linux-kernel - Re: [PATCH v3 5/6] spi: rockchip: Suspend and resume the bus during NOIRQ_SYSTEM_SLEEP

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 for Android: free password hash cracker in your pocket

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

Message-ID: <Zsj_CJR56TcahXBO@google.com>
Date: Fri, 23 Aug 2024 14:28:40 -0700
From: Brian Norris <briannorris@...omium.org>
To: Ondřej Jirman <megi@....cz>,
	Jon Lin <jon.lin@...k-chips.com>, broonie@...nel.org,
	heiko@...ech.de, linux-spi@...r.kernel.org,
	linux-arm-kernel@...ts.infradead.org,
	linux-rockchip@...ts.infradead.org, linux-kernel@...r.kernel.org,
	shengfei Xu <xsf@...k-chips.com>
Subject: Re: [PATCH v3 5/6] spi: rockchip: Suspend and resume the bus during
 NOIRQ_SYSTEM_SLEEP_PM ops

On Tue, Jun 21, 2022 at 05:42:18PM +0200, Ondřej Jirman wrote:
> On Wed, Feb 16, 2022 at 09:40:27AM +0800, Jon Lin wrote:
> > From: shengfei Xu <xsf@...k-chips.com>
> > 
> > the wakeup interrupt handler which is guaranteed not to run while
> > @resume noirq() is being executed. the patch can help to avoid the
> > wakeup source try to access spi when the spi is in suspend mode.
> 
> This patch causes oops on suspend every single time, because it tries to disable
> already disabled clocks (one disable in runtime PM suspend, and other one in
> system suspend). It also fails to properly handle errors from clk_prepare_enable
> in rockchip_spi_resume, potentially causing even more clock enable/disable
> imballance issues.
> 
> Please send a revert and figure out a better fix for the original issue.

Did anyone ever resolve this? I still see this issue on 6.10.6.

IIUC, we can actually do a partial revert -- we *should* be using
pm_runtime_force_{suspend,resume}() (so that we coordinate clk-disable,
etc. with the runtime PM state), but we also want to run in the noirq
phase.

If I don't see anyone else's solution or input, I'll plan on submitting
a partial revert, which seems to test out OK for me.

Brian

P.S. One can work around this problem by disabling runtime PM for the
controllers:

for i in /sys/bus/platform/drivers/rockchip-spi/*/power/control; do
  echo on >$i
done

But obviously that's not ideal.