> On Nov 7, 2024, at 2:41 AM, Jan Kara <jack@xxxxxxx> wrote: > > On Wed 06-11-24 20:40:50, Amir Goldstein wrote: >> On Thu, Oct 31, 2024 at 2:52 AM Song Liu <songliubraving@xxxxxxxx> wrote: >>>> Alternately, maybe there is some way to designate that an entire >>>> vfsmount is a child of a watched (or ignored) directory? >>>> >>>>> @Christian, I would like to know your thoughts on this (walking up the >>>>> directory tree in fanotify fastpath handler). It can be expensive for >>>>> very very deep subtree. >>>>> >>>> >>>> I'm not Christian, but I'll make the case for it. It's basically a >>>> bunch of pointer chasing. That's probably not "cheap", but if you can >>>> do it under RCU it might not be too awful. It might still suck with >>>> really deep paths, but this is a sample module. It's not expected that >>>> everyone will want to use it anyway. >>> >>> Thanks for the suggestion! I will try to do it under RCU. >>> >> >> That's the cost of doing a subtree filter. >> Not sure how it could be avoided? > > Yes. For a real solution (not this example), we'd probably have to limit > the parent walk to say 16 steps and if we can reach neither root nor our > dir in that number of steps, we'll just chicken out and pass the event to > userspace to deal with it. This way the kernel filter will deal with most > cases anyway and we won't risk livelocking or too big performance overhead > of the filter. > > For this example, I think using fs/dcache.c:is_subdir() will be OK for > demonstration purposes. Thanks for the pointer! It does look like a good solution for this example. Song
