[PATCH] drm/virtio: populate .set_busid callback

Mon Oct 3 17:08:02 UTC 2016

On 10/03/16 18:43, Emil Velikov wrote:
> Earlier commit was removing all the users of drm_platform_set_busid and
> removed the virtio hunk (which uses the PCI version of the function) by
> mistake.
> 
> This in itself resulted in user space receiving an incorrect value for
> the busid, which in the case below lead to the failure to detect
> the (correct) device and ultimately the X server failing to start.
> 
> Fixes: a325725633c ("drm: Lobotomize set_busid nonsense for !pci
> drivers")
> Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
> Cc: Laszlo Ersek <lersek at redhat.com>
> Cc: stable at vger.kernel.org
> Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1366842
> Reported-by: Laszlo Ersek <lersek at redhat.com>
> Signed-off-by: Emil Velikov <emil.l.velikov at gmail.com>
> ---
> Since I'm not 100% sure if we can get into .set_busid() as pci_dev is
> NULL (for the virtio-vga 'vs' virgio-gpu-pci case) I've left the local
> wrapper.
> 
> Note: patch is against mainline (refs/tags/v4.8) but should apply for
> drm-next/others.
> ---
>  drivers/gpu/drm/virtio/virtgpu_drm_bus.c | 10 ++++++++++
>  drivers/gpu/drm/virtio/virtgpu_drv.c     |  1 +
>  drivers/gpu/drm/virtio/virtgpu_drv.h     |  1 +
>  3 files changed, 12 insertions(+)
> 
> diff --git a/drivers/gpu/drm/virtio/virtgpu_drm_bus.c b/drivers/gpu/drm/virtio/virtgpu_drm_bus.c
> index 7f0e93f87..88a3916 100644
> --- a/drivers/gpu/drm/virtio/virtgpu_drm_bus.c
> +++ b/drivers/gpu/drm/virtio/virtgpu_drm_bus.c
> @@ -27,6 +27,16 @@
>  
>  #include "virtgpu_drv.h"
>  
> +int drm_virtio_set_busid(struct drm_device *dev, struct drm_master *master)
> +{
> +	struct pci_dev *pdev = dev->pdev;
> +
> +	if (pdev) {
> +		return drm_pci_set_busid(dev, master);
> +	}
> +	return 0;
> +}
> +
>  static void virtio_pci_kick_out_firmware_fb(struct pci_dev *pci_dev)
>  {
>  	struct apertures_struct *ap;
> diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.c b/drivers/gpu/drm/virtio/virtgpu_drv.c
> index c13f70c..5820b702 100644
> --- a/drivers/gpu/drm/virtio/virtgpu_drv.c
> +++ b/drivers/gpu/drm/virtio/virtgpu_drv.c
> @@ -117,6 +117,7 @@ static const struct file_operations virtio_gpu_driver_fops = {
>  
>  static struct drm_driver driver = {
>  	.driver_features = DRIVER_MODESET | DRIVER_GEM | DRIVER_PRIME | DRIVER_RENDER | DRIVER_ATOMIC,
> +	.set_busid = drm_virtio_set_busid,
>  	.load = virtio_gpu_driver_load,
>  	.unload = virtio_gpu_driver_unload,
>  	.open = virtio_gpu_driver_open,
> diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.h b/drivers/gpu/drm/virtio/virtgpu_drv.h
> index b18ef31..acf556a 100644
> --- a/drivers/gpu/drm/virtio/virtgpu_drv.h
> +++ b/drivers/gpu/drm/virtio/virtgpu_drv.h
> @@ -49,6 +49,7 @@
>  #define DRIVER_PATCHLEVEL 1
>  
>  /* virtgpu_drm_bus.c */
> +int drm_virtio_set_busid(struct drm_device *dev, struct drm_master *master);
>  int drm_virtio_init(struct drm_driver *driver, struct virtio_device *vdev);
>  
>  struct virtio_gpu_object {
> 

I've come up with the identical patch (code-wise). I haven't posted it yet because:
- I'd like to reproduce the original issue with the fresh 4.8 kernel, using the F24 config (so the build is taking forever :( :( :(), 
- I'd like to test the patch too.

