Skip to content

Revert "fuse: invalidate the page cache after direct write"#105

Merged
bsbernd merged 1 commit into
DDNStorage:redfs-ubuntu-noble-6.8.0-58.60from
cding-ddn:fix-revert-aio-f6de786c
Mar 2, 2026
Merged

Revert "fuse: invalidate the page cache after direct write"#105
bsbernd merged 1 commit into
DDNStorage:redfs-ubuntu-noble-6.8.0-58.60from
cding-ddn:fix-revert-aio-f6de786c

Conversation

@cding-ddn
Copy link
Copy Markdown
Collaborator

This reverts commit f6de786.
f6de786 causes deadlock on coroutine based fuse daemon when running AIO + buffered I/O mixed workload.

@cding-ddn cding-ddn requested review from bsbernd, hbirth and yongzech March 2, 2026 11:35
@bsbernd
Copy link
Copy Markdown
Collaborator

bsbernd commented Mar 2, 2026

@cding-ddn I don't think the commit message is right. Let's say one runs single thread fuse, why shouldn't it be an issue then?

@bsbernd bsbernd merged commit 78b63e3 into DDNStorage:redfs-ubuntu-noble-6.8.0-58.60 Mar 2, 2026
@cding-ddn
Copy link
Copy Markdown
Collaborator Author

@cding-ddn I don't think the commit message is right. Let's say one runs single thread fuse, why shouldn't it be an issue then?

You are right. single thread fuse will fail too. I emphasize coroutine mainly to distinguish it from multi-threading. coroutine/reactor-based fuse could fail as that's what we have tested (though they don't have to fail if they call reply_data in a separated thread) . I would assume multi-thread based fuse works - otherwise the author of f6de786 will not pass his test case.

I realized that I need to revert it from other branches too, going to do that.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants