On Sun, Mar 17, 2024 at 9:31 PM Miklos Szeredi <miklos@xxxxxxxxxx> wrote:
>
> On Sun, 17 Mar 2024 at 14:54, Amir Goldstein <amir73il@xxxxxxxxx> wrote:>
>
> > I see that you decided to drop the waiting for parallel dio logic
> > in the final version. Decided to simply or found a problem?
>
> I don't think I dropped this. Which one are you thinking?
My bad. just looked at the wrong diff.
>
> > Also, FUSE passthrough patch 9/9 ("fuse: auto-invalidate inode
> > attributes in passthrough mode") was not included in the final version,
> > so these fstests are failing on non uptodate size/mtime/ctime after
> > mapped write:
> >
> > Failures: generic/080 generic/120 generic/207 generic/215
> >
> > Was it left out by mistake or for a reason?
>
> Yes, this was deliberate. For plain read/write it's simply
> unnecessary, since it should get 100% hit rate on the c/mtime tests.
>
> For mmap we do need something, and that something might be best done
> by looking at backing inode times. But even in that case I'd just
> compare the times from a previous version of the backing inode to the
> current version, instead of comparing with the cached values.
>
Well, looking ahead to implementing more passthrough ops, I have
a use case for passthrough of getattr(), so if that will be a possible
configuration in the future, maybe we won't need to worry about
the cached values at all?
Thanks,
Amir.