Obviously, testing your patch is just as good as testing my (identical) patch, it's just that I'll have to respond with a Tested-by.

However, my commit message is different; what do you think of it?

--------
commit a005c99587ce52b60cfae897071f124cc5867347
Author: Laszlo Ersek <lersek at redhat.com>
Date:   Mon Oct 3 17:59:14 2016 +0200

    drm: virtio: reinstate drm_virtio_set_busid()

    Before commit a325725633c2 ("drm: Lobotomize set_busid nonsense for !pci
    drivers"), several DRM drivers for platform devices used to expose an
    explicit "drm_driver.set_busid" callback, invariably backed by
    drm_platform_set_busid().

    Commit a325725633c2 removed drm_platform_set_busid(), along with the
    referring .set_busid field initializations. This was justified because
    interchangeable functionality had been implemented in drm_dev_alloc() /
    drm_dev_init(), which DRM_IOCTL_SET_VERSION would rely on going forward.

    However, commit a325725633c2 also removed drm_virtio_set_busid(), for
    which the same consolidation was not appropriate: this .set_busid callback
    had been implemented with drm_pci_set_busid(), and not
    drm_platform_set_busid(). The error regressed Xorg/xserver on QEMU's
    "virtio-vga" card; the drmGetBusid() function from libdrm would no longer
    return stable PCI identifiers like "pci:0000:00:02.0", but rather unstable
    platform ones like "virtio0".

    Reinstate drm_virtio_set_busid() with judicious use of

      git checkout -p a325725633c2^ -- drivers/gpu/drm/virtio

    Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
    Cc: Daniel Vetter <daniel.vetter at intel.com>
    Cc: David Airlie <airlied at redhat.com>
    Cc: Emil Velikov <emil.l.velikov at gmail.com>
    Cc: Gerd Hoffmann <kraxel at redhat.com>
    Cc: Gustavo Padovan <gustavo.padovan at collabora.co.uk>
    Cc: Hans de Goede <hdegoede at redhat.com>
    Cc: Joachim Frieben <jfrieben at hotmail.com>
    Cc: stable at vger.kernel.org
    Reported-by: Joachim Frieben <jfrieben at hotmail.com>
    Fixes: a325725633c26aa66ab940f762a6b0778edf76c0
    Ref: https://bugzilla.redhat.com/show_bug.cgi?id=1366842
    Signed-off-by: Laszlo Ersek <lersek at redhat.com>

 drivers/gpu/drm/virtio/virtgpu_drm_bus.c | 10 ++++++++++
 drivers/gpu/drm/virtio/virtgpu_drv.c     |  1 +
 drivers/gpu/drm/virtio/virtgpu_drv.h     |  1 +
 3 files changed, 12 insertions(+)
--------

I obviously don't "insist" that you take my commit message :), but you might want to scavenge it for some bits. You might not as well. :)

Regarding the local wrapper around drm_pci_set_busid(): that's not because of virtio-vga vs. virtio-gpu-pci. Instead, it's due to the virtio-mmio vs. virtio-pci transports that are possible for virtio devices. Theoretically at least, a virtio-gpu device could be exposed by QEMU as a memory mapped device in arm/arm64 guests. In that case the "platform bus ID" would simply be the base address of the virtio-mmio register block.

In practice, virtio-gpu over virtio-mmio would be an extremely outlandish configuration (noone does / should do that, plain and simple). Which is why implementing actual "platform bus ID" logic for the virtio-mmio transport case is unnecessary. It's just that the kernel shouldn't crash even if someone tries virtio-gpu over virtio-mmio.

So, the wrapper function (with the pdev nullity check) should be preserved, even if not entirely for the reason you suspected.

Anyway, my laptop seems to have ceased emitting sounds resembling a gas turbine, which I'll take as the completion of the 4.8 kernel build. Will be back soon with test results...

Thanks!
Laszlo