CoreSight July 2024

coresight@lists.linaro.org

8 participants
36 discussions

[PATCH v4 0/3] coresight: Add static trace id support

by Mao Jinlong

Some HW has static trace id which cannot be changed via software programming. For this case, configure the trace id in device tree with "arm,static-trace-id = <xxx>", and call coresight_trace_id_get_static_system_id with the trace id value in device probe function. The id will be reserved for the HW all the time if the device is probed. Changes since V3: 1. Adda new API function int coresight_trace_id_get_system_static_id(int trace_id). 2. Use the term "static trace id" for these devices where the hardware sets a non-programmable trace ID. Changes since V2: 1. Change "trace-id" to "arm,trace-id". 2. Add trace id flag for getting preferred id or ODD id. Changes since V1: 1. Add argument to coresight_trace_id_get_system_id for preferred id instead of adding new function coresight_trace_id_reserve_system_id. 2. Add constraint to trace-id in dt-binding file. Mao Jinlong (3): dt-bindings: arm: Add arm,trace-id for coresight dummy source coresight: Add support to get static id for system trace sources coresight: dummy: Add static trace id support for dummy source .../sysfs-bus-coresight-devices-dummy-source | 15 +++++ .../arm/arm,coresight-dummy-source.yaml | 6 ++ drivers/hwtracing/coresight/coresight-dummy.c | 59 +++++++++++++++++-- .../hwtracing/coresight/coresight-platform.c | 26 ++++++++ .../hwtracing/coresight/coresight-trace-id.c | 38 ++++++++---- .../hwtracing/coresight/coresight-trace-id.h | 9 +++ include/linux/coresight.h | 1 + 7 files changed, 140 insertions(+), 14 deletions(-) create mode 100644 Documentation/ABI/testing/sysfs-bus-coresight-devices-dummy-source -- 2.41.0

2 weeks, 1 day

[PATCH v6 00/17] coresight: Use per-sink trace ID maps for Perf sessions

by James Clark

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 v5: * Hide queue number printout behind -v option * Style change in cs_etm__process_aux_output_hw_id() * Move new format enum to an earlier commit to reduce churn 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 | 629 +++++++++++------- tools/perf/util/cs-etm.h | 12 +- 25 files changed, 650 insertions(+), 475 deletions(-) create mode 100644 tools/perf/util/cs-etm-decoder/cs-etm-min-version.h -- 2.34.1

1 month, 3 weeks

[PATCH] coresight: tmc: sg: Do not leak sg_table

by Suzuki K Poulose

Running perf with cs_etm on Juno triggers the following kmemleak warning ! :~# cat /sys/kernel/debug/kmemleak unreferenced object 0xffffff8806b6d720 (size 96): comm "perf", pid 562, jiffies 4297810960 hex dump (first 32 bytes): 38 d8 13 07 88 ff ff ff 00 d0 9e 85 c0 ff ff ff 8............... 00 10 00 88 c0 ff ff ff 00 f0 ff f7 ff 00 00 00 ................ backtrace (crc 1dbf6e00): [<ffffffc08107381c>] kmemleak_alloc+0xbc/0xd8 [<ffffffc0802f9798>] kmalloc_trace_noprof+0x220/0x2e8 [<ffffffc07bb71948>] tmc_alloc_sg_table+0x48/0x208 [coresight_tmc] [<ffffffc07bb71cbc>] tmc_etr_alloc_sg_buf+0xac/0x240 [coresight_tmc] [<ffffffc07bb72538>] tmc_alloc_etr_buf.constprop.0+0x1f0/0x260 [coresight_tmc] [<ffffffc07bb7280c>] alloc_etr_buf.constprop.0.isra.0+0x74/0xa8 [coresight_tmc] [<ffffffc07bb72950>] tmc_alloc_etr_buffer+0x110/0x260 [coresight_tmc] [<ffffffc07bb38afc>] etm_setup_aux+0x204/0x3b0 [coresight] [<ffffffc08025837c>] rb_alloc_aux+0x20c/0x318 [<ffffffc08024dd84>] perf_mmap+0x2e4/0x7a0 [<ffffffc0802cceb0>] mmap_region+0x3b0/0xa08 [<ffffffc0802cd8a8>] do_mmap+0x3a0/0x500 [<ffffffc080295328>] vm_mmap_pgoff+0x100/0x1d0 [<ffffffc0802cadf8>] ksys_mmap_pgoff+0xb8/0x110 [<ffffffc080020688>] __arm64_sys_mmap+0x38/0x58 [<ffffffc080028fc0>] invoke_syscall.constprop.0+0x58/0x100 This due to the fact that we do not free the "sg_table" itself while freeing up the SG table and data pages. Fix this by freeing the sg_table in tmc_free_sg_table(). Fixes: 99443ea19e8b ("coresight: Add generic TMC sg table framework") Cc: Mike Leach <mike.leach(a)linaro.org> Cc: James Clark <james.clark(a)arm.com> Signed-off-by: Suzuki K Poulose <suzuki.poulose(a)arm.com> --- drivers/hwtracing/coresight/coresight-tmc-etr.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/hwtracing/coresight/coresight-tmc-etr.c b/drivers/hwtracing/coresight/coresight-tmc-etr.c index e75428fa1592..610ad51cda65 100644 --- a/drivers/hwtracing/coresight/coresight-tmc-etr.c +++ b/drivers/hwtracing/coresight/coresight-tmc-etr.c @@ -261,6 +261,7 @@ void tmc_free_sg_table(struct tmc_sg_table *sg_table) { tmc_free_table_pages(sg_table); tmc_free_data_pages(sg_table); + kfree(sg_table); } EXPORT_SYMBOL_GPL(tmc_free_sg_table); @@ -342,7 +343,6 @@ struct tmc_sg_table *tmc_alloc_sg_table(struct device *dev, rc = tmc_alloc_table_pages(sg_table); if (rc) { tmc_free_sg_table(sg_table); - kfree(sg_table); return ERR_PTR(rc); } -- 2.34.1

