Re: [PATCH] drm/vmwgfx: remove DRM_ERROR message, stops log spamming

From: Daniel Vetter <daniel@ffwll.ch>
To: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>,
	kernel-janitors@vger.kernel.org,
	"Linux-Kernel@Vger. Kernel. Org" <linux-kernel@vger.kernel.org>,
	ML dri-devel <dri-devel@lists.freedesktop.org>,
	VMware Graphics <linux-graphics-maintainer@vmware.com>,
	Colin Ian King <colin.king@canonical.com>
Subject: Re: [PATCH] drm/vmwgfx: remove DRM_ERROR message, stops log spamming
Date: Wed, 20 Sep 2017 18:33:48 +0000	[thread overview]
Message-ID: <20170920183348.GD25248@dvetter-linux.ger.corp.intel.com> (raw)
In-Reply-To: <CACvgo53QktiPssai5hzoMP=Sy4Wtbrti+fOTHE4pGZXCZbUryA@mail.gmail.com>

On Tue, Sep 12, 2017 at 06:54:45PM +0100, Emil Velikov wrote:
> On 12 September 2017 at 18:47, Colin Ian King <colin.king@canonical.com> wrote:
> > On 12/09/17 18:42, Thomas Hellstrom wrote:
> >> Hi, Colin,
> >>
> >> On 09/12/2017 07:35 PM, Colin King wrote:
> >>> From: Colin Ian King <colin.king@canonical.com>
> >>>
> >>> mmap'ing the device multiple times will spam the kernel log with the
> >>> DRM_ERROR message about illegal mmap'ing the old fifo space.
> >> How are you hitting this? Multiple mappings should be fine as long as
> >> mapping offsets are correct,
> >> so hitting this message should indicate that the user-space app is doing
> >> something seriously wrong, and
> >> having it present in the log should probably help more than it hurts.
> >>
> >> /Thomas
> >
> > Good question.  I hit similar issues with the drm qxl driver when
> > running some kernel regression tests with stress-ng [1]. I realize this
> > is an artificial test scenario so it is definitely not a typical
> > use-case, however, sync the illegal mmapping will return -EINVAL the
> > application will pick up that this is an error without the need of
> > spotting it in the kernel log. And a user space application can perform
> > many millions of these invalid mmaps causing kernel log spamming.
> >
> FWIW I'm the one to "blame" here - pointing Colin to drop the message.
> 
> Two reasons come to mind:
>  - there is a unwritten rule that roughly says "user input should not
> cause kernel log spam"
>  - out of all the DRM drivers only QXL and VMWGFX print a message,
> with a patch addressing the former

Maybe we should make this a written rule by patching
Documentation/drivers/gpu? Would definitely make sense as part of this
patch series.

Thanks, Daniel
> 
> HTH
> Emil
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/dri-devel

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch