[LU-10403] sanityn test_32b: @@@@@@ FAIL: lockless i/o works when disabled Created: 18/Dec/17  Updated: 18/Dec/17  Resolved: 18/Dec/17

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Bug Priority: Major
Reporter: Mikhail Pershin Assignee: James A Simmons
Resolution: Duplicate Votes: 0
Labels: None

Issue Links:
Related
is related to LU-9019 Migrate lustre to standard 64 bit tim... Resolved
Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

After the last update this test starts to fail:
https://testing.hpdd.intel.com/test_sets/fa6fd9d6-e372-11e7-8027-52540065bddc
https://testing.hpdd.intel.com/test_sets/8d0371ee-e32e-11e7-9840-52540065bddc

and more



 Comments   
Comment by Mikhail Pershin [ 18/Dec/17 ]

I've done git bisect and it shows commit 4f2a5d5887492da9abe320074511811415e0a06c as source of problem. This is change https://review.whamcloud.com/#/c/30063/ from LU-9019.

This change passed own tests so it seems that issue is result of interaction with some other patch, anyway, reverting of this particular patch solves the problem

Comment by Peter Jones [ 18/Dec/17 ]

Should have been fixed by revert of offending patch

Comment by James A Simmons [ 18/Dec/17 ]

Oh that is strange. I will track down what the conflict is. Thanks.

Comment by Bob Glossman (Inactive) [ 18/Dec/17 ]

suspect it may be due to time64_t being signed while cfs_time_t is unsigned

Comment by James A Simmons [ 18/Dec/17 ]

So I attempted to reproduce the problem locally and I can't. It passes sanityn 32b. I have a theory why it fails. Using seconds is to large of a time gap. The pages are not being purged because the time that passes is less than one second. I bet if I move to nanoseconds this problem will go away.

Comment by Mikhail Pershin [ 18/Dec/17 ]

James, I see that each run in my local vm, so you may be right about time gap.

Generated at Sat Feb 10 02:34:45 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.