Re: [PATCH 6.1 14/29] xfs: use xfs_defer_pending objects to recover intent items

This sounds good to me.

Greg, can we drop the following patches?

'[PATCH 6.1 14/29] xfs: use xfs_defer_pending objects to recover'
'[PATCH 6.1 15/29] xfs: pass the xfs_defer_pending object to iop_recover'
'[PATCH 6.1 16/29] xfs: transfer recovered intent item ownership in
->iop_recover'

Thanks,
leah

On Mon, Mar 24, 2025 at 1:53 AM Fedor Pchelkin <[email protected]> wrote:
>
> On Sun, 23. Mar 17:29, Leah Rumancik wrote:
> > Okay so a summary from my understanding, correct me if I'm wrong:
> >
> > 03f7767c9f612 introduced the issue in both 6.1 and 6.6.
> >
> > On mainline, this is resolved by e5f1a5146ec3. This commit is painful
> > to apply to 6.1 but does apply to 6.6 along with the rest of the
> > patchset it was a part of (which is the set you just sent out for
> > 6.6).
>
> Yeah, that's all correct.
>
> >
> > With the stable branches we try to balance the risk of introducing new
> > bugs via huge fixes with the benefit of the fix itself. Especially if
> > the patches don't apply cleanly, it might not be worth the risk and
> > effort to do the porting. Hmm, since it seems like we might not even
> > end up taking 03f7767c9f6120 to stable, I'd propose we just drop
> > 03f7767c9f6120 for now. If the rest of the subsequent patches in the
> > original set apply cleanly, I don't think we need to drop them all. We
> > can then try to fix the UAF with a more targeted approach in a later
> > patch instead of via direct cherry-picks.
> >
> >  What do you think?
>
> 03f7767c9f6120 is '[PATCH 6.1 14/29] xfs: use xfs_defer_pending objects to recover'
>
> Two subsequent patches depend on it logically so should also be dropped:
>
> '[PATCH 6.1 15/29] xfs: pass the xfs_defer_pending object to iop_recover'
> '[PATCH 6.1 16/29] xfs: transfer recovered intent item ownership in ->iop_recover'
>
>
> On the other side, '[PATCH 6.1 13/29] xfs: don't leak recovered attri intent items'
> which is at the start of the original patchset [1] looks OK to be taken.
> It's rather aside from the subsequent rework patches and fixes a pinpoint
> bug.
>
> [1]: https://lore.kernel.org/linux-xfs/170191741007.1195961.10092536809136830257.stg-ugh@frogsfrogsfrogs/
>
>
> So I've tried the current xfs backport series with three dropped commits:
>
> [PATCH 6.1 14/29] xfs: use xfs_defer_pending objects to recover
> [PATCH 6.1 15/29] xfs: pass the xfs_defer_pending object to iop_recover
> [PATCH 6.1 16/29] xfs: transfer recovered intent item ownership in ->iop_recover
>
> (everything before and after that still applies cleanly and touches
> other things)
>
> and no regressions seen on my side.