May 2011 13:11:12 +0000]
x86: Remove warning and warning_symbol from struct stacktrace_ops

Both warning and warning_symbol are nowhere used.
Let's get rid of them.

Signed-off-by: Richard Weinberger <richard@nod.at>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Soeren Sandmann Pedersen <ssp@redhat.com>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: x86 <x86@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Link: http://lkml.kernel.org/r/1305205872-10321-2-git-send-email-richard@nod.at
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>

8 years agoperf probe: Fix the missed parameter initialization
Lin Ming [Fri, 29 Apr 2011 08:41:57 +0000]
perf probe: Fix the missed parameter initialization

pubname_callback_param::found should be initialized to 0 in
fastpath lookup, the structure is on the stack and
uninitialized otherwise.

Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Link: http://lkml.kernel.org/r/1304066518-30420-2-git-send-email-ming.m.lin@intel.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agoMerge commit 'v2.6.39-rc7' into perf/core
Ingo Molnar [Tue, 10 May 2011 15:05:24 +0000]
Merge commit 'v2.6.39-rc7' into perf/core

Merge reason: pull in the latest fixes.

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

8 years agoLinux 2.6.39-rc7
Linus Torvalds [Tue, 10 May 2011 02:33:54 +0000]
Linux 2.6.39-rc7

8 years agovm: fix vm_pgoff wrap in upward expansion
Hugh Dickins [Tue, 10 May 2011 00:44:42 +0000]
vm: fix vm_pgoff wrap in upward expansion

Commit a626ca6a6564 ("vm: fix vm_pgoff wrap in stack expansion") fixed
the case of an expanding mapping causing vm_pgoff wrapping when you had
downward stack expansion.  But there was another case where IA64 and
PA-RISC expand mappings: upward expansion.

This fixes that case too.

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoMerge branch 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/keith...
Linus Torvalds [Mon, 9 May 2011 23:59:51 +0000]
Merge branch 'drm-intel-fixes' of git://git./linux/kernel/git/keithp/linux-2.6

* 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/keithp/linux-2.6:
  drm/i915/lvds: Only act on lid notify when the device is on
  drm/i915: fix intel_crtc_clock_get pipe reads after "cleanup cleanup"
  drm/i915: Only enable the plane after setting the fb base (pre-ILK)
  drm/i915/dp: Be paranoid in case we disable a DP before it is attached
  drm/i915: Release object along create user fb error path

8 years agoDon't lock guardpage if the stack is growing up
Mikulas Patocka [Mon, 9 May 2011 11:01:09 +0000]
Don't lock guardpage if the stack is growing up

Linux kernel excludes guard page when performing mlock on a VMA with
down-growing stack. However, some architectures have up-growing stack
and locking the guard page should be excluded in this case too.

This patch fixes lvm2 on PA-RISC (and possibly other architectures with
up-growing stack). lvm2 calculates number of used pages when locking and
when unlocking and reports an internal error if the numbers mismatch.

[ Patch changed fairly extensively to also fix /proc/<pid>/maps for the
  grows-up case, and to move things around a bit to clean it all up and
  share the infrstructure with the /proc bits.

  Tested on ia64 that has both grow-up and grow-down segments  - Linus ]

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Tested-by: Tony Luck <tony.luck@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mjg59/platf...
Linus Torvalds [Mon, 9 May 2011 19:00:49 +0000]
Merge branch 'for_linus' of git://git./linux/kernel/git/mjg59/platform-drivers-x86

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mjg59/platform-drivers-x86:
  eeepc-laptop: Use ACPI handle to identify rfkill port
  [PATCH] sony-laptop: limit brightness range to DSDT provided ones
  sony-laptop: report failures on setting LCD brightness
  thinkpad-acpi: module autoloading for newer Lenovo ThinkPads.

