linux-kernel - Re: [PATCH v1 0/6] Fix RK3588 GPU domain

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <vh5gsouseahs6ddauzevcdtcjutl35dcyoc5cv4pvn56lyyilv@rekmpklpbbyh>
Date: Thu, 19 Sep 2024 11:05:35 +0200
From: Sebastian Reichel <sebastian.reichel@...labora.com>
To: Ulf Hansson <ulf.hansson@...aro.org>
Cc: Rob Herring <robh@...nel.org>, 
	Krzysztof Kozlowski <krzk+dt@...nel.org>, Conor Dooley <conor+dt@...nel.org>, 
	Heiko Stuebner <heiko@...ech.de>, Elaine Zhang <zhangqing@...k-chips.com>, 
	Adrián Martínez Larumbe <adrian.larumbe@...labora.com>, Boris Brezillon <boris.brezillon@...labora.com>, 
	devicetree@...r.kernel.org, linux-rockchip@...ts.infradead.org, linux-kernel@...r.kernel.org, 
	linux-pm@...r.kernel.org, kernel@...labora.com
Subject: Re: [PATCH v1 0/6] Fix RK3588 GPU domain

Hi,

On Fri, Sep 13, 2024 at 01:59:10PM GMT, Ulf Hansson wrote:
> On Tue, 10 Sept 2024 at 20:05, Sebastian Reichel
> <sebastian.reichel@...labora.com> wrote:
> > I got a report, that the Linux kernel crashes on Rock 5B when the panthor
> > driver is loaded late after booting. The crash starts with the following
> > shortened error print:
> >
> > rockchip-pm-domain fd8d8000.power-management:power-controller: failed to set domain 'gpu', val=0
> > rockchip-pm-domain fd8d8000.power-management:power-controller: failed to get ack on domain 'gpu', val=0xa9fff
> > SError Interrupt on CPU4, code 0x00000000be000411 -- SError
> >
> > This series first does some cleanups in the Rockchip power domain
> > driver and changes the driver, so that it no longer tries to continue
> > when it fails to enable a domain. This gets rid of the SError interrupt
> > and long backtraces. But the kernel still hangs when it fails to enable
> > a power domain. I have not done further analysis to check if that can
> > be avoided.
> >
> > Last but not least this provides a fix for the GPU power domain failing
> > to get enabled - after some testing from my side it seems to require the
> > GPU voltage supply to be enabled.
> >
> > I'm not really happy about the hack to get a regulator for a sub-node
> > in the 5th patch, which I took over from the Mediatek driver. But to
> > get things going and open a discussion around it I thought it would be
> > best to send a first version as soon as possible.
> 
> That creates a circular dependency from the fw_devlink point of view.

Yes.

> I assume that isn't a problem and fw_devlink takes care of this, so
> the GPU power domain still can probe?

This has been tested on Radxa Rock 5B and RK3588 EVB1. It properly
probes the GPU power domain and fixes late probing of the GPU driver :)

> Other than this, I think this looks okay to me.

I will send a V2 with the minor things pointed out.

Greetings,

-- Sebastian

Download attachment "signature.asc" of type "application/pgp-signature" (834 bytes)