openwrt/staging/blogic.git
9 years agoperf/x86/intel: Add basic Haswell LBR call stack support
Yan, Zheng [Wed, 5 Nov 2014 02:56:00 +0000 (21:56 -0500)]
perf/x86/intel: Add basic Haswell LBR call stack support

Haswell has a new feature that utilizes the existing LBR facility to
record call chains. To enable this feature, bits (JCC, NEAR_IND_JMP,
NEAR_REL_JMP, FAR_BRANCH, EN_CALLSTACK) in LBR_SELECT must be set to 1,
bits (NEAR_REL_CALL, NEAR-IND_CALL, NEAR_RET) must be cleared. Due to
a hardware bug of Haswell, this feature doesn't work well with
FREEZE_LBRS_ON_PMI.

When the call stack feature is enabled, the LBR stack will capture
unfiltered call data normally, but as return instructions are executed,
the last captured branch record is flushed from the on-chip registers
in a last-in first-out (LIFO) manner. Thus, branch information relative
to leaf functions will not be captured, while preserving the call stack
information of the main line execution path.

This patch defines a separate lbr_sel map for Haswell. The map contains
a new entry for the call stack feature.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: eranian@google.com
Cc: jolsa@redhat.com
Link: http://lkml.kernel.org/r/1415156173-10035-5-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86/intel: Use context switch callback to flush LBR stack
Yan, Zheng [Wed, 5 Nov 2014 02:55:59 +0000 (21:55 -0500)]
perf/x86/intel: Use context switch callback to flush LBR stack

Previous commit introduces context switch callback, its function
overlaps with the flush branch stack callback. So we can use the
context switch callback to flush LBR stack.

This patch adds code that uses the flush branch callback to
flush the LBR stack when task is being scheduled in. The callback
is enabled only when there are events use the LBR hardware. This
patch also removes all old flush branch stack code.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: eranian@google.com
Cc: jolsa@redhat.com
Link: http://lkml.kernel.org/r/1415156173-10035-4-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Introduce pmu context switch callback
Yan, Zheng [Wed, 5 Nov 2014 02:55:58 +0000 (21:55 -0500)]
perf: Introduce pmu context switch callback

The callback is invoked when process is scheduled in or out.
It provides mechanism for later patches to save/store the LBR
stack. For the schedule in case, the callback is invoked at
the same place that flush branch stack callback is invoked.
So it also can replace the flush branch stack callback. To
avoid unnecessary overhead, the callback is enabled only when
there are events use the LBR stack.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: eranian@google.com
Cc: jolsa@redhat.com
Link: http://lkml.kernel.org/r/1415156173-10035-3-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86/intel: Reduce lbr_sel_map[] size
Yan, Zheng [Wed, 5 Nov 2014 02:55:57 +0000 (21:55 -0500)]
perf/x86/intel: Reduce lbr_sel_map[] size

The index of lbr_sel_map is bit value of perf branch_sample_type.
PERF_SAMPLE_BRANCH_MAX is 1024 at present, so each lbr_sel_map uses
4096 bytes. By using bit shift as index, we can reduce lbr_sel_map
size to 40 bytes. This patch defines 'bit shift' for branch types,
and use 'bit shift' to define lbr_sel_maps.

Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Stephane Eranian <eranian@google.com>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: jolsa@redhat.com
Cc: linux-api@vger.kernel.org
Link: http://lkml.kernel.org/r/1415156173-10035-2-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86/amd/ibs: Convert force_ibs_eilvt_setup() to void
Aravind Gopalakrishnan [Fri, 23 Jan 2015 18:19:35 +0000 (12:19 -0600)]
perf/x86/amd/ibs: Convert force_ibs_eilvt_setup() to void

The caller of force_ibs_eilvt_setup() is ibs_eilvt_setup()
which does not care about the return values.

So mark it void and clean up the return statements.

Signed-off-by: Aravind Gopalakrishnan <aravind.gopalakrishnan@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <hpa@zytor.com>
Cc: <paulus@samba.org>
Cc: <tglx@linutronix.de>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1422037175-20957-1-git-send-email-aravind.gopalakrishnan@amd.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Update userspace page info for software event
Shaohua Li [Thu, 5 Feb 2015 23:55:32 +0000 (15:55 -0800)]
perf: Update userspace page info for software event

For hardware events, the userspace page of the event gets updated in
context switches, so if we read the timestamp in the page, we get
fresh info.

For software events, this is missing currently. This patch makes the
behavior consistent.

With this patch, we can implement clock_gettime(THREAD_CPUTIME) with
PERF_COUNT_SW_DUMMY in userspace as suggested by Andy and Peter. Code
like this:

  if (pc->cap_user_time) {
do {
seq = pc->lock;
barrier();

running = pc->time_running;
cyc = rdtsc();
time_mult = pc->time_mult;
time_shift = pc->time_shift;
time_offset = pc->time_offset;

barrier();
} while (pc->lock != seq);

quot = (cyc >> time_shift);
rem = cyc & ((1 << time_shift) - 1);
delta = time_offset + quot * time_mult +
((rem * time_mult) >> time_shift);

running += delta;
return running;
  }

I tried it on a busy system, the userspace page updating doesn't
have noticeable overhead.

Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/aa2dd2e4f1e9f2225758be5ba00f14d6909a8ce1.1423180257.git.shli@fb.com
[ Improved the changelog. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Update shadow timestamp before add event
Shaohua Li [Thu, 5 Feb 2015 23:55:31 +0000 (15:55 -0800)]
perf: Update shadow timestamp before add event

Update the shadow timestamp before start event, because .add might
use the timestamp.

Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul Mackerras <paulus@samba.org>
Link: http://lkml.kernel.org/r/9cd0276d6a047cb7c2885994f25e3a1f7c8c28af.1423180257.git.shli@fb.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86/intel/uncore: Delete an unnecessary check before pci_dev_put() call
Markus Elfring [Tue, 3 Feb 2015 11:40:54 +0000 (12:40 +0100)]
perf/x86/intel/uncore: Delete an unnecessary check before pci_dev_put() call

The pci_dev_put() function tests whether its argument is NULL and then
returns immediately. Thus the test around the call is not needed.

This issue was detected by using the Coccinelle software.

Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/54D0B59C.2060106@users.sourceforge.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86: Add /sys/devices/cpu/rdpmc=2 to allow rdpmc for all tasks
Andy Lutomirski [Fri, 24 Oct 2014 22:58:13 +0000 (15:58 -0700)]
perf/x86: Add /sys/devices/cpu/rdpmc=2 to allow rdpmc for all tasks

While perfmon2 is a sufficiently evil library (it pokes MSRs
directly) that breaking it is fair game, it's still useful, so we
might as well try to support it.  This allows users to write 2 to
/sys/devices/cpu/rdpmc to disable all rdpmc protection so that hack
like perfmon2 can continue to work.

At some point, if perf_event becomes fast enough to replace
perfmon2, then this can go.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/caac3c1c707dcca48ecbc35f4def21495856f479.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86: Only allow rdpmc if a perf_event is mapped
Andy Lutomirski [Fri, 24 Oct 2014 22:58:12 +0000 (15:58 -0700)]
perf/x86: Only allow rdpmc if a perf_event is mapped

We currently allow any process to use rdpmc.  This significantly
weakens the protection offered by PR_TSC_DISABLED, and it could be
helpful to users attempting to exploit timing attacks.

Since we can't enable access to individual counters, use a very
coarse heuristic to limit access to rdpmc: allow access only when
a perf_event is mmapped.  This protects seccomp sandboxes.

There is plenty of room to further tighen these restrictions.  For
example, this allows rdpmc for any x86_pmu event, but it's only
useful for self-monitoring tasks.

As a side effect, cap_user_rdpmc will now be false for AMD uncore
events.  This isn't a real regression, since .event_idx is disabled
for these events anyway for the time being.  Whenever that gets
re-added, the cap_user_rdpmc code can be adjusted or refactored
accordingly.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/a2bdb3cf3a1d70c26980d7c6dddfbaa69f3182bf.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Pass the event to arch_perf_update_userpage()
Andy Lutomirski [Fri, 24 Oct 2014 22:58:11 +0000 (15:58 -0700)]
perf: Pass the event to arch_perf_update_userpage()

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/0fea9a7fac3c1eea86cb0a5954184e74f4213666.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Add pmu callbacks to track event mapping and unmapping
Andy Lutomirski [Fri, 24 Oct 2014 22:58:10 +0000 (15:58 -0700)]
perf: Add pmu callbacks to track event mapping and unmapping

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/266afcba1d1f91ea5501e4e16e94bbbc1a9339b6.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agox86: Add a comment clarifying LDT context switching
Andy Lutomirski [Fri, 24 Oct 2014 22:58:09 +0000 (15:58 -0700)]
x86: Add a comment clarifying LDT context switching

The code is correct, but only for a rather subtle reason.  This
confused me for quite a while when I read switch_mm, so clarify the
code to avoid confusing other people, too.

TBH, I wouldn't be surprised if this code was only correct by
accident.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/0db86397f968996fb772c443c251415b0b430ddd.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agox86: Store a per-cpu shadow copy of CR4
Andy Lutomirski [Fri, 24 Oct 2014 22:58:08 +0000 (15:58 -0700)]
x86: Store a per-cpu shadow copy of CR4

Context switches and TLB flushes can change individual bits of CR4.
CR4 reads take several cycles, so store a shadow copy of CR4 in a
per-cpu variable.

To avoid wasting a cache line, I added the CR4 shadow to
cpu_tlbstate, which is already touched in switch_mm.  The heaviest
users of the cr4 shadow will be switch_mm and __switch_to_xtra, and
__switch_to_xtra is called shortly after switch_mm during context
switch, so the cacheline is likely to be hot.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/3a54dd3353fffbf84804398e00dfdc5b7c1afd7d.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agox86: Clean up cr4 manipulation
Andy Lutomirski [Fri, 24 Oct 2014 22:58:07 +0000 (15:58 -0700)]
x86: Clean up cr4 manipulation

CR4 manipulation was split, seemingly at random, between direct
(write_cr4) and using a helper (set/clear_in_cr4).  Unfortunately,
the set_in_cr4 and clear_in_cr4 helpers also poke at the boot code,
which only a small subset of users actually wanted.

