Skip to content

cloud-kernel
cloud-kernel

openanolis / cloud-kernel
1 年多 前同步成功

通知 161
Star 36
Fork 7
cloud-kernel
cloud-kernel
切换分支/标签
  1. 07 6月, 2015 5 次提交
    • Y
      perf/x86/intel: Drain the PEBS buffer during context switches · 9c964efa
      由 Yan, Zheng 提交于 
      Flush the PEBS buffer during context switches if PEBS interrupt threshold
is larger than one. This allows perf to supply TID for sample outputs.
Signed-off-by: NYan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: NKan Liang <kan.liang@intel.com>
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@infradead.org
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/1430940834-8964-6-git-send-email-kan.liang@intel.comSigned-off-by: NIngo Molnar <mingo@kernel.org>
      9c964efa
    • Y
      perf/x86/intel: Implement batched PEBS interrupt handling (large PEBS interrupt threshold) · 3569c0d7
      由 Yan, Zheng 提交于 
      PEBS always had the capability to log samples to its buffers without
an interrupt. Traditionally perf has not used this but always set the
PEBS threshold to one.

For frequently occurring events (like cycles or branches or load/store)
this in term requires using a relatively high sampling period to avoid
overloading the system, by only processing PMIs. This in term increases
sampling error.

For the common cases we still need to use the PMI because the PEBS
hardware has various limitations. The biggest one is that it can not
supply a callgraph. It also requires setting a fixed period, as the
hardware does not support adaptive period. Another issue is that it
cannot supply a time stamp and some other options. To supply a TID it
requires flushing on context switch. It can however supply the IP, the
load/store address, TSX information, registers, and some other things.

So we can make PEBS work for some specific cases, basically as long as
you can do without a callgraph and can set the period you can use this
new PEBS mode.

The main benefit is the ability to support much lower sampling period
(down to -c 1000) without extensive overhead.

One use cases is for example to increase the resolution of the c2c tool.
Another is double checking when you suspect the standard sampling has
too much sampling error.

Some numbers on the overhead, using cycle soak, comparing the elapsed
time from "kernbench -M -H" between plain (threshold set to one) and
multi (large threshold).

The test command for plain:
  "perf record --time -e cycles:p -c $period -- kernbench -M -H"

The test command for multi:
  "perf record --no-time -e cycles:p -c $period -- kernbench -M -H"

( The only difference of test command between multi and plain is time
  stamp options. Since time stamp is not supported by large PEBS
  threshold, it can be used as a flag to indicate if large threshold is
  enabled during the test. )

	period    plain(Sec)  multi(Sec)  Delta
	10003     32.7        16.5        16.2
	20003     30.2        16.2        14.0
	40003     18.6        14.1        4.5
	80003     16.8        14.6        2.2
	100003    16.9        14.1        2.8
	800003    15.4        15.7        -0.3
	1000003   15.3        15.2        0.2
	2000003   15.3        15.1        0.1

With periods below 100003, plain (threshold one) cause much more
overhead. With 10003 sampling period, the Elapsed Time for multi is
even 2X faster than plain.
Signed-off-by: NYan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: NKan Liang <kan.liang@intel.com>
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@infradead.org
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/1430940834-8964-5-git-send-email-kan.liang@intel.comSigned-off-by: NIngo Molnar <mingo@kernel.org>
      3569c0d7
    • Y
      perf/x86/intel: Handle multiple records in the PEBS buffer · 21509084
      由 Yan, Zheng 提交于 
      When the PEBS interrupt threshold is larger than one record and the
machine supports multiple PEBS events, the records of these events are
mixed up and we need to demultiplex them.

Demuxing the records is hard because the hardware is deficient. The
hardware has two issues that, when combined, create impossible
scenarios to demux.

