perf data/pipe handling improvements
Status from this message: Applied to perf-tools-next, thanks!
Get series with b4
b4 am [email protected]Versions newest first · linked by author and title
- v4 Latest8 patches 1 replyApplied
- v37 patches 0 repliesSuperseded
- v27 patches 5 repliesSuperseded
Earlier versions · 1
| Version | Posted | Patches | Replies | Status |
|---|---|---|---|---|
| v1 | 5 patches | 2 replies | Superseded |
Patches in v4 Applied
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/8 | perf clockid: Add missing includetools/ | +2 | 0 replies | |
| 2/8 | perf header: Add utility to convert feature number to a stringtools/ tools/ | +9 | 0 replies | |
| 3/8 | perf header: Properly warn/print when libtraceevent/libbpf support is missingtools/ | +47 | 0 replies | |
| 4/8 | perf session: Extra logging for failed to process eventstools/ | +20 | 0 replies | |
| 5/8 | perf header: Refactor pipe mode end marker handlingtools/ tools/ tools/ +5 more | +75 | 0 replies | |
| 6/8 | perf ordered-events: Event processing consistency with the regular readertools/ | +1 | 0 replies | |
| 7/8 | perf evsel: Make unknown event names more uniquetools/ | +4 | 0 replies | |
| 8/8 | perf data convert ctf: Pipe mode improvementstools/ | +59 | 0 replies |
Reviews trailers per patch
No Reviewed-by, Acked-by or Tested-by trailers for v4 yet.
Cover letter [PATCH v4 0/8] perf data/pipe handling improvements
I'm looking at improving the perf data converter with files generated in pipe mode. In pipe mode I found the feature handling for the last feature marker was problematic. If a new feature was added then the old marker looks like the new feature. These changes do some minor logging and build fixes, but they also change the last feature handling in pipe mode so that the marker is treated as a marker and not a feature that is broken.
Two additional fixes are added in v2. The first is making ordered events handle positive process event return values in a consistent manner with the reader that ignores positive values. Not handling this properly breaks with tracing data events in pipe mode. The second set of fixes are to the perf data converter --to-ctf output where generating the CTF events is deferred until the event desc feature or tracing data events occur in pipe mode.
v4: Address Sashiko nits, primarily extra bound checks. Add patch to
better warn about libtraceevent/libbpf support being missing.
v3: Adjust comment for event setup that in pipe mode the initial one
does nothing (Namhyung).
https://lore.kernel.org/lkml/[email protected]/
v2: Two additional fixes, bring back --header-only early exit
(Namhyung) and some comment/whitespace nits.
https://lore.kernel.org/linux-perf-users/[email protected]/
v1: https://lore.kernel.org/lkml/[email protected]/
Ian Rogers (8):
perf clockid: Add missing include
perf header: Add utility to convert feature number to a string perf header: Properly warn/print when libtraceevent/libbpf support is
missing
perf session: Extra logging for failed to process events perf header: Refactor pipe mode end marker handling
perf ordered-events: Event processing consistency with the regular
reader
perf evsel: Make unknown event names more unique
perf data convert ctf: Pipe mode improvements
tools/perf/builtin-annotate.c | 11 +-- tools/perf/builtin-report.c | 27 +++--- tools/perf/builtin-script.c | 11 +-- tools/perf/util/clockid.h | 3 +- tools/perf/util/data-convert-bt.c | 72 ++++++++++++-- tools/perf/util/data-convert-json.c | 12 +-- tools/perf/util/evsel.c | 7 +- tools/perf/util/header.c | 145 +++++++++++++++++++++------- tools/perf/util/header.h | 6 +- tools/perf/util/intel-tpebs.c | 11 +-- tools/perf/util/ordered-events.c | 2 +- tools/perf/util/session.c | 28 ++++-- 12 files changed, 217 insertions(+), 118 deletions(-)
Signature · 1 line
2.53.0.1118.gaef5881109-goog