Re: [PATCH 2/6] ARM: xen: Register with kernel restart handler

[Thierry Reding]

On Thu, Jun 03, 2021 at 03:03:01PM +0100, Lee Jones wrote:
> On Thu, 03 Jun 2021, Russell King (Oracle) wrote:
> 
> > On Thu, Jun 03, 2021 at 09:48:59AM -0400, Boris Ostrovsky wrote:
> > > On 6/3/21 9:38 AM, Lee Jones wrote:
> > > > On Thu, 03 Jun 2021, Guenter Roeck wrote:
> > > >
> > > >> On Thu, Jun 03, 2021 at 01:43:36PM +0100, Lee Jones wrote:
> > > >>> On Tue, 15 Oct 2019 at 15:52, Thierry Reding <thierry.reding@xxxxxxxxx>
> > > >>> wrote:
> > > >>>
> > > >>>> From: Guenter Roeck <linux@xxxxxxxxxxxx>
> > > >>>>
> > > >>>> Register with kernel restart handler instead of setting arm_pm_restart
> > > >>>> directly.
> > > >>>>
> > > >>>> Select a high priority of 192 to ensure that default restart handlers
> > > >>>> are replaced if Xen is running.
> > > >>>>
> > > >>>> Acked-by: Arnd Bergmann <arnd@xxxxxxxx>
> > > >>>> Reviewed-by: Wolfram Sang <wsa+renesas@xxxxxxxxxxxxxxxxxxxx>
> > > >>>> Reviewed-by: Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>
> > > >>>> Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> > > >>>> Signed-off-by: Thierry Reding <treding@xxxxxxxxxx>
> > > >>>> ---
> > > >>>>  arch/arm/xen/enlighten.c | 12 ++++++++++--
> > > >>>>  1 file changed, 10 insertions(+), 2 deletions(-)
> > > >>>>
> > > >>> This patch does appear to be useful.
> > > >>>
> > > >>> Is this just being solved in downstream trees at the moment?
> > > >>>
> > > >>> It would be nice if we could relinquish people of this burden and get it
> > > >>> into Mainline finally.
> > > >>>
> > > >> There must have been half a dozen attempts to send this patch series
> > > >> upstream. I have tried, and others have tried. Each attempt failed with
> > > >> someone else objecting for non-technical reasons (such as "we need more
> > > >> reviews") or no reaction at all, and maintainers just don't pick it up.
> > > > Looking at the *-by tag list above, I think we have enough quality
> > > > reviews to take this forward.
> > > >
> > > >> So, yes, this patch series can only be found in downstream trees,
> > > >> and it seems pointless to submit it yet again.
> > > > IMHO, it's unfair to burden multiple downstream trees with this purely
> > > > due to poor or nervy maintainership.  Functionality as broadly useful
> > > > as this should be merged and maintained in Mainline.
> > > >
> > > > OOI, who is blocking?  As I see it, we have 2 of the key maintainers
> > > > in the *-by list.  With those on-board, it's difficult to envisage
> > > > what the problem is.
> > > 
> > > 
> > > Stefano (who is ARM Xen maintainer) left Citrix a while ago. He is at sstabellini@xxxxxxxxxx (which I added to the CC line).
> > 
> > Stefano already reviewed this patch, which is part of a larger series
> > that primarily touches 32-bit ARM code, but also touches 64-bit ARM
> > code as well.
> > 
> > As I said in my previous reply, I don't see that there's any problem
> > with getting these patches merged had the usual processes been
> > followed - either ending up in the patch system, or the pull request
> > being sent to me directly.
> > 
> > Sadly, the pull request was sent to the arm-soc people excluding me,
> > I happened to notice it and requested to see the patches that were
> > being asked to be pulled (since I probably couldn't find them)...
> > and it then took two further weeks before the patches were posted,
> > which I then missed amongst all the other email.
> > 
> > It's a process failure and unfortunate timing rather than anything
> > malicious.
> 
> Understood.
> 
> Is there anything I can do to help this forward?
> 
> I can either collect and re-submit the patches to the MLs if that
> makes people's lives any easier.  Or if one of the original submitters
> wish to retain responsibility, I have no qualms with that either.

I had stumbled across these patches from Guenter when I was looking to
solve a reboot/power-off issue on one of the boards that I was working
on. This was supposed to be preparatory work to get rid of the global
function pointers that drivers are simply overwriting, and the goal had
been to add a "system power" framework that would allow drivers to
register a chip structure to provide a bit more "formality" than
overwriting function pointers or registering notifier callbacks.

There ended up not being any interest in such a subsystem, so I wanted
to at least get this prep work in, because it is at least a bit of an
improvement.

I vaguely recall that Arnd or perhaps Olof had mentioned that they'd
pull these patches, but the timing was bad, so they asked me to remind
them after the merge window. By the time we had gotten through the merge
window, I probably had gotten sidetracked and forgot...

Feel free to give this a shot. This series itself is still useful, in my
opinion.

Thierry

Attachment: signature.asc
Description: PGP signature