Skip to contentSkip to conversation

LTP Release

LTP Cyril Hrubis 7 participants 16 messages Sep 6, 2019 – Sep 26, 2019

lore
Share & tools
lore
#0 lore raw

Hi!
As usually we should start prepare for a release somewhere at the end of the September.

I will try to review as much patches as possible next week, then I would like to start the pre-release testing. So if there is something that really should go in before the release let me know.

I would personally like to get the eBPF tests in, it would be nice if anyone could ack the v5.

Signature · 2 lines
Cyril Hrubis
[email protected]
#1 lore raw

Hello,

Quoted 10 lines · Cyril Hrubis

Hi!
As usually we should start prepare for a release somewhere at the end of the September.

I will try to review as much patches as possible next week, then I would like to start the pre-release testing. So if there is something that really should go in before the release let me know.

I would personally like to get the eBPF tests in, it would be nice if anyone could ack the v5.

I would also like bpf_prog02 and the capability API unless someone has serious problems with these.

Quoted 4 lines
Signature · 3 lines
-- 
Cyril Hrubis
[email protected]
Signature · 2 lines
Thank you,
Richard.
#2 lore raw
In reply to Cyril Hrubis
Quoted 8 lines · Cyril Hrubis

Hi!
As usually we should start prepare for a release somewhere at the end of the September.

I will try to review as much patches as possible next week, then I would like to start the pre-release testing. So if there is something that really should go in before the release let me know.

The one to print errno number.
[PATCH] tst_res: Print errno number in addition to error name

Some test failures:
==============

acct02 (s390x):
acct02.c:174: INFO: Verifying using 'struct acct_v3' acct02.c:140: INFO: Number of accounting file entries tested: 2 acct02.c:145: FAIL: acct() wrote incorrect file contents!

timer_create01 (s390x, ppc64le, aarch64):
timer_create01.c:48: INFO: Testing notification type: SIGEV_THREAD_ID timer_create01.c:93: PASS: Timer successfully created for CLOCK_REALTIME timer_create01.c:93: PASS: Timer successfully created for CLOCK_MONOTONIC timer_create01.c:93: PASS: Timer successfully created for CLOCK_PROCESS_CPUTIME_ID
timer_create01.c:93: PASS: Timer successfully created for CLOCK_THREAD_CPUTIME_ID
timer_create01.c:93: PASS: Timer successfully created for CLOCK_BOOTTIME timer_create01.c:87: FAIL: Failed to create timer for CLOCK_BOOTTIME_ALARM:
???
timer_create01.c:87: FAIL: Failed to create timer for CLOCK_REALTIME_ALARM:
???
timer_create01.c:93: PASS: Timer successfully created for CLOCK_TAI

Signature · 5 lines
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20190909/8fb3bd75/attachment-0001.htm>
#6 lore raw
In reply to Cyril Hrubis
Quoted 7 lines · Cyril Hrubis

Hi!
As usually we should start prepare for a release somewhere at the end of the September.

I will try to review as much patches as possible next week, then I would like to start the pre-release testing. So if there is something that really should go in before the release let me know.

Hi Cyril

I think my patch[1][2] to fix statx04 and ueven03 should be merged before the release. Also my patch[3] to fix cgroup case should be reviewed if you have free time.

[1]https://patchwork.ozlabs.org/patch/1149792
[2]https://patchwork.ozlabs.org/patch/1158763
[3]https://patchwork.ozlabs.org/patch/1113680

Thanks
Yang Xu

Quoted 4 lines

I would personally like to get the eBPF tests in, it would be nice if anyone could ack the v5.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20190910/5e36f5a6/attachment.htm>

#3 lore raw
In reply to Li Wang

Hi!

Quoted 2 lines

The one to print errno number.
[PATCH] tst_res: Print errno number in addition to error name

Pushed the patch as it was, I do not think that reordering the statements was important enough.

Quoted 7 lines

Some test failures:
==============

acct02 (s390x):
acct02.c:174: INFO: Verifying using 'struct acct_v3' acct02.c:140: INFO: Number of accounting file entries tested: 2 acct02.c:145: FAIL: acct() wrote incorrect file contents!

I can reproduce it here as well, no idea what is wrong there. I guess that modifying the test to be more verbose, i.e. to print if we haven't matched the pid or if the contents of the structure were wrong would be a good start.

