linux-kernel - Re: [PATCH] nouveau: pci: quiesce GPU on shutdown

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

Message-ID: <19bde331fbd.dd44bec734977.5910036498943757747@linux.beauty>
Date: Wed, 21 Jan 2026 09:37:31 +0800
From: Li Chen <me@...ux.beauty>
To: "Danilo Krummrich" <dakr@...nel.org>
Cc: "Dave Airlie" <airlied@...il.com>, "Lyude Paul" <lyude@...hat.com>,
	"Maarten Lankhorst" <maarten.lankhorst@...ux.intel.com>,
	"Maxime Ripard" <mripard@...nel.org>,
	"Thomas Zimmermann" <tzimmermann@...e.de>,
	"Simona Vetter" <simona@...ll.ch>,
	"dri-devel" <dri-devel@...ts.freedesktop.org>,
	"nouveau" <nouveau@...ts.freedesktop.org>,
	"linux-kernel" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] nouveau: pci: quiesce GPU on shutdown

Hi Danilo

 > On Wed Jan 21, 2026 at 1:50 AM CET, Li Chen wrote:
 > >  > > +       usleep_range(200, 400);\
 > >  > 
 > >  > Why is this needed? it at least needs a comment.
 > >
 > > This patch is needed for kexec/KHO/LUO reboot: firmware reset is skipped, so the GPU may remain in an active/DMA 
 > > state and the next kernel can fail to re-init nouveau. We can’t call the full DRM/TTM teardown from ->shutdown() because
 > > userspace may still hold DRM fds during shutdown, which triggers WARNs.
 > >
 > > So the shutdown hook reuses the suspend/quiesce path (safe with open fds) and then powers down the PCI function (D3hot)
 > > to leave the device in a consistent state for the next kernel.
 > 
 > I think the question was about the usleep_range() specifically. :)
 >

Thanks for the hint. 

That usleep_range(200, 400) is just to give the PCI power transition time to settle before an immediate kexec jump.

 It’s mirroring the existing nouveau_pmops_suspend() behavior, which already does udelay(200) right after pci_set_power_state(..., PCI_D3hot). In ->shutdown() we’re allowed to
 sleep, so I used usleep_range() instead of a busy-wait udelay().
 
Regards,
Li