2 months, 1 week

[PATCH v2 0/3] Add source filtering for multi-port output

by Tao Zhang

In our hardware design, by combining a funnel and a replicator, it implement a hardware device with one-to-one correspondence between output ports and input ports. The programming usage on this device is the same as funnel. The software uses a funnel and a static replicator to implement the driver of this device. Since original funnels only support a single output connection and original replicator only support a single input connection, the code needs to be modified to support this new feature. The following is a typical topology diagram of multi-port output mechanism. |----------| |---------| |----------| |---------| | TPDM 0 | | Source0 | | Source 1 | | TPDM 1 | |----------| |---------| |----------| |---------| | | | | | | | | | --------- | | | | | | | | | | | | | | | \-------------/ ---------------------- | \ Funnel 0 / | | ----------- | ------------------------------ | | | | | | \------------------/ \ Funnel 1 / ----| \--------------/ | | |----> Combine a funnel and a | | static replicator /-----------------\ | / replicator 0 \ ----| /---------------------\ | | | | | |-----------| | |---------| | | |TPDM0 |TPDM1 | \-----------------/ | \ TPDA 0 / | \-------------/ | | | | |Source0/1 | \-------------------------------/ \ Funnel 2 / \---------------------------/ Changes in V2: 1. Change the reference for endpoint property in dt-binding. -- Krzysztof Kozlowski 2. Change the property name "filter_src" to "filter-src". -- Krzysztof Kozlowski 3. Fix the errors in running 'make dt_binding_check'. -- Rob Herring 4. Pass in the source parameter instead of path. -- Suzuki K Poulose 5. Reset the "filter_src_dev" if the "src" csdev is being removed. -- Suzuki K Poulose 6. Add a warning if the "filter_src_dev" is of not the type DEV_TYPE_SOURCE. -- Suzuki K Poulose 7. Optimize the procedure for handling all possible cases. -- Suzuki K Poulose Changes in V1: 1. Add a static replicator connect to a funnel to implement the correspondence between the output ports and the input ports on funnels. -- Suzuki K Poulose 2. Add filter_src_dev and filter_src_dev phandle to "coresight_connection" struct, and populate them if there is one. -- Suzuki K Poulose 3. To look at the phandle and then fixup/remove the filter_src device in fixup/remove connections. -- Suzuki K Poulose 4. When TPDA reads DSB/CMB element size, it is implemented by looking up filter src device in the connections. -- Suzuki K Poulose Tao Zhang (3): dt-bindings: arm: qcom,coresight-static-replicator: Add property for source filtering coresight: Add source filtering for multi-port output coresight-tpda: Optimize the function of reading element size .../arm/arm,coresight-static-replicator.yaml | 19 ++- drivers/hwtracing/coresight/coresight-core.c | 125 ++++++++++++++++-- .../hwtracing/coresight/coresight-platform.c | 18 +++ drivers/hwtracing/coresight/coresight-tpda.c | 3 +- include/linux/coresight.h | 5 + 5 files changed, 155 insertions(+), 15 deletions(-) -- 2.17.1

2 months, 3 weeks

Re: [PATCH] perf scripts python arm-cs-trace-disasm.py: Skip disasm if address continuity is broken

