On Tue, 2025-06-03 at 19:29 -0700, Jakub Kicinski wrote: > On Tue, 03 Jun 2025 07:27:11 -0400 Jeff Layton wrote: > > For those just joining in, this series adds a new top-level > > "ref_tracker" debugfs directory, and has each ref_tracker_dir register a > > file in there as part of its initialization. It also adds the ability to > > register a symlink with a more human-usable name that points to the > > file, and does some general cleanup of how the ref_tracker object names > > are handled. > > > > This reposting is mostly to address Krzysztof's comments. I've dropped > > the i915 patch, and rebased the rest of the series on top. > > > > Note that I still see debugfs: warnings in the i915 driver even when we > > gate the registration of the debugfs file on the classname pointer being > > NULL. Here is a CI report from v12: > > > > > > https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_148490v8/bat-arls-6/igt@i915_selftest@l...@workarounds.html > > > > I think the i915 driver is doing something it shouldn't with these > > objects. They seem to be initialized more than once, which could lead > > to leaked ref_tracker objects. It would be good for one of the i915 > > maintainers to comment on whether this is a real problem. > > I still see the fs crashes: > https://netdev-3.bots.linux.dev/vmksft-packetdrill-dbg/results/149560/2-tcp-slow-start-slow-start-app-limited-pkt/stderr > I'll hide this series from patchwork for now. We will pull from Linus > on Thu, I'll reactivate it and let's see if the problem was already > fixed.
Sorry, I never got any mail about those failures. I would have looked at that sooner. It looks like the last netns reference can be put in a rcu callback? That makes sense. I think ref_tracker_dir_exit() has to remain safe to call from any context. Perhaps we can defer the dentry deletion to the system_wq? We can drop this series for now. I'll have to think about this. Thanks, -- Jeff Layton <jlay...@kernel.org>
