[LU-10769] osd-zfs deadlock on osd_object::oo_guard

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.11.0, Lustre 2.10.4
• Affects Version/s: Lustre 2.10.0, Lustre 2.11.0
• Severity: 2
• Assignee: nasf (Inactive)
• Reporter: nasf (Inactive)

Description

Consider the following scenarios:

1) The Thread1 calls osd_attr_set() to set flags on the object. The osd_attr_set() will call the osd_xattr_get() with holding the read mode semaphore on the object::oo_guard.

2) The Thread2 calls osd_declare_destroy() to destroy the object, it will down_write() on the object::oo_gurad, but be blocked by the Thread1's granted read mode semaphore.

3) The osd_xattr_get() triggered by the osd_xattr_set() will also down_read() on the object::oo_guard. But it will be blocked by the Thread2's pending down_write() request.

Then the Thread1 and the Thread2 deadlock.

Comments

[Gerrit Updater]

John L. Hammond ([e-mail]) merged in patch https://review.whamcloud.com/31514/
Subject: LU-10769 osd-zfs: fix deadlock on osd_object::oo_guard
Project: fs/lustre-release
Branch: b2_10
Current Patch Set:
Commit: 0e51abb8a512213b58579e66a9d74133001877ff

[Peter Jones]

Landed for 2.11

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/31511/
Subject: LU-10769 osd-zfs: fix deadlock on osd_object::oo_guard
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 095643bbd1bb7c70061eaddf7b6413c429322b31