by James Clark

On 24/07/2024 7:38 am, Ganapatrao Kulkarni wrote: > > > On 23-07-2024 09:16 pm, James Clark wrote: >> >> >> On 23/07/2024 4:26 pm, Ganapatrao Kulkarni wrote: >>> >>> >>> On 23-07-2024 06:40 pm, James Clark wrote: >>>> >>>> >>>> On 22/07/2024 11:02 am, Ganapatrao Kulkarni wrote: >>>>> >>>>> Hi James, >>>>> >>>>> On 19-07-2024 08:09 pm, James Clark wrote: >>>>>> >>>>>> >>>>>> On 19/07/2024 10:26 am, Ganapatrao Kulkarni wrote: >>>>>>> To generate the instruction tracing, script uses 2 contiguous >>>>>>> packets >>>>>>> address range. If there a continuity brake due to discontiguous >>>>>>> branch >>>>>>> address, it is required to reset the tracing and start tracing >>>>>>> with the >>>>>>> new set of contiguous packets. >>>>>>> >>>>>>> Adding change to identify the break and complete the remaining >>>>>>> tracing >>>>>>> of current packets and restart tracing from new set of packets, if >>>>>>> continuity is established. >>>>>>> >>>>>> >>>>>> Hi Ganapatrao, >>>>>> >>>>>> Can you add a before and after example of what's changed to the >>>>>> commit message? It wasn't immediately obvious to me if this is >>>>>> adding missing output, or it was correcting the tail end of the >>>>>> output that was previously wrong. >>>>> >>>>> It is adding tail end of the trace as well avoiding the segfault of >>>>> the perf application. With out this change the perf segfaults with >>>>> as below log >>>>> >>>>> >>>>> ./perf script >>>>> --script=python:./scripts/python/arm-cs-trace-disasm.py -- -d >>>>> objdump -k ../../vmlinux -v $* > dump >>>>> objdump: error: the stop address should be after the start address >>>>> Traceback (most recent call last): >>>>> File "./scripts/python/arm-cs-trace-disasm.py", line 271, in >>>>> process_event >>>>> print_disam(dso_fname, dso_vm_start, start_addr, stop_addr) >>>>> File "./scripts/python/arm-cs-trace-disasm.py", line 105, in >>>>> print_disam >>>>> for line in read_disam(dso_fname, dso_start, start_addr, >>>>> stop_addr): >>>>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ >>>>> File "./scripts/python/arm-cs-trace-disasm.py", line 99, in >>>>> read_disam >>>>> disasm_output = check_output(disasm).decode('utf-8').split('\n') >>>>> ^^^^^^^^^^^^^^^^^^^^ >>>>> File "/usr/lib64/python3.12/subprocess.py", line 466, in >>>>> check_output >>>>> return run(*popenargs, stdout=PIPE, timeout=timeout, check=True, >>>>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ >>>>> File "/usr/lib64/python3.12/subprocess.py", line 571, in run >>>>> raise CalledProcessError(retcode, process.args, >>>>> subprocess.CalledProcessError: Command '['objdump', '-d', '-z', >>>>> '--start-address=0xffff80008125b758', >>>>> '--stop-address=0xffff80008125a934', '../../vmlinux']' returned >>>>> non-zero exit status 1. >>>>> Fatal Python error: handler_call_die: problem in Python trace event >>>>> handler >>>>> Python runtime state: initialized >>>>> >>>>> Current thread 0x0000ffffb05054e0 (most recent call first): >>>>> <no Python frame> >>>>> >>>>> Extension modules: perf_trace_context, systemd._journal, >>>>> systemd._reader, systemd.id128, report._py3report, _dbus_bindings, >>>>> problem._py3abrt (total: 7) >>>>> Aborted (core dumped) >>>>> >>>>>> >>>>>>> Signed-off-by: Ganapatrao Kulkarni >>>>>>> <gankulkarni(a)os.amperecomputing.com> >>>>>>> --- >>>>>>> tools/perf/scripts/python/arm-cs-trace-disasm.py | 10 ++++++++++ >>>>>>> 1 file changed, 10 insertions(+) >>>>>>> >>>>>>> diff --git a/tools/perf/scripts/python/arm-cs-trace-disasm.py >>>>>>> b/tools/perf/scripts/python/arm-cs-trace-disasm.py >>>>>>> index d973c2baed1c..ad10cee2c35e 100755 >>>>>>> --- a/tools/perf/scripts/python/arm-cs-trace-disasm.py >>>>>>> +++ b/tools/perf/scripts/python/arm-cs-trace-disasm.py >>>>>>> @@ -198,6 +198,10 @@ def process_event(param_dict): >>>>>>> cpu_data[str(cpu) + 'addr'] = addr >>>>>>> return >>>>>>> + if (cpu_data.get(str(cpu) + 'ip') == None): >>>>>>> + cpu_data[str(cpu) + 'ip'] = ip >>>>>>> + >>>>>> >>>>>> Do you need to write into the global cpu_data here? Doesn't it get >>>>>> overwritten after you load it back into 'prev_ip' >>>>> >>>>> No, the logic is same as holding the addr of previous packet. >>>>> Saving the previous packet saved ip in to prev_ip before >>>>> overwriting with the current packet. >>>> >>>> It's not exactly the same logic as holding the addr of the previous >>>> sample. For addr, we return on the first None, with your change we >>>> now "pretend" that the second one is also the previous one: >>>> >>>> if (cpu_data.get(str(cpu) + 'addr') == None): >>>> cpu_data[str(cpu) + 'addr'] = addr >>>> return <----------------------------sample 0 return >>>> >>>> if (cpu_data.get(str(cpu) + 'ip') == None): >>>> cpu_data[str(cpu) + 'ip'] = ip <---- sample 1 save but no return >>>> >>>> Then for sample 1 'prev_ip' is actually now the 'current' IP: >>> >>> Yes, it is dummy for first packet. Added anticipating that we wont >>> hit the discontinuity for the first packet itself. >>> >>> Can this be changed to more intuitive like below? >>> >>> diff --git a/tools/perf/scripts/python/arm-cs-trace-disasm.py >>> b/tools/perf/scripts/python/arm-cs-trace-disasm.py >>> index d973c2baed1c..d49f5090059f 100755 >>> --- a/tools/perf/scripts/python/arm-cs-trace-disasm.py >>> +++ b/tools/perf/scripts/python/arm-cs-trace-disasm.py >>> @@ -198,6 +198,8 @@ def process_event(param_dict): >>> cpu_data[str(cpu) + 