8 years agodrm/i915/lvds: Only act on lid notify when the device is on
Alex Williamson [Thu, 21 Apr 2011 22:08:14 +0000]
drm/i915/lvds: Only act on lid notify when the device is on

If we're using vga switcheroo, the device may be turned off
and poking it can return random state. This provokes an OOPS fixed
separately by 8ff887c847 (drm/i915/dp: Be paranoid in case we disable a
DP before it is attached). Trying to use and respond to events on a
device that has been turned off by the user is in principle a silly thing
to do.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org
Signed-off-by: Keith Packard <keithp@keithp.com>

8 years agodrm/i915: fix intel_crtc_clock_get pipe reads after "cleanup cleanup"
Chris Wilson [Fri, 22 Apr 2011 21:17:21 +0000]
drm/i915: fix intel_crtc_clock_get pipe reads after "cleanup cleanup"

Despite the fixes in 548f245ba6a31 (drm/i915: fix per-pipe reads after
"cleanup"), we missed one neighbouring read that was mistakenly replaced
with the reg value in 9db4a9c (drm/i915: cleanup per-pipe reg usage).
This was preventing us from correctly determining the mode the BIOS left
the panel in for machines that neither have an OpRegion nor access to
the VBT, (e.g. the EeePC 700).

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: stable@kernel.org
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Keith Packard <keithp@keithp.com>

8 years agodrm/i915: Only enable the plane after setting the fb base (pre-ILK)
Chris Wilson [Tue, 19 Apr 2011 20:14:14 +0000]
drm/i915: Only enable the plane after setting the fb base (pre-ILK)

When enabling the plane, it is helpful to have already pointed that
plane to valid memory or else we may incur the wrath of a PGTBL_ER.
This code preserved the behaviour from the bad old days for unknown
reasons...

Found by assert_fb_bound_for_plane().

References: https://bugs.freedesktop.org/show_bug.cgi?id=36246
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Keith Packard <keithp@keithp.com>

8 years agoMerge branch 'fix/asoc' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
Linus Torvalds [Mon, 9 May 2011 16:13:10 +0000]
Merge branch 'fix/asoc' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'fix/asoc' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
  ASoC: Fix CODEC DAI names for Goni
  ASoC: Fix CODEC name in Goni
  davinci-mcasp: fix _CBM_CFS pin directions
  davinci-mcasp: fix _CBM_CFS hw_params
  davinci-mcasp: use bitfield definitions for PDIR
  ASoC: davinci-mcasp: correct tdm_slots limit

8 years agoMerge branch 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Mon, 9 May 2011 16:09:04 +0000]
Merge branch 'drm-fixes' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm/radeon/kms: add pci id to acer travelmate quirk for 5730
  drm/radeon: fix order of doing things in radeon_crtc_cursor_set
  drm: mm: fix debug output
  drm/radeon/kms: ATPX switcheroo fixes
  drm/nouveau: Fix a crash at card takedown for NV40 and older cards

8 years agoMerge branch 'hpfs'
Linus Torvalds [Mon, 9 May 2011 16:07:55 +0000]
Merge branch 'hpfs'

* hpfs:
  HPFS: Remove unused variable
  HPFS: Move declaration up, so that there are no out-of-scope pointers
  HPFS: Fix some unaligned accesses
  HPFS: Fix endianity. Make hpfs work on big-endian machines
  HPFS: Implement fsync for hpfs
  HPFS: Fix a bug that filesystem was not marked dirty when remounting it
  HPFS: Restrict uid and gid to 16-bit values
  HPFS: When marking or clearing the dirty bit, sync the filesystem
  HPFS: Use types with defined width
  HPFS: Remove mark_inode_dirty
  HPFS: Remove CR/LF conversion option
  HPFS: Remove remaining locks
  HPFS: Introduce a global mutex and lock it on every callback from VFS.
  HPFS: Make HPFS compile on preempt and SMP

8 years agoHPFS: Remove unused variable
Mikulas Patocka [Sun, 8 May 2011 18:44:46 +0000]
HPFS: Remove unused variable

Remove unused variable

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Move declaration up, so that there are no out-of-scope pointers
Mikulas Patocka [Sun, 8 May 2011 18:44:38 +0000]
HPFS: Move declaration up, so that there are no out-of-scope pointers

Move declaration up, so that there are no out-of-scope pointers

Reported-by: Jesper Juhl <jj@chaosbits.net>
Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Fix some unaligned accesses
Mikulas Patocka [Sun, 8 May 2011 18:44:32 +0000]
HPFS: Fix some unaligned accesses

Fix some unaligned accesses

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Fix endianity. Make hpfs work on big-endian machines
Mikulas Patocka [Sun, 8 May 2011 18:44:26 +0000]
HPFS: Fix endianity. Make hpfs work on big-endian machines

Fix endianity. Make hpfs work on big-endian machines.