The first issue is that the 'status' field of the PEBS record is a copy
of the GLOBAL_STATUS MSR at PEBS assist time. To see why this is a
problem let us first describe the regular PEBS cycle:

A) the CTRn value reaches 0:
  - the corresponding bit in GLOBAL_STATUS gets set
  - we start arming the hardware assist
  < some unspecified amount of time later -- this could cover multiple
    events of interest >

B) the hardware assist is armed, any next event will trigger it

C) a matching event happens:
  - the hardware assist triggers and generates a PEBS record
    this includes a copy of GLOBAL_STATUS at this moment
  - if we auto-reload we (re)set CTRn
  - we clear the relevant bit in GLOBAL_STATUS

Now consider the following chain of events:

  A0, B0, A1, C0

The event generated for counter 0 will include a status with counter 1
set, even though its not at all related to the record. A similar thing
can happen with a !PEBS event if it just happens to overflow at the
right moment.

The second issue is that the hardware will only emit one record for two
or more counters if the event that triggers the assist is 'close'. The
'close' can be several cycles. In some cases even the complete assist,
if the event is something that doesn't need retirement.

For instance, consider this chain of events:

  A0, B0, A1, B1, C01

Where C01 is an event that triggers both hardware assists, we will
generate but a single record, but again with both counters listed in the
status field.

This time the record pertains to both events.

Note that these two cases are different but undistinguishable with the
data as generated. Therefore demuxing records with multiple PEBS bits
(we can safely ignore status bits for !PEBS counters) is impossible.

Furthermore we cannot emit the record to both events because that might
cause a data leak -- the events might not have the same privileges -- so
what this patch does is discard such events.

The assumption/hope is that such discards will be rare.

Here lists some possible ways you may get high discard rate.

  - when you count the same thing multiple times. But it is not a useful
    configuration.
  - you can be unfortunate if you measure with a userspace only PEBS
    event along with either a kernel or unrestricted PEBS event. Imagine
    the event triggering and setting the overflow flag right before
    entering the kernel. Then all kernel side events will end up with
    multiple bits set.
Signed-off-by: NYan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: NKan Liang <kan.liang@intel.com>
[ Changelog improvements. ]
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@infradead.org
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/1430940834-8964-4-git-send-email-kan.liang@intel.comSigned-off-by: NIngo Molnar <mingo@kernel.org>
Signed-off-by: NIngo Molnar <mingo@kernel.org>
      21509084
    • Y
      perf/x86/intel: Introduce setup_pebs_sample_data() · 43cf7631
      由 Yan, Zheng 提交于 
      Move code that sets up the PEBS sample data to a separate function.
Signed-off-by: NYan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: NKan Liang <kan.liang@intel.com>
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@infradead.org
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/1430940834-8964-3-git-send-email-kan.liang@intel.comSigned-off-by: NIngo Molnar <mingo@kernel.org>
      43cf7631
    • Y
      perf/x86/intel: Use the PEBS auto reload mechanism when possible · 851559e3
      由 Yan, Zheng 提交于 
      When a fixed period is specified, this patch makes perf use the PEBS
auto reload mechanism. This makes normal profiling faster, because
it avoids one costly MSR write in the PMI handler.

However, the reset value will be loaded by hardware assist. There is a
small delay compared to the previous non-auto-reload mechanism. The
delay time is arbitrary, but very small. The assist cost is 400-800
cycles, assuming common cases with everything cached. The minimum period
the patch currently uses is 10000. In that extreme case it can be ~10%
if cycles are used.
Signed-off-by: NYan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: NKan Liang <kan.liang@intel.com>
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: acme@infradead.org
Cc: eranian@google.com
Link: http://lkml.kernel.org/r/1430940834-8964-2-git-send-email-kan.liang@intel.comSigned-off-by: NIngo Molnar <mingo@kernel.org>
      851559e3
  3. 27 5月, 2015 1 次提交
    • P
      perf/x86: Fix event/group validation · b371b594
      由 Peter Zijlstra 提交于 
      Commit 43b45780 ("perf/x86: Reduce stack usage of
x86_schedule_events()") violated the rule that 'fake' scheduling; as
used for event/group validation; should not change the event state.

This went mostly un-noticed because repeated calls of
x86_pmu::get_event_constraints() would give the same result. And
x86_pmu::put_event_constraints() would mostly not do anything.