This patch replaces all cr4 access in functions that don't leave cr4
exactly the way they found it with new helpers cr4_set_bits,
cr4_clear_bits, and cr4_set_bits_and_update_boot.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Vince Weaver <vince@deater.net>
Cc: "hillf.zj" <hillf.zj@alibaba-inc.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/495a10bdc9e67016b8fd3945700d46cfd5c12c2f.1414190806.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge branch 'x86/asm' into perf/x86, to avoid conflicts with upcoming patches
Ingo Molnar [Wed, 4 Feb 2015 08:01:12 +0000 (09:01 +0100)]
Merge branch 'x86/asm' into perf/x86, to avoid conflicts with upcoming patches

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Decouple unthrottling and rotating
Mark Rutland [Wed, 7 Jan 2015 15:01:54 +0000 (15:01 +0000)]
perf: Decouple unthrottling and rotating

Currently the adjusments made as part of perf_event_task_tick() use the
percpu rotation lists to iterate over any active PMU contexts, but these
are not used by the context rotation code, having been replaced by
separate (per-context) hrtimer callbacks. However, some manipulation of
the rotation lists (i.e. removal of contexts) has remained in
perf_rotate_context(). This leads to the following issues:

* Contexts are not always removed from the rotation lists. Removal of
  PMUs which have been placed in rotation lists, but have not been
  removed by a hrtimer callback can result in corruption of the rotation
  lists (when memory backing the context is freed).

  This has been observed to result in hangs when PMU drivers built as
  modules are inserted and removed around the creation of events for
  said PMUs.

* Contexts which do not require rotation may be removed from the
  rotation lists as a result of a hrtimer, and will not be considered by
  the unthrottling code in perf_event_task_tick.

This patch fixes the issue by updating the rotation ist when events are
scheduled in/out, ensuring that each rotation list stays in sync with
the HW state. As each event holds a refcount on the module of its PMU,
this ensures that when a PMU module is unloaded none of its CPU contexts
can be in a rotation list. By maintaining a list of perf_event_contexts
rather than perf_event_cpu_contexts, we don't need separate paths to
handle the cpu and task contexts, which also makes the code a little
simpler.

As the rotation_list variables are not used for rotation, these are
renamed to active_ctx_list, which better matches their current function.
perf_pmu_rotate_{start,stop} are renamed to
perf_pmu_ctx_{activate,deactivate}.

Reported-by: Johannes Jensen <johannes.jensen@arm.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Will Deacon <Will.Deacon@arm.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20150129134511.GR17721@leverpostej
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Drop module reference on event init failure
Mark Rutland [Wed, 7 Jan 2015 14:56:51 +0000 (14:56 +0000)]
perf: Drop module reference on event init failure

When initialising an event, perf_init_event will call try_module_get() to
ensure that the PMU's module cannot be removed for the lifetime of the
event, with __free_event() dropping the reference when the event is
finally destroyed. If something fails after the event has been
initialised, but before the event is installed, perf_event_alloc will
drop the reference on the module.

However, if we fail to initialise an event for some reason (e.g. we ask
an uncore PMU to perform sampling, and it refuses to initialise the
event), we do not drop the refcount. If we try to open such a bogus
event without a precise IDR type, we will loop over each PMU in the pmus
list, incrementing each of their refcounts without decrementing them.

This patch adds a module_put when pmu->event_init(event) fails, ensuring
that the refcounts are balanced in failure cases. As the innards of the
precise and search based initialisation look very similar, this logic is
hoisted out into a new helper function. While the early return for the
failed try_module_get is removed from the search case, this is handled
by the remaining return when ret is not -ENOENT.

Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1420642611-22667-1-git-send-email-mark.rutland@arm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Use POLLIN instead of POLL_IN for perf poll data in flag
Jiri Olsa [Wed, 28 Jan 2015 17:54:38 +0000 (18:54 +0100)]
perf: Use POLLIN instead of POLL_IN for perf poll data in flag

Currently we flag available data (via poll syscall) on perf fd with
POLL_IN macro, which is normally used for SIGIO interface.

We've been lucky, because POLLIN (0x1) is subset of POLL_IN (0x20001)
and sys_poll (do_pollfd function) cut the extra bit out (0x20000).

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1422467678-22341-1-git-send-email-jolsa@kernel.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Fix put_event() ctx lock
Peter Zijlstra [Thu, 29 Jan 2015 13:44:34 +0000 (14:44 +0100)]
perf: Fix put_event() ctx lock

So what I suspect; but I'm in zombie mode today it seems; is that while
I initially thought that it was impossible for ctx to change when
refcount dropped to 0, I now suspect its possible.

Note that until perf_remove_from_context() the event is still active and
visible on the lists. So a concurrent sys_perf_event_open() from another
task into this task can race.

Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Stephane Eranian <eranian@gmail.com>
Cc: mark.rutland@arm.com
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20150129134434.GB26304@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Fix move_group() order
Peter Zijlstra (Intel) [Tue, 27 Jan 2015 10:53:12 +0000 (11:53 +0100)]
perf: Fix move_group() order

Jiri reported triggering the new WARN_ON_ONCE in event_sched_out over
the weekend:

  event_sched_out.isra.79+0x2b9/0x2d0
  group_sched_out+0x69/0xc0
  ctx_sched_out+0x106/0x130
  task_ctx_sched_out+0x37/0x70
  __perf_install_in_context+0x70/0x1a0
  remote_function+0x48/0x60
  generic_exec_single+0x15b/0x1d0
  smp_call_function_single+0x67/0xa0
  task_function_call+0x53/0x80
  perf_install_in_context+0x8b/0x110

I think the below should cure this; if we install a group leader it
will iterate the (still intact) group list and find its siblings and
try and install those too -- even though those still have the old
event->ctx -- in the new ctx.

Upon installing the first group sibling we'd try and schedule out the
group and trigger the above warn.

Fix this by installing the group leader last, installing siblings
would have no effect, they're not reachable through the group lists
and therefore we don't schedule them.

Also delay resetting the state until we're absolutely sure the events
are quiescent.

Reported-by: Jiri Olsa <jolsa@redhat.com>
Reported-by: vincent.weaver@maine.edu
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20150126162639.GA21418@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Fix event->ctx locking
Peter Zijlstra [Fri, 23 Jan 2015 11:24:14 +0000 (12:24 +0100)]
perf: Fix event->ctx locking

There have been a few reported issues wrt. the lack of locking around
changing event->ctx. This patch tries to address those.

It avoids the whole rwsem thing; and while it appears to work, please
give it some thought in review.

What I did fail at is sensible runtime checks on the use of
event->ctx, the RCU use makes it very hard.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20150123125834.209535886@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf: Add a bit of paranoia
Peter Zijlstra [Fri, 23 Jan 2015 10:20:10 +0000 (11:20 +0100)]
perf: Add a bit of paranoia

Add a few WARN()s to catch things that should never happen.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20150123125834.150481799@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'v3.19-rc7' into perf/core, to merge fixes before applying new changes
Ingo Molnar [Wed, 4 Feb 2015 06:58:29 +0000 (07:58 +0100)]
Merge tag 'v3.19-rc7' into perf/core, to merge fixes before applying new changes

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'pr-20150201-x86-entry' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Tue, 3 Feb 2015 11:24:08 +0000 (12:24 +0100)]
Merge tag 'pr-20150201-x86-entry' of git://git./linux/kernel/git/luto/linux into x86/asm

Pull "x86: Entry cleanups and a bugfix for 3.20" from Andy Lutomirski:

 " This fixes a bug in the RCU code I added in ist_enter.  It also includes
   the sysret stuff discussed here:

     http://lkml.kernel.org/g/cover.1421453410.git.luto%40amacapital.net "

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'pr-20150201-x86-vdso' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Tue, 3 Feb 2015 11:22:47 +0000 (12:22 +0100)]
Merge tag 'pr-20150201-x86-vdso' of git://git./linux/kernel/git/luto/linux into x86/asm

Pull VDSO fix fro Andy Lutomirski:

 "x86, vdso: One trivial last-minute VDSO build improvement

  Andrey noticed that the VDSO build wasn't cleaning itself up.  This
  one-liner fixes it."

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'v3.19-rc7' into x86/asm, to refresh the branch before pulling in new changes
Ingo Molnar [Tue, 3 Feb 2015 11:22:18 +0000 (12:22 +0100)]
Merge tag 'v3.19-rc7' into x86/asm, to refresh the branch before pulling in new changes

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoLinux 3.19-rc7
Linus Torvalds [Mon, 2 Feb 2015 04:07:21 +0000 (20:07 -0800)]
Linux 3.19-rc7

9 years agoMerge tag 'armsoc-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
Linus Torvalds [Sun, 1 Feb 2015 21:20:47 +0000 (13:20 -0800)]
Merge tag 'armsoc-for-linus' of git://git./linux/kernel/git/arm/arm-soc

Pull ARM SoC fixes from Olof Johansson:
 "One more week's worth of fixes.  Worth pointing out here are:

   - A patch fixing detaching of iommu registrations when a device is
     removed -- earlier the ops pointer wasn't managed properly
   - Another set of Renesas boards get the same GIC setup fixup as
     others have in previous -rcs
   - Serial port aliases fixups for sunxi.  We did the same to tegra but
     we caught that in time before the merge window due to more machines
     being affected.  Here it took longer for anyone to notice.
   - A couple more DT tweaks on sunxi
   - A follow-up patch for the mvebu coherency disabling in last -rc
     batch"

* tag 'armsoc-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc:
  arm: dma-mapping: Set DMA IOMMU ops in arm_iommu_attach_device()
  ARM: shmobile: r8a7790: Instantiate GIC from C board code in legacy builds
  ARM: shmobile: r8a73a4: Instantiate GIC from C board code in legacy builds
  ARM: mvebu: don't set the PL310 in I/O coherency mode when I/O coherency is disabled
  ARM: sunxi: dt: Fix aliases
  ARM: dts: sun4i: Add simplefb node with de_fe0-de_be0-lcd0-hdmi pipeline
  ARM: dts: sun6i: ippo-q8h-v5: Fix serial0 alias
  ARM: dts: sunxi: Fix usb-phy support for sun4i/sun5i

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sun, 1 Feb 2015 21:16:40 +0000 (13:16 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input layer updates from Dmitry Torokhov:
 "Just a few quirks for PS/2 this time"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: elantech - add more Fujtisu notebooks to force crc_enabled
  Input: i8042 - add noloop quirk for Medion Akoya E7225 (MD98857)
  Input: synaptics - adjust min/max for Lenovo ThinkPad X1 Carbon 2nd

9 years agosched: don't cause task state changes in nested sleep debugging
Linus Torvalds [Sun, 1 Feb 2015 20:23:32 +0000 (12:23 -0800)]
sched: don't cause task state changes in nested sleep debugging

Commit 8eb23b9f35aa ("sched: Debug nested sleeps") added code to report
on nested sleep conditions, which we generally want to avoid because the
inner sleeping operation can re-set the thread state to TASK_RUNNING,
but that will then cause the outer sleep loop not actually sleep when it
calls schedule.