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Implement fsync for hpfs
Mikulas Patocka [Sun, 8 May 2011 18:44:19 +0000]
HPFS: Implement fsync for hpfs

Implement fsync for hpfs.

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Fix a bug that filesystem was not marked dirty when remounting it
Mikulas Patocka [Sun, 8 May 2011 18:44:08 +0000]
HPFS: Fix a bug that filesystem was not marked dirty when remounting it

Fix a bug that filesystem was not marked dirty when remounting it

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Restrict uid and gid to 16-bit values
Mikulas Patocka [Sun, 8 May 2011 18:44:00 +0000]
HPFS: Restrict uid and gid to 16-bit values

Restrict uid and gid to 16-bit values.

HPFS stores only 2 bytes in the EAs.

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: When marking or clearing the dirty bit, sync the filesystem
Mikulas Patocka [Sun, 8 May 2011 18:43:41 +0000]
HPFS: When marking or clearing the dirty bit, sync the filesystem

When marking or clearing the dirty bit, sync the filesystem

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Use types with defined width
Mikulas Patocka [Sun, 8 May 2011 18:43:34 +0000]
HPFS: Use types with defined width

Use types with defined width

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Remove mark_inode_dirty
Mikulas Patocka [Sun, 8 May 2011 18:43:27 +0000]
HPFS: Remove mark_inode_dirty

Remove mark_inode_dirty

HPFS doesn't use kernel's dirty inode indicator anyway because
writing an inode requires directory's mutex.

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Remove CR/LF conversion option
Mikulas Patocka [Sun, 8 May 2011 18:43:19 +0000]
HPFS: Remove CR/LF conversion option

Remove CR/LF conversion option

It is unused anyway. It was used on 2.2 kernels or so.

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Remove remaining locks
Mikulas Patocka [Sun, 8 May 2011 18:43:06 +0000]
HPFS: Remove remaining locks

Remove remaining locks

Because of a new global per-fs lock, no other locks are needed

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Introduce a global mutex and lock it on every callback from VFS.
Mikulas Patocka [Sun, 8 May 2011 18:42:54 +0000]
HPFS: Introduce a global mutex and lock it on every callback from VFS.

Introduce a global mutex and lock it on every callback from VFS.

Performance doesn't matter, reviewing the whole code for locking correctness
would be too complicated, so simply lock it all.

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoHPFS: Make HPFS compile on preempt and SMP
Mikulas Patocka [Sun, 8 May 2011 18:42:44 +0000]
HPFS: Make HPFS compile on preempt and SMP

Make HPFS compile on preempt and SMP

Signed-off-by: Mikulas Patocka <mikulas@artax.karlin.mff.cuni.cz>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoeeepc-laptop: Use ACPI handle to identify rfkill port
Matthew Garrett [Mon, 9 May 2011 14:44:01 +0000]
eeepc-laptop: Use ACPI handle to identify rfkill port

The ACPI notification we get from rfkill events on these machines gives
us all the information we need to identify the port that's changed. Do
so rather than assuming that it's always bus 1.

Signed-off-by: Matthew Garrett <mjg@redhat.com>

8 years ago[PATCH] sony-laptop: limit brightness range to DSDT provided ones
Mattia Dongili [Mon, 9 May 2011 14:20:29 +0000]
[PATCH] sony-laptop: limit brightness range to DSDT provided ones

The new style brightness control provides an operating range of 9 values
(seems consistent over a large number of models sharing the same
brightness control methods).
Read and use the minimum and maximum values to limit the backlight
interface between those boundaries.

Signed-off-by: Mattia Dongili <malattia@linux.it>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

8 years agosony-laptop: report failures on setting LCD brightness
Mattia Dongili [Tue, 5 Apr 2011 14:38:36 +0000]
sony-laptop: report failures on setting LCD brightness

Check if we were successful in setting the requested brightness and
report failure in that case.

Signed-off-by: Mattia Dongili <malattia@linux.it>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

8 years agothinkpad-acpi: module autoloading for newer Lenovo ThinkPads.
Manoj Iyer [Sun, 8 May 2011 22:04:29 +0000]
thinkpad-acpi: module autoloading for newer Lenovo ThinkPads.

The newer Lenovo ThinkPads have HKEY HID of LEN0068 instead
of IBM0068. Added new HID so that thinkpad_acpi module will
auto load on these newer Lenovo ThinkPads.