Quoted 6 lines

timer_create01 (s390x, ppc64le, aarch64):
timer_create01.c:48: INFO: Testing notification type: SIGEV_THREAD_ID timer_create01.c:93: PASS: Timer successfully created for CLOCK_REALTIME timer_create01.c:93: PASS: Timer successfully created for CLOCK_MONOTONIC timer_create01.c:93: PASS: Timer successfully created for CLOCK_PROCESS_CPUTIME_ID

Quoted 8 lines

timer_create01.c:93: PASS: Timer successfully created for CLOCK_THREAD_CPUTIME_ID
timer_create01.c:93: PASS: Timer successfully created for CLOCK_BOOTTIME timer_create01.c:87: FAIL: Failed to create timer for CLOCK_BOOTTIME_ALARM:
???
timer_create01.c:87: FAIL: Failed to create timer for CLOCK_REALTIME_ALARM:
???
timer_create01.c:93: PASS: Timer successfully created for CLOCK_TAI

I'm pretty sure that we concluded that this is a kernel bug since it returns kernel internal errno.

Signature · 2 lines
Cyril Hrubis
[email protected]
#8 lore raw
In reply to Cyril Hrubis

Hi Cyril,

Quoted 7 lines

Hi!
As usually we should start prepare for a release somewhere at the end of the September.

I will try to review as much patches as possible next week, then I would like to start the pre-release testing. So if there is something that really should go in before the release let me know.

I'd like to have shell timeout fixes + enhancements [1].
I hope I'll send route shell tests rewrite v5, tst_res return exit value [2] is a preparation for it (if acceptable).

Kind regards,
Petr

[1] https://patchwork.ozlabs.org/project/ltp/list/?series=130562&state=*
[2] https://patchwork.ozlabs.org/patch/1161773/

#4 lore raw

[LTP] acct02 failure [was: Re: LTP Release]

In reply to Cyril Hrubis

Hi,

Quoted 9 lines
Quoted 4 lines

acct02 (s390x):
acct02.c:174: INFO: Verifying using 'struct acct_v3' acct02.c:140: INFO: Number of accounting file entries tested: 2 acct02.c:145: FAIL: acct() wrote incorrect file contents!

I can reproduce it here as well, no idea what is wrong there. I guess that modifying the test to be more verbose, i.e. to print if we haven't matched the pid or if the contents of the structure were wrong would be a good start.

I discovered failure on x86_64 on kernel 5.3.0-rc6-3.g8b30bf3-default (openSUSE) and 5.2.9 (Debian), which or might not be related.

Kind regards,
Petr

#9 lore raw
In reply to Cyril Hrubis

Hi!
I did a few testruns and so far it looks good.

I've send one patch for acct02, which is mostly cosmetic, that I would like to get merged.

I would still consider merging the eBPF regression test, does anybody have a strong opinion on this?

Apart from this preadv203 sometimes fails to trigger EAGAIN on virtual machines, but I doubt we can fix this before the release, I will try to do so later on.

The rest seems to be good, i.e. failures does not seem to be bugs in tests.

What is the status from rest of you? Is there anything that should be considered for the release? Is something failing unexpectedly?

Signature · 2 lines
Cyril Hrubis
[email protected]
#10 lore raw

----- Original Message -----

Quoted 13 lines

Hi!
I did a few testruns and so far it looks good.

I've send one patch for acct02, which is mostly cosmetic, that I would like to get merged.

I would still consider merging the eBPF regression test, does anybody have a strong opinion on this?

Apart from this preadv203 sometimes fails to trigger EAGAIN on virtual machines, but I doubt we can fix this before the release, I will try to do so later on.

Quoted 5 lines

The rest seems to be good, i.e. failures does not seem to be bugs in tests.

What is the status from rest of you? Is there anything that should be considered for the release? Is something failing unexpectedly?

I'm thinking if we should patch BPF tests to increase max lock memory to +1MB. On Fedora30 default limit is '64', and I found couple systems which by default come very close to that limit right after boot. So even one iteration of BPF tests will hit a failure.