However, that's actually valid traditional behavior, with the inner
sleep being some fairly rare case (like taking a sleeping lock that
normally doesn't actually need to sleep).

And the debug code would actually change the state of the task to
TASK_RUNNING internally, which makes that kind of traditional and
working code not work at all, because now the nested sleep doesn't just
sometimes cause the outer one to not block, but will cause it to happen
every time.

In particular, it will cause the cardbus kernel daemon (pccardd) to
basically busy-loop doing scheduling, converting a laptop into a heater,
as reported by Bruno Prémont.  But there may be other legacy uses of
that nested sleep model in other drivers that are also likely to never
get converted to the new model.

This fixes both cases:

 - don't set TASK_RUNNING when the nested condition happens (note: even
   if WARN_ONCE() only _warns_ once, the return value isn't whether the
   warning happened, but whether the condition for the warning was true.
   So despite the warning only happening once, the "if (WARN_ON(..))"
   would trigger for every nested sleep.

 - in the cases where we knowingly disable the warning by using
   "sched_annotate_sleep()", don't change the task state (that is used
   for all core scheduling decisions), instead use '->task_state_change'
   that is used for the debugging decision itself.

(Credit for the second part of the fix goes to Oleg Nesterov: "Can't we
avoid this subtle change in behaviour DEBUG_ATOMIC_SLEEP adds?" with the
suggested change to use 'task_state_change' as part of the test)

Reported-and-bisected-by: Bruno Prémont <bonbons@linux-vserver.org>
Tested-by: Rafael J Wysocki <rjw@rjwysocki.net>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>,
Cc: Ilya Dryomov <ilya.dryomov@inktank.com>,
Cc: Mike Galbraith <umgwanakikbuti@gmail.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Hurley <peter@hurleysoftware.com>,
Cc: Davidlohr Bueso <dave@stgolabs.net>,
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoInput: elantech - add more Fujtisu notebooks to force crc_enabled
Rainer Koenig [Tue, 27 Jan 2015 23:15:11 +0000 (15:15 -0800)]
Input: elantech - add more Fujtisu notebooks to force crc_enabled

Add two more Fujitsu LIFEBOOK models that also ship with the Elantech
touchpad and don't work with crc_disabled to the quirk list.

Signed-off-by: Rainer Koenig <Rainer.Koenig@ts.fujitsu.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
9 years agoMerge tag 'renesas-soc-fixes3-for-v3.19' of git://git.kernel.org/pub/scm/linux/kernel...
Olof Johansson [Sun, 1 Feb 2015 16:51:12 +0000 (08:51 -0800)]
Merge tag 'renesas-soc-fixes3-for-v3.19' of git://git./linux/kernel/git/horms/renesas into fixes

Merge "Third Round of Renesas ARM Based SoC Fixes for v3.19" from Simon Horman:

* Instantiate GIC from C board code in legacy builds on r8a7790 and r8a73a4

* tag 'renesas-soc-fixes3-for-v3.19' of git://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas:
  ARM: shmobile: r8a7790: Instantiate GIC from C board code in legacy builds
  ARM: shmobile: r8a73a4: Instantiate GIC from C board code in legacy builds

Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agox86_64, entry: Remove the syscall exit audit and schedule optimizations
Andy Lutomirski [Mon, 7 Jul 2014 18:37:17 +0000 (11:37 -0700)]
x86_64, entry: Remove the syscall exit audit and schedule optimizations

We used to optimize rescheduling and audit on syscall exit.  Now
that the full slow path is reasonably fast, remove these
optimizations.  Syscall exit auditing is now handled exclusively by
syscall_trace_leave.

This adds something like 10ns to the previously optimized paths on
my computer, presumably due mostly to SAVE_REST / RESTORE_REST.

I think that we should eventually replace both the syscall and
non-paranoid interrupt exit slow paths with a pair of C functions
along the lines of the syscall entry hooks.

Link: http://lkml.kernel.org/r/22f2aa4a0361707a5cfb1de9d45260b39965dead.1421453410.git.luto@amacapital.net
Acked-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
9 years agox86_64, entry: Use sysret to return to userspace when possible
Andy Lutomirski [Tue, 22 Jul 2014 19:46:50 +0000 (12:46 -0700)]
x86_64, entry: Use sysret to return to userspace when possible

The x86_64 entry code currently jumps through complex and
inconsistent hoops to try to minimize the impact of syscall exit
work.  For a true fast-path syscall, almost nothing needs to be
done, so returning is just a check for exit work and sysret.  For a
full slow-path return from a syscall, the C exit hook is invoked if
needed and we join the iret path.

Using iret to return to userspace is very slow, so the entry code
has accumulated various special cases to try to do certain forms of
exit work without invoking iret.  This is error-prone, since it
duplicates assembly code paths, and it's dangerous, since sysret
can malfunction in interesting ways if used carelessly.  It's
also inefficient, since a lot of useful cases aren't optimized
and therefore force an iret out of a combination of paranoia and
the fact that no one has bothered to write even more asm code
to avoid it.

I would argue that this approach is backwards.  Rather than trying
to avoid the iret path, we should instead try to make the iret path
fast.  Under a specific set of conditions, iret is unnecessary.  In
particular, if RIP==RCX, RFLAGS==R11, RIP is canonical, RF is not
set, and both SS and CS are as expected, then
movq 32(%rsp),%rsp;sysret does the same thing as iret.  This set of
conditions is nearly always satisfied on return from syscalls, and
it can even occasionally be satisfied on return from an irq.

Even with the careful checks for sysret applicability, this cuts
nearly 80ns off of the overhead from syscalls with unoptimized exit
work.  This includes tracing and context tracking, and any return
that invokes KVM's user return notifier.  For example, the cost of
getpid with CONFIG_CONTEXT_TRACKING_FORCE=y drops from ~360ns to
~280ns on my computer.

This may allow the removal and even eventual conversion to C
of a respectable amount of exit asm.

This may require further tweaking to give the full benefit on Xen.

It may be worthwhile to adjust signal delivery and exec to try hit
the sysret path.

This does not optimize returns to 32-bit userspace.  Making the same
optimization for CS == __USER32_CS is conceptually straightforward,
but it will require some tedious code to handle the differences
between sysretl and sysexitl.

Link: http://lkml.kernel.org/r/71428f63e681e1b4aa1a781e3ef7c27f027d1103.1421453410.git.luto@amacapital.net
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
9 years agox86, traps: Fix ist_enter from userspace
Andy Lutomirski [Sat, 31 Jan 2015 12:53:53 +0000 (04:53 -0800)]
x86, traps: Fix ist_enter from userspace

context_tracking_user_exit() has no effect if in_interrupt() returns true,
so ist_enter() didn't work.  Fix it by calling exception_enter(), and thus
context_tracking_user_exit(), before incrementing the preempt count.

This also adds an assertion that will catch the problem reliably if
CONFIG_PROVE_RCU=y to help prevent the bug from being reintroduced.

Link: http://lkml.kernel.org/r/261ebee6aee55a4724746d0d7024697013c40a08.1422709102.git.luto@amacapital.net
Fixes: 959274753857 x86, traps: Track entry into and exit from IST context
Reported-and-tested-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
9 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sat, 31 Jan 2015 18:34:25 +0000 (10:34 -0800)]
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "i2c driver bugfixes (s3c2410, slave-eeprom, sh_mobile), size
  regression "bugfix" (i2c slave), documentation bugfix (st).

  Also, one documentation update (da9063), so some devicetrees can now
  be verified"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: sh_mobile: terminate DMA reads properly
  i2c: Only include slave support if selected
  i2c: s3c2410: fix ABBA deadlock by keeping clock prepared
  i2c: slave-eeprom: fix boundary check when using sysfs
  i2c: st: Rename clock reference to something that exists
  DT: i2c: Add devices handled by the da9063 MFD driver

9 years agoMerge tag 'char-misc-3.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Sat, 31 Jan 2015 03:49:44 +0000 (19:49 -0800)]
Merge tag 'char-misc-3.19-rc7' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are two tiny patches, one fixing up the drivers/Kconfig file, and
  one adding a MAINTAINERS entry for the UIO git tree"

* tag 'char-misc-3.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  drivers/Kconfig: remove duplicate entry for soc
  MAINTAINERS: add git url entry for UIO

9 years agoMerge tag 'staging-3.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sat, 31 Jan 2015 03:44:56 +0000 (19:44 -0800)]
Merge tag 'staging-3.19-rc7' of git://git./linux/kernel/git/gregkh/staging

Pull staging tree fixes from Greg KH:
 "Here are two tiny staging tree fixes.  One for the nvec driver to
  resolve a reported problem, and one to add a MAINTAINERS entry for the
  Android drivers"

* tag 'staging-3.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  MAINTAINERS: add Android driver entries
  staging: nvec: specify a platform-device base id

9 years agoMerge tag 'usb-3.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sat, 31 Jan 2015 03:35:35 +0000 (19:35 -0800)]
Merge tag 'usb-3.19-rc7' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some small USB fixes and quirk additions for 3.19-rc7.