Acked-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Cc: stable@kernel.org
Signed-off-by: Manoj Iyer <manoj.iyer@canonical.com>
Signed-off-by: Andy Lutomirski <luto@mit.edu>
Signed-off-by: Matthew Garrett <mjg@redhat.com>

8 years agodrm/radeon/kms: add pci id to acer travelmate quirk for 5730
Alex Deucher [Wed, 4 May 2011 15:41:47 +0000]
drm/radeon/kms: add pci id to acer travelmate quirk for 5730

Fixes:
https://bugzilla.kernel.org/show_bug.cgi?id=34082

Reported by: Sampo Laaksonen <zhamahn@gmail.com>
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agodrm/radeon: fix order of doing things in radeon_crtc_cursor_set
Ilija Hadzic [Thu, 5 May 2011 00:15:03 +0000]
drm/radeon: fix order of doing things in radeon_crtc_cursor_set

 if object pin or object lookup in radeon_cursor_set fail, the function
 could leave inconsistent mouse width and hight values in radeon_crtc
 fixed by moving cursor width and height assignments after all
 checks have passed

Signed-off-by: Ilija Hadzic <ihadzic@research.bell-labs.com>
Reviewed-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agodrm: mm: fix debug output
Daniel Vetter [Fri, 6 May 2011 21:47:53 +0000]
drm: mm: fix debug output

The looping helper didn't do anything due to a superficial
semicolon. Furthermore one of the two dump functions suffered
from copy&paste fail.

While staring at the code I've also noticed that the replace
helper (currently unused) is a bit broken.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agoMerge remote branch 'nouveau/drm-nouveau-fixes' of /ssd/git/drm-nouveau-next into...
Dave Airlie [Sun, 8 May 2011 23:14:38 +0000]
Merge remote branch 'nouveau/drm-nouveau-fixes' of /ssd/git/drm-nouveau-next into drm-fixes

* 'nouveau/drm-nouveau-fixes' of /ssd/git/drm-nouveau-next:
  drm/nouveau: Fix a crash at card takedown for NV40 and older cards

8 years agodrm/radeon/kms: ATPX switcheroo fixes
Alex Deucher [Fri, 6 May 2011 05:42:49 +0000]
drm/radeon/kms: ATPX switcheroo fixes

When we switch the display mux, also switch
the i2c mux.  Also use the start and finish
methods to let the sbios know that the switch
is happening.

Should fix:
https://bugs.freedesktop.org/show_bug.cgi?id=35398

Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agodrm/nouveau: Fix a crash at card takedown for NV40 and older cards
Jimmy Rentz [Sun, 17 Apr 2011 20:15:09 +0000]
drm/nouveau: Fix a crash at card takedown for NV40 and older cards

NV40 and older cards (pre NV50) reserve a vram bo for the vga memory at
card init. This bo is then freed at card shutdown.  The problem is that
the ttm bo vram manager was already freed. So a crash occurs when the
vga bo is freed. The fix is to free the vga bo prior to freeing the ttm
bo vram manager. There might be other solutions but this seemed the
simplest to me.

Signed-off-by: Jimmy Rentz <jb17bsome@gmail.com>
Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

8 years agoMerge branch 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 7 May 2011 20:17:37 +0000]
Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  perf tools: Makefile: Use gcc to determine ARCH
  perf events, x86: Fix Intel Nehalem and Westmere last level cache event definitions
  hw_breakpoints, powerpc: Fix CONFIG_HAVE_HW_BREAKPOINT off-case in ptrace_set_debugreg()
  sh, hw_breakpoints: Fix racy access to ptrace breakpoints
  arm, hw_breakpoints: Fix racy access to ptrace breakpoints
  powerpc, hw_breakpoints: Fix racy access to ptrace breakpoints
  x86, hw_breakpoints: Fix racy access to ptrace breakpoints
  ptrace: Prepare to fix racy accesses on task breakpoints

8 years agoperf tools: Makefile: Use gcc to determine ARCH
Lin Ming [Sat, 7 May 2011 04:41:14 +0000]
perf tools: Makefile: Use gcc to determine ARCH

The original Makefile uses "uname -m" to determine ARCH.
This causes problem on x86 when compile perf tool on 32 bit
userspace with a 64 bit kernel.

 bench/../../../arch/x86/lib/memcpy_64.S: Assembler messages:
 bench/../../../arch/x86/lib/memcpy_64.S:28: Error: bad register name `%rdi'

This is because "uname -m" returns x86_64 and memcpy_64.S is
included in 32 bit build.