'addr'] = addr >>> return >>> >>> + if (cpu_data.get(str(cpu) + 'ip') != None): >>> + prev_ip = cpu_data[str(cpu) + 'ip'] >>> >>> if (options.verbose == True): >>> print("Event type: %s" % name) >>> @@ -243,12 +245,18 @@ def process_event(param_dict): >>> >>> # Record for previous sample packet >>> cpu_data[str(cpu) + 'addr'] = addr >>> + cpu_data[str(cpu) + 'ip'] = stop_addr >>> >>> # Handle CS_ETM_TRACE_ON packet if start_addr=0 and stop_addr=4 >>> if (start_addr == 0 and stop_addr == 4): >>> print("CPU%d: CS_ETM_TRACE_ON packet is inserted" % >>> cpu) >>> return >>> >>> + if (stop_addr < start_addr and prev_ip != 0): >>> + # Continuity of the Packets broken, set start_addr to >>> previous >>> + # packet ip to complete the remaining tracing of the >>> address range. >>> + start_addr = prev_ip >>> + >>> if (start_addr < int(dso_start) or start_addr > int(dso_end)): >>> print("Start address 0x%x is out of range [ 0x%x .. >>> 0x%x ] for dso %s" % (start_addr, int(dso_start), int(dso_end), dso)) >>> return >>> >>> Without this patch below is the failure log(with segfault) for >>> reference. >>> >>> [root@sut01sys-r214 perf]# timeout 4s ./perf record -e cs_etm// -C 1 >>> dd if=/dev/zero of=/dev/null >>> [ perf record: Woken up 1 times to write data ] >>> [ perf record: Captured and wrote 1.087 MB perf.data ] >>> [root@sut01sys-r214 perf]# ./perf script >>> --script=python:./scripts/python/arm-cs-trace-disasm.py -- -d objdump >>> -k ../../vmlinux -v $* > dump >>> objdump: error: the stop address should be after the start address >>> Traceback (most recent call last): >>> File "./scripts/python/arm-cs-trace-disasm.py", line 271, in >>> process_event >>> print_disam(dso_fname, dso_vm_start, start_addr, stop_addr) >>> File "./scripts/python/arm-cs-trace-disasm.py", line 105, in >>> print_disam >>> for line in read_disam(dso_fname, dso_start, start_addr, >>> stop_addr): >>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ >>> File "./scripts/python/arm-cs-trace-disasm.py", line 99, in >>> read_disam >>> disasm_output = check_output(disasm).decode('utf-8').split('\n') >>> ^^^^^^^^^^^^^^^^^^^^ >>> File "/usr/lib64/python3.12/subprocess.py", line 466, in check_output >>> return run(*popenargs, stdout=PIPE, timeout=timeout, check=True, >>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ >>> File "/usr/lib64/python3.12/subprocess.py", line 571, in run >>> raise CalledProcessError(retcode, process.args, >>> subprocess.CalledProcessError: Command '['objdump', '-d', '-z', >>> '--start-address=0xffff80008125b758', >>> '--stop-address=0xffff80008125a934', '../../vmlinux']' returned >>> non-zero exit status 1. >>> Fatal Python error: handler_call_die: problem in Python trace event >>> handler >>> Python runtime state: initialized >>> >>> Current thread 0x0000ffffb90d54e0 (most recent call first): >>> <no Python frame> >>> >>> Extension modules: perf_trace_context, systemd._journal, >>> systemd._reader, systemd.id128, report._py3report, _dbus_bindings, >>> problem._py3abrt (total: 7) >>> Aborted (core dumped) >>> >>> >>> dump snippet: >>> ============ >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff80008030cb00 phys_addr: >>> 0x0000000000000000 ip: 0xffff800080313f0c pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff800080313f04 <__perf_event_header__init_id+0x4c>: >>> ffff800080313f04: 36100094 tbz w20, #2, >>> ffff800080313f14 <__perf_event_header__init_id+0x5c> >>> ffff800080313f08: f941e6a0 ldr x0, [x21, #968] >>> ffff800080313f0c: d63f0000 blr x0 >>> perf 12720/12720 [0001] 5986.372298040 >>> __perf_event_header__init_id+0x54 >>> .../coresight/linux/kernel/events/core.c 586 return >>> event->clock(); >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff8000801bb4a8 phys_addr: >>> 0x0000000000000000 ip: 0xffff80008030cb0c pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff80008030cb00 <local_clock>: >>> ffff80008030cb00: d503233f paciasp >>> ffff80008030cb04: a9bf7bfd stp x29, x30, >>> [sp, #-16]! >>> ffff80008030cb08: 910003fd mov x29, sp >>> ffff80008030cb0c: 97faba67 bl ffff8000801bb4a8 >>> <sched_clock> >>> perf 12720/12720 [0001] 5986.372298040 >>> local_clock+0xc ...t/linux/./include/linux/sched/clock.h 64 return >>> sched_clock(); >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff80008125a8a8 phys_addr: >>> 0x0000000000000000 ip: 0xffff8000801bb4c8 pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff8000801bb4a8 <sched_clock>: >>> ffff8000801bb4a8: d503233f paciasp >>> ffff8000801bb4ac: a9be7bfd stp x29, x30, >>> [sp, #-32]! >>> ffff8000801bb4b0: 910003fd mov x29, sp >>> ffff8000801bb4b4: a90153f3 stp x19, x20, >>> [sp, #16] >>> ffff8000801bb4b8: d5384113 mrs x19, sp_el0 >>> ffff8000801bb4bc: b9401260 ldr w0, [x19, #16] >>> ffff8000801bb4c0: 11000400 add w0, w0, #0x1 >>> ffff8000801bb4c4: b9001260 str w0, [x19, #16] >>> ffff8000801bb4c8: 94427cf8 bl ffff80008125a8a8 >>> <sched_clock_noinstr> >>> perf 12720/12720 [0001] 5986.372298040 >>> sched_clock+0x20 ...sight/linux/kernel/time/sched_clock.c 105 ns = >>> sched_clock_noinstr(); >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff80008125b758 phys_addr: >>> 0x0000000000000000 ip: 0xffff80008125a8e4 pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff80008125a8a8 <sched_clock_noinstr>: >>> ffff80008125a8a8: d503233f paciasp >>> ffff80008125a8ac: a9bc7bfd stp