  All have been in linux-next for a while with no reported problems"

* tag 'usb-3.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  USB: Add OTG PET device to TPL
  usb-storage/SCSI: blacklist FUA on JMicron 152d:2566 USB-SATA controller
  uas: Add no-report-opcodes quirk for Simpletech devices with id 4971:8017
  storage: Revise/fix quirk for 04E6:000F SCM USB-SCSI converter
  usb: phy: never defer probe in non-OF case
  usb: dwc2: call dwc2_is_controller_alive() under spinlock

9 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Fri, 30 Jan 2015 22:34:55 +0000 (14:34 -0800)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Ingo Molnar:
 "Mostly tooling fixes, but also an event groups fix, two PMU driver
  fixes and a CPU model variant addition"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf: Tighten (and fix) the grouping condition
  perf/x86/intel: Add model number for Airmont
  perf/rapl: Fix crash in rapl_scale()
  perf/x86/intel/uncore: Move uncore_box_init() out of driver initialization
  perf probe: Fix probing kretprobes
  perf symbols: Introduce 'for' method to iterate over the symbols with a given name
  perf probe: Do not rely on map__load() filter to find symbols
  perf symbols: Introduce method to iterate symbols ordered by name
  perf symbols: Return the first entry with a given name in find_by_name method
  perf annotate: Fix memory leaks in LOCK handling
  perf annotate: Handle ins parsing failures
  perf scripting perl: Force to use stdbool
  perf evlist: Remove extraneous 'was' on error message

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Fri, 30 Jan 2015 22:25:52 +0000 (14:25 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fix from Chris Mason:
 "We have one more fix for btrfs in my for-linus branch - this was a bug
  in the new raid5/6 scrubbing support"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  btrfs: fix raid56 scrub failed in xfstests btrfs/072

9 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs
Linus Torvalds [Fri, 30 Jan 2015 21:46:04 +0000 (13:46 -0800)]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs

Pull quota and UDF fix from Jan Kara:
 "A fix for UDF to properly free preallocated blocks and a fix for quota
  so that Q_GETQUOTA quotactl reports correct numbers for XFS filesystem
  (and similarly Q_XGETQUOTA quotactl works properly for other
  filesystems)"

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  quota: Switch ->get_dqblk() and ->set_dqblk() to use bytes as space units
  udf: Release preallocation on last writeable close

9 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Fri, 30 Jan 2015 18:45:24 +0000 (10:45 -0800)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fixes from Paolo Bonzini:
 "The ARM changes are largish, but not too scary.  And a simple fix for
  x86 (bug introduced in 3.19)"

(Paolo sayus these are the "Final" fixes. We'll see).

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86: check LAPIC presence when building apic_map
  arm/arm64: KVM: Use kernel mapping to perform invalidation on page fault
  arm/arm64: KVM: Invalidate data cache on unmap
  arm/arm64: KVM: Use set/way op trapping to track the state of the caches

9 years agoMerge tag 'iommu-fixes-v3.19-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 30 Jan 2015 18:41:26 +0000 (10:41 -0800)]
Merge tag 'iommu-fixes-v3.19-rc6' of git://git./linux/kernel/git/joro/iommu

Pull IOMMU fixes from Joerg Roedel:
 "Two small fixes for the Tegra GART IOMMU driver:

   - provide a .map_sg function for iommu_ops
   - do not register Tegra GART driver as a workaround because of issues
     with it when used from DRM code"

* tag 'iommu-fixes-v3.19-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/tegra: gart: Provide default ->map_sg() callback
  iommu/tegra: gart: Do not register with bus

9 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 30 Jan 2015 18:34:24 +0000 (10:34 -0800)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull intel and dp mst drm fixes from Dave Airlie:
 "Intel had a few more fixes lined up and no point me sitting on them,
  along with a DP MST fix from Rob for a race at undock + vt switch"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm: fix fb-helper vs MST dangling connector ptrs (v2)
  drm/i915: BDW Fix Halo PCI IDs marked as ULT.
  drm/i915: Fix and clean BDW PCH identification
  drm/i915: Only fence tiled region of object.
  drm/i915: fix inconsistent brightness after resume
  drm/i915: Init PPGTT before context enable

9 years agoarc: mm: Fix build failure
Guenter Roeck [Fri, 30 Jan 2015 03:15:33 +0000 (19:15 -0800)]
arc: mm: Fix build failure

Fix misspelled define.

Fixes: 33692f27597f ("vm: add VM_FAULT_SIGSEGV handling support")
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoMerge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Fri, 30 Jan 2015 18:26:32 +0000 (19:26 +0100)]
Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/core

Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:

 User visible changes:

  - Show precise number of samples in at the end of a 'record' session, if
    processing build ids, since we will then traverse the whole perf.data file
    and see all the PERF_RECORD_SAMPLE records, otherwise stop showing the
    previous off-base heuristicly counted number of "samples"  (Namhyung Kim).

  - Support to read compressed module from build-id cache (Namhyung Kim)

 Infrastructure changes:

  - Cache eh/debug frame offset for dwarf unwind (Namhyung Kim)

  - Set header version correctly in all cases (Namhyung Kim)

  - Set attr.task bit for a tracking event, to be consistent (Namhyung Kim)
    perf tools: Use perf_data_file__fd() consistently
    perf symbols: Convert lseek + read to pread

  - Don't rely on malloc working for sz 0, fixing another problem when
    using uClibc (Vineet Gupta)

  - Provide stub for missing pthread_attr_setaffinity_np for libcs where this
    is not available, such as uClibc (Vineet Gupta)

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoi2c: sh_mobile: terminate DMA reads properly
Wolfram Sang [Thu, 29 Jan 2015 17:33:16 +0000 (18:33 +0100)]
i2c: sh_mobile: terminate DMA reads properly

DMA read requests could miss proper termination, so two more bytes would
have been read via PIO overwriting the end of the buffer with wrong
data. Make DMA stop handling more readable while we are here.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
9 years agoKVM: x86: check LAPIC presence when building apic_map
Radim Krčmář [Thu, 29 Jan 2015 21:33:35 +0000 (22:33 +0100)]
KVM: x86: check LAPIC presence when building apic_map

We forgot to re-check LAPIC after splitting the loop in commit
173beedc1601 (KVM: x86: Software disabled APIC should still deliver
NMIs, 2014-11-02).

Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
Fixes: 173beedc1601f51dae9d579aa7a414c5aa8f700b
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoMerge tag 'kvm-arm-fixes-3.19-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Paolo Bonzini [Fri, 30 Jan 2015 08:41:38 +0000 (09:41 +0100)]
Merge tag 'kvm-arm-fixes-3.19-2' of git://git./linux/kernel/git/kvmarm/kvmarm into kvm-master

Second round of fixes for KVM/ARM for 3.19.

Fixes memory corruption issues on APM platforms and swapping issues on
DMA-coherent systems.

9 years agoMerge tag 'drm-intel-fixes-2015-01-29' of git://anongit.freedesktop.org/drm-intel...
Dave Airlie [Fri, 30 Jan 2015 03:32:24 +0000 (13:32 +1000)]
Merge tag 'drm-intel-fixes-2015-01-29' of git://anongit.freedesktop.org/drm-intel into drm-fixes

misc i915 fixes, mostly all stable material as well.

* tag 'drm-intel-fixes-2015-01-29' of git://anongit.freedesktop.org/drm-intel:
  drm/i915: BDW Fix Halo PCI IDs marked as ULT.
  drm/i915: Fix and clean BDW PCH identification
  drm/i915: Only fence tiled region of object.
  drm/i915: fix inconsistent brightness after resume
  drm/i915: Init PPGTT before context enable

9 years agodrm: fix fb-helper vs MST dangling connector ptrs (v2)
Rob Clark [Mon, 26 Jan 2015 15:11:08 +0000 (10:11 -0500)]
drm: fix fb-helper vs MST dangling connector ptrs (v2)

VT switch back/forth from console to xserver (for example) has potential
to go horribly wrong if a dynamic DP MST connector ends up in the saved
modeset that is restored when switching back to fbcon.

When removing a dynamic connector, don't forget to clean up the saved
state.

v1: original
v2: null out set->fb if no more connectors to avoid making i915 cranky

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1184968
Cc: stable@vger.kernel.org #v3.17+
Signed-off-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoMerge tag 'dm-3.19-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
Linus Torvalds [Thu, 29 Jan 2015 23:41:27 +0000 (15:41 -0800)]
Merge tag 'dm-3.19-fixes-3' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:
 "One stable fix for a dm-cache 3.19-rc6 regression and one stable fix
  for dm-thin:

   - fix DM cache metadata open/lookup error paths to properly use
     ERR_PTR and IS_ERR (fixes: 3.19-rc6 "stable" commit 9b1cc9f251)

   - fix DM thin-provisioning to disallow userspace from sending
     messages to the thin-pool if the pool is in READ_ONLY or FAIL mode
     since no metadata changes are allowed in these modes"

* tag 'dm-3.19-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm thin: don't allow messages to be sent to a pool target in READ_ONLY or FAIL mode
  dm cache: fix missing ERR_PTR returns and handling

9 years agoMerge tag 'nfs-for-3.19-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Linus Torvalds [Thu, 29 Jan 2015 23:18:12 +0000 (15:18 -0800)]
Merge tag 'nfs-for-3.19-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client bugfixes from Trond Myklebust:
 "Highlights include:

   - Stable fix for a NFSv4.1 Oops on mount
   - Stable fix for an O_DIRECT deadlock condition
   - Fix an issue with submounted volumes and fake duplicate inode
     numbers"

* tag 'nfs-for-3.19-3' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  NFS: Fix use of nfs_attr_use_mounted_on_fileid()
  NFSv4.1: Fix an Oops in nfs41_walk_client_list
  nfs: fix dio deadlock when O_DIRECT flag is flipped

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Thu, 29 Jan 2015 23:13:56 +0000 (15:13 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

Pull Ceph fixes from Sage Weil:
 "These paches from Ilya finally squash a race condition with layered
  images that he's been chasing for a while"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: drop parent_ref in rbd_dev_unprobe() unconditionally
  rbd: fix rbd_dev_parent_get() when parent_overlap == 0

9 years agoarm/arm64: KVM: Use kernel mapping to perform invalidation on page fault
Marc Zyngier [Mon, 5 Jan 2015 21:13:24 +0000 (21:13 +0000)]
arm/arm64: KVM: Use kernel mapping to perform invalidation on page fault

When handling a fault in stage-2, we need to resync I$ and D$, just
to be sure we don't leave any old cache line behind.

