Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Laurent Pinchart <laurent.pinchart@ideasonboard.com>
To:     Nicholas Mc Guire <hofrat@osadl.org>
Cc:     Tomi Valkeinen <tomi.valkeinen@ti.com>,
        David Airlie <airlied@linux.ie>,
        Sebastian Reichel <sre@kernel.org>,
        Peter Ujfalusi <peter.ujfalusi@ti.com>,
        "Andrew F. Davis" <afd@ti.com>, dri-devel@lists.freedesktop.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] drm/omap: dss: do not allow devm_kasprintf() to fail
Date:   Fri, 23 Nov 2018 16:39:59 +0200
Message-ID: <3550363.0Jm6vJdJjr@avalon>
Organization: Ideas on Board Oy
In-Reply-To: <1542974495-12387-1-git-send-email-hofrat@osadl.org>
References: <1542974495-12387-1-git-send-email-hofrat@osadl.org>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="us-ascii"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hello Nicholas,

On Friday, 23 November 2018 14:01:35 EET Nicholas Mc Guire wrote:
> omapdss_display_init() is called by multiple drivers and does not expect
> a return value so without changing all call-sites the low-probability
> failure of devm_kasprintf() can not be reported up the call stack. As
> the amount allocated here is very small (<= 16 bytes) and it is an
> initialization function that most likely will be called during system
> initialization it should be OK to use __GFP_NOFAIL here to prevent
> devm_kasprintf() from returning NULL.
> 
> Signed-off-by: Nicholas Mc Guire <hofrat@osadl.org>
> Fixes: 36c61ae2b755 ("drm/omap: dss: Remove display ordering from
> dss/display.c")
> ---
> 
> Problem located with experimental coccinelle script
> 
> While the use of __GFP_NOFAIL is to be limited (small infrequent
> allocations) this case does seems suitable as it is rare and small
> (initialization) .As all the current drivers using
> omapdss_display_init() currently seem not to initialize dssdev->name
> prior to calling omapdss_display_init() and the unlikely failure
> case can not be reasonably responded (returns void) not allowing
> a allocation failure here should be acceptable.

Given that my plan is to eventually drop omapdss_display_init(), this looks 
fine to me.

Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>

> Patch was compile tested with: omap2plus_defconfig (implies OMAP_DSS_BASE=m)
> 
> Patch is against 4.20-rc3 (localversion-next is next-20181123)
> 
>  drivers/gpu/drm/omapdrm/dss/display.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/omapdrm/dss/display.c
> b/drivers/gpu/drm/omapdrm/dss/display.c index 34b2a4e..7dbe874 100644
> --- a/drivers/gpu/drm/omapdrm/dss/display.c
> +++ b/drivers/gpu/drm/omapdrm/dss/display.c
> @@ -45,7 +45,8 @@ void omapdss_display_init(struct omap_dss_device *dssdev)
>  	of_property_read_string(dssdev->dev->of_node, "label", &dssdev->name);
> 
>  	if (dssdev->name == NULL)
> -		dssdev->name = devm_kasprintf(dssdev->dev, GFP_KERNEL,
> +		dssdev->name = devm_kasprintf(dssdev->dev,
> +					      GFP_KERNEL | __GFP_NOFAIL,
>  					      "display%u", id);
>  }
>  EXPORT_SYMBOL_GPL(omapdss_display_init);

-- 
Regards,

Laurent Pinchart