Re: [PATCH] ext4: allow readdir()'s of large empty directories to be interrupted

[Theodore Ts'o <tytso@mit.edu>]

On Thu, Apr 14, 2016 at 07:01:14PM -0700, Linus Torvalds wrote:
> On Thu, Apr 14, 2016 at 1:37 PM, Theodore Ts'o <tytso@mit.edu> wrote:
> >
> >  So Linus's proposal to
> > add "if (signal_pending(current)) return -EINTR;" to filldir64() would
> > probably cause more than a few userspace regressions.
> 
> I don't think you actually read or understood my proposal.
> 
> The proposal added it to inside the if-statement in
> 
>         dirent = buf->previous;
>         if (dirent) {
> +               if (signal_pending(current)) return -EINTR;
> 
> and that actually guarantees that readdir() _never_ returns -EINTR,
> because there is at least one entry that got filled out (the previous
> one filled in, now pointed to be "dirent").

Sorry, yes I didn't understand that was what you were getting at.
That makes a lot of sense.  Thanks!!

					- Ted