That's very good, except that we do so using the *user* address.
Under heavy load (swapping like crazy), we may end up in a situation
where the page gets mapped in stage-2 while being unmapped from
userspace by another CPU.

At that point, the DC/IC instructions can generate a fault, which
we handle with kvm->mmu_lock held. The box quickly deadlocks, user
is unhappy.

Instead, perform this invalidation through the kernel mapping,
which is guaranteed to be present. The box is much happier, and so
am I.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
9 years agoarm/arm64: KVM: Invalidate data cache on unmap
Marc Zyngier [Fri, 19 Dec 2014 16:48:06 +0000 (16:48 +0000)]
arm/arm64: KVM: Invalidate data cache on unmap

Let's assume a guest has created an uncached mapping, and written
to that page. Let's also assume that the host uses a cache-coherent
IO subsystem. Let's finally assume that the host is under memory
pressure and starts to swap things out.

Before this "uncached" page is evicted, we need to make sure
we invalidate potential speculated, clean cache lines that are
sitting there, or the IO subsystem is going to swap out the
cached view, loosing the data that has been written directly
into memory.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
9 years agoarm/arm64: KVM: Use set/way op trapping to track the state of the caches
Marc Zyngier [Fri, 19 Dec 2014 16:05:31 +0000 (16:05 +0000)]
arm/arm64: KVM: Use set/way op trapping to track the state of the caches

Trying to emulate the behaviour of set/way cache ops is fairly
pointless, as there are too many ways we can end-up missing stuff.
Also, there is some system caches out there that simply ignore
set/way operations.

So instead of trying to implement them, let's convert it to VA ops,
and use them as a way to re-enable the trapping of VM ops. That way,
we can detect the point when the MMU/caches are turned off, and do
a full VM flush (which is what the guest was trying to do anyway).

This allows a 32bit zImage to boot on the APM thingy, and will
probably help bootloaders in general.

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
9 years agoperf symbols: Convert lseek + read to pread
Namhyung Kim [Thu, 29 Jan 2015 20:02:01 +0000 (17:02 -0300)]
perf symbols: Convert lseek + read to pread

When dso_cache__read() is called, it reads data from the given offset
using lseek + normal read syscall.  It can be combined to a single pread
syscall.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1422518843-25818-40-git-send-email-namhyung@kernel.org
[ Fixed it up when cherry picking it from the multi threaded patchkit ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Use perf_data_file__fd() consistently
Namhyung Kim [Thu, 29 Jan 2015 08:06:48 +0000 (17:06 +0900)]
perf tools: Use perf_data_file__fd() consistently

Do not reference file->fd directly since we want hide the
implementation details from outside for possible future changes.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1422518843-25818-8-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf symbols: Support to read compressed module from build-id cache
Namhyung Kim [Thu, 29 Jan 2015 08:06:42 +0000 (17:06 +0900)]
perf symbols: Support to read compressed module from build-id cache

The commit c00c48fc6e6e ("perf symbols: Preparation for compressed
kernel module support") added support for compressed kernel modules but
it only supports system path DSOs.  When a dso is read from build-id
cache, its filename doesn't end with ".gz" but has build-id.  In this
case, we should fallback to the original dso->name.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1422518843-25818-2-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf evsel: Set attr.task bit for a tracking event
Namhyung Kim [Thu, 29 Jan 2015 08:06:46 +0000 (17:06 +0900)]
perf evsel: Set attr.task bit for a tracking event

The perf_event_attr.task bit is to track task (fork and exit) events but
it missed to be set by perf_evsel__config().  While it was not a problem
in practice since setting other bits (comm/mmap) ended up being in same
result, it'd be good to set it explicitly anyway.

The attr->task is to track task related events (fork/exit) only but
other meta events like comm and mmap[2] also needs the task events.  So
setting attr->comm and/or attr->mmap causes the kernel emits the task
events anyway.  So the attr->task is only meaningful when other bits are
off but I'd like to set it for completeness.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1422518843-25818-6-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf header: Set header version correctly
Namhyung Kim [Thu, 29 Jan 2015 08:06:45 +0000 (17:06 +0900)]
perf header: Set header version correctly

When check_magic_endian() is called, it checks the magic number in the
perf data file to determine version and endianness.  But if it uses a
same endian the verison number wasn't updated and makes confusion.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1422518843-25818-5-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoMerge tag 'sound-3.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Thu, 29 Jan 2015 19:48:36 +0000 (11:48 -0800)]
Merge tag 'sound-3.19-rc7' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "This batch ended up being larger than wished, but there is nothing to
  worry too much there.

  Most of commits are for ASoC, a compress NULL dereference fix, a fix
  for probe error handling, and the rest are device-specific fixes.  In
  addition, we have a fix for a long-standing but of seq-dummy driver,
  which just cuts off the buggy part in the end"

* tag 'sound-3.19-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: seq-dummy: remove deadlock-causing events on close
  ASoC: omap-mcbsp: Correct CBM_CFS dai format configuration
  ASoC: soc-compress.c: fix NULL dereference
  ASoC: rt286: set the same format for dac and adc
  ASoC: wm8904: fix runtime warning
  ASoC: simple-card: Fix crash in asoc_simple_card_unref()
  ASoC: fsl: imx-wm8962: Set the card owner field
  ASoC: pcm512x: Fix DSP program selection
  ASoC: rt5677: Modify the behavior that updates the PLL parameter.
  ASoC: fsl_ssi: Fix irq error check
  ASoC: rockchip: i2s: applys rate symmetry for CPU DAI
  ASoC: Intel: Add NULL checks for the stream pointer
  ASoC: wm8960: Fix capture sample rate from 11250 to 11025
  ASoC: adi: Add missing return statement.
  ASoC: Intel: Don't change offset of block allocator during fixed allocate
  ASoC: ts3a227e: Check and report jack status at probe
  ASoC: fsl_esai: Fix incorrect xDC field width of xCCR registers

9 years agoMerge tag 'pinctrl-v3.19-4' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Thu, 29 Jan 2015 19:43:43 +0000 (11:43 -0800)]
Merge tag 'pinctrl-v3.19-4' of git://git./linux/kernel/git/linusw/linux-pinctrl

Pull final pin control fix from Linus Walleij:
 "A late pin control fix for the v3.19 series: The AT91 gpio controller
  would miss wakeup events, this single fix make it work properly"

[ "Final"? Yeah, I'll believe that once I've actually released 3.19 ;)   - Linus ]

* tag 'pinctrl-v3.19-4' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: at91: allow to have disabled gpio bank

9 years agoperf record: Show precise number of samples
Namhyung Kim [Thu, 29 Jan 2015 08:06:44 +0000 (17:06 +0900)]
perf record: Show precise number of samples

After perf record finishes, it prints file size and number of samples in
the file but this info is wrong since it assumes typical sample size of
24 bytes and divides file size by the value.

However as we post-process recorded samples for build-id, it can show
correct number like below.  If build-id post-processing is not requested
just omit the wrong number of samples.

  $ perf record noploop 1
    [ perf record: Woken up 1 times to write data ]
    [ perf record: Captured and wrote 0.159 MB perf.data (3989 samples) ]

  $ perf report --stdio -n
  # To display the perf.data header info, please use --header/--header-only options.
  #
  # Samples: 3K of event 'cycles'
  # Event count (approx.): 3771330663
  #
  # Overhead       Samples  Command  Shared Object     Symbol
  # ........  ............  .......  ................  ..........................
  #
      99.90%          3982  noploop  noploop           [.] main
       0.09%             1  noploop  ld-2.17.so        [.] _dl_check_map_versions
       0.01%             1  noploop  [kernel.vmlinux]  [k] setup_arg_pages
       0.00%             5  noploop  [kernel.vmlinux]  [k] intel_pmu_enable_all

Reported-by: Milian Wolff <mail@milianw.de>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1422518843-25818-4-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf tools: Do not use __perf_session__process_events() directly
Namhyung Kim [Thu, 29 Jan 2015 08:06:43 +0000 (17:06 +0900)]
perf tools: Do not use __perf_session__process_events() directly

It's only used for perf record to process build-id because its file size
it's not fixed at this time due to remaining header features.

However data offset and size is available so that we can use the
perf_session__process_events() once we set the file size as the current
offset like for now.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1422518843-25818-3-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf callchain: Cache eh/debug frame offset for dwarf unwind
Namhyung Kim [Thu, 29 Jan 2015 08:07:21 +0000 (17:07 +0900)]
perf callchain: Cache eh/debug frame offset for dwarf unwind

When libunwind tries to resolve callchains it needs to know the offset
of .eh_frame_hdr or .debug_frame to access the dso.

Since it will always return the same result for a given DSO, just cache
the result as an optimization.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1422518843-25818-41-git-send-email-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agovm: make stack guard page errors return VM_FAULT_SIGSEGV rather than SIGBUS
Linus Torvalds [Thu, 29 Jan 2015 19:15:17 +0000 (11:15 -0800)]
vm: make stack guard page errors return VM_FAULT_SIGSEGV rather than SIGBUS

The stack guard page error case has long incorrectly caused a SIGBUS
rather than a SIGSEGV, but nobody actually noticed until commit
fee7e49d4514 ("mm: propagate error from stack expansion even for guard
page") because that error case was never actually triggered in any
normal situations.

Now that we actually report the error, people noticed the wrong signal
that resulted.  So far, only the test suite of libsigsegv seems to have
actually cared, but there are real applications that use libsigsegv, so
let's not wait for any of those to break.

Reported-and-tested-by: Takashi Iwai <tiwai@suse.de>
Tested-by: Jan Engelhardt <jengelh@inai.de>
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com> # "s390 still compiles and boots"
Cc: linux-arch@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoarm: dma-mapping: Set DMA IOMMU ops in arm_iommu_attach_device()
Laurent Pinchart [Fri, 23 Jan 2015 14:21:49 +0000 (16:21 +0200)]
arm: dma-mapping: Set DMA IOMMU ops in arm_iommu_attach_device()

Commit 4bb25789ed28228a ("arm: dma-mapping: plumb our iommu mapping ops
into arch_setup_dma_ops") moved the setting of the DMA operations from
arm_iommu_attach_device() to arch_setup_dma_ops() where the DMA
operations to be used are selected based on whether the device is
connected to an IOMMU. However, the IOMMU detection scheme requires the
IOMMU driver to be ported to the new IOMMU of_xlate API. As no driver
has been ported yet, this effectively breaks all IOMMU ARM users that
depend on the IOMMU being handled transparently by the DMA mapping API.

Fix this by restoring the setting of DMA IOMMU ops in
arm_iommu_attach_device() and splitting the rest of the function into a
new internal __arm_iommu_attach_device() function, called by
arch_setup_dma_ops().

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Tested-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agovm: add VM_FAULT_SIGSEGV handling support
Linus Torvalds [Thu, 29 Jan 2015 18:51:32 +0000 (10:51 -0800)]
vm: add VM_FAULT_SIGSEGV handling support

The core VM already knows about VM_FAULT_SIGBUS, but cannot return a
"you should SIGSEGV" error, because the SIGSEGV case was generally
handled by the caller - usually the architecture fault handler.

That results in lots of duplication - all the architecture fault
handlers end up doing very similar "look up vma, check permissions, do
retries etc" - but it generally works.  However, there are cases where
the VM actually wants to SIGSEGV, and applications _expect_ SIGSEGV.

In particular, when accessing the stack guard page, libsigsegv expects a
SIGSEGV.  And it usually got one, because the stack growth is handled by
that duplicated architecture fault handler.

However, when the generic VM layer started propagating the error return
from the stack expansion in commit fee7e49d4514 ("mm: propagate error
from stack expansion even for guard page"), that now exposed the
existing VM_FAULT_SIGBUS result to user space.  And user space really
expected SIGSEGV, not SIGBUS.

To fix that case, we need to add a VM_FAULT_SIGSEGV, and teach all those
duplicate architecture fault handlers about it.  They all already have
the code to handle SIGSEGV, so it's about just tying that new return
value to the existing code, but it's all a bit annoying.

This is the mindless minimal patch to do this.  A more extensive patch
would be to try to gather up the mostly shared fault handling logic into
one generic helper routine, and long-term we really should do that
cleanup.

Just from this patch, you can generally see that most architectures just
copied (directly or indirectly) the old x86 way of doing things, but in
the meantime that original x86 model has been improved to hold the VM
semaphore for shorter times etc and to handle VM_FAULT_RETRY and other
"newer" things, so it would be a good idea to bring all those
improvements to the generic case and teach other architectures about
them too.

Reported-and-tested-by: Takashi Iwai <tiwai@suse.de>
Tested-by: Jan Engelhardt <jengelh@inai.de>
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com> # "s390 still compiles and boots"
Cc: linux-arch@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoARM: shmobile: r8a7790: Instantiate GIC from C board code in legacy builds
Magnus Damm [Thu, 29 Jan 2015 07:25:32 +0000 (16:25 +0900)]
ARM: shmobile: r8a7790: Instantiate GIC from C board code in legacy builds

As of commit 9a1091ef0017c40a ("irqchip: gic: Support hierarchy irq
domain."), the Lager legacy board support is known to be broken.

The IRQ numbers of the GIC are now virtual, and no longer match the
hardcoded hardware IRQ numbers in the legacy platform board code.

To fix this issue specific to non-multiplatform r8a7790 and Lager:
 1) Instantiate the GIC from platform board code and also
 2) Skip over the DT arch timer as well as
 3) Force delay setup based on DT CPU frequency

With these 3 fixes in place interrupts on Lager are now unbroken.

Partially based on legacy GIC fix by Geert Uytterhoeven, thanks to
him for the initial work.

Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
9 years agox86, vdso: teach 'make clean' remove vdso64 binaries
Andrey Skvortsov [Wed, 28 Jan 2015 14:04:27 +0000 (17:04 +0300)]
x86, vdso: teach 'make clean' remove vdso64 binaries

After 'make clean' vdso64.so and vdso64.dbg.so were left in arch/x86/vdso/.

Link: http://lkml.kernel.org/r/1422453867-17326-1-git-send-email-andrej.skvortzov@gmail.com
Signed-off-by: Andrey Skvortsov <andrej.skvortzov@gmail.com>
Signed-off-by: Andy Lutomirski <luto@amacapital.net>
9 years agoARM: shmobile: r8a73a4: Instantiate GIC from C board code in legacy builds
Magnus Damm [Wed, 28 Jan 2015 12:46:45 +0000 (21:46 +0900)]
ARM: shmobile: r8a73a4: Instantiate GIC from C board code in legacy builds

As of commit 9a1091ef0017c40a ("irqchip: gic: Support hierarchy irq
domain."), the APE6EVM legacy board support is known to be broken.

The IRQ numbers of the GIC are now virtual, and no longer match the
hardcoded hardware IRQ numbers in the legacy platform board code.

To fix this issue specific to non-muliplatform r8a73a4 and APE6EVM:
 1) Instantiate the GIC from platform board code and also
 2) Skip over the DT arch timer as well as
 3) Force delay setup based on DT CPU frequency

With these 3 fixes in place interrupts on APE6EVM are now unbroken.

Partially based on legacy GIC fix by Geert Uytterhoeven, thanks to
him for the initial work.

Signed-off-by: Magnus Damm <damm+renesas@opensource.se>
Acked-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
9 years agoMerge tag 'mvebu-fixes-3.19-6' of git://git.infradead.org/linux-mvebu into fixes
Olof Johansson [Wed, 28 Jan 2015 22:59:33 +0000 (14:59 -0800)]
Merge tag 'mvebu-fixes-3.19-6' of git://git.infradead.org/linux-mvebu into fixes

Merge "mvebu-fixes-6" from Andrew Lunn:

The previous fix for Armada XP, disabling I/O coherency, broke Armada
375/38x.  Only switch the PL310 to I/O coherent mode if I/O coherency
is enabled.

* tag 'mvebu-fixes-3.19-6' of git://git.infradead.org/linux-mvebu:
  ARM: mvebu: don't set the PL310 in I/O coherency mode when I/O coherency is disabled

Signed-off-by: Olof Johansson <olof@lixom.net>
9 years agoARM: mvebu: don't set the PL310 in I/O coherency mode when I/O coherency is disabled
Thomas Petazzoni [Wed, 28 Jan 2015 11:55:45 +0000 (12:55 +0100)]
ARM: mvebu: don't set the PL310 in I/O coherency mode when I/O coherency is disabled

Since commit f2c3c67f00 (merge commit that adds commit "ARM: mvebu:
completely disable hardware I/O coherency"), we disable I/O coherency
on Armada EBU platforms.

However, we continue to initialize the coherency fabric, because this
coherency fabric is needed on Armada XP for inter-CPU
coherency. Unfortunately, due to this, we also continued to execute
the coherency fabric initialization code for Armada 375/38x, which
switched the PL310 into I/O coherent mode. This has the effect of
disabling the outer cache sync operation: this is needed when I/O
coherency is enabled to work around a PCIe/L2 deadlock. But obviously,
when I/O coherency is disabled, having the outer cache sync operation
is crucial.

Therefore, this commit fixes the armada_375_380_coherency_init() so
that the PL310 is switched to I/O coherent mode only if I/O coherency
is enabled.

Without this fix, all devices using DMA are broken on Armada 375/38x.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Acked-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Tested-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Cc: <stable@vger.kernel.org> # v3.8+
9 years agoperf tools: Provide stub for missing pthread_attr_setaffinity_np
Vineet Gupta [Tue, 13 Jan 2015 13:43:24 +0000 (19:13 +0530)]
perf tools: Provide stub for missing pthread_attr_setaffinity_np

uClibc Linuxthreads.old doesn't support the pthread_attr_setaffinity_np()
functioo:

   ----------------->8-----------------------
  CC       bench/futex-hash.o
  CC       bench/futex-wake.o
bench/futex-hash.c: In function 'bench_futex_hash':
bench/futex-hash.c:161:3: error: implicit declaration of function
'pthread_attr_setaffinity_np' [-Werror=implicit-function-declaration]
   ret = pthread_attr_setaffinity_np(&thread_attr, sizeof(cpu_set_t),
&cpu);
   ^
bench/futex-hash.c:161:3: error: nested extern declaration of
'pthread_attr_setaffinity_np' [-Werror=nested-externs]
   ----------------->8-----------------------

So introduce a test to check that and if not available provide a stub.

