lkml.org 
[lkml]   [2012]   [Jun]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    From
    Subject[RFCv5 00/19] perf: Add backtrace post dwarf unwind
    Date
    hi,
    besides fixing several issues, going back to the original design
    because the last one was considered too generic.. now we have:

    sample_regs_user - != 0 triggers the user level regs dump
    sample_stack_user - != 0 triggers the user stack dump

    We can allway extend this in future with new mask and flags
    for IRQ/PEBS regs.

    patches available also as tarball in here:
    http://people.redhat.com/~jolsa/perf_post_unwind_v5.tar.bz2

    v5 changes:
    patch 1/19 - having just one enum set of the perf registers
    patch 2/19 - using for_each_set_bit for scanning the mask
    - single regs enum for both 32 and 64 bits versions
    - using regs mask != 0 trigger to trigger the regs dump
    patch 5/19 - adding perf_output_skip so we can skip undumped part of the stack in RB
    patch 6/19 - using stack size != 0 trigger to trigger the stack dump
    - do not zero the memory for non retrieved part of the stack dump
    patch 7/19 - adding exclude_callchain_kernel attribute
    patch 8/19 - this could be taken without the rest of the series

    v4 changes:
    - no real change from v3, just rebase
    - v3 patch 06/17 got already merged

    v3 changes:
    patch 01/17
    - added HAVE_PERF_REGS config option
    patch 02/17, 04/17
    - regs and stack perf interface is more general now
    patch 06/17
    - unrelated online fix for i386 compilation
    patch 16/17
    - few namespace fixies

    ---
    Adding the post unwinding user stack backtrace using dwarf unwind
    via libunwind. The original work was done by Frederic. I mostly took
    his patches and make them compile in current kernel code plus I added
    some stuff here and there.

    The main idea is to store user registers and portion of user
    stack when the sample data during the record phase. Then during
    the report, when the data is presented, perform the actual dwarf
    dwarf unwind.

    attached patches:
    01/19 perf: Unified API to record selective sets of arch registers
    02/19 perf: Add ability to attach user level registers dump to sample
    03/19 perf, x86: Add copy_from_user_nmi_nochk for best effort copy
    04/19 perf: Factor __output_copy to be usable with specific copy function
    05/19 perf: Add perf_output_skip function to skip bytes in sample
    06/19 perf: Add ability to attach user stack dump to sample
    07/19 perf: Add attribute to filter out callchains
    08/19 perf, tool: Remove unsused evsel parameter from machine__resolve_callchain
    09/19 perf, tool: Factor DSO symtab types to generic binary types
    10/19 perf, tool: Add interface to read DSO image data
    11/19 perf, tool: Add '.note' check into search for NOTE section
    12/19 perf, tool: Back [vdso] DSO with real data
    13/19 perf, tool: Add interface to arch registers sets
    14/19 perf, tool: Add libunwind dependency for dwarf cfi unwinding
    15/19 perf, tool: Support user regs and stack in sample parsing
    16/19 perf, tool: Support for dwarf cfi unwinding on post processing
    17/19 perf, tool: Support for dwarf mode callchain on perf record
    18/19 perf, tool: Add dso data caching
    19/19 perf, tool: Add dso data caching tests


    I tested on Fedora. There was not much gain on i386, because the
    binaries are compiled with frame pointers. Thought the dwarf
    backtrace is more accurade and unwraps calls in more details
    (functions that do not set the frame pointers).

    I could see some improvement on x86_64, where I got full backtrace
    where current code could got just the first address out of the
    instruction pointer.

    Example on x86_64:
    [dwarf]
    perf record -g -e syscalls:sys_enter_write date

    100.00% date libc-2.14.90.so [.] __GI___libc_write
    |
    --- __GI___libc_write
    _IO_file_write@@GLIBC_2.2.5
    new_do_write
    _IO_do_write@@GLIBC_2.2.5
    _IO_file_overflow@@GLIBC_2.2.5
    0x4022cd
    0x401ee6
    __libc_start_main
    0x4020b9


    [frame pointer]
    perf record -g fp -e syscalls:sys_enter_write date

    100.00% date libc-2.14.90.so [.] __GI___libc_write
    |
    --- __GI___libc_write

    Also I tested on coreutils binaries mainly, but I could see
    getting wider backtraces with dwarf unwind for more complex
    application like firefox.

    The unwind should go throught [vdso] object. I haven't studied
    the [vsyscall] yet, so not sure there.

    Attached patches should work on both x86 and x86_64. I did
    some initial testing so far.

    The unwind backtrace can be interrupted by following reasons:
    - bug in unwind information of processed shared library
    - bug in unwind processing code (most likely ;) )
    - insufficient dump stack size
    - wrong register value - x86_64 does not store whole
    set of registers when in exception, but so far
    it looks like RIP and RSP should be enough

    thanks for comments,
    jirka
    ---
    arch/Kconfig | 6 +
    arch/x86/Kconfig | 1 +
    arch/x86/include/asm/perf_event.h | 2 +
    arch/x86/include/asm/perf_regs.h | 34 ++
    arch/x86/include/asm/uaccess.h | 2 +
    arch/x86/kernel/Makefile | 2 +
    arch/x86/kernel/perf_regs.c | 91 ++++
    arch/x86/lib/usercopy.c | 15 +-
    include/linux/perf_event.h | 24 +-
    include/linux/perf_regs.h | 19 +
    kernel/events/callchain.c | 25 +-
    kernel/events/core.c | 132 +++++-
    kernel/events/internal.h | 69 ++-
    kernel/events/ring_buffer.c | 10 +-
    tools/perf/Makefile | 45 ++-
    tools/perf/arch/x86/Makefile | 3 +
    tools/perf/arch/x86/include/perf_regs.h | 80 +++
    tools/perf/arch/x86/util/unwind.c | 111 ++++
    tools/perf/builtin-record.c | 86 +++-
    tools/perf/builtin-report.c | 24 +-
    tools/perf/builtin-script.c | 56 ++-
    tools/perf/builtin-test.c | 7 +-
    tools/perf/builtin-top.c | 7 +-
    tools/perf/config/feature-tests.mak | 25 +
    tools/perf/perf.h | 9 +-
    tools/perf/util/annotate.c | 2 +-
    tools/perf/util/dso-test-data.c | 154 ++++++
    tools/perf/util/event.h | 15 +-
    tools/perf/util/evlist.c | 16 +
    tools/perf/util/evlist.h | 2 +
    tools/perf/util/evsel.c | 35 ++-
    tools/perf/util/include/linux/compiler.h | 1 +
    tools/perf/util/map.c | 23 +-
    tools/perf/util/map.h | 9 +-
    tools/perf/util/perf_regs.h | 14 +
    tools/perf/util/python.c | 3 +-
    .../perf/util/scripting-engines/trace-event-perl.c | 3 +-
    .../util/scripting-engines/trace-event-python.c | 3 +-
    tools/perf/util/session.c | 110 ++++-
    tools/perf/util/session.h | 17 +-
    tools/perf/util/symbol.c | 435 +++++++++++++---
    tools/perf/util/symbol.h | 52 ++-
    tools/perf/util/trace-event-scripting.c | 3 +-
    tools/perf/util/trace-event.h | 5 +-
    tools/perf/util/unwind.c | 565 ++++++++++++++++++++
    tools/perf/util/unwind.h | 34 ++
    tools/perf/util/vdso.c | 90 +++
    tools/perf/util/vdso.h | 8 +
    48 files changed, 2278 insertions(+), 206 deletions(-)


    \
     
     \ /
      Last update: 2012-06-11 16:25    [W:0.041 / U:0.728 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site