Neil Brown writes via Kernel.org Bugzilla:
I think this bug was fixed in Linux v6.7 by
Commit ca1d36b82394 ("mm: shrinker: make global slab shrink lockless")
The shrinker_rwsem has been replaced by shrinker_mutex and is no longer held while shrinking slabs. So if nfs4_evict_inode() blocks on a non-responsive server it may block one vmscan thread, but shouldn't cause any blockage outside of that filesystem.
You were using v5.10 which is 4 years old next week. Please re-test with a newer kernel.
(And I think we *should* consider loop-back NFS to be supported in the sense that we will make a reasonable effort to fix any problems that are reported)
View: https://bugzilla.kernel.org/show_bug.cgi?id=219550#c6
You can reply to this message to join the discussion.
--
Deet-doot-dot, I am a bot.
Kernel.org Bugzilla (bugspray 0.1-dev)