Commit e979121b ("perf/x86/intel: Implement cross-HT corruption
bug workaround") made the situation much worse by actually setting the
event->hw.constraint value to NULL, so when validation and actual
scheduling interact we get NULL ptr derefs.

Fix it by removing the constraint pointer from the event and move it
back to an array, this time in cpuc instead of on the stack.

validate_group()
  x86_schedule_events()
    event->hw.constraint = c; # store

      <context switch>
        perf_task_event_sched_in()
          ...
            x86_schedule_events();
              event->hw.constraint = c2; # store

              ...

              put_event_constraints(event); # assume failure to schedule
                intel_put_event_constraints()
                  event->hw.constraint = NULL;

      <context switch end>

    c = event->hw.constraint; # read -> NULL

    if (!test_bit(hwc->idx, c->idxmsk)) # <- *BOOM* NULL deref

This in particular is possible when the event in question is a
cpu-wide event and group-leader, where the validate_group() tries to
add an event to the group.
Reported-by: NVince Weaver <vincent.weaver@maine.edu>
Signed-off-by: NPeter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Hunter <ahh@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Maria Dimakopoulou <maria.n.dimakopoulou@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 43b45780 ("perf/x86: Reduce stack usage of x86_schedule_events()")
Fixes: e979121b ("perf/x86/intel: Implement cross-HT corruption bug workaround")
Signed-off-by: NIngo Molnar <mingo@kernel.org>
      b371b594
  5. 17 4月, 2015 1 次提交
  7. 02 4月, 2015 2 次提交
  9. 16 1月, 2015 1 次提交
  11. 18 11月, 2014 1 次提交
    • D
      x86: Remove arbitrary instruction size limit in instruction decoder · 6ba48ff4
      由 Dave Hansen 提交于 
      The current x86 instruction decoder steps along through the
instruction stream but always ensures that it never steps farther
than the largest possible instruction size (MAX_INSN_SIZE).

The MPX code is now going to be doing some decoding of userspace
instructions.  We copy those from userspace in to the kernel and
they're obviously completely untrusted coming from userspace.  In
addition to the constraint that instructions can only be so long,
we also have to be aware of how long the buffer is that came in
from userspace.  This _looks_ to be similar to what the perf and
kprobes is doing, but it's unclear to me whether they are
affected.

The whole reason we need this is that it is perfectly valid to be
executing an instruction within MAX_INSN_SIZE bytes of an
unreadable page. We should be able to gracefully handle short
reads in those cases.

This adds support to the decoder to record how long the buffer
being decoded is and to refuse to "validate" the instruction if
we would have gone over the end of the buffer to decode it.

The kprobes code probably needs to be looked at here a bit more
carefully.  This patch still respects the MAX_INSN_SIZE limit
there but the kprobes code does look like it might be able to
be a bit more strict than it currently is.
Signed-off-by: NDave Hansen <dave.hansen@linux.intel.com>
Acked-by: NJim Keniston <jkenisto@us.ibm.com>
Acked-by: NMasami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: x86@kernel.org
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Cc: Anil S Keshavamurthy <anil.s.keshavamurthy@intel.com>
Cc: "David S. Miller" <davem@davemloft.net>
Link: http://lkml.kernel.org/r/20141114153957.E6B01535@viggo.jf.intel.comSigned-off-by: NThomas Gleixner <tglx@linutronix.de>
      6ba48ff4
  13. 16 11月, 2014 3 次提交
  15. 09 9月, 2014 1 次提交
  17. 27 8月, 2014 1 次提交
    • C
      x86: Replace __get_cpu_var uses · 89cbc767
      由 Christoph Lameter 提交于 
      __get_cpu_var() is used for multiple purposes in the kernel source. One of
them is address calculation via the form &__get_cpu_var(x).  This calculates
the address for the instance of the percpu variable of the current processor
based on an offset.

Other use cases are for storing and retrieving data from the current
processors percpu area.  __get_cpu_var() can be used as an lvalue when
writing data or on the right side of an assignment.

