LBR call stack only has user-space callchains. It is output in the
PERF_SAMPLE_BRANCH_STACK data format. For kernel callchains, it's
still in the form of PERF_SAMPLE_CALLCHAIN.
The perf tool has to handle both data sources to construct a
complete callstack.
For the "perf report -D" option, both lbr and fp information will be
displayed.
A new call chain recording option "lbr" is introduced into the perf
tool for LBR call stack. The user can use --call-graph lbr to get
the call stack information from hardware.
Here are some examples.
When profiling bc(1) on Fedora 19:
echo 'scale=2000; 4*a(1)' > cmd; perf record --call-graph lbr bc -l < cmd
If enabling LBR, perf report output looks like:
50.36% bc bc [.] bc_divide
|
--- bc_divide
execute
run_code
yyparse
main
__libc_start_main
_start
33.66% bc bc [.] _one_mult
|
--- _one_mult
bc_divide
execute
run_code
yyparse
main
__libc_start_main
_start
7.62% bc bc [.] _bc_do_add
|
--- _bc_do_add
|
|--99.89%-- 0x2000186a8
--0.11%-- [...]
6.83% bc bc [.] _bc_do_sub
|
--- _bc_do_sub
|
|--99.94%-- bc_add
| execute
| run_code
| yyparse
| main
| __libc_start_main
| _start
--0.06%-- [...]
0.46% bc libc-2.17.so [.] __memset_sse2
|
--- __memset_sse2
|
|--54.13%-- bc_new_num
| |
| |--51.00%-- bc_divide
| | execute
| | run_code
| | yyparse
| | main
| | __libc_start_main
| | _start
| |
| |--30.46%-- _bc_do_sub
| | bc_add
| | execute
| | run_code
| | yyparse
| | main
| | __libc_start_main
| | _start
| |
| --18.55%-- _bc_do_add
| bc_add
| execute
| run_code
| yyparse
| main
| __libc_start_main
| _start
|
--45.87%-- bc_divide
execute
run_code
yyparse
main
__libc_start_main
_start
If using FP, perf report output looks like:
echo 'scale=2000; 4*a(1)' > cmd; perf record --call-graph fp bc -l < cmd
50.49% bc bc [.] bc_divide
|
--- bc_divide
33.57% bc bc [.] _one_mult
|
--- _one_mult
7.61% bc bc [.] _bc_do_add
|
--- _bc_do_add
0x2000186a8
6.88% bc bc [.] _bc_do_sub
|
--- _bc_do_sub
0.42% bc libc-2.17.so [.] __memcpy_ssse3_back
|
--- __memcpy_ssse3_back
If using LBR, perf report -D output looks like:
3458145275743 0x2fd750 [0xd8]: PERF_RECORD_SAMPLE(IP, 0x2): 9748/9748: 0x408ea8 period: 609644 addr: 0
... LBR call chain: nr:8
..... 0: fffffffffffffe00
..... 1: 0000000000408e50
..... 2: 000000000040a458
..... 3: 000000000040562e
..... 4: 0000000000408590
..... 5: 00000000004022c0
..... 6: 00000000004015dd
..... 7: 0000003d1cc21b43
... FP chain: nr:2
..... 0: fffffffffffffe00
..... 1: 0000000000408ea8
... thread: bc:9748
...... dso: /usr/bin/bc
The LBR call stack has the following known limitations:
- Zero length calls are not filtered out by the hardware
- Exception handing such as setjmp/longjmp will have calls/returns not
match
- Pushing different return address onto the stack will have
calls/returns not match
- If callstack is deeper than the LBR, only the last entries are
captured
Tested-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Kan Liang <kan.liang@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Borislav Petkov <bp@suse.de>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Simon Que <sque@chromium.org>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1420482185-29830-3-git-send-email-kan.liang@intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
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>
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>
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>
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>
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>
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>
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
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
Commit 8eb23b9f35 ("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>
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>
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>
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>
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: 9592747538 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>
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
