[P9][Ubuntu 16.04.03] Additional PMU fixes for P9

[bugproxy]

== Comment: #0 - Shriya R. Kulkarni <email address hidden> - 2017-08-16 04:07:45 ==
Please integrate the following PMU fixes for P9.

93fc5ca9a0048c ('powerpc/perf: Cleanup of PM_BR_CMPL vs. PM_BRU_CMPL in Power9 event list')
91e0bd1e62519b ('powerpc/perf: Add PM_LD_MISS_L1 and PM_BR_2PATH to power9 event list')
70a7e720998d5b ('powerpc/perf: Factor out PPMU_ONLY_COUNT_RUN check code from power8')
7aa345d84245a7 ('powerpc/perf: Update default sdar_mode value for power9')
101dd590a7fa37 ('powerpc/perf: Avoid spurious PMU interrupts after idle')
bdd21ddb919d28 ('powerpc/perf: Avoid spurious PMU interrupts after idle on Power9')
3f0bd8dad0db73 ('powerpc/perf: Add POWER9 alternate PM_RUN_CYC and PM_RUN_INST_CMPL events')
20dd4c624d2515 ('powerpc/perf: Fix SDAR_MODE value for continous sampling on Power9')
24bedcb7c81137 ('powerpc/perf: Fix branch event code for power9')
8c218578fcbbbd ('powerpc/perf: Fix Power9 test_adder fields')

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-08-21 01:45 EDT-------
These additional patches are also needed.

f2080b9ac3c45 ('powerpc/perf: Add Power8 mem_access event to sysfs')
d148c94c27a87 ('powerpc/perf: Support to export SIERs bit in Power9')
453ce7a9432ee ('powerpc/perf: Support to export SIERs bit in Power8')
170a315f41c64 ('powerpc/perf: Support to export MMCRA[TEC*] field to userspace')
79e96f8f930d4 ('powerpc/perf: Export memory hierarchy info to user space')
8c5073db0ee68 ('powerpc/perf: Define big-endian version of perf_mem_data_src')

These are all upstreamed.

------- Comment From <email address hidden> 2017-08-21 02:56 EDT-------
marking it external ...

[Manoj Iyer (manjo)]

IBM, How does this bug affect package perforate ? These patches seem to be kernel patches.

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-08-28 14:42 EDT-------
(In reply to comment #7)
> IBM, How does this bug affect package perforate ? These patches seem to be
> kernel patches.

Looks like that was a mistake in the mirror request. These are all kernel/perf changes. I am not sure how to fix the mirrored package from our bugzilla side.. can this please be corrected from the launchpad side? Thanks!

[Joseph Salisbury (jsalisbury)]

It looks like these patches are all in linux-next. Can you post a comment here when they land in mainline?

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-09-11 17:23 EDT-------
Looks like all but one of the patches are now in mainline:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=93fc5ca9a0048c
powerpc/perf: Cleanup of PM_BR_CMPL vs. PM_BRU_CMPL in Power9 event list
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=91e0bd1e62519b
powerpc/perf: Add PM_LD_MISS_L1 and PM_BR_2PATH to power9 event list
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=70a7e720998d5b
powerpc/perf: Factor out PPMU_ONLY_COUNT_RUN check code from power8
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=7aa345d84245a7
powerpc/perf: Update default sdar_mode value for power9
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=101dd590a7fa37
powerpc/perf: Avoid spurious PMU interrupts after idle
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=3f0bd8dad0db73
powerpc/perf: Add POWER9 alternate PM_RUN_CYC and PM_RUN_INST_CMPL events
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=20dd4c624d2515
powerpc/perf: Fix SDAR_MODE value for continous sampling on Power9
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=24bedcb7c81137
powerpc/perf: Fix branch event code for power9
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8c218578fcbbbd
powerpc/perf: Fix Power9 test_adder fields
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=f2080b9ac3c45
powerpc/perf: Add Power8 mem_access event to sysfs
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=d148c94c27a87
powerpc/perf: Support to export SIERs bit in Power9
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=453ce7a9432ee
powerpc/perf: Support to export SIERs bit in Power8
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=170a315f41c64
powerpc/perf: Support to export MMCRA[TEC*] field to userspace
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=79e96f8f930d4
powerpc/perf: Export memory hierarchy info to user space
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=8c5073db0ee68
powerpc/perf: Define big-endian version of perf_mem_data_src

But this one is still in linux-next:

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?id=bdd21ddb919d28
powerpc/perf: Avoid spurious PMU interrupts after idle on Power9

Shriya / Madhavan,
Do you still expect that last one to be pulled in soon?