# lscpu 
Architecture:        x86_64
CPU op-mode(s):      32-bit, 64-bit
Byte Order:          Little Endian
Address sizes:       40 bits physical, 48 bits virtual
CPU(s):              16
On-line CPU(s) list: 0-15
Thread(s) per core:  2
Core(s) per socket:  4
Socket(s):           2
NUMA node(s):        2
Vendor ID:           GenuineIntel
CPU family:          6
Model:               26
Model name:          Intel(R) Xeon(R) CPU           E5530  @ 2.40GHz
Stepping:            5
CPU MHz:             1596.091
CPU max MHz:         2395.0000
CPU min MHz:         1596.0000
BogoMIPS:            4788.26
Virtualization:      VT-x
L1d cache:           32K
L1i cache:           32K
L2 cache:            256K
L3 cache:            8192K
NUMA node0 CPU(s):   0,2,4,6,8,10,12,14
NUMA node1 CPU(s):   1,3,5,7,9,11,13,15
Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc cpuid aperfmperf pni dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm dca sse4_1 sse4_2 popcnt lahf_lm pti tpr_shadow vnmi flexpriority ept vpid dtherm ida
# uname -r
5.3.0-351c8a0.cki
# cat /etc/redhat-release 
Fedora release 30 (Thirty)
Quoted 8 lines

--
Cyril Hrubis
[email protected]

--
Mailing list info: https://lists.linux.it/listinfo/ltp

#11 lore raw

Hi!

Quoted 15 lines
Quoted 15 lines

I did a few testruns and so far it looks good.

I've send one patch for acct02, which is mostly cosmetic, that I would like to get merged.

I would still consider merging the eBPF regression test, does anybody have a strong opinion on this?

Apart from this preadv203 sometimes fails to trigger EAGAIN on virtual machines, but I doubt we can fix this before the release, I will try to do so later on.

The rest seems to be good, i.e. failures does not seem to be bugs in tests.

Quoted 7 lines
Quoted 2 lines

What is the status from rest of you? Is there anything that should be considered for the release? Is something failing unexpectedly?

I'm thinking if we should patch BPF tests to increase max lock memory to +1MB. On Fedora30 default limit is '64', and I found couple systems which by default come very close to that limit right after boot. So even one iteration of BPF tests will hit a failure.

I've seen them fail on ppc64le, because of 64k pages as well. Maybe we shouldn't workaround the failure but rather print a TINFO message that the limit is too low and the test is likely to fail. Then I will write the email to kernel developers once the release is done and we will do something about them later on.

Signature · 2 lines
Cyril Hrubis
[email protected]
#12 lore raw

----- Original Message -----

Quoted 24 lines

Hi!

Quoted 15 lines
Quoted 15 lines

I did a few testruns and so far it looks good.

I've send one patch for acct02, which is mostly cosmetic, that I would like to get merged.

I would still consider merging the eBPF regression test, does anybody have a strong opinion on this?

Apart from this preadv203 sometimes fails to trigger EAGAIN on virtual machines, but I doubt we can fix this before the release, I will try to do so later on.

The rest seems to be good, i.e. failures does not seem to be bugs in tests.

Quoted 7 lines
Quoted 2 lines

What is the status from rest of you? Is there anything that should be considered for the release? Is something failing unexpectedly?

I'm thinking if we should patch BPF tests to increase max lock memory to +1MB. On Fedora30 default limit is '64', and I found couple systems which by default come very close to that limit right after boot. So even one iteration of BPF tests will hit a failure.

Quoted 5 lines

I've seen them fail on ppc64le, because of 64k pages as well. Maybe we shouldn't workaround the failure but rather print a TINFO message that the limit is too low and the test is likely to fail. Then I will write the email to kernel developers once the release is done and we will do something about them later on.

This appears to be known, and some tools like perf started auto-bumping limit:
https://lkml.org/lkml/2019/7/17/714

I'd prefer LTP does that too, we can always revert later when it's fixed/changed on kernel side.

Quoted 5 lines

--
Cyril Hrubis
[email protected]

#15 lore raw
In reply to Cyril Hrubis
Quoted 13 lines · Cyril Hrubis

Hi!
I did a few testruns and so far it looks good.

I've send one patch for acct02, which is mostly cosmetic, that I would like to get merged.

I would still consider merging the eBPF regression test, does anybody have a strong opinion on this?

Apart from this preadv203 sometimes fails to trigger EAGAIN on virtual machines, but I doubt we can fix this before the release, I will try to do so later on.

Quoted 6 lines

The rest seems to be good, i.e. failures does not seem to be bugs in tests.

What is the status from rest of you? Is there anything that should be considered for the release? Is something failing unexpectedly?

