linux-kernel - Re: [PATCH] staging: media: tegra-video: Remove the use of dev_err

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: <jodxtve3lnv4hc2f2cm25ckdj6ukfzsbqjzpxuyqwx2qmsw27w@emg722bix2a5>
Date: Tue, 19 Aug 2025 13:13:09 +0200
From: Uwe Kleine-König <u.kleine-koenig@...libre.com>
To: Luca Ceresoli <luca.ceresoli@...tlin.com>
Cc: Xichao Zhao <zhao.xichao@...o.com>, thierry.reding@...il.com, 
	jonathanh@...dia.com, skomatineni@...dia.com, mchehab@...nel.org, 
	gregkh@...uxfoundation.org, linux-media@...r.kernel.org, linux-tegra@...r.kernel.org, 
	linux-staging@...ts.linux.dev, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] staging: media: tegra-video: Remove the use of
 dev_err_probe()

Hello,

On Tue, Aug 19, 2025 at 12:28:30PM +0200, Luca Ceresoli wrote:
> +Uwe, author of 2f3cfd2f4b7c ("driver core: Make dev_err_probe() silent
> for -ENOMEM").

The main purpose of that commit was to make dev_err_probe() silent if
the error code might be ENOMEM. If it's sensible to call dev_err_probe()
if you know that the error code is ENOMEM is subjective. On one hand
it's explicitly a noop then, but on the other hand it might serve as
code documentation and make it explicit that error messaging is
consistent.

There was also a recent discussion about:

	ret = devm_add_action(...)
	if (ret)
		return dev_err_probe(..., ret, ...);

which is a case of "ret can only be 0 or -ENOMEM", but that's harder to
spot than the patched code below.
		
> On Tue, 19 Aug 2025 17:23:30 +0800
> Xichao Zhao <zhao.xichao@...o.com> wrote:
> 
> > The dev_err_probe() doesn't do anything when error is '-ENOMEM'. Therefore,
> > remove the useless call to dev_err_probe(), and just return the value instead.
> > 
> > Signed-off-by: Xichao Zhao <zhao.xichao@...o.com>
> > ---
> >  drivers/staging/media/tegra-video/tegra20.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/staging/media/tegra-video/tegra20.c b/drivers/staging/media/tegra-video/tegra20.c
> > index 7b8f8f810b35..ee65e89c119c 100644
> > --- a/drivers/staging/media/tegra-video/tegra20.c
> > +++ b/drivers/staging/media/tegra-video/tegra20.c
> > @@ -255,7 +255,7 @@ static int tegra20_channel_host1x_syncpt_init(struct tegra_vi_channel *chan)
> >  
> >  	out_sp = host1x_syncpt_request(&vi->client, HOST1X_SYNCPT_CLIENT_MANAGED);
> >  	if (!out_sp)
> > -		return dev_err_probe(vi->dev, -ENOMEM, "failed to request syncpoint\n");
> > +		return -ENOMEM;
> 
> Thanks for your patch!
> 
> Reviewed-by: Luca Ceresoli <luca.ceresoli@...tlin.com>
> 
> About dev_err_probe(), I wonder whether it should mention the -ENOMEM
> exception in kerneldoc. Is it part of the API "contract" it provides?

In general you're not supposed to wail about -ENOMEM because such a
problem is already "loud". I claim it's fixed that dev_err_probe() is
silent for -ENOMEM. If you consider it helpful to add it to the
documentation, add it.

In this case I agree that dropping dev_err_probe() is sane. In the
devm_add_action() case above I tend to keep it.

Best regards
Uwe

Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)