Reported-by: Riccardo Magliocchetti <riccardo.magliocchetti@gmail.com>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Link: http://lkml.kernel.org/r/1304743274.3132.17.camel@localhost
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6
Linus Torvalds [Fri, 6 May 2011 22:32:41 +0000]
Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  cifs: handle errors from coalesce_t2
  cifs: refactor mid finding loop in cifs_demultiplex_thread
  cifs: sanitize length checking in coalesce_t2 (try #3)
  cifs: check for bytes_remaining going to zero in CIFS_SessSetup
  cifs: change bleft in decode_unicode_ssetup back to signed type

8 years agoRegression: partial revert "tracing: Remove lock_depth from event entry"
Arjan van de Ven [Fri, 6 May 2011 03:55:18 +0000]
Regression: partial revert "tracing: Remove lock_depth from event entry"

This partially reverts commit e6e1e2593592a8f6f6380496655d8c6f67431266.

That commit changed the structure layout of the trace structure, which
in turn broke PowerTOP (1.9x generation) quite badly.

I appreciate not wanting to expose the variable in question, and
PowerTOP was not using it, so I've replaced the variable with just a
padding field - that way if in the future a new field is needed it can
just use this padding field.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoMerge branch 'perf/stat' into perf/core
Ingo Molnar [Fri, 6 May 2011 19:07:33 +0000]
Merge branch 'perf/stat' into perf/core

Merge reason: the perf stat improvements are tested and ready now.

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

8 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6
Linus Torvalds [Fri, 6 May 2011 17:01:54 +0000]
Merge git://git./linux/kernel/git/jejb/scsi-rc-fixes-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6:
  [SCSI] fix oops in scsi_run_queue()

8 years agoValidate size of EFI GUID partition entries.
Timo Warns [Fri, 6 May 2011 11:47:35 +0000]
Validate size of EFI GUID partition entries.

Otherwise corrupted EFI partition tables can cause total confusion.

Signed-off-by: Timo Warns <warns@pre-sense.de>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoperf events, x86: Fix Intel Nehalem and Westmere last level cache event definitions
Peter Zijlstra [Fri, 22 Apr 2011 22:57:42 +0000]
perf events, x86: Fix Intel Nehalem and Westmere last level cache event definitions

The Intel Nehalem offcore bits implemented in:

  e994d7d23a0b: perf: Fix LLC-* events on Intel Nehalem/Westmere

... are wrong: they implemented _ACCESS as _HIT and counted OTHER_CORE_HIT* as
MISS even though its clearly documented as an L3 hit ...

Fix them and the Westmere definitions as well.

Cc: Andi Kleen <ak@linux.intel.com>
Cc: Lin Ming <ming.m.lin@intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/1299119690-13991-3-git-send-email-ming.m.lin@intel.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>

8 years agohw_breakpoints, powerpc: Fix CONFIG_HAVE_HW_BREAKPOINT off-case in ptrace_set_debugreg()
Frederic Weisbecker [Thu, 5 May 2011 23:53:18 +0000]
hw_breakpoints, powerpc: Fix CONFIG_HAVE_HW_BREAKPOINT off-case in ptrace_set_debugreg()

We make use of ptrace_get_breakpoints() / ptrace_put_breakpoints() to
protect ptrace_set_debugreg() even if CONFIG_HAVE_HW_BREAKPOINT if off.
However in this case, these APIs are not implemented.

To fix this, push the protection down inside the relevant ifdef.
Best would be to export the code inside
CONFIG_HAVE_HW_BREAKPOINT into a standalone function to cleanup
the ifdefury there and call the breakpoint ref API inside. But
as it is more invasive, this should be rather made in an -rc1.

Fixes this build error:

  arch/powerpc/kernel/ptrace.c:1594: error: implicit declaration of function 'ptrace_get_breakpoints' make[2]: ***

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: LPPC <linuxppc-dev@lists.ozlabs.org>
Cc: Prasad <prasad@linux.vnet.ibm.com>
Cc: v2.6.33.. <stable@kernel.org>
Link: http://lkml.kernel.org/r/1304639598-4707-1-git-send-email-fweisbec@gmail.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>