On Tue, 10 May 2022 at 09:33, ronnie sahlberg <ronniesahlberg@xxxxxxxxx> wrote:
>
> On Fri, 6 May 2022 at 16:21, Enzo Matsumiya <ematsumiya@xxxxxxx> wrote:
> >
> > On 05/04, Ronnie Sahlberg wrote:
> > >+struct cached_dirents {
> > >+ bool is_valid:1;
> > >+ bool is_failed:1;
> >
> > Does it make sense to have both? Do you expect a situation where
> > is_valid && is_failed happens? From the patch, I don't see such case and
> > the code could be adjusted to use !is_valid where appropriate.
> > But let me know if I missed something.
>
> The reason to have both is to handle cases where we fail populating
> the cache partway through readdir.
>
> The idea is that is_valid is set once we have fully populated the
> cache successfully
> and is_failed is set on failure during the population of the cache and
> once it is set there is no point in even trying to add new entries to
> the cache.
What I mean is that we have three states and thus need two booleans:
1, successfully scanned the whole directory and cached all entries
2, in the process of scanning the directory and no failure yet
3, in the process of scanning the direcotry but we have already
encountered a failure
>
> >
> > This is just a cosmetic nitpick, but other than that,
> >
> > Reviewed-by: Enzo Matsumiya <ematsumiya@xxxxxxx>
> >
> >
> > Cheers,
> >
> > Enzo
