Re: [Lsf-pc] [LSF/MM/BPF TOPIC] sorting pathwalk semantics out

Re: [Lsf-pc] [LSF/MM/BPF TOPIC] sorting pathwalk semantics out

[Amir Goldstein]

Al,

Just making this public as requested, so potential attendees can chime in.
I suppose this was your intention.

Thanks,
Amir.

On Fri, Feb 14, 2020 at 7:09 AM Al Viro <viro@zeniv.linux.org.uk> wrote:
>
>         There had been quite a bit of activity around that area
> (openat2() series, for example) and IMO the complexity in the damn
> thing is getting close to critical - we are not quite at the point
> where the average amount of bugs introduced by fixing a bug is greater
> than one, but it's uncomfortably close.  Quite a few places are
> far too subtle and convoluted (anything related to do_last(), for
> one thing).
>
>         We need to get that mess under control.  I have a (still
> growing) series massaging it to somewhat saner shape, but there
> are some interesting dark corners that need to be sorted out (situation
> around mount traps, for example).
>
>         We definitely need to settle on some description of semantics.
> Preferably with a set of litmus tests, similar to what memory model
> folks had been doing.
>
>         I would like to catch at least the people who'd been active
> around that area.  Miklos Szeredi (due to atomic_open involvement,
> if nothing else), David Howells and Ian Kent (mount traps, among
> other things), Eric Biederman (userns fallouts of all sorts),
> Aleksa Sarai (openat2 - most recent large changes in the whole
> thing), Jeff Layton (revalidation et.al.)...
> _______________________________________________
> Lsf-pc mailing list
> Lsf-pc@lists.linux-foundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/lsf-pc