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1421156604-30603-6-git-send-email-vgupta@synopsys.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agoperf evsel: Don't rely on malloc working for sz 0
Vineet Gupta [Tue, 13 Jan 2015 13:43:23 +0000 (19:13 +0530)]
perf evsel: Don't rely on malloc working for sz 0

When running perf on ARC (uClibc based userspace), ran into this issue
   ------------->8----------------
[ARCLinux]$ ./perf record ls
bin             etc             perf            sys
debug           init            perf.data       tmp
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.001 MB perf.data (~24 samples) ]

[ARCLinux]$ ./perf report
incompatible file format (rerun with -v to learn more)
   ------------->8----------------

The problem happens in the following call stack when zalloc is called
with size zero

glibc default / uClibc with MALLOC_GLIBC_COMPAT are OK, but not if that
config option is not enabled.

  cmd_report
     perf_session__new
perf_session__open
    perf_session__read_header
read_attr(fd, header, &f_attr)
nr_ids = f_attr.ids.size / sizeof(u64); <-- 0
perf_evsel__alloc_id(vsel, 1, nr_ids)
zalloc(ncpus * nthreads * sizeof(u64)) <-- 0

header.c: read_attr()

(gdb) p *f_attr
$17 = {
  attr = {
    type = 0,
    size = 96,
    config = 0,
    {
      sample_period = 4000,
      sample_freq = 4000
    },
...
  ids = {
    offset = 104,
    size = 0      <------
  }
}

Signed-off-by: Vineet Gupta <vgupta@synopsys.com>
Suggested-by: Namhyung Kim <namhyung@kernel.org>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexey Brodkin <Alexey.Brodkin@synopsys.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1421156604-30603-5-git-send-email-vgupta@synopsys.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
9 years agodm thin: don't allow messages to be sent to a pool target in READ_ONLY or FAIL mode
Joe Thornber [Mon, 26 Jan 2015 11:38:21 +0000 (11:38 +0000)]
dm thin: don't allow messages to be sent to a pool target in READ_ONLY or FAIL mode

You can't modify the metadata in these modes.  It's better to fail these
messages immediately than let the block-manager deny write locks on
metadata blocks.  Otherwise these failed metadata changes will trigger
'needs_check' to get set in the metadata superblock -- requiring repair
using the thin_check utility.

Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
9 years agodm cache: fix missing ERR_PTR returns and handling
Joe Thornber [Wed, 28 Jan 2015 12:07:46 +0000 (12:07 +0000)]
dm cache: fix missing ERR_PTR returns and handling

Commit 9b1cc9f251 ("dm cache: share cache-metadata object across
inactive and active DM tables") mistakenly ignored the use of ERR_PTR
returns.  Restore missing IS_ERR checks and ERR_PTR returns where
appropriate.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Joe Thornber <ejt@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Cc: stable@vger.kernel.org
9 years agoMerge branch 'perf/hw_breakpoints' into perf/core
Ingo Molnar [Wed, 28 Jan 2015 14:48:59 +0000 (15:48 +0100)]
Merge branch 'perf/hw_breakpoints' into perf/core

The new hw_breakpoint bits are now ready for v3.20, merge them
into the main branch, to avoid conflicts.

Conflicts:
tools/perf/Documentation/perf-record.txt

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Wed, 28 Jan 2015 14:43:15 +0000 (15:43 +0100)]
Merge tag 'perf-core-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/core

Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo:

"
   User visible changes:

   - Enable sampling loads and stores simultaneously in 'perf mem' (Stephane Eranian)

   - 'perf diff' output improvements (Namhyung Kim)

   - Fix error reporting for evsel pgfault constructor (Arnaldo Carvalho de Melo)

   Infrastructure changes:

   - Move debugfs sterrno like method to tools/lib/ so that it may be used by
     other tools, as 'perf probe' will be soon (Arnaldo Carvalho de Melo)

   - Introduce function fro deleting/removing hist_entry to avoid code duplication
     (Arnaldo Carvalho de Melo)

   - Support parsing parameterized events (Cody P Schafer)

   - Add support for IP address formats in libtraceevent (David Ahern)

   - Fix typo in sample-parsing.c 'perf test' entry (Rasmus Villemoes)

   - Remove some unused functions from color.c (Rickard Strandqvist)
"

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge branch 'perf/urgent' into perf/core, to pick up fixes
Ingo Molnar [Wed, 28 Jan 2015 14:42:56 +0000 (15:42 +0100)]
Merge branch 'perf/urgent' into perf/core, to pick up fixes

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'perf-urgent-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Wed, 28 Jan 2015 14:40:30 +0000 (15:40 +0100)]
Merge tag 'perf-urgent-for-mingo' of git://git./linux/kernel/git/acme/linux into perf/urgent

Pull perf/urgent fixes from Arnaldo Carvalho de Melo:

" User visible fixes:

  - Fix probing at function return (Namhyumg Kim)

  Developer visible fixes:

  - Symbol processing changes necessary for fixing support for
    kretprobes in 'perf probe' (Namhyung Kim, Arnaldo Carvalho de Melo)

  - Annotation memory leaks and instruction parsing fixes (Rabin Vincent)

  - Fix perl build on ARM64 (Wang Nam)
"

Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge tag 'pr-20150114-x86-entry' of git://git.kernel.org/pub/scm/linux/kernel/git...
Ingo Molnar [Wed, 28 Jan 2015 14:32:03 +0000 (15:32 +0100)]
Merge tag 'pr-20150114-x86-entry' of git://git./linux/kernel/git/luto/linux into x86/asm

Pull x86/entry enhancements from Andy Lutomirski:

" This is my accumulated x86 entry work, part 1, for 3.20.  The meat
  of this is an IST rework.  When an IST exception interrupts user
  space, we will handle it on the per-thread kernel stack instead of
  on the IST stack.  This sounds messy, but it actually simplifies the
  IST entry/exit code, because it eliminates some ugly games we used
  to play in order to handle rescheduling, signal delivery, etc on the
  way out of an IST exception.

  The IST rework introduces proper context tracking to IST exception
  handlers.  I haven't seen any bug reports, but the old code could
  have incorrectly treated an IST exception handler as an RCU extended
  quiescent state.

  The memory failure change (included in this pull request with
  Borislav and Tony's permission) eliminates a bunch of code that
  is no longer needed now that user memory failure handlers are
  called in process context.

  Finally, this includes a few on Denys' uncontroversial and Obviously
  Correct (tm) cleanups.

  The IST and memory failure changes have been in -next for a while.

  LKML references:

  IST rework:
  http://lkml.kernel.org/r/cover.1416604491.git.luto@amacapital.net

  Memory failure change:
  http://lkml.kernel.org/r/54ab2ffa301102cd6e@agluck-desk.sc.intel.com

  Denys' cleanups:
  http://lkml.kernel.org/r/1420927210-19738-1-git-send-email-dvlasenk@redhat.com
"

This tree semantically depends on and is based on the following RCU commit:

  734d16801349 ("rcu: Make rcu_nmi_enter() handle nesting")

... and for that reason won't be pushed upstream before the RCU bits hit Linus's tree.

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoMerge commit 3669ef9fa7d3 ("x86, tls: Interpret an all-zero struct user_desc as ...
Ingo Molnar [Wed, 28 Jan 2015 14:30:32 +0000 (15:30 +0100)]
Merge commit 3669ef9fa7d3 ("x86, tls: Interpret an all-zero struct user_desc as 'no segment'") into x86/asm

Pick up the latestest asm fixes before advancing it any further.

Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agorbd: drop parent_ref in rbd_dev_unprobe() unconditionally
Ilya Dryomov [Mon, 19 Jan 2015 09:06:14 +0000 (12:06 +0300)]
rbd: drop parent_ref in rbd_dev_unprobe() unconditionally

This effectively reverts the last hunk of 392a9dad7e77 ("rbd: detect
when clone image is flattened").

The problem with parent_overlap != 0 condition is that it's possible
and completely valid to have an image with parent_overlap == 0 whose
parent state needs to be cleaned up on unmap.  The next commit, which
drops the "clone image now standalone" logic, opens up another window
of opportunity to hit this, but even without it

    # cat parent-ref.sh
    #!/bin/bash
    rbd create --image-format 2 --size 1 foo
    rbd snap create foo@snap
    rbd snap protect foo@snap
    rbd clone foo@snap bar
    rbd resize --allow-shrink --size 0 bar
    rbd resize --size 1 bar
    DEV=$(rbd map bar)
    rbd unmap $DEV

leaves rbd_device/rbd_spec/etc and rbd_client along with ceph_client
hanging around.

My thinking behind calling rbd_dev_parent_put() unconditionally is that
there shouldn't be any requests in flight at that point in time as we
are deep into unmap sequence.  Hence, even if rbd_dev_unparent() caused
by flatten is delayed by in-flight requests, it will have finished by
the time we reach rbd_dev_unprobe() caused by unmap, thus turning
unconditional rbd_dev_parent_put() into a no-op.

Fixes: http://tracker.ceph.com/issues/10352
Cc: stable@vger.kernel.org # 3.11+
Signed-off-by: Ilya Dryomov <idryomov@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Alex Elder <elder@linaro.org>
9 years agorbd: fix rbd_dev_parent_get() when parent_overlap == 0
Ilya Dryomov [Mon, 19 Jan 2015 15:13:43 +0000 (18:13 +0300)]
rbd: fix rbd_dev_parent_get() when parent_overlap == 0

The comment for rbd_dev_parent_get() said

    * We must get the reference before checking for the overlap to
    * coordinate properly with zeroing the parent overlap in
    * rbd_dev_v2_parent_info() when an image gets flattened.  We
    * drop it again if there is no overlap.

but the "drop it again if there is no overlap" part was missing from
the implementation.  This lead to absurd parent_ref values for images
with parent_overlap == 0, as parent_ref was incremented for each
img_request and virtually never decremented.

Fix this by leveraging the fact that refresh path calls
rbd_dev_v2_parent_info() under header_rwsem and use it for read in
rbd_dev_parent_get(), instead of messing around with atomics.  Get rid
of barriers in rbd_dev_v2_parent_info() while at it - I don't see what
they'd pair with now and I suspect we are in a pretty miserable
situation as far as proper locking goes regardless.

