Re: [PATCH] arm64/io: Don't use WZR in writel

From: AngeloGioacchino Del Regno <kholk11@gmail.com>
To: Marc Gonzalez <marc.w.gonzalez@free.fr>
Cc: Mark Rutland <mark.rutland@arm.com>,
	Marc Zyngier <marc.zyngier@arm.com>,
	MSM <linux-arm-msm@vger.kernel.org>,
	Will Deacon <will.deacon@arm.com>,
	Bjorn Andersson <bjorn.andersson@linaro.org>,
	Jeffrey Hugo <jhugo@codeaurora.org>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Robin Murphy <robin.murphy@arm.com>,
	Linux ARM <linux-arm-kernel@lists.infradead.org>
Subject: Re: [PATCH] arm64/io: Don't use WZR in writel
Date: Sun, 5 May 2019 20:05:44 +0200	[thread overview]
Message-ID: <CAK7fi1Y64uawt5VFqec9KzcytxW0bKOHAECv6gQ4yS1=jedg0Q@mail.gmail.com> (raw)
In-Reply-To: <CAK7fi1b5PP+ToK5fS6xEvGaNF=43=OtA8=5KhypTcFeryWSr9Q@mail.gmail.com>

orry again for the double email, but we've just discovered another
issue with using the zero register for writel.
On the downstream kernel we are observing an entire userspace
crash when closing the camera on * at least * MSM8996, SDM630
and SDM636. We're not sure that the issue is present on MSM8998
and we should probably test MSM8956 as well.

I know, the testing is not really complete, but I thought to urgently
write this to you guys as soon as the thing got discovered.

I'm mostly sure that this is not happening because of the IOMMU again
but instead because of something within the camera drivers, probably
some writel to the ISP or to the VFE is the cause.

Last thing, yes - this was discovered on downstream (kernel 4.9) and
not on mainline, but the IOMMU thing was also discovered there, so
something tells me that it's gonna be affecting the camera drivers that
we currently have on mainline as well, or future developments of it.
Especially if it's another hypervisor crap.

I'm going to do some more research about that camera thing downstream
ASAP and if you want I will keep you informed.

Have a nice day,
Angelo

Il giorno sab 4 mag 2019 alle ore 15:35 AngeloGioacchino Del Regno
<kholk11@gmail.com> ha scritto:
>
> Sorry, I've replied yesterday but I just realized that I did click on the wrong
> button and the email got sent only to Bjorn. My bad.
> Resending to all... this is the original text:
>
> For me, the [1] solution is working fine on qcom SDM630 (Xperia XA2),
> MSM8998 (Xperia XZ Premium) and MSM8996 (Xperia X Performance).
> I couldn't test on others that I have for time reasons, but I think that it's
> not even needed.
>
> By the way, I suggest to clearly document the fact that (from what I
> understand, at least) we can write whatever value we want to that
> register, as the change as it is may confuse some developers around.
>
> In any case... if you want, you can also include my:
>
> Tested-by: AngeloGioacchino Del Regno <kholk11@gmail.com>
>
> Il giorno ven 3 mag 2019 alle ore 15:07 Marc Gonzalez
> <marc.w.gonzalez@free.fr> ha scritto:
> >
> > On 03/05/2019 14:48, Robin Murphy wrote:
> >
> > > Anyway, I'll clean up my patch and post it properly - thanks to you and
> > > Bjorn for testing.
> >
> > Cool. Thanks!
> >
> > AngeloGioacchino, are you still monitoring this thread?
> >
> > On which qcom platform(s) did you run into the issue?
> > (Robin's work-around has been tested on msm8996 and msm8998.)
> >
> > Regards.