x29, x30, >>> [sp, #-64]! >>> ffff80008125a8b0: 910003fd mov x29, sp >>> ffff80008125a8b4: a90153f3 stp x19, x20, >>> [sp, #16] >>> ffff80008125a8b8: b000e354 adrp x20, >>> ffff800082ec3000 <tick_bc_dev+0x140> >>> ffff80008125a8bc: 910d0294 add x20, x20, >>> #0x340 >>> ffff80008125a8c0: a90363f7 stp x23, x24, >>> [sp, #48] >>> ffff80008125a8c4: 91002297 add x23, x20, #0x8 >>> ffff80008125a8c8: 52800518 mov w24, #0x28 >>> // #40 >>> ffff80008125a8cc: a9025bf5 stp x21, x22, >>> [sp, #32] >>> ffff80008125a8d0: b9400296 ldr w22, [x20] >>> ffff80008125a8d4: 120002d5 and w21, w22, #0x1 >>> ffff80008125a8d8: 9bb87eb5 umull x21, w21, w24 >>> ffff80008125a8dc: 8b1502f3 add x19, x23, x21 >>> ffff80008125a8e0: f9400e60 ldr x0, [x19, #24] >>> ffff80008125a8e4: d63f0000 blr x0 >>> perf 12720/12720 [0001] 5986.372298040 >>> sched_clock_noinstr+0x3c ...sight/linux/kernel/time/sched_clock.c >>> 93 cyc = (rd->read_sched_clock() - rd->epoch_cyc) & >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff8000801bb4cc phys_addr: >>> 0x0000000000000000 ip: 0xffff80008125a930 pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> >>> >>> With fix: >>> ========= >>> >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff80008030cb00 phys_addr: >>> 0x0000000000000000 ip: 0xffff800080313f0c pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff800080313f04 <__perf_event_header__init_id+0x4c>: >>> ffff800080313f04: 36100094 tbz w20, #2, >>> ffff800080313f14 <__perf_event_header__init_id+0x5c> >>> ffff800080313f08: f941e6a0 ldr x0, [x21, #968] >>> ffff800080313f0c: d63f0000 blr x0 >>> perf 12720/12720 [0001] 5986.372298040 >>> __perf_event_header__init_id+0x54 >>> .../coresight/linux/kernel/events/core.c 586 return >>> event->clock(); >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff8000801bb4a8 phys_addr: >>> 0x0000000000000000 ip: 0xffff80008030cb0c pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff80008030cb00 <local_clock>: >>> ffff80008030cb00: d503233f paciasp >>> ffff80008030cb04: a9bf7bfd stp x29, x30, >>> [sp, #-16]! >>> ffff80008030cb08: 910003fd mov x29, sp >>> ffff80008030cb0c: 97faba67 bl ffff8000801bb4a8 >>> <sched_clock> >>> perf 12720/12720 [0001] 5986.372298040 >>> local_clock+0xc ...t/linux/./include/linux/sched/clock.h 64 return >>> sched_clock(); >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff80008125a8a8 phys_addr: >>> 0x0000000000000000 ip: 0xffff8000801bb4c8 pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff8000801bb4a8 <sched_clock>: >>> ffff8000801bb4a8: d503233f paciasp >>> ffff8000801bb4ac: a9be7bfd stp x29, x30, >>> [sp, #-32]! >>> ffff8000801bb4b0: 910003fd mov x29, sp >>> ffff8000801bb4b4: a90153f3 stp x19, x20, >>> [sp, #16] >>> ffff8000801bb4b8: d5384113 mrs x19, sp_el0 >>> ffff8000801bb4bc: b9401260 ldr w0, [x19, #16] >>> ffff8000801bb4c0: 11000400 add w0, w0, #0x1 >>> ffff8000801bb4c4: b9001260 str w0, [x19, #16] >>> ffff8000801bb4c8: 94427cf8 bl ffff80008125a8a8 >>> <sched_clock_noinstr> >>> perf 12720/12720 [0001] 5986.372298040 >>> sched_clock+0x20 ...sight/linux/kernel/time/sched_clock.c 105 ns = >>> sched_clock_noinstr(); >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff80008125b758 phys_addr: >>> 0x0000000000000000 ip: 0xffff80008125a8e4 pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff80008125a8a8 <sched_clock_noinstr>: >>> ffff80008125a8a8: d503233f paciasp >>> ffff80008125a8ac: a9bc7bfd stp x29, x30, >>> [sp, #-64]! >>> ffff80008125a8b0: 910003fd mov x29, sp >>> ffff80008125a8b4: a90153f3 stp x19, x20, >>> [sp, #16] >>> ffff80008125a8b8: b000e354 adrp x20, >>> ffff800082ec3000 <tick_bc_dev+0x140> >>> ffff80008125a8bc: 910d0294 add x20, x20, >>> #0x340 >>> ffff80008125a8c0: a90363f7 stp x23, x24, >>> [sp, #48] >>> ffff80008125a8c4: 91002297 add x23, x20, #0x8 >>> ffff80008125a8c8: 52800518 mov w24, #0x28 >>> // #40 >>> ffff80008125a8cc: a9025bf5 stp x21, x22, >>> [sp, #32] >>> ffff80008125a8d0: b9400296 ldr w22, [x20] >>> ffff80008125a8d4: 120002d5 and w21, w22, #0x1 >>> ffff80008125a8d8: 9bb87eb5 umull x21, w21, w24 >>> ffff80008125a8dc: 8b1502f3 add x19, x23, x21 >>> ffff80008125a8e0: f9400e60 ldr x0, [x19, #24] >>> ffff80008125a8e4: d63f0000 blr x0 >> >> It looks like the disassembly now assumes this BLR wasn't taken. We go >> from ffff80008125a8e4 straight through to ... >> >>> perf 12720/12720 [0001] 5986.372298040 >>> sched_clock_noinstr+0x3c ...sight/linux/kernel/time/sched_clock.c >>> 93 cyc = (rd->read_sched_clock() - rd->epoch_cyc) & >>> Event type: branches >>> Sample = { cpu: 0001 addr: 0xffff8000801bb4cc phys_addr: >>> 0x0000000000000000 ip: 0xffff80008125a930 pid: 12720 tid: 12720 >>> period: 1 time: 5986372298040 } >>> ffff80008125a8e8 <sched_clock_noinstr+0x40>: >>> ffff80008125a8e8: f8756ae3 ldr x3, [x23, x21] >> >> ffff80008125a8e4 which is just the previous one +4. Isn't your issue >> actually a decode issue in Perf itself? Why is there a discontinuity >> without branch samples being generated where either the source or >> destination address is 0? >> >> What are your record options to create this issue? As I mentioned in >> the previous reply I haven't been able to reproduce it. > > I am using below perf record command. > > timeout 4s ./perf record -e cs_etm// -C 1 dd if=/dev/zero of=/dev/null > Thanks I managed to reproduce it. I'll take a look to see if I think the issue is somewhere else. >> > > Thanks, > Ganapat

