Re: [PATCH 5/5] Change vfs_mkdir() to unlock on failure.
Al Viro <[email protected]>
| Newsgroups | gmane.comp.file-systems.ecryptfs.general,gmane.linux.file-systems,gmane.comp.file-systems.coda.general,gmane.linux.nfs,gmane.linux.file-systems.union,gmane.linux.kernel |
|---|---|
| Message-ID | <20250609053442.GC299672@ZenIV> |
On Mon, Jun 09, 2025 at 03:22:00PM +1000, NeilBrown wrote: > On Mon, 09 Jun 2025, Al Viro wrote: > > On Mon, Jun 09, 2025 at 09:09:37AM +1000, NeilBrown wrote: > > > Proposed changes to directory-op locking will lock the dentry rather > > > than the whole directory. So the dentry will need to be unlocked. > > > > Please, repost your current proposal _before_ that one goes anywhere. > > > > I've posted my proposal for the new API. This makes the value of the > vfs_mkdir() change clear (I hope). > > Would you also like me to post the patches which introduce the new > locking scheme? Yes, seeing that the rest does not make much sense without that. I would really like a description of that locking scheme as well, TBH, but if you prefer to start with the patches, then so be it. I can't promise a response tonight - going down in an hour or so and I'd like to do enough reordering of #work.mount to be able to post the initial variant of at least some of that in the morning...