mbox series

[00/35] perf intel-pt: Add support for tracing virtual machine user space on the host

Message ID 20220711093218.10967-1-adrian.hunter@intel.com (mailing list archive)
Headers show
Series perf intel-pt: Add support for tracing virtual machine user space on the host | expand

Message

Adrian Hunter July 11, 2022, 9:31 a.m. UTC
Hi

Here are patches to support decoding an Intel PT trace that contains data
from virtual machine userspace.

This is done by adding functionality to perf inject to be able to inject
sideband events needed for decoding, into the perf.data file recorded on
the host.  That is, inject events from a perf.data file recorded in a
virtual machine into a perf.data file recorded on the host at the same
time.

For more details, see the example in the documentation added in the last
patch.

Note there was already support for tracing virtual machines kernel-only:

 https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/perf/Documentation/perf-intel-pt.txt?h=v5.19-rc1#n1221
 
or the special case of tracing KVM self tests:

 https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/perf/Documentation/perf-intel-pt.txt?h=v5.19-rc1#n1403

For general information about Intel PT also see the wiki page:

 https://perf.wiki.kernel.org/index.php/Perf_tools_support_for_Intel%C2%AE_Processor_Trace

The patches fall into 5 groups:
 1. the first patch is a fix
 2. the next 22 patches are preparation
 3. the main patch is "perf inject: Add support for injecting guest
 sideband events"
 4. 3 more preparation patches
 5. Intel PT decoding changes

The patches are mostly small except for "perf inject: Add support for
injecting guest sideband events".  However the code there adds new
functionality, does not affect existing functionality and is consequently
pretty self-contained.


Adrian Hunter (35):
      perf tools: Fix dso_id inode generation comparison
      perf tools: Export dsos__for_each_with_build_id()
      perf ordered_events: Add ordered_events__last_flush_time()
      perf tools: Export perf_event__process_finished_round()
      perf tools: Factor out evsel__id_hdr_size()
      perf tools: Add perf_event__synthesize_id_sample()
      perf script: Add --dump-unsorted-raw-trace option
      perf buildid-cache: Add guestmount'd files to the build ID cache
      perf buildid-cache: Do not require purge files to also be in the file system
      perf tools: Add machine_pid and vcpu to id_index
      perf session: Create guest machines from id_index
      perf tools: Add guest_cpu to hypervisor threads
      perf tools: Add machine_pid and vcpu to perf_sample
      perf tools: Use sample->machine_pid to find guest machine
      perf script: Add machine_pid and vcpu
      perf dlfilter: Add machine_pid and vcpu
      perf auxtrace: Add machine_pid and vcpu to auxtrace_error
      perf script python: Add machine_pid and vcpu
      perf script python: intel-pt-events: Add machine_pid and vcpu
      perf tools: Remove also guest kcore_dir with host kcore_dir
      perf tools: Make has_kcore_dir() work also for guest kcore_dir
      perf tools: Automatically use guest kcore_dir if present
      perf tools: Add reallocarray_as_needed()
      perf inject: Add support for injecting guest sideband events
      perf machine: Use realloc_array_as_needed() in machine__set_current_tid()
      perf tools: Handle injected guest kernel mmap event
      perf tools: Add perf_event__is_guest()
      perf intel-pt: Remove guest_machine_pid
      perf intel-pt: Add some more logging to intel_pt_walk_next_insn()
      perf intel-pt: Track guest context switches
      perf intel-pt: pt disable sync switch
      perf intel-pt: Determine guest thread from guest sideband
      perf intel-pt: Add machine_pid and vcpu to auxtrace_error
      perf intel-pt: Use guest pid/tid etc in guest samples
      perf intel-pt: Add documentation for tracing guest machine user space