3 months

[PATCH v4 0/2] coresight: core: Add device name support

by Mao Jinlong

With current design, the name of the non-cpu bounded coresight component is the device type with the number. And with 'ls' command we can get the register address of the component. But from these information, we can't know what the HW or system the component belongs to. Add device-name in DT to support it. cti_sys0 -> ../../../devices/platform/soc(a)0/138f0000.cti/cti_sys0 cti_sys1 -> ../../../devices/platform/soc(a)0/13900000.cti/cti_sys1 tpdm0 -> ../../../devices/platform/soc(a)0/10b0d000.tpdm/tpdm0 tpdm1 -> ../../../devices/platform/soc(a)0/10c28000.tpdm/tpdm1 tpdm2 -> ../../../devices/platform/soc(a)0/10c29000.tpdm/tpdm2 Change since V3: 1. Change device-name to arm,cs-dev-name. 2. Add arm,cs-dev-name to only CTI and sources' dt-binding. Change since V2: 1. Fix the error in coresight core. drivers/hwtracing/coresight/coresight-core.c:1775:7: error: assigning to 'char *' from 'const char *' discards qualifiers 2. Fix the warning when run dtbinding check. Documentation/devicetree/bindings/arm/arm,coresight-cpu-debug.yaml: device-name: missing type definition Change since V1: 1. Change coresight-name to device name. 2. Add the device-name in coresight dt bindings. Mao Jinlong (2): coresight: core: Add device name support dt-bindings: arm: Add device-name in the coresight components .../bindings/arm/arm,coresight-catu.yaml | 6 +++ .../bindings/arm/arm,coresight-cpu-debug.yaml | 6 +++ .../bindings/arm/arm,coresight-cti.yaml | 6 +++ .../arm/arm,coresight-dummy-sink.yaml | 6 +++ .../arm/arm,coresight-dummy-source.yaml | 6 +++ .../arm/arm,coresight-dynamic-funnel.yaml | 6 +++ .../arm/arm,coresight-dynamic-replicator.yaml | 6 +++ .../bindings/arm/arm,coresight-etb10.yaml | 6 +++ .../bindings/arm/arm,coresight-etm.yaml | 6 +++ .../arm/arm,coresight-static-funnel.yaml | 6 +++ .../arm/arm,coresight-static-replicator.yaml | 6 +++ .../bindings/arm/arm,coresight-stm.yaml | 6 +++ .../bindings/arm/arm,coresight-tmc.yaml | 6 +++ .../bindings/arm/arm,coresight-tpiu.yaml | 6 +++ .../bindings/arm/qcom,coresight-tpda.yaml | 6 +++ .../bindings/arm/qcom,coresight-tpdm.yaml | 6 +++ drivers/hwtracing/coresight/coresight-core.c | 37 ++++++++++--------- .../hwtracing/coresight/coresight-platform.c | 31 ++++++++++++++++ include/linux/coresight.h | 3 +- 19 files changed, 149 insertions(+), 18 deletions(-) Mao Jinlong (2): dt-bindings: arm: Add device-name in the coresight components coresight: core: Add device name support .../bindings/arm/arm,coresight-cti.yaml | 6 +++ .../arm/arm,coresight-dummy-source.yaml | 6 +++ .../bindings/arm/arm,coresight-stm.yaml | 6 +++ .../bindings/arm/qcom,coresight-tpdm.yaml | 6 +++ drivers/hwtracing/coresight/coresight-core.c | 37 ++++++++++--------- .../hwtracing/coresight/coresight-platform.c | 30 +++++++++++++++ include/linux/coresight.h | 3 +- 7 files changed, 76 insertions(+), 18 deletions(-) -- 2.41.0

