Re: [PATCH 0/2] proc: proc_setattr for /proc/$PID/net

[Willy Tarreau]

On Sun, Jul 09, 2023 at 07:57:27PM +0200, Thomas Weißschuh wrote:
> Hi Willy,
> 
> On 2023-07-09 19:27:53+0200, Willy Tarreau wrote:
> > On Sun, Jul 09, 2023 at 07:10:58PM +0200, Thomas Weißschuh wrote:
> > > On 2023-07-09 11:29:47+0200, Willy Tarreau wrote:
> > > > On Fri, Jun 30, 2023 at 10:06:09PM +0800, Zhangjin Wu wrote:
> > > >> [..]
> > > > 
> > > > Now queued, thanks!
> > > > Willy
> > > 
> > > Don't we need an Ack from the fs maintainers for the patch to
> > > fs/proc/proc_net.c ?
> > > 
> > > Personally I expected this series to go in via the fs tree because of
> > > that patch.
> > 
> > Gasp! You're totally right, I confused it with a test only changing
> > the nolibc-test file, as the chmod_net test appeared as a dependency!
> > Let me drop it from the series and push again.
> 
> I think if this patch now also goes in via both the nolibc/rcu trees and
> the fs tree it would not be great.
>
> The best way forward would probably for you to rebase your tree on top
> of mainline after the fs tree has introduced both patches of the series
> into Linus' tree and then you can drop your copy of the test removal.

Yeah I agree.

> I want to keep both patches together because I expect the fs change to
> be backported and if it is backported on its own it will break
> nolibc-test in those trees.

OK but we can also fix the test regardless, and mark it for backport, no ?

> But maybe I'm overthinking it, nobody is running nolibc-test on
> non-mainline kernels anyways and both patches can be split.

I agree that we shouldn't grant too much importance to this test ;-)
I'm regularly seeing Sasha propose them for backports and am thinking
"ok it cannot hurt but I'm not convinced anyone will notice the fix".

> If they are to be kept together and go via fs an Ack on the nolibc-test
> patch is probably needed, too.

OK. Let's first see if someone from FS agrees on the change.

Thanks for the clarification,
Willy