 tools/lib/perf/include/internal/evsel.h            |    4 +
 tools/lib/perf/include/perf/event.h                |    7 +
 tools/perf/Documentation/perf-dlfilter.txt         |   22 +
 tools/perf/Documentation/perf-inject.txt           |   17 +
 tools/perf/Documentation/perf-intel-pt.txt         |  181 +++-
 tools/perf/Documentation/perf-script.txt           |   10 +-
 tools/perf/builtin-inject.c                        | 1043 +++++++++++++++++++-
 tools/perf/builtin-script.c                        |   19 +
 tools/perf/include/perf/perf_dlfilter.h            |    8 +
 tools/perf/scripts/python/intel-pt-events.py       |   32 +-
 tools/perf/util/auxtrace.c                         |   30 +-
 tools/perf/util/auxtrace.h                         |    4 +
 tools/perf/util/build-id.c                         |   80 +-
 tools/perf/util/build-id.h                         |   16 +-
 tools/perf/util/data.c                             |   43 +-
 tools/perf/util/data.h                             |    1 +
 tools/perf/util/dlfilter.c                         |    2 +
 tools/perf/util/dso.h                              |    6 +
 tools/perf/util/dsos.c                             |   10 +-
 tools/perf/util/event.h                            |   23 +
 tools/perf/util/evlist.c                           |   42 +-
 tools/perf/util/evsel.c                            |   27 +
 tools/perf/util/evsel.h                            |    2 +
 tools/perf/util/intel-pt.c                         |  183 +++-
 tools/perf/util/machine.c                          |   41 +-
 tools/perf/util/machine.h                          |    2 +
 tools/perf/util/ordered-events.h                   |    6 +
 .../util/scripting-engines/trace-event-python.c    |   15 +-
 tools/perf/util/session.c                          |  111 ++-
 tools/perf/util/session.h                          |    4 +
 tools/perf/util/symbol.c                           |    6 +-
 tools/perf/util/synthetic-events.c                 |   98 +-
 tools/perf/util/synthetic-events.h                 |    2 +
 tools/perf/util/thread.c                           |    1 +
 tools/perf/util/thread.h                           |    1 +
 tools/perf/util/util.c                             |   70 +-
 tools/perf/util/util.h                             |   15 +
 37 files changed, 2029 insertions(+), 155 deletions(-)


Regards
Adrian

Comments

Arnaldo Carvalho de Melo July 18, 2022, 3:28 p.m. UTC | #1
Em Mon, Jul 11, 2022 at 12:31:43PM +0300, Adrian Hunter escreveu:
> Hi
> 
> Here are patches to support decoding an Intel PT trace that contains data
> from virtual machine userspace.
> 
> This is done by adding functionality to perf inject to be able to inject
> sideband events needed for decoding, into the perf.data file recorded on
> the host.  That is, inject events from a perf.data file recorded in a
> virtual machine into a perf.data file recorded on the host at the same
> time.
> 
> For more details, see the example in the documentation added in the last
> patch.
> 
> Note there was already support for tracing virtual machines kernel-only:
> 
>  https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/perf/Documentation/perf-intel-pt.txt?h=v5.19-rc1#n1221
>  
> or the special case of tracing KVM self tests:
> 
>  https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/tools/perf/Documentation/perf-intel-pt.txt?h=v5.19-rc1#n1403
> 
> For general information about Intel PT also see the wiki page:
> 
>  https://perf.wiki.kernel.org/index.php/Perf_tools_support_for_Intel%C2%AE_Processor_Trace
> 
> The patches fall into 5 groups:
>  1. the first patch is a fix
>  2. the next 22 patches are preparation
>  3. the main patch is "perf inject: Add support for injecting guest
>  sideband events"
>  4. 3 more preparation patches
>  5. Intel PT decoding changes
> 
> The patches are mostly small except for "perf inject: Add support for
> injecting guest sideband events".  However the code there adds new
> functionality, does not affect existing functionality and is consequently
> pretty self-contained.

Applied locally, going thru tests.