[Joseph Salisbury (jsalisbury)]

I built a Zesty test kernel with the sixteen patches mentioned in comment #5. The test kernel can be downloaded from:

http://kernel.ubuntu.com/~jsalisbury/lp1711104/

Can you test this kernel and see if it resolves this bug?

Thanks in advance!

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-09-14 01:51 EDT-------
(In reply to comment #10)
> Looks like all but one of the patches are now in mainline:
>
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=93fc5ca9a0048c
> powerpc/perf: Cleanup of PM_BR_CMPL vs. PM_BRU_CMPL in Power9 event list
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=91e0bd1e62519b
> powerpc/perf: Add PM_LD_MISS_L1 and PM_BR_2PATH to power9 event list
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=70a7e720998d5b
> powerpc/perf: Factor out PPMU_ONLY_COUNT_RUN check code from power8
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=7aa345d84245a7
> powerpc/perf: Update default sdar_mode value for power9
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=101dd590a7fa37
> powerpc/perf: Avoid spurious PMU interrupts after idle
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=3f0bd8dad0db73
> powerpc/perf: Add POWER9 alternate PM_RUN_CYC and PM_RUN_INST_CMPL events
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=20dd4c624d2515
> powerpc/perf: Fix SDAR_MODE value for continous sampling on Power9
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=24bedcb7c81137
> powerpc/perf: Fix branch event code for power9
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=8c218578fcbbbd
> powerpc/perf: Fix Power9 test_adder fields
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=f2080b9ac3c45
> powerpc/perf: Add Power8 mem_access event to sysfs
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=d148c94c27a87
> powerpc/perf: Support to export SIERs bit in Power9
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=453ce7a9432ee
> powerpc/perf: Support to export SIERs bit in Power8
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=170a315f41c64
> powerpc/perf: Support to export MMCRA[TEC*] field to userspace
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=79e96f8f930d4
> powerpc/perf: Export memory hierarchy info to user space
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/
> ?id=8c5073db0ee68
> powerpc/perf: Define big-endian version of perf_mem_data_src
>
>
> But this one is still in linux-next:
>
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/
> ?id=bdd21ddb919d28
> powerpc/perf: Avoid spurious PMU interrupts after idle on Power9
>
> Shriya / Madhavan,
> Do you still expect that last one to be pulled in soon?

Looks like

https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/commit/?id=09539f9b123652e969894d6299ae0df2fe12cb5d

this is missing. Kindly can you pull this also.

[Joseph Salisbury (jsalisbury)]

The test kernel posted in comment #6 had the following commit from linux-next:

https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?id=bdd21ddb919d28
powerpc/perf: Avoid spurious PMU interrupts after idle on Power9

That commit seems like it does the same thing as commit 09539f9, which you asked for in comment #7. Commit 09539f9 was added to mainline as of v4.13-rc4. Does that mean you no longer need commit bdd21ddb919d in linux-next?

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-09-19 00:55 EDT-------
Yes commit bdd21ddb919d28 (-next) and 09539f9 (upstream) are same. Just wanted to confirm whether one of these is pulled into the ubuntu kernel (http://kernel.ubuntu.com/~jsalisbury/lp1711104/ ) or the in the latest kernel.

[Joseph Salisbury (jsalisbury)]

Yes, that commit is in the test kernel posted at:
http://kernel.ubuntu.com/~jsalisbury/lp1711104/

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-12-11 10:56 EDT-------
If we do not see any answer on this test for more than 1 week, let's reject it.

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2017-12-13 13:32 EDT-------
(In reply to comment #17)
> Yes, that commit is in the test kernel posted at:
> http://kernel.ubuntu.com/~jsalisbury/lp1711104/

Tested the kernel given in the above link :

uname -a : Linux ltc-wspoon12 4.10.0-33-generic #37~lp1711104 SMP Wed Sep 13 19:55:56 UTC 2017 ppc64le ppc64le ppc64le GNU/Linux
Hardware : Witherspoon + DD2.1
FW : 1742Eg

root@ltc-wspoon12:~# cat /etc/os-release
NAME="Ubuntu"
VERSION="16.04.3 LTS (Xenial Xerus)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu 16.04.3 LTS"
VERSION_ID="16.04"
HOME_URL="http://www.ubuntu.com/"
SUPPORT_URL="http://help.ubuntu.com/"
BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"
VERSION_CODENAME=xenial
UBUNTU_CODENAME=xenial