__get_cpu_var() is defined as :

#define __get_cpu_var(var) (*this_cpu_ptr(&(var)))

__get_cpu_var() always only does an address determination. However, store
and retrieve operations could use a segment prefix (or global register on
other platforms) to avoid the address calculation.

this_cpu_write() and this_cpu_read() can directly take an offset into a
percpu area and use optimized assembly code to read and write per cpu
variables.

This patch converts __get_cpu_var into either an explicit address
calculation using this_cpu_ptr() or into a use of this_cpu operations that
use the offset.  Thereby address calculations are avoided and less registers
are used when code is generated.

Transformations done to __get_cpu_var()

1. Determine the address of the percpu instance of the current processor.

	DEFINE_PER_CPU(int, y);
	int *x = &__get_cpu_var(y);

    Converts to

	int *x = this_cpu_ptr(&y);

2. Same as #1 but this time an array structure is involved.

	DEFINE_PER_CPU(int, y[20]);
	int *x = __get_cpu_var(y);

    Converts to

	int *x = this_cpu_ptr(y);

3. Retrieve the content of the current processors instance of a per cpu
variable.

	DEFINE_PER_CPU(int, y);
	int x = __get_cpu_var(y)

   Converts to

	int x = __this_cpu_read(y);

4. Retrieve the content of a percpu struct

	DEFINE_PER_CPU(struct mystruct, y);
	struct mystruct x = __get_cpu_var(y);

   Converts to

	memcpy(&x, this_cpu_ptr(&y), sizeof(x));

5. Assignment to a per cpu variable

	DEFINE_PER_CPU(int, y)
	__get_cpu_var(y) = x;

   Converts to

	__this_cpu_write(y, x);

6. Increment/Decrement etc of a per cpu variable

	DEFINE_PER_CPU(int, y);
	__get_cpu_var(y)++

   Converts to

	__this_cpu_inc(y)

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: x86@kernel.org
Acked-by: NH. Peter Anvin <hpa@linux.intel.com>
Acked-by: NIngo Molnar <mingo@kernel.org>
Signed-off-by: NChristoph Lameter <cl@linux.com>
Signed-off-by: NTejun Heo <tj@kernel.org>
      89cbc767
  19. 13 8月, 2014 4 次提交
    • S
      perf/x86: Clean up __intel_pmu_pebs_event() code · c8aab2e0
      由 Stephane Eranian 提交于 
      This patch makes the code more readable. It also renames
precise_store_data_hsw() to precise_datala_hsw() because
the function is called for both loads and stores on HSW.
The patch also gets rid of the hardcoded store events
codes in that same function.
Signed-off-by: NStephane Eranian <eranian@google.com>
Signed-off-by: NPeter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1407785233-32193-5-git-send-email-eranian@google.com
Cc: ak@linux.intel.com
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Signed-off-by: NIngo Molnar <mingo@kernel.org>
      c8aab2e0
    • S
      perf/x86: Fix data source encoding issues for load latency/precise store · 770eee1f
      由 Stephane Eranian 提交于 
      This patch fixes issues introuduce by Andi's previous patch 'Revamp PEBS'
series.

This patch fixes the following:

 - precise_store_data_hsw() encode the mem op type whenever we can
 - precise_store_data_hsw set the default data source correctly

 - 0 is not a valid init value for data source. Define PERF_MEM_NA as the
   default value

This bug was actually introduced by

    commit 722e76e6
    Author: Stephane Eranian <eranian@google.com>
    Date:   Thu May 15 17:56:44 2014 +0200

        fix Haswell precise store data source encoding
Signed-off-by: NStephane Eranian <eranian@google.com>
Signed-off-by: NPeter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1407785233-32193-4-git-send-email-eranian@google.com
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: ak@linux.intel.com
Signed-off-by: NIngo Molnar <mingo@kernel.org>
      770eee1f
    • A
      perf/x86: Don't mark DataLA addresses as store · f3908b8c
      由 Andi Kleen 提交于 
      Haswell supports reporting the data address for a range
of PEBS events, including:

	UOPS_RETIRED.ALL
	MEM_UOPS_RETIRED.STLB_MISS_LOADS
	MEM_UOPS_RETIRED.STLB_MISS_STORES
	MEM_UOPS_RETIRED.LOCK_LOADS
	MEM_UOPS_RETIRED.SPLIT_LOADS
	MEM_UOPS_RETIRED.SPLIT_STORES
	MEM_UOPS_RETIRED.ALL_LOADS
	MEM_UOPS_RETIRED.ALL_STORES
	MEM_LOAD_UOPS_RETIRED.L1_HIT
	MEM_LOAD_UOPS_RETIRED.L2_HIT
	MEM_LOAD_UOPS_RETIRED.L3_HIT
	MEM_LOAD_UOPS_RETIRED.L1_MISS
	MEM_LOAD_UOPS_RETIRED.L2_MISS
	MEM_LOAD_UOPS_RETIRED.L3_MISS
	MEM_LOAD_UOPS_RETIRED.HIT_LFB
	MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS
	MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT
	MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM
	MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_NONE
	MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM

This facility was already enabled earlier with the original Haswell
perf changes.

However these addresses were always reports as stores by perf, which is wrong,
as they could be loads too.  The hardware does not distinguish loads and stores
for these instructions, so there's no (cheap) way for the profiler
to find out.

Change the type to PERF_MEM_OP_NA instead.
Signed-off-by: NAndi Kleen <ak@linux.intel.com>
Reviewed-by: NStephane Eranian <eranian@google.com>
Signed-off-by: NPeter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Link: http://lkml.kernel.org/r/1407785233-32193-3-git-send-email-eranian@google.comSigned-off-by: NIngo Molnar <mingo@kernel.org>
      f3908b8c
    • A
      perf/x86: Revamp PEBS event selection · 86a04461
      由 Andi Kleen 提交于 
      The basic idea is that it does not make sense to list all PEBS
events individually. The list is very long, sometimes outdated
and the hardware doesn't need it. If an event does not support
PEBS it will just not count, there is no security issue.

We need to only list events that something special, like
supporting load or store addresses.

This vastly simplifies the PEBS event selection. It also
speeds up the scheduling because the scheduler doesn't
have to walk as many constraints.

Bugs fixed:

 - We do not allow setting forbidden flags with PEBS anymore
   (SDM 18.9.4), except for the special cycle event.
   This is done using a new constraint macro that also
   matches on the event flags.

 - Correct DataLA and load/store/na flags reporting on Haswell
   [Requires a followon patch]

 - We did not allow all PEBS events on Haswell:
   We were missing some valid subevents in d1-d2 (MEM_LOAD_UOPS_RETIRED.*,
   MEM_LOAD_UOPS_RETIRED_L3_HIT_RETIRED.*)

This includes the changes proposed by Stephane earlier and obsoletes
his patchkit (except for some changes on pre Sandy Bridge/Silvermont
CPUs)

I only did Sandy Bridge and Silvermont and later so far, mostly because these
are the parts I could directly confirm the hardware behavior with hardware
architects. Also I do not believe the older CPUs have any
missing events in their PEBS list, so there's no pressing
need to change them.

I did not implement the flag proposed by Peter to allow
setting forbidden flags. If really needed this could
be implemented on to of this patch.

v2: Fix broken store events on SNB/IVB (Stephane Eranian)
v3: More fixes. Rename some arguments (Stephane Eranian)
v4: List most Haswell events individually again to report
memory operation type correctly.
Add new flags to describe load/store/na for datala.
Update description.
Signed-off-by: NAndi Kleen <ak@linux.intel.com>
Reviewed-by: NStephane Eranian <eranian@google.com>
Signed-off-by: NPeter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1407785233-32193-2-git-send-email-eranian@google.com
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Kan Liang <kan.liang@intel.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Maria Dimakopoulou <maria.n.dimakopoulou@gmail.com>
Cc: Mark Davies <junk@eslaf.co.uk>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: NIngo Molnar <mingo@kernel.org>
      86a04461
  21. 16 7月, 2014 1 次提交
  23. 19 5月, 2014 1 次提交
  25. 06 11月, 2013 1 次提交
  27. 16 10月, 2013 1 次提交
    • P
      perf/x86: Optimize intel_pmu_pebs_fixup_ip() · 9536c8d2
      由 Peter Zijlstra 提交于 
      There's been reports of high NMI handler overhead, highlighted by
such kernel messages:

  [ 3697.380195] perf samples too long (10009 > 10000), lowering kernel.perf_event_max_sample_rate to 13000
  [ 3697.389509] INFO: NMI handler (perf_event_nmi_handler) took too long to run: 9.331 msecs

Don Zickus analyzed the source of the overhead and reported:

 > While there are a few places that are causing latencies, for now I focused on
 > the longest one first.  It seems to be 'copy_user_from_nmi'
 >
 > intel_pmu_handle_irq ->
 >	intel_pmu_drain_pebs_nhm ->
 >		__intel_pmu_drain_pebs_nhm ->
 >			__intel_pmu_pebs_event ->
 >				intel_pmu_pebs_fixup_ip ->
 >					copy_from_user_nmi
 >
 > In intel_pmu_pebs_fixup_ip(), if the while-loop goes over 50, the sum of
 > all the copy_from_user_nmi latencies seems to go over 1,000,000 cycles
 > (there are some cases where only 10 iterations are needed to go that high
 > too, but in generall over 50 or so).  At this point copy_user_from_nmi
 > seems to account for over 90% of the nmi latency.

The solution to that is to avoid having to call copy_from_user_nmi() for
every instruction.

Since we already limit the max basic block size, we can easily
pre-allocate a piece of memory to copy the entire thing into in one
go.

Don reported this test result:

 > Your patch made a huge difference in improvement.  The
 > copy_from_user_nmi() no longer hits the million of cycles.  I still
 > have a batch of 100,000-300,000 cycles.  My longest NMI paths used
 > to be dominated by copy_from_user_nmi, now it is not (I have to dig
 > up the new hot path).
Reported-and-tested-by: NDon Zickus <dzickus@redhat.com>
Cc: jmario@redhat.com
Cc: acme@infradead.org
Cc: dave.hansen@linux.intel.com
Cc: eranian@google.com
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: NPeter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/20131016105755.GX10651@twins.programming.kicks-ass.netSigned-off-by: NIngo Molnar <mingo@kernel.org>
      9536c8d2
  29. 04 10月, 2013 1 次提交
  31. 20 9月, 2013 2 次提交
  33. 14 9月, 2013 1 次提交
  35. 13 9月, 2013 2 次提交
  37. 02 9月, 2013 2 次提交
  39. 27 6月, 2013 1 次提交
  41. 19 6月, 2013 3 次提交
  43. 01 4月, 2013 3 次提交
  45. 21 3月, 2013 1 次提交
    • S
      perf/x86: Fix uninitialized pt_regs in intel_pmu_drain_bts_buffer() · 0e48026a
      由 Stephane Eranian 提交于 
      This patch fixes an uninitialized pt_regs struct in drain BTS
function. The pt_regs struct is propagated all the way to the
code_get_segment() function from perf_instruction_pointer()
and may get garbage.

We cannot simply inherit the actual pt_regs from the interrupt
because BTS must be flushed on context-switch or when the
associated event is disabled. And there we do not have a pt_regs
handy.

Setting pt_regs to all zeroes may not be the best option but it
is not clear what else to do given where the drain_bts_buffer()
is called from.

In V2, we move the memset() later in the code to avoid doing it
when we end up returning early without doing the actual BTS
processing. Also dropped the reg.val initialization because it
is redundant with the memset() as suggested by PeterZ.
Signed-off-by: NStephane Eranian <eranian@google.com>
Acked-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: peterz@infradead.org
Cc: sqazi@google.com
Cc: ak@linux.intel.com
Cc: jolsa@redhat.com
Link: http://lkml.kernel.org/r/20130319151038.GA25439@quadSigned-off-by: NIngo Molnar <mingo@kernel.org>
      0e48026a