On Wed, Aug 07, 2024 at 01:43:48PM +0100, Al Viro wrote: > On Wed, Aug 07, 2024 at 11:50:50AM +0200, Mateusz Guzik wrote: > > > tripping ip: > > vfs_tmpfile+0x162/0x230: > > fsnotify_parent at include/linux/fsnotify.h:81 > > (inlined by) fsnotify_file at include/linux/fsnotify.h:131 > > (inlined by) fsnotify_open at include/linux/fsnotify.h:401 > > (inlined by) vfs_tmpfile at fs/namei.c:3781 > > Try this for incremental; missed the fact that finish_open() is > used by ->tmpfile() instances, not just ->atomic_open(). > > Al, crawling back to sleep... I _really_ hate ->atomic_open() calling conventions; FWIW, I suspect that in the current form this series is OK, but only because none of the existing instances call finish_open() on a preexisting aliases found by d_splice_alias(). And control flow in the instances (especially the cleanup paths) is bloody awful... We never got it quite right, and while the previous iterations of the calling conventions for that methods had been worse, it's still nasty in the current form ;-/ Oh, well - review of those has been long overdue.