Re: [PATCH] firmware: arm_scmi: Relax BASE protocol sanity checks on protocol list
From: Sudeep Holla
Date: Mon Jun 06 2022 - 11:11:00 EST
On Mon, Jun 06, 2022 at 04:55:10PM +0200, Heiko Stübner wrote:
> Am Montag, 6. Juni 2022, 16:43:05 CEST schrieb Sudeep Holla:
> > On Mon, Jun 06, 2022 at 02:31:48PM +0100, Cristian Marussi wrote:
> > > On Mon, Jun 06, 2022 at 02:59:10PM +0200, Michael Riesch wrote:
> > > > Hi Cristian,
> > > >
> > > Hi Michael,
> > >
> > > > On 5/23/22 19:15, Cristian Marussi wrote:
> > > > > Even though malformed replies from firmware must be treated carefully to
> > > > > avoid memory corruption Kernel side, some out-of-spec SCMI replies can
> > > > > be tolerated to avoid breaking existing deployed system, as long as they
> > > > > won't cause memory issues.
> > > > >
> > > > > Reported-by: Nicolas Frattaroli <frattaroli.nicolas@xxxxxxxxx>
> > > > > Cc: Etienne Carriere <etienne.carriere@xxxxxxxxxx>
> > > > > Cc: Sudeep Holla <sudeep.holla@xxxxxxx>
> > > > > Signed-off-by: Cristian Marussi <cristian.marussi@xxxxxxx>
> > > >
> > > > Thanks a lot, without this fix the Mali G52 GPU won't probe on my RK3568
> > > > EVB1 in vanilla v5.19-rc1.
> > > >
> > >
> > > Yes, the break was reported on -next and today it appeared in 5.19-rc1.
> > > A proper FW fix is also up for review by Etienne but in the meantime
> > > this tries to limit damages relaxing a bit the checks.
> > >
> > > > I guess this patch should have a Fixes: tag, right?
> > > >
> > >
> > > It has not a Fixes tag because the issue was introduced in 5.19-rc1 and the
> > > fix will go in with the next round of v5.19 fixes by Sudeep (AFAIU) so it
> > > will be solved within the v5.19 cycle and I thought the Fixes tag was
> > > not needed in this case (I could be wrong...)
> >
> > Correct, if for some reason, we can't push this before v5.19, then fixes
> > tag needs to added. I will add that then, but for now let us target getting
> > it in before v5.19
>
> hmm, I'd disagree for the generalization.
>
> While true that is not 100% necessary to be present in all cases, so
> definitly no reason for a new version when applied to the same -rc series,
> having the Fixes tag not only clearly marks the patch as such, but also
> allows people reading either mailing lists or the later the git history
> to actually see where the issue started. So I really think it is a
> nice-to-have in most cases.
Absolutely agreed in general. But I asked Cristian to not add as we are work
around the firmware bug by relaxing the checks in the kernel. This is not
fixing anything in the original commit IMO, it is just that the original
commit highlighted the firmware issue on that system.
--
Regards,
Sudeep