3 months

Re: [PATCH v2 1/4] perf auxtrace: Iterate all AUX events when finish reading

by Arnaldo Carvalho de Melo

On Mon, Jul 29, 2024 at 09:48:57AM +0100, Leo Yan wrote: > On 7/25/2024 4:40 PM, Adrian Hunter wrote: > > On 25/07/24 09:46, Leo Yan wrote: > >> When finished to read AUX trace data from mmaped buffer, based on the > > > > Here and elsewhere 'mmapped' is more common than 'mmaped' so maybe: > > > > mmaped -> mmapped > > Will fix. > > >> AUX buffer index the core layer needs to search the corresponding PMU > >> event and re-enable it to continue tracing. > >> > >> However, current code only searches the first AUX event. It misses to > >> search other enabled AUX events, thus, it returns failure if the buffer > >> index does not belong to the first AUX event. > >> > >> This patch extends the auxtrace_record__read_finish() function to > >> search for every enabled AUX events, so all the mmaped buffer indexes > >> can be covered. > > > > Looking at this again, a couple more things came to mind - see below. > > > >> > >> Signed-off-by: Leo Yan <leo.yan(a)arm.com> > >> --- > >> tools/perf/util/auxtrace.c | 25 ++++++++++++++++++++----- > >> 1 file changed, 20 insertions(+), 5 deletions(-) > >> > >> diff --git a/tools/perf/util/auxtrace.c b/tools/perf/util/auxtrace.c > >> index b99e72f7da88..61835a6a9ea3 100644 > >> --- a/tools/perf/util/auxtrace.c > >> +++ b/tools/perf/util/auxtrace.c > >> @@ -670,18 +670,33 @@ static int evlist__enable_event_idx(struct evlist *evlist, struct evsel *evsel, > >> int auxtrace_record__read_finish(struct auxtrace_record *itr, int idx) > >> { > >> struct evsel *evsel; > >> + int ret = -EINVAL; > >> > >> - if (!itr->evlist || !itr->pmu) > >> + if (!itr->evlist) > >> return -EINVAL; > >> > >> evlist__for_each_entry(itr->evlist, evsel) { > >> - if (evsel->core.attr.type == itr->pmu->type) { > >> + if (evsel__is_aux_event(evsel)) { > >> if (evsel->disabled) > >> - return 0; > >> - return evlist__enable_event_idx(itr->evlist, evsel, idx); > >> + continue; > > > > That will make the evsel->disabled case an error, which > > might be a problem. Possibly the event can be disabled > > (e.g. via control fifo) but still have data to read out. > > If so, we need to extend evlist__enable_event_idx() or create a new function > to check if the memory index is belonged to an evsel. If the idx is found for > an evsel, then we can check the evsel->disabled flag. > > >> + /* > >> + * Multiple AUX events might be opened in a session. > >> + * Bail out for success case as the AUX event has been > >> + * found and enabled, otherwise, continue to check if > >> + * the next AUX event can cover the mmaped buffer with > >> + * 'idx'. > >> + */ > > > > Thinking about this some more, raised some questions: > > > > How do we know there is only one AUX event per mmap? > > On my test platform, there have two Arm SPE events, the first one's cpumask is > 2-5 and the second SPE's cpumask is 6-7. The AUX events do not intersect CPU > maps. Therefore, a mmapped AUX buffer only binds to a AUX event. > > I think we can add a checking in the function record__auxtrace_init(). After > it calls auxtrace_record__init(), we can report failure if the AUX events have > the overlapped cpumask. > > > They would have to be on different CPUs for that to be true? > > Yes. > > > And wouldn't --per-thread have all AUX events in every mmap? > > Before I roughly tested '--per-thread' mode and did not see issue. But this > time I tested '--per-thread' mode and found the failure by the kernel checking > [1] - it does not allow the different AUX events to bind to the same FD. > > Here I need to dig a bit for two options, either we need to fix the perf > tool to open multiple AUX events for '--per-thread' mode, or remove the > kernel's checking to allow different AUX events to bind to same FD. > > Thanks a lot for review! Ok, the two patches from Adrian are in, I'll now wait for you to refresh this series, Thanks, - Arnaldo