- Arnaldo
 
> 
> Adrian Hunter (35):
>       perf tools: Fix dso_id inode generation comparison
>       perf tools: Export dsos__for_each_with_build_id()
>       perf ordered_events: Add ordered_events__last_flush_time()
>       perf tools: Export perf_event__process_finished_round()
>       perf tools: Factor out evsel__id_hdr_size()
>       perf tools: Add perf_event__synthesize_id_sample()
>       perf script: Add --dump-unsorted-raw-trace option
>       perf buildid-cache: Add guestmount'd files to the build ID cache
>       perf buildid-cache: Do not require purge files to also be in the file system
>       perf tools: Add machine_pid and vcpu to id_index
>       perf session: Create guest machines from id_index
>       perf tools: Add guest_cpu to hypervisor threads
>       perf tools: Add machine_pid and vcpu to perf_sample
>       perf tools: Use sample->machine_pid to find guest machine
>       perf script: Add machine_pid and vcpu
>       perf dlfilter: Add machine_pid and vcpu
>       perf auxtrace: Add machine_pid and vcpu to auxtrace_error
>       perf script python: Add machine_pid and vcpu
>       perf script python: intel-pt-events: Add machine_pid and vcpu
>       perf tools: Remove also guest kcore_dir with host kcore_dir
>       perf tools: Make has_kcore_dir() work also for guest kcore_dir
>       perf tools: Automatically use guest kcore_dir if present
>       perf tools: Add reallocarray_as_needed()
>       perf inject: Add support for injecting guest sideband events
>       perf machine: Use realloc_array_as_needed() in machine__set_current_tid()
>       perf tools: Handle injected guest kernel mmap event
>       perf tools: Add perf_event__is_guest()
>       perf intel-pt: Remove guest_machine_pid
>       perf intel-pt: Add some more logging to intel_pt_walk_next_insn()
>       perf intel-pt: Track guest context switches
>       perf intel-pt: pt disable sync switch
>       perf intel-pt: Determine guest thread from guest sideband
>       perf intel-pt: Add machine_pid and vcpu to auxtrace_error
>       perf intel-pt: Use guest pid/tid etc in guest samples
>       perf intel-pt: Add documentation for tracing guest machine user space
> 
>  tools/lib/perf/include/internal/evsel.h            |    4 +
>  tools/lib/perf/include/perf/event.h                |    7 +
>  tools/perf/Documentation/perf-dlfilter.txt         |   22 +
>  tools/perf/Documentation/perf-inject.txt           |   17 +
>  tools/perf/Documentation/perf-intel-pt.txt         |  181 +++-
>  tools/perf/Documentation/perf-script.txt           |   10 +-
>  tools/perf/builtin-inject.c                        | 1043 +++++++++++++++++++-
>  tools/perf/builtin-script.c                        |   19 +
>  tools/perf/include/perf/perf_dlfilter.h            |    8 +
>  tools/perf/scripts/python/intel-pt-events.py       |   32 +-
>  tools/perf/util/auxtrace.c                         |   30 +-
>  tools/perf/util/auxtrace.h                         |    4 +
>  tools/perf/util/build-id.c                         |   80 +-
>  tools/perf/util/build-id.h                         |   16 +-
>  tools/perf/util/data.c                             |   43 +-
>  tools/perf/util/data.h                             |    1 +
>  tools/perf/util/dlfilter.c                         |    2 +
>  tools/perf/util/dso.h                              |    6 +
>  tools/perf/util/dsos.c                             |   10 +-
>  tools/perf/util/event.h                            |   23 +
>  tools/perf/util/evlist.c                           |   42 +-
>  tools/perf/util/evsel.c                            |   27 +
>  tools/perf/util/evsel.h                            |    2 +
>  tools/perf/util/intel-pt.c                         |  183 +++-
>  tools/perf/util/machine.c                          |   41 +-
>  tools/perf/util/machine.h                          |    2 +
>  tools/perf/util/ordered-events.h                   |    6 +
>  .../util/scripting-engines/trace-event-python.c    |   15 +-
>  tools/perf/util/session.c                          |  111 ++-
>  tools/perf/util/session.h                          |    4 +
>  tools/perf/util/symbol.c                           |    6 +-
>  tools/perf/util/synthetic-events.c                 |   98 +-
>  tools/perf/util/synthetic-events.h                 |    2 +
>  tools/perf/util/thread.c                           |    1 +
>  tools/perf/util/thread.h                           |    1 +
>  tools/perf/util/util.c                             |   70 +-
>  tools/perf/util/util.h                             |   15 +
>  37 files changed, 2029 insertions(+), 155 deletions(-)
> 
> 
> Regards
> Adrian