Re: [GIT PULL] io_uring updates for 5.14-rc1

[Linus Torvalds <torvalds@linux-foundation.org>]

On Fri, Jul 2, 2021 at 4:32 AM Dmitry Kadashev <dkadashev@gmail.com> wrote:
>
> The problem is we have to keep the filenames around for retries on
> ESTALE. It's not consumed by __filename_create() on success. So it's not
> as simple as setting the name to NULL after calling __filename_create().

I wonder if the semantics couldn't be that __filename_create() never
eats the name, and filename_create() keeps the old semantics?

You kind of made it go halfway, with filename_create() eating it only
on success.

That would make the filename_create() wrapper much simpler too, ie
we'd have it be just

    struct dentry *res = __filename_create(dfd, name, path, lookup_flags);
    putname(name);
    return res;

so it would remove a lot of conditionals, and leave it to callers
whether they want to keep the name live or not.

But I didn't check all the other cases, so maybe that causes its own
set of inconveniences..

            Linus