On Fri, Mar 18, 2022 at 6:06 PM Amir Goldstein <amir73il@xxxxxxxxx> wrote:
>
> > > > So far my thinking is that we either follow the path of possibly generating
> > > > more events than necessary (i.e., any merge of two masks that do not both
> > > > have FAN_MARK_VOLATILE set will clear FAN_MARK_VOLATILE)
>
> I agree that would provide predictable behavior which is also similar to
> that of _SURV_MODIFY.
> But IMO, this is very weird to explain/document in the wider sense.
> However, if we only document that
> "FAN_MARK_VOLATILE cannot be set on an existing mark and any update
> of the mask without FAN_MARK_VOLATILE clears that flag"
> (i.e. we make _VOLATILE imply the _CREATE behavior)
> then the merge logic is the same as you suggested, but easier to explain.
>
[...]
>
> To summarize my last proposal:
>
> 1. On fanotify_mark() with FAN_MARK_VOLATILE
> 1.a. If the mark is new, the HAS_IREF flag is not set and no ihold()
> 1.b. If mark already exists without HAS_IREF flag, mask is updated
> 1.c. If mark already exists with HAS_IREF flag, mark add fails with EEXISTS
>
> 2. On fanotify_mark() without FAN_MARK_VOLATILE
> 2.a. If the mark is new or exists without HAS_IREF, the HAS_IREF flag
> is set and ihold()
> 2.b. If mark already exists with HAS_IREF flag, mask is updated
>
> Do we have a winner?
>
FYI, I've implemented the above and pushed to branch fan_evictable.
Yes, I also changed the name of the flag to be more coherent with the
documented behavior:
fanotify: add support for "evictable" inode marks
When an inode mark is created with flag FAN_MARK_EVICTABLE, it will not
pin the marked inode to inode cache, so when inode is evicted from cache
due to memory pressure, the mark will be lost.
When an inode mark with flag FAN_MARK_EVICATBLE is updated without using
this flag, the marked inode is pinned to inode cache.
When an inode mark is updated with flag FAN_MARK_EVICTABLE but an
existing mark already has the inode pinned, the mark update fails with
error EEXIST.
I also took care of avoiding direct reclaim deadlocks from fanotify_add_mark().
If you agree to the proposed UAPI I will post v2 patches.
Thanks,
Amir.
