[PATCH v5 00/17] coresight: Use per-sink trace ID maps for Perf sessions

From: James Clark
Date: Fri Jul 12 2024 - 06:22:55 EST


This will allow sessions with more than CORESIGHT_TRACE_IDS_MAX ETMs
as long as there are fewer than that many ETMs connected to each sink.

Each sink owns its own trace ID map, and any Perf session connecting to
that sink will allocate from it, even if the sink is currently in use by
other users. This is similar to the existing behavior where the dynamic
trace IDs are constant as long as there is any concurrent Perf session
active. It's not completely optimal because slightly more IDs will be
used than necessary, but the optimal solution involves tracking the PIDs
of each session and allocating ID maps based on the session owner. This
is difficult to do with the combination of per-thread and per-cpu modes
and some scheduling issues. The complexity of this isn't likely to worth
it because even with multiple users they'd just see a difference in the
ordering of ID allocations rather than hitting any limits (unless the
hardware does have too many ETMs connected to one sink).

Per-thread mode works but only until there are any overlapping IDs, at
which point Perf will error out. Both per-thread mode and sysfs mode are
left to future changes, but both can be added on top of this initial
implementation and only sysfs mode requires further driver changes.

The HW_ID version field hasn't been bumped in order to not break Perf
which already has an error condition for other values of that field.
Instead a new minor version has been added which signifies that there
are new fields but the old fields are backwards compatible.

Changes since v4:

* Fix compilation failure when TRACE_ID_DEBUG is set
* Expand comment about not freeing individual trace IDs in
free_event_data()

Changes since v3:

* Fix issue where trace IDs were overwritten by possibly invalid ones
by Perf in unformatted mode. Now the HW_IDs are also used for
unformatted mode unless the kernel didn't emit any.
* Add a commit to check the OpenCSD version.
* Add a commit to not save invalid IDs in the Perf header.
* Replace cs_etm_queue's formatted and formatted_set members with a
single enum which is easier to use.
* Drop CORESIGHT_TRACE_ID_UNUSED_FLAG as it's no longer needed.
* Add a commit to print the queue number in the raw dump.
* Don't assert on the number of unformatted decoders if decoders == 0.


Changes since v2:

* Rebase on coresight-next 6.10-rc2 (b9b25c8496).
* Fix double free of csdev if device registration fails.
* Fix leak of coresight_trace_id_perf_start() if trace ID allocation
fails.
* Don't resend HW_ID for sink changes in per-thread mode. The existing
CPU field on AUX records can be used to track this instead.
* Tidy function doc for coresight_trace_id_release_all()
* Drop first two commits now that they are in coresight-next
* Add a commit to make the trace ID spinlock local to the map

Changes since V1:

* Rename coresight_device.perf_id_map to perf_sink_id_map.
* Instead of outputting a HW_ID for each reachable ETM, output
the sink ID and continue to output only the HW_ID once for
each mapping.
* Keep the first two Perf patches so that it applies cleanly
on coresight-next, although they have been applied on perf-tools-next
* Add new *_map() functions to the trace ID public API instead of
modifying existing ones.
* Collapse "coresight: Pass trace ID map into source enable" into
"coresight: Use per-sink trace ID maps for Perf sessions" because the
first commit relied on the default map being accessible which is no
longer necessary due to the previous bullet point.

James Clark (17):
perf: cs-etm: Create decoders after both AUX and HW_ID search passes
perf: cs-etm: Allocate queues for all CPUs
perf: cs-etm: Move traceid_list to each queue
perf: cs-etm: Create decoders based on the trace ID mappings
perf: cs-etm: Only save valid trace IDs into files
perf: cs-etm: Support version 0.1 of HW_ID packets
perf: cs-etm: Print queue number in raw trace dump
perf: cs-etm: Add runtime version check for OpenCSD
coresight: Remove unused ETM Perf stubs
coresight: Clarify comments around the PID of the sink owner
coresight: Move struct coresight_trace_id_map to common header
coresight: Expose map arguments in trace ID API
coresight: Make CPU id map a property of a trace ID map
coresight: Use per-sink trace ID maps for Perf sessions
coresight: Remove pending trace ID release mechanism
coresight: Emit sink ID in the HW_ID packets
coresight: Make trace ID map spinlock local to the map

drivers/hwtracing/coresight/coresight-core.c | 37 +-
drivers/hwtracing/coresight/coresight-dummy.c | 3 +-
.../hwtracing/coresight/coresight-etm-perf.c | 43 +-
.../hwtracing/coresight/coresight-etm-perf.h | 18 -
.../coresight/coresight-etm3x-core.c | 9 +-
.../coresight/coresight-etm4x-core.c | 9 +-
drivers/hwtracing/coresight/coresight-priv.h | 1 +
drivers/hwtracing/coresight/coresight-stm.c | 3 +-
drivers/hwtracing/coresight/coresight-sysfs.c | 3 +-
.../hwtracing/coresight/coresight-tmc-etr.c | 5 +-
drivers/hwtracing/coresight/coresight-tmc.h | 5 +-
drivers/hwtracing/coresight/coresight-tpdm.c | 3 +-
.../hwtracing/coresight/coresight-trace-id.c | 138 ++--
.../hwtracing/coresight/coresight-trace-id.h | 70 +-
include/linux/coresight-pmu.h | 17 +-
include/linux/coresight.h | 21 +-
tools/build/feature/test-libopencsd.c | 4 +-
tools/include/linux/coresight-pmu.h | 17 +-
tools/perf/Makefile.config | 2 +-
tools/perf/arch/arm/util/cs-etm.c | 11 +-
.../perf/util/cs-etm-decoder/cs-etm-decoder.c | 49 +-
.../perf/util/cs-etm-decoder/cs-etm-decoder.h | 3 +-
.../util/cs-etm-decoder/cs-etm-min-version.h | 13 +
tools/perf/util/cs-etm.c | 625 +++++++++++-------
tools/perf/util/cs-etm.h | 12 +-
25 files changed, 645 insertions(+), 476 deletions(-)
create mode 100644 tools/perf/util/cs-etm-decoder/cs-etm-min-version.h

--
2.34.1