On Wed, Sep 17, 2025 at 3:39 PM Max Kellermann max.kellermann@ionos.com wrote:
On Wed, Sep 17, 2025 at 3:14 PM Mateusz Guzik mjguzik@gmail.com wrote:
Does the patch convert literally all iput calls within ceph into the async variant? I would be worried that mandatory deferral of literally all final iputs may be a regression from perf standpoint.
ok, in that case i have no further commentary
I don't think this affects performance at all. It almost never happens that the last reference gets dropped by somebody other than dcache (which only happens under memory pressure).
Well only changing the problematic consumers as opposed *everyone* should be the end of it.
(Forgot to reply to this part) No, I changed just the ones that are called from Writeback+Messenger.
I don't think this affects performance at all. It almost never happens that the last reference gets dropped by somebody other than dcache (which only happens under memory pressure). It was very difficult to reproduce this bug:
- "echo 2 >drop_caches" in a loop
- a kernel patch that adds msleep() to several functions
- another kernel patch that allows me to disconnect the Ceph server via ioctl
The latter was to free inode references that are held by Ceph caps. For this deadlock to occur, all references other than writeback/messenger must be gone already. (It did happen on our production servers, crashing all of them a few days ago causing a major service outage, but apparently in all these years we're the first ones to observe this deadlock bug.)
This makes sense to me.
The VFS layer is hopefully going to get significantly better assert coverage, so I expect this kind of trouble will be reported on without having to actually run into it. Presumably including yet-to-be-discovered deadlocks. ;)