3 months

Re: [PATCH V9 02/13] perf/core: Add aux_pause, aux_resume, aux_start_paused

by Arnaldo Carvalho de Melo

On Fri, Jul 26, 2024 at 07:00:55PM +0300, Adrian Hunter wrote: > On 26/07/24 17:41, Arnaldo Carvalho de Melo wrote: > > On Thu, Jul 18, 2024 at 06:06:16PM +0300, Adrian Hunter wrote: > >> On 18/07/24 15:58, Peter Zijlstra wrote: > >>> On Thu, Jul 18, 2024 at 01:51:26PM +0200, Peter Zijlstra wrote: > >>> > >>>> OK, let me do that and make a few more edits and see if I can stare at > >>>> that next patch some. > >>> > >>> I pushed out a stack of patches into queue.git perf/core > >>> Could you please double check I didn't wreck anything? > >> > >> Looks fine, and seems to work OK in a brief test. > >> > >> Thank you! :-) > > > > So should I go ahead and pick the tooling patches since the kernel bits > > are merged? > > Not exactly merged. Probably should wait until it is in tip at least. Ok, so I get just these, as you asked on another message: acme@x1:~/git/perf-tools-next$ git log --oneline -5 perf-tools-next/tmp.perf-tools-next 9140fec01b2de8d3 (HEAD -> perf-tools-next, perf-tools-next/tmp.perf-tools-next, perf-tools-next.korg/tmp.perf-tools-next, number/perf-tools-next, acme.korg/tmp.perf-tools-next) perf tools: Enable evsel__is_aux_event() to work for S390_CPUMSF c3b7dba6ea81a5b1 perf tools: Enable evsel__is_aux_event() to work for ARM/ARM64 866400c0b08ef9d9 perf scripts python cs-etm: Restore first sample log in verbose mode 08ee74eb03e37191 perf cs-etm: Output 0 instead of 0xdeadbeef when exception packets are flushed c22dd7ec2b2808b2 perf inject: Convert comma to semicolon acme@x1:~/git/perf-tools-next$ It'll go to perf-tools-next once 6.11-rc1 is out. - Arnaldo

3 months

Re: [PATCH V9 06/13] perf tools: Enable evsel__is_aux_event() to work for S390_CPUMSF

by Arnaldo Carvalho de Melo

On Tue, Jul 23, 2024 at 12:09:46PM +0300, Adrian Hunter wrote: > On 23/07/24 11:26, Leo Yan wrote: > > On 7/15/2024 5:07 PM, Adrian Hunter wrote: > >> > >> evsel__is_aux_event() identifies AUX area tracing selected events. > >> > >> S390_CPUMSF uses a raw event type (PERF_TYPE_RAW - refer > >> s390_cpumsf_evsel_is_auxtrace()) not a PMU type value that could be checked > >> in evsel__is_aux_event(). However it sets needs_auxtrace_mmap (refer > >> auxtrace_record__init()), so check that first. > >> > >> Currently, the features that use evsel__is_aux_event() are used only by > >> Intel PT, but that may change in the future. > >> > >> Signed-off-by: Adrian Hunter <adrian.hunter(a)intel.com> > >> Acked-by: Ian Rogers <irogers(a)google.com> > >> Reviewed-by: Andi Kleen <ak(a)linux.intel.com> > > > > Reviewed-by: Leo Yan <leo.yan(a)arm.com> > > Namhyung, could we get patches 5 and 6 from this series > applied? They are independent and Leo's new patch set is > dependent on patch 5. I'm getting those two into tmp.perf-tools-next, probably should pick the rest as peterz has the kernel bits already in his queue.git/perf/core, right? - Arnaldo > > > >> --- > >> tools/perf/util/pmu.c | 6 +++++- > >> 1 file changed, 5 insertions(+), 1 deletion(-) > >> > >> diff --git a/tools/perf/util/pmu.c b/tools/perf/util/pmu.c > >> index 986166bc7c78..0b38c51bd6eb 100644 > >> --- a/tools/perf/util/pmu.c > >> +++ b/tools/perf/util/pmu.c > >> @@ -1199,8 +1199,12 @@ void perf_pmu__warn_invalid_formats(struct perf_pmu *pmu) > >> > >> bool evsel__is_aux_event(const struct evsel *evsel) > >> { > >> - struct perf_pmu *pmu = evsel__find_pmu(evsel); > >> + struct perf_pmu *pmu; > >> > >> + if (evsel->needs_auxtrace_mmap) > >> + return true; > >> + > >> + pmu = evsel__find_pmu(evsel); > >> return pmu && pmu->auxtrace; > >> } > >> > >> -- > >> 2.34.1 > >> > >>

3 months

Re: [PATCH V9 02/13] perf/core: Add aux_pause, aux_resume, aux_start_paused

by Arnaldo Carvalho de Melo

On Thu, Jul 18, 2024 at 06:06:16PM +0300, Adrian Hunter wrote: > On 18/07/24 15:58, Peter Zijlstra wrote: > > On Thu, Jul 18, 2024 at 01:51:26PM +0200, Peter Zijlstra wrote: > > > >> OK, let me do that and make a few more edits and see if I can stare at > >> that next patch some. > > > > I pushed out a stack of patches into queue.git perf/core > > Could you please double check I didn't wreck anything? > > Looks fine, and seems to work OK in a brief test. > > Thank you! :-) So should I go ahead and pick the tooling patches since the kernel bits are merged? - Arnaldo

3 months

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

CoreSight July 2024