Cc: stable@vger.kernel.org # 3.11+
Signed-off-by: Ilya Dryomov <idryomov@redhat.com>
Reviewed-by: Josh Durgin <jdurgin@redhat.com>
Reviewed-by: Alex Elder <elder@linaro.org>
9 years agoperf: Tighten (and fix) the grouping condition
Peter Zijlstra [Fri, 23 Jan 2015 10:19:48 +0000 (11:19 +0100)]
perf: Tighten (and fix) the grouping condition

The fix from 9fc81d87420d ("perf: Fix events installation during
moving group") was incomplete in that it failed to recognise that
creating a group with events for different CPUs is semantically
broken -- they cannot be co-scheduled.

Furthermore, it leads to real breakage where, when we create an event
for CPU Y and then migrate it to form a group on CPU X, the code gets
confused where the counter is programmed -- triggered in practice
as well by me via the perf fuzzer.

Fix this by tightening the rules for creating groups. Only allow
grouping of counters that can be co-scheduled in the same context.
This means for the same task and/or the same cpu.

Fixes: 9fc81d87420d ("perf: Fix events installation during moving group")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20150123125834.090683288@infradead.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86/intel: Add model number for Airmont
Kan Liang [Thu, 22 Jan 2015 07:50:53 +0000 (07:50 +0000)]
perf/x86/intel: Add model number for Airmont

Intel Airmont supports the same architectural and non-architectural
performance monitoring events as Silvermont.

Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/1421913053-99803-1-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/rapl: Fix crash in rapl_scale()
Stephane Eranian [Thu, 22 Jan 2015 20:38:34 +0000 (21:38 +0100)]
perf/rapl: Fix crash in rapl_scale()

This patch fixes a systematic crash in rapl_scale()
due to an invalid pointer.

The bug was introduced by commit:

  89cbc76768c2 ("x86: Replace __get_cpu_var uses")

The fix is simple. Just put the parenthesis where it needs
to be, i.e., around rapl_pmu. To my surprise, the compiler
was not complaining about passing an integer instead of a
pointer.

Reported-by: Vince Weaver <vincent.weaver@maine.edu>
Tested-by: Vince Weaver <vincent.weaver@maine.edu>
Fixes: 89cbc76768c2 ("x86: Replace __get_cpu_var uses")
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: cl@linux.com
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20150122203834.GA10228@thinkpad
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoperf/x86/intel/uncore: Move uncore_box_init() out of driver initialization
Kan Liang [Tue, 20 Jan 2015 04:54:25 +0000 (04:54 +0000)]
perf/x86/intel/uncore: Move uncore_box_init() out of driver initialization

There were some issues about the uncore driver tried to access
non-existing boxes, which caused boot crashes. These issues have
been all fixed. But we should avoid boot failures if that ever
happens again.

This patch intends to prevent this kind of potential issues.
It moves uncore_box_init out of driver initialization. The box
will be initialized when it's first enabled.

Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1421729665-5912-1-git-send-email-kan.liang@intel.com
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
9 years agoquota: Switch ->get_dqblk() and ->set_dqblk() to use bytes as space units
Jan Kara [Thu, 9 Oct 2014 14:03:13 +0000 (16:03 +0200)]
quota: Switch ->get_dqblk() and ->set_dqblk() to use bytes as space units

Currently ->get_dqblk() and ->set_dqblk() use struct fs_disk_quota which
tracks space limits and usage in 512-byte blocks. However VFS quotas
track usage in bytes (as some filesystems require that) and we need to
somehow pass this information. Upto now it wasn't a problem because we
didn't do any unit conversion (thus VFS quota routines happily stuck
number of bytes into d_bcount field of struct fd_disk_quota). Only if
you tried to use Q_XGETQUOTA or Q_XSETQLIM for VFS quotas (or Q_GETQUOTA
/ Q_SETQUOTA for XFS quotas), you got bogus results. Hardly anyone
tried this but reportedly some Samba users hit the problem in practice.
So when we want interfaces compatible we need to fix this.

We bite the bullet and define another quota structure used for passing
information from/to ->get_dqblk()/->set_dqblk. It's somewhat sad we have
to have more conversion routines in fs/quota/quota.c and another copying
of quota structure slows down getting of quota information by about 2%
but it seems cleaner than overloading e.g. units of d_bcount to bytes.

CC: stable@vger.kernel.org
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
9 years agoudf: Release preallocation on last writeable close
Jan Kara [Wed, 28 Jan 2015 07:38:20 +0000 (08:38 +0100)]
udf: Release preallocation on last writeable close

Commit 6fb1ca92a640 "udf: Fix race between write(2) and close(2)"
changed the condition when preallocation is released. The idea was that
we don't want to release the preallocation for an inode on close when
there are other writeable file descriptors for the inode. However the
condition was written in the opposite way so we released preallocation
only if there were other writeable file descriptors. Fix the problem by
changing the condition properly.

CC: stable@vger.kernel.org
Fixes: 6fb1ca92a6409a9d5b0696447cd4997bc9aaf5a2
Reported-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Jan Kara <jack@suse.cz>
9 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Wed, 28 Jan 2015 03:12:38 +0000 (19:12 -0800)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "This feels larger than I'd like but its for three reasons.

   a) amdkfd finalising the API more, this is a new feature introduced
      last merge window, and I'd prefer to make the tweaks to the API
      before it first gets into a stable release.

   b) radeon regression required splitting an internal API to fix
      properly, so it just changed a few more lines

   c) vmwgfx fix changes a lock from a mutex->spin lock, this is fallout
      from the new sleep checking.

  Otherwise there is just some tda998x fixes"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/radeon: Remove rdev->gart.pages_addr array
  drm/radeon: Restore GART table contents after pinning it in VRAM v3
  drm/radeon: Split off gart_get_page_entry ASIC hook from set_page_entry
  drm/amdkfd: Fix bug in call to init_pipelines()
  drm/amdkfd: Fix bug in pipelines initialization
  drm/radeon: Don't increment pipe_id in kgd_init_pipeline
  drm/i2c: tda998x: set the CEC I2C address based on the slave I2C address
  drm/vmwgfx: Replace the hw mutex with a hw spinlock
  drm/amdkfd: Allow user to limit only queues per device
  drm/amdkfd: PQM handle queue creation fault
  drm: tda998x: Fix EDID read timeout on HDMI connect
  drm: tda998x: Protect the page register

9 years agobtrfs: fix raid56 scrub failed in xfstests btrfs/072
Gui Hecheng [Fri, 9 Jan 2015 01:39:40 +0000 (09:39 +0800)]
btrfs: fix raid56 scrub failed in xfstests btrfs/072

The xfstests btrfs/072 reports uncorrectable read errors in dmesg,
because scrub forgets to use commit_root for parity scrub routine
and scrub attempts to scrub those extents items whose contents are
not fully on disk.

To fix it, we just add the @search_commit_root flag back.

Signed-off-by: Gui Hecheng <guihc.fnst@cn.fujitsu.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Reviewed-by: Miao Xie <miaoxie@huawei.com>
Signed-off-by: Chris Mason <clm@fb.com>
9 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Tue, 27 Jan 2015 21:55:36 +0000 (13:55 -0800)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Don't OOPS on socket AIO, from Christoph Hellwig.

 2) Scheduled scans should be aborted upon RFKILL, from Emmanuel
    Grumbach.

 3) Fix sleep in atomic context in kvaser_usb, from Ahmed S Darwish.

 4) Fix RCU locking across copy_to_user() in bpf code, from Alexei
    Starovoitov.

 5) Lots of crash, memory leak, short TX packet et al bug fixes in
    sh_eth from Ben Hutchings.

 6) Fix memory corruption in SCTP wrt.  INIT collitions, from Daniel
    Borkmann.

 7) Fix return value logic for poll handlers in netxen, enic, and bnx2x.
    From Eric Dumazet and Govindarajulu Varadarajan.

 8) Header length calculation fix in mac80211 from Fred Chou.

 9) mv643xx_eth doesn't handle highmem correctly in non-TSO code paths.
    From Ezequiel Garcia.

10) udp_diag has bogus logic in it's hash chain skipping, copy same fix
    tcp diag used.  From Herbert Xu.

11) amd-xgbe programs wrong rx flow control register, from Thomas
    Lendacky.

12) Fix race leading to use after free in ping receive path, from Subash
    Abhinov Kasiviswanathan.

13) Cache redirect routes otherwise we can get a heavy backlog of rcu
    jobs liberating DST_NOCACHE entries.  From Hannes Frederic Sowa.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (48 commits)
  net: don't OOPS on socket aio
  stmmac: prevent probe drivers to crash kernel
  bnx2x: fix napi poll return value for repoll
  ipv6: replacing a rt6_info needs to purge possible propagated rt6_infos too
  sh_eth: Fix DMA-API usage for RX buffers
  sh_eth: Check for DMA mapping errors on transmit
  sh_eth: Ensure DMA engines are stopped before freeing buffers
  sh_eth: Remove RX overflow log messages
  ping: Fix race in free in receive path
  udp_diag: Fix socket skipping within chain
  can: kvaser_usb: Fix state handling upon BUS_ERROR events
  can: kvaser_usb: Retry the first bulk transfer on -ETIMEDOUT
  can: kvaser_usb: Send correct context to URB completion
  can: kvaser_usb: Do not sleep in atomic context
  ipv4: try to cache dst_entries which would cause a redirect
  samples: bpf: relax test_maps check
  bpf: rcu lock must not be held when calling copy_to_user()
  net: sctp: fix slab corruption from use after free on INIT collisions
  net: mv643xx_eth: Fix highmem support in non-TSO egress path
  sh_eth: Fix serialisation of interrupt disable with interrupt & NAPI handlers
  ...

9 years agonet: don't OOPS on socket aio
Christoph Hellwig [Tue, 27 Jan 2015 20:25:33 +0000 (12:25 -0800)]
net: don't OOPS on socket aio

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
9 years agostmmac: prevent probe drivers to crash kernel
Andy Shevchenko [Tue, 27 Jan 2015 16:38:03 +0000 (18:38 +0200)]
stmmac: prevent probe drivers to crash kernel

In the case when alloc_netdev fails we return NULL to a caller. But there is no
check for NULL in the probe drivers. This patch changes NULL to an error
pointer. The function description is amended to reflect what we may get
returned.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>