linux-kernel - Re: [PATCH 3/4] media: dw100: Fix kernel oops with PREEMPT

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: <20260105235921.GI10026@pendragon.ideasonboard.com>
Date: Tue, 6 Jan 2026 01:59:21 +0200
From: Laurent Pinchart <laurent.pinchart@...asonboard.com>
To: Nicolas Dufresne <nicolas@...fresne.ca>
Cc: Stefan Klug <stefan.klug@...asonboard.com>,
	Xavier Roumegue <xavier.roumegue@....nxp.com>,
	Mauro Carvalho Chehab <mchehab@...nel.org>,
	Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
	Clark Williams <clrkwllms@...nel.org>,
	Steven Rostedt <rostedt@...dmis.org>, linux-media@...r.kernel.org,
	linux-kernel@...r.kernel.org, linux-rt-devel@...ts.linux.dev
Subject: Re: [PATCH 3/4] media: dw100: Fix kernel oops with PREEMPT_RT enabled

On Mon, Jan 05, 2026 at 02:02:38PM -0500, Nicolas Dufresne wrote:
> Le lundi 05 janvier 2026 à 12:35 +0100, Stefan Klug a écrit :
> > On kernels with PREEMPT_RT enabled, a "BUG: scheduling while atomic"
> > kernel oops occurs inside dw100_irq_handler -> vb2_buffer_done. This is
> > because vb2_buffer_done takes a spinlock which is not allowed within
> > interrupt context on PREEMPT_RT.
> > 
> > Fix that by making the irq handler threaded. The threaded interrupt
> > handling might cause the interrupt line to be disabled a little longer
> > than before. As the line is not shared, this has no negative side
> > effects.
> 
> That's interesting, do you plan to update more drivers ? There is a lot of m2m
> using hard IRQ to minimize the idle time (save a context switch), but RT support
> might be more worth then that.

This is a part of PREEMPT_RT that puzzles me. By turning regular
spinlocks into mutexes, RT seems to break drivers that use those
spinlocks in hard IRQ handlers. That's a very large number of drivers
given how widespread regular spinlock usage is. Do drivers need to be
manually converted to either raw spinlocks or threaded IRQ handlers ?
What about non-RT kernels, how can a driver avoid the thread scheduling
penalty in those cases, do they need to manually select between
request_irq() and request_threaded_irq() based on if RT is enabled ?
This puzzles me, it feels like I must be missing something.

> > Signed-off-by: Stefan Klug <stefan.klug@...asonboard.com>
> 
> Reviewed-by: Nicolas Dufresne <nicolas.dufresne@...labora.com>
> 
> > ---
> >  drivers/media/platform/nxp/dw100/dw100.c | 5 +++--
> >  1 file changed, 3 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/media/platform/nxp/dw100/dw100.c
> > b/drivers/media/platform/nxp/dw100/dw100.c
> > index
> > 8a421059a1c9b55f514a29d3c2c5a6ffb76e0a64..4f5ef70e5f4a052fb5f208e35f8785f9d30d
> > c54e 100644
> > --- a/drivers/media/platform/nxp/dw100/dw100.c
> > +++ b/drivers/media/platform/nxp/dw100/dw100.c
> > @@ -1600,8 +1600,9 @@ static int dw100_probe(struct platform_device *pdev)
> >  
> >  	pm_runtime_put_sync(&pdev->dev);
> >  
> > -	ret = devm_request_irq(&pdev->dev, irq, dw100_irq_handler, IRQF_ONESHOT,
> > -			       dev_name(&pdev->dev), dw_dev);
> > +	ret = devm_request_threaded_irq(&pdev->dev, irq, NULL,
> > +					dw100_irq_handler, IRQF_ONESHOT,
> > +					dev_name(&pdev->dev), dw_dev);
> >  	if (ret < 0) {
> >  		dev_err(&pdev->dev, "Failed to request irq: %d\n", ret);
> >  		goto err_pm;

-- 
Regards,

Laurent Pinchart