[LU-13745] tasks hang with copy_file_range: ll_file_splice_read()

• Type: Bug
• Resolution: Unresolved
• Priority: Major
• Labels: always_except
• Severity: 3
• Assignee: Zhenyu Xu
• Reporter: Zhenyu Xu

Description

copy_file_range(2) reads from a Lustre file hang.

With no .copy_file_range VFS API implemented it calls do_splice_direct()->splice_direct_to_actor()->do_splice_to()->ll_file_splice_read().

While the call chain of ll_file_splice_read()->ll_file_io_generic()->generic_file_splice_read()->ll_file_read_iter()->ll_file_io_generic().

And that would try to get LDLM lock twice in ll_file_io_generic(), so that hang ensued.

Issue Links

• is related to: LU-12425 Add test for splice_read
• is related to: LU-14060 sanity test_426 (splice-test) issues
• is related to: LU-14045 Fix O_DIRECT and encrypted files

Comments

[Peter Jones]

The fix itself has landed for 2.14. All that remains tracked by this ticket is a test. Are there still plans to land that test imminently or can we either abandon that changeset /move it to a new JIRA?

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/40396/
Subject: LU-13745 pcc: fall back normal splice read for detached file
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: cca45ad8aeaa8e124e9e48361bf7cff89a035f82

[Gerrit Updater]

Yingjin Qian ([e-mail]) uploaded a new patch: https://review.whamcloud.com/40396
Subject: LU-13745 pcc: fall back normal splice read for detached file
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 7b45c8d9c77a9c45862bd61ea05b8c46117cffa4

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/40366/
Subject: LU-13745 tests: skip sanity test_426 for 4.15+
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: f8a8d3f83db67be9dcc724ff49757cce81b13a5e

[Gerrit Updater]

John L. Hammond ([e-mail]) uploaded a new patch: https://review.whamcloud.com/40366
Subject: LU-13745 tests: skip sanity test_426 for 4.15+
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: a8a6a55a1e4e11ee3b61aa7e230d752b5c1a476a

[Bruno Faccini (Inactive)]

adilger, after I have rebased my patch on top of change #40326, it appears that same crash during sanity/test_426 still occurs with Ubuntu Client because the used Kernel version is 4.15 so sub-test is not skipped !!

[John Hammond]

This is also failing on Ubuntu 18.04 which is using 4.15.0-72-generic. See https://testing.whamcloud.com/test_sets/bdfb8c4b-a6a2-493a-ab57-6a9923f96e7c.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/40326/
Subject: LU-13745 tests: skip sanity test_426 for 4.18+
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 010425898fa4b2abc6325a8073e20cb994ce7947

[Bruno Faccini (Inactive)]

adilger this may be of interest, but I think my patch https://review.whamcloud.com/35856 has also failed 100% of "test review-ldiskfs-ubuntu on CentOS 7.8/x86_64, Ubuntu 18.04/x86_64" stage I have attempted, for the same sanity/test_426 crash on Client side that you have provided the significant stack in LU-14045. So it should be more a Kernel v4.x related issue than an arch related one.

[Wang Shilong (Inactive)]

adilger That means the newly added test is helpful and something we need fix still