No more failure from my side except the known issues(timer_create01, acct02). It's time to release a new package since now is already ending in September.

Signature · 5 lines
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20190926/26a8384c/attachment.htm>
#13 lore raw
In reply to Jan Stancek

Hi!

Quoted 11 lines
Quoted 10 lines
Quoted 4 lines

I'm thinking if we should patch BPF tests to increase max lock memory to +1MB. On Fedora30 default limit is '64', and I found couple systems which by default come very close to that limit right after boot. So even one iteration of BPF tests will hit a failure.

I've seen them fail on ppc64le, because of 64k pages as well. Maybe we shouldn't workaround the failure but rather print a TINFO message that the limit is too low and the test is likely to fail. Then I will write the email to kernel developers once the release is done and we will do something about them later on.

Quoted 5 lines

This appears to be known, and some tools like perf started auto-bumping limit:
https://lkml.org/lkml/2019/7/17/714

I'd prefer LTP does that too, we can always revert later when it's fixed/changed on kernel side.

Agreed then, will you prepare the patch?

Signature · 2 lines
Cyril Hrubis
[email protected]
#14 lore raw

----- Original Message -----

Quoted 13 lines

Hi!

Quoted 12 lines
Quoted 11 lines
Quoted 5 lines

I'm thinking if we should patch BPF tests to increase max lock memory to +1MB. On Fedora30 default limit is '64', and I found couple systems which by default come very close to that limit right after boot.
So even one iteration of BPF tests will hit a failure.

I've seen them fail on ppc64le, because of 64k pages as well. Maybe we shouldn't workaround the failure but rather print a TINFO message that the limit is too low and the test is likely to fail. Then I will write the email to kernel developers once the release is done and we will do something about them later on.

Quoted 8 lines
Quoted 6 lines

This appears to be known, and some tools like perf started auto-bumping limit:
https://lkml.org/lkml/2019/7/17/714

I'd prefer LTP does that too, we can always revert later when it's fixed/changed on kernel side.

Agreed then, will you prepare the patch?

Sure, I'll post it today.

Quoted 5 lines

--
Cyril Hrubis
[email protected]

#5 lore raw
In reply to Cyril Hrubis
Quoted 35 lines · Cyril Hrubis

Hi!

Quoted 2 lines

The one to print errno number.
[PATCH] tst_res: Print errno number in addition to error name

Pushed the patch as it was, I do not think that reordering the statements was important enough.

Quoted 7 lines

Some test failures:
==============

acct02 (s390x):
acct02.c:174: INFO: Verifying using 'struct acct_v3' acct02.c:140: INFO: Number of accounting file entries tested: 2 acct02.c:145: FAIL: acct() wrote incorrect file contents!

I can reproduce it here as well, no idea what is wrong there. I guess that modifying the test to be more verbose, i.e. to print if we haven't matched the pid or if the contents of the structure were wrong would be a good start.

Quoted 6 lines

timer_create01 (s390x, ppc64le, aarch64):
timer_create01.c:48: INFO: Testing notification type: SIGEV_THREAD_ID timer_create01.c:93: PASS: Timer successfully created for CLOCK_REALTIME timer_create01.c:93: PASS: Timer successfully created for CLOCK_MONOTONIC timer_create01.c:93: PASS: Timer successfully created for CLOCK_PROCESS_CPUTIME_ID

Quoted 8 lines

timer_create01.c:93: PASS: Timer successfully created for CLOCK_THREAD_CPUTIME_ID
timer_create01.c:93: PASS: Timer successfully created for CLOCK_BOOTTIME timer_create01.c:87: FAIL: Failed to create timer for CLOCK_BOOTTIME_ALARM:
???
timer_create01.c:87: FAIL: Failed to create timer for CLOCK_REALTIME_ALARM:
???
timer_create01.c:93: PASS: Timer successfully created for CLOCK_TAI

Quoted 3 lines

I'm pretty sure that we concluded that this is a kernel bug since it returns kernel internal errno.

And a pull request has already been sent to Linus to fix this. [1]

Now, the test also needs to check for EOPNOTSUPP instead of only EINVAL. I'll prepare a patch.

[1] https://lore.kernel.org/lkml/156864062018.3407.10822767012822306757.tglx@nanos.tec.linutronix.de/

Quoted 6 lines
Signature · 5 lines
Cyril Hrubis
[email protected]

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp