[LU-2723] a client hangs forever in osc_extent_wait() on directio Created: 31/Jan/13  Updated: 23/Apr/13  Resolved: 05/Feb/13

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.4.0
Fix Version/s: Lustre 2.4.0

Type: Bug Priority: Minor
Reporter: Vitaly Fertman Assignee: WC Triage
Resolution: Fixed Votes: 0
Labels: patch

Issue Links:
Duplicate
is duplicated by LU-2643 osc_cache_writeback_range() ASSERTION... Resolved
Severity: 3
Rank (Obsolete): 6617

 Description   

PID: 6774 TASK: ffff8803fc62e0c0 CPU: 1 COMMAND: "fsx-linux-aio"
#0 [ffff8803de9334e8] schedule at ffffffff8141e405
#1 [ffff8803de9335c0] cfs_waitq_wait at ffffffffa01a66ee [libcfs]
#2 [ffff8803de9335d0] osc_extent_wait at ffffffffa0725aec [osc]
#3 [ffff8803de933740] osc_cache_wait_range at ffffffffa0726167 [osc]
#4 [ffff8803de9338d0] osc_io_fsync_end at ffffffffa0717a79 [osc]
#5 [ffff8803de933910] cl_io_end at ffffffffa03048b2 [obdclass]
#6 [ffff8803de933940] lov_io_end_wrapper at ffffffffa07c26f9 [lov]
#7 [ffff8803de933960] lov_io_fsync_end at ffffffffa07c407b [lov]
#8 [ffff8803de933990] cl_io_end at ffffffffa03048b2 [obdclass]
#9 [ffff8803de9339c0] cl_io_loop at ffffffffa0308c4a [obdclass]
#10 [ffff8803de9339f0] cl_sync_file_range at ffffffffa08aacbb [lustre]
#11 [ffff8803de933a60] ll_writepages at ffffffffa08d0164 [lustre]
#12 [ffff8803de933aa0] do_writepages at ffffffff810cd631
#13 [ffff8803de933ab0] __filemap_fdatawrite_range at ffffffff810c5b65
#14 [ffff8803de933b10] filemap_write_and_wait_range at ffffffff810c5bca
#15 [ffff8803de933b40] generic_file_buffered_write at ffffffff810c5e23
#16 [ffff8803de933c10] __generic_file_aio_write at ffffffff810c6d37
#17 [ffff8803de933cc0] generic_file_aio_write at ffffffff810c6e42
#18 [ffff8803de933d00] vvp_io_write_start at ffffffffa08fd4f6 [lustre]
#19 [ffff8803de933d40] cl_io_start at ffffffffa0304a1a [obdclass]
#20 [ffff8803de933d70] cl_io_loop at ffffffffa0308c3c [obdclass]
#21 [ffff8803de933da0] ll_file_io_generic at ffffffffa08a9cc5 [lustre]
#22 [ffff8803de933e10] ll_file_aio_write at ffffffffa08aa01b [lustre]
#23 [ffff8803de933e60] aio_rw_vect_retry at ffffffff811426d4
#24 [ffff8803de933e90] aio_run_iocb at ffffffff81143ffa
#25 [ffff8803de933ec0] sys_io_submit at ffffffff81144e38
#26 [ffff8803de933f80] system_call_fastpath at ffffffff8100305b
RIP: 00007f7f5d49a6f7 RSP: 00007fff771a51a8 RFLAGS: 00010217
RAX: 00000000000000d1 RBX: ffffffff8100305b RCX: 00000000000000e2
RDX: 00007fff771a5150 RSI: 0000000000000001 RDI: 00007f7f5d8b0000
RBP: 0000000000001000 R8: 0000000000000000 R9: 0000000000000003
R10: 000000000000000f R11: 0000000000000202 R12: 00007fff771a5600
R13: 0000000000404658 R14: 0000000000001000 R15: 0000000000001000
ORIG_RAX: 00000000000000d1 CS: 0033 SS: 002b

 Comments   
Comment by Vitaly Fertman [ 31/Jan/13 ]

If dio overlaps with mapped region, first overlapped pages are sent by generic_file_direct_write->filemap_write_and_wait_range->writepages. After that the dio pages are sent by write_begin/write_end, writepages.
However, write_end->osc_page_cache_add does not release extent because the IO is not considered as synchronous. as the result, the following writepages->osc_cache_writeback_range() does not send pages are the extent is still ACTIVE and osc_cache_wait_range() hangs forever.

Comment by Vitaly Fertman [ 31/Jan/13 ]

http://review.whamcloud.com/5224

Comment by Peter Jones [ 05/Feb/13 ]

Landed for 2.4

Generated at Sat Feb 10 01:27:39 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.