Test results :
========
Testing looks good and results are as expected.

root@ltc-wspoon12:~# ./perf stat -e r4d05e -- sleep 1

Performance counter stats for 'sleep 1':

95,723 r4d05e

1.001016884 seconds time elapsed

root@ltc-wspoon12:~# ./perf stat -e r4d05e -- sleep 2

Performance counter stats for 'sleep 2':

95,859 r4d05e

2.000996600 seconds time elapsed

----------------

root@ltc-wspoon12:~# ./perf stat -e r3e054 -- sleep 3

Performance counter stats for 'sleep 3':

8,973 r3e054

3.001030248 seconds time elapsed

root@ltc-wspoon12:~# ./perf stat -e r400f0 -- sleep 3

Performance counter stats for 'sleep 3':

8,905 r400f0

3.000995167 seconds time elapsed

--------------

root@ltc-wspoon12:~# ./perf stat -e r200f4 -- sleep 3

Performance counter stats for 'sleep 3':

1,411,106 r200f4

3.000942725 seconds time elapsed

root@ltc-wspoon12:~# ^C
root@ltc-wspoon12:~# ./perf stat -e r400fa -- sleep 3

Performance counter stats for 'sleep 3':

544,977 r400fa

3.000928921 seconds time elapsed

------------------

root@ltc-wspoon12:~# ./perf stat -e r500fa -- sleep 3

Performance counter stats for 'sleep 3':

550,846 r500fa

3.000949027 seconds time elapsed

---------------------

root@ltc-wspoon12:~/ebizzy-0.3# ./perf record -e r67340101ea -d --weight ./ebizzy -m -n 100 -P -R -s 5120 -S 50 -t 3
^C[ perf record: Woken up 93 times to write data ]
[ perf record: Captured and wrote 24.162 MB perf.data (32508 samples) ]

[Joseph Salisbury (jsalisbury)]

The fix for this bug made it into Zesty via bug 1716491.

[Joseph Salisbury (jsalisbury)]

Can you confirm that Zesty with the latest updates and at least the Ubuntu-4.10.0-38 resolves this bug?

[bugproxy (bugproxy)]

------- Comment From <email address hidden> 2018-01-03 01:34 EDT-------
Tested on Ubuntu 16.04.03 , kernel : 4.10.0-43-generic
--------------------------------------------------------------------------------------
Hardware : P9 Witherspoon + DD2.1

Testing looks fine.

root@ltc-wspoon12:~# uname -a
Linux ltc-wspoon12 4.10.0-43-generic #47~16.04.1-Ubuntu SMP Wed Dec 13 19:21:53 UTC 2017 ppc64le ppc64le ppc64le GNU/Linux

Results :
=====
root@ltc-wspoon12:~# perf stat -e r4d05e -- sleep 1

Performance counter stats for 'sleep 1':

98,250 r4d05e

1.000861597 seconds time elapsed

root@ltc-wspoon12:~# perf stat -e r4d05e -- sleep 2

Performance counter stats for 'sleep 2':

98,603 r4d05e

2.000723944 seconds time elapsed

root@ltc-wspoon12:~# perf stat -e r3e054 -- sleep 3

Performance counter stats for 'sleep 3':

9,167 r3e054

3.000673888 seconds time elapsed

root@ltc-wspoon12:~# perf stat -e r400f0 -- sleep 3

Performance counter stats for 'sleep 3':

9,143 r400f0

3.000728772 seconds time elapsed

root@ltc-wspoon12:~# perf stat -e r200f4 -- sleep 3

Performance counter stats for 'sleep 3':

751,349 r200f4

3.000665969 seconds time elapsed

root@ltc-wspoon12:~# perf stat -e r400fa -- sleep 3

Performance counter stats for 'sleep 3':

563,389 r400fa

3.000760674 seconds time elapsed

root@ltc-wspoon12:~# perf stat -e r500fa -- sleep 3

Performance counter stats for 'sleep 3':

563,902 r500fa

3.000681230 seconds time elapsed

root@ltc-wspoon12:~# perf record -e r67340101ea ls
perf.data
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.006 MB perf.data (5 samples) ]