NACK: [SRU][Bionic][PATCH 0/2] Cutting and Pasting files from NFS sec=sys to NFS sec=krb5p causes NFS to hang

[Matthew Ruffell]

On Fri, Jul 17, 2020 at 11:51 AM Matthew Ruffell
<matthew.ruffell at canonical.com> wrote:
>
> >
> > I think we may need to apply also:
> >
> >  29e7ca715f2a ("NFS: Fix handling of reply page vector")
> >
> > that seems to fix a bug introduced by 02ef04e432ba. If that's the case
> > could you re-send the patch set including also this fix.
>
> Thanks Andrea, yes, I agree we need this patch. Sorry for missing it!
>
> I will build a fresh test kernel, test it and resend the patch set.
>
Okay, I built a test kernel based on:

- Ubuntu-4.15.0-111-generic:
  + cf500bac8fd4 "SUNRPC: Introduce rpc_prepare_reply_pages()"
  + 02ef04e432ba "NFS: Account for XDR pad of buf->pages"
  + 29e7ca715f2a "NFS: Fix handling of reply page vector"

The most curious thing happens - the issue returns and is reproducible. It seems
the fixup commit either breaks things, or what I believed to be fixes don't
actually fix the problem, but only cover it up.

I am going to investigate this more deeply, and will resubmit for the next
SRU cycle once I have determined what is happening.

Engineering notes:

If I build a kernel based on:
- 29e7ca715f2a "NFS: Fix handling of reply page vector"

It works.

If I build a kernel based on:
- 02ef04e432ba "NFS: Account for XDR pad of buf->pages"

It works.

If I build a kernel based on:

- 02ef04e432ba "NFS: Account for XDR pad of buf->pages"
  + 29e7ca715f2a "NFS: Fix handling of reply page vector"

It fails, and the issue returns.

Therefore, something additionally got fixed between 02ef04e432ba..29e7ca715f2a
which might be the actual root cause.

Looks like I'm going to have to bisect again.