Re: [PATCH v7] perf: Sharing PMU counters across compatible events

[Jiri Olsa]

On Fri, Nov 22, 2019 at 07:50:06PM +0000, Song Liu wrote:
> 
> 
> > On Nov 22, 2019, at 11:33 AM, Jiri Olsa <jolsa@xxxxxxxxxx> wrote:
> > 
> > On Fri, Nov 15, 2019 at 03:55:04PM -0800, Song Liu wrote:
> >> This patch tries to enable PMU sharing. When multiple perf_events are
> >> counting the same metric, they can share the hardware PMU counter. We
> >> call these events as "compatible events".
> >> 
> >> The PMU sharing are limited to events within the same perf_event_context
> >> (ctx). When a event is installed or enabled, search the ctx for compatible
> >> events. This is implemented in perf_event_setup_dup(). One of these
> >> compatible events are picked as the master (stored in event->dup_master).
> >> Similarly, when the event is removed or disabled, perf_event_remove_dup()
> >> is used to clean up sharing.
> >> 
> >> A new state PERF_EVENT_STATE_ENABLED is introduced for the master event.
> >> This state is used when the slave event is ACTIVE, but the master event
> >> is not.
> >> 
> >> On the critical paths (add, del read), sharing PMU counters doesn't
> >> increase the complexity. Helper functions event_pmu_[add|del|read]() are
> >> introduced to cover these cases. All these functions have O(1) time
> >> complexity.
> >> 
> >> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> >> Cc: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
> >> Cc: Jiri Olsa <jolsa@xxxxxxxxxx>
> >> Cc: Alexey Budankov <alexey.budankov@xxxxxxxxxxxxxxx>
> >> Cc: Namhyung Kim <namhyung@xxxxxxxxxx>
> >> Cc: Tejun Heo <tj@xxxxxxxxxx>
> >> Signed-off-by: Song Liu <songliubraving@xxxxxx>
> >> 
> >> ---
> >> Changes in v7:
> >> Major rewrite to avoid allocating extra master event.
> > 
> > hi,
> > what is this based on? I can't apply it on tip/master:
> > 
> > 	Applying: perf: Sharing PMU counters across compatible events
> > 	error: patch failed: include/linux/perf_event.h:722
> > 	error: include/linux/perf_event.h: patch does not apply
> > 	Patch failed at 0001 perf: Sharing PMU counters across compatible events
> > 	hint: Use 'git am --show-current-patch' to see the failed patch
> > 	When you have resolved this problem, run "git am --continue".
> > 	If you prefer to skip this patch, run "git am --skip" instead.
> > 	To restore the original branch and stop patching, run "git am --abort".
> 

hi,
I'm getting warning below when running 'perf test',
not sure what's the reason yet..

jirka


---


[  230.228358] WARNING: CPU: 29 PID: 2133 at kernel/events/core.c:3069 __perf_event_enable+0x1d3/0x220^M
[  230.237395] Modules linked in: intel_rapl_msr intel_rapl_common skx_edac nfit x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel kvm ipmi_ssif irqbypass crct10dif_pclmul crc32_pclmul ghash_clmulni_intel intel_cstate intel_uncore ipmi_si dell_smbios intel_rapl_perf iTCO_wdt wmi_bmof dell_wmi_descriptor ipmi_devintf mei_me iTCO_vendor_support dcdbas mei i2c_i801 lpc_ich wmi ipmi_msghandler acpi_power_meter xfs libcrc32c mgag200 drm_kms_helper i2c_algo_bit drm_vram_helper ttm drm megaraid_sas tg3 crc32c_intel^M
[  230.282700] CPU: 29 PID: 2133 Comm: perf Not tainted 5.4.0-rc7+ #37^M
[  230.288964] Hardware name: Dell Inc. PowerEdge R440/08CYF7, BIOS 1.7.0 12/14/2018^M
[  230.296444] RIP: 0010:__perf_event_enable+0x1d3/0x220^M
[  230.301496] Code: 0f b6 d2 83 c2 01 48 83 b8 b0 00 00 00 00 74 15 5b 44 09 f2 5d 4c 89 ef 41 5c 41 5d 41 5e 41 5f e9 72 fd ff ff 83 ca 08 eb e6 <0f> 0b eb ba 48 8b 45 10 4c 8d 78 f0 4c 39 fd 74 83 49 8b 87 98 00^M
[  230.320242] RSP: 0018:ffff9b2d87a93bd0 EFLAGS: 00010086^M
[  230.325468] RAX: ffff88ad77312010 RBX: ffff88ad77312000 RCX: 0000000000000000^M
[  230.332600] RDX: 0000000000000000 RSI: 0000000000000000 RDI: ffff88ad77312000^M
[  230.339731] RBP: ffff88ad77312000 R08: ffff88ad77312000 R09: ffff88ad77312000^M
[  230.346863] R10: 0000000000000000 R11: 0000000000000000 R12: ffff88ad84b07100^M
[  230.353997] R13: ffff88ad8f9af700 R14: 0000000000000003 R15: ffff88ad77312000^M
[  230.361130] FS:  00007fca82c1f100(0000) GS:ffff88ad8f980000(0000) knlGS:0000000000000000^M
[  230.369217] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033^M
[  230.374961] CR2: 00007fca82c1cff0 CR3: 00000017f65a2003 CR4: 00000000007606e0^M
[  230.382095] DR0: 0000000000b65268 DR1: 0000000000000000 DR2: 0000000000000000^M
[  230.389227] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000600^M
[  230.396358] PKRU: 55555554^M
[  230.399071] Call Trace:^M
[  230.401538]  event_function+0x85/0xc0^M
[  230.405211]  ? ctx_resched+0xc0/0xc0^M
[  230.408789]  remote_function+0x3e/0x50^M
[  230.412547]  generic_exec_single+0x91/0xd0^M
[  230.416652]  ? ctx_resched+0xc0/0xc0^M
[  230.420230]  smp_call_function_single+0xd1/0x110^M
[  230.424852]  task_function_call+0x45/0x70^M
[  230.428872]  ? event_sync_dup_count+0x90/0x90^M
[  230.433239]  event_function_call+0x8f/0x110^M
[  230.437425]  ? ctx_resched+0xc0/0xc0^M
[  230.441003]  ? ring_buffer_attach+0x186/0x1b0^M
[  230.445362]  ? _perf_event_disable+0x50/0x50^M
[  230.449636]  perf_event_for_each_child+0x34/0x80^M
[  230.454256]  ? _perf_event_disable+0x50/0x50^M
[  230.458530]  _perf_ioctl+0x24b/0x700^M
[  230.462118]  ? mem_cgroup_commit_charge+0xcb/0x1a0^M
[  230.466918]  ? __handle_mm_fault+0xd49/0x1ac0^M
[  230.471279]  ? _cond_resched+0x15/0x30^M
[  230.475037]  perf_ioctl+0x3d/0x60^M
[  230.478361]  do_vfs_ioctl+0x405/0x660^M
[  230.482032]  ksys_ioctl+0x5e/0x90^M
[  230.485351]  ? __x64_sys_fcntl+0x84/0xb0^M
[  230.489276]  __x64_sys_ioctl+0x16/0x20^M
[  230.493033]  do_syscall_64+0x5b/0x180^M
[  230.496708]  entry_SYSCALL_64_after_hwframe+0x44/0xa9^M