On Thu, Mar 24, 2022 at 11:19:57AM +0800, Yang Jihong wrote:
> Sysfs show formats of files in /sys/devices/cpu/events/ are not unified,
> some end with "\n", and some do not. Modify sysfs show format of events
> defined by EVENT_ATTR_STR to end with "\n".
Did you test all the userspace that consumes these fields to make sure
none of them break? I suppose it's mostly perf tool, but I'm fairly sure
there's others out there as well.
Hi Peter,
On 2022/3/24 18:11, Peter Zijlstra wrote:
> On Thu, Mar 24, 2022 at 11:19:57AM +0800, Yang Jihong wrote:
>> Sysfs show formats of files in /sys/devices/cpu/events/ are not unified,
>> some end with "\n", and some do not. Modify sysfs show format of events
>> defined by EVENT_ATTR_STR to end with "\n".
>
> Did you test all the userspace that consumes these fields to make sure
> none of them break? I suppose it's mostly perf tool, but I'm fairly sure
> there's others out there as well.
>
Yes, I tested "perf record" and "perf stat" commands on my machine
against the modified events, and the results are as follows:
$ perf stat -e
topdown-fetch-bubbles,topdown-recovery-bubbles,topdown-slots-issued,topdown-slots-retired,topdown-total-slots
-a sleep 1
Performance counter stats for 'system wide':
2356235 topdown-fetch-bubbles # 1211.9% frontend
bound
226132 topdown-recovery-bubbles # 182.9% bad
speculation
1073595 topdown-slots-issued # -1780.4% backend bound
944133 topdown-slots-retired # 485.6% retiring
194420 topdown-total-slots
0.987394388 seconds time elapsed
$ perf record -e
topdown-fetch-bubbles,topdown-recovery-bubbles,topdown-slots-issued,topdown-slots-retired,topdown-total-slots
-a sleep 1
Lowering default frequency rate from 4000 to 3000.
Please consider tweaking /proc/sys/kernel/perf_event_max_sample_rate.
Couldn't synthesize bpf events.
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.300 MB perf.data (2620 samples) ]
$ perf evlist
topdown-fetch-bubbles
topdown-recovery-bubbles
topdown-slots-issued
topdown-slots-retired
topdown-total-slots
dummy:HG
.
Thanks,
Jihong
> .
>
On Fri, Mar 25, 2022 at 09:38:47AM +0800, Yang Jihong wrote:
> Hi Peter,
>
> On 2022/3/24 18:11, Peter Zijlstra wrote:
> > On Thu, Mar 24, 2022 at 11:19:57AM +0800, Yang Jihong wrote:
> > > Sysfs show formats of files in /sys/devices/cpu/events/ are not unified,
> > > some end with "\n", and some do not. Modify sysfs show format of events
> > > defined by EVENT_ATTR_STR to end with "\n".
> >
> > Did you test all the userspace that consumes these fields to make sure
> > none of them break? I suppose it's mostly perf tool, but I'm fairly sure
> > there's others out there as well.
> >
> Yes, I tested "perf record" and "perf stat" commands on my machine against
> the modified events, and the results are as follows:
Fair enough, I'll queue it for after -rc1.