Re: [PATCH v3 2/2] drm: remove drm_device_is_unplugged and related code

[David Herrmann <dh.herrmann@gmail.com>]

Hi

On Fri, Feb 12, 2016 at 8:50 PM, Haixia Shi <hshi@chromium.org> wrote:
> v1: Remove the general drm_device_is_unplugged() checks, and move the unplugged
> flag handling logic into drm/udl. In general we want to keep driver-specific
> logic out of common drm code.
>
> v2: Based on discussion with Stephane and David, drop most of the unplugged
> flag handling logic in drm/udl except for udl_detect() and udl_fb_open().
> The intention is to treat the device removal as a connector-unplug, and kep
> the UDL device fully functional.
>
> v3: Based on feedback from David, entirely drop the "unplugged" flag and all
> related code. There is no need to check for the unplugged flag as the existing
> udl_usb_disconnect() behavior already ensures the controller is removed, and
> all code paths that uses the usb-device are not reachable after unplug.
>
> When a UDL monitor is unplugged, we need to still treat it as a fully
> functional device which just happens to have its connector unplugged.
> This allows user-space to properly deallocate fbs and dumb buffers
> before closing the device.
>
> This drops the "unplugged" flag hack, which puts the device in a
> non-functional state after USB unplug and rejects most operations on
> the device such as ioctls with error -ENODEV.
>
> Signed-off-by: Haixia Shi <hshi@chromium.org>
> Reviewed-by: Stéphane Marchesin <marcheu@chromium.org>
> Cc: David Herrmann <dh.herrmann@gmail.com>
> Cc: Daniel Vetter <daniel@ffwll.ch>
> ---
>  drivers/gpu/drm/drm_drv.c           |  6 ------
>  drivers/gpu/drm/drm_fops.c          |  2 --
>  drivers/gpu/drm/drm_gem.c           |  3 ---
>  drivers/gpu/drm/drm_ioctl.c         |  3 ---
>  drivers/gpu/drm/drm_vm.c            |  3 ---
>  drivers/gpu/drm/udl/udl_connector.c |  2 --
>  drivers/gpu/drm/udl/udl_fb.c        |  6 ------
>  include/drm/drmP.h                  | 14 --------------
>  8 files changed, 39 deletions(-)
>
> diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
> index 167c8d3..f93ee12 100644
> --- a/drivers/gpu/drm/drm_drv.c
> +++ b/drivers/gpu/drm/drm_drv.c
> @@ -376,9 +376,6 @@ struct drm_minor *drm_minor_acquire(unsigned int minor_id)
>
>         if (!minor) {
>                 return ERR_PTR(-ENODEV);
> -       } else if (drm_device_is_unplugged(minor->dev)) {
> -               drm_dev_unref(minor->dev);
> -               return ERR_PTR(-ENODEV);
>         }
>
>         return minor;
> @@ -464,9 +461,6 @@ void drm_unplug_dev(struct drm_device *dev)
>         drm_minor_unregister(dev, DRM_MINOR_CONTROL);
>
>         mutex_lock(&drm_global_mutex);
> -
> -       drm_device_set_unplugged(dev);
> -
>         if (dev->open_count == 0) {
>                 drm_put_dev(dev);
>         }
> diff --git a/drivers/gpu/drm/drm_fops.c b/drivers/gpu/drm/drm_fops.c
> index 1ea8790..b4332d4 100644
> --- a/drivers/gpu/drm/drm_fops.c
> +++ b/drivers/gpu/drm/drm_fops.c
> @@ -497,8 +497,6 @@ int drm_release(struct inode *inode, struct file *filp)
>
>         if (!--dev->open_count) {
>                 retcode = drm_lastclose(dev);
> -               if (drm_device_is_unplugged(dev))
> -                       drm_put_dev(dev);

Who frees the udl device now? What code-path is responsible to call
drm_dev_unregister() for udl devices? This hunk looks bogus to me.
Please elaborate.

>         }
>         mutex_unlock(&drm_global_mutex);
>
> diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
> index 2e8c77e..c622e32 100644
> --- a/drivers/gpu/drm/drm_gem.c
> +++ b/drivers/gpu/drm/drm_gem.c
> @@ -900,9 +900,6 @@ int drm_gem_mmap(struct file *filp, struct vm_area_struct *vma)
>         struct drm_vma_offset_node *node;
>         int ret;
>
> -       if (drm_device_is_unplugged(dev))
> -               return -ENODEV;
> -
>         drm_vma_offset_lock_lookup(dev->vma_offset_manager);
>         node = drm_vma_offset_exact_lookup_locked(dev->vma_offset_manager,
>                                                   vma->vm_pgoff,
> diff --git a/drivers/gpu/drm/drm_ioctl.c b/drivers/gpu/drm/drm_ioctl.c
> index 8ce2a0c..f959074 100644
> --- a/drivers/gpu/drm/drm_ioctl.c
> +++ b/drivers/gpu/drm/drm_ioctl.c
> @@ -695,9 +695,6 @@ long drm_ioctl(struct file *filp,
>
>         dev = file_priv->minor->dev;
>
> -       if (drm_device_is_unplugged(dev))
> -               return -ENODEV;
> -
>         is_driver_ioctl = nr >= DRM_COMMAND_BASE && nr < DRM_COMMAND_END;
>
>         if (is_driver_ioctl) {
> diff --git a/drivers/gpu/drm/drm_vm.c b/drivers/gpu/drm/drm_vm.c
> index f90bd5f..3a68be4 100644
> --- a/drivers/gpu/drm/drm_vm.c
> +++ b/drivers/gpu/drm/drm_vm.c
> @@ -657,9 +657,6 @@ int drm_legacy_mmap(struct file *filp, struct vm_area_struct *vma)
>         struct drm_device *dev = priv->minor->dev;
>         int ret;
>
> -       if (drm_device_is_unplugged(dev))
> -               return -ENODEV;
> -
>         mutex_lock(&dev->struct_mutex);
>         ret = drm_mmap_locked(filp, vma);
>         mutex_unlock(&dev->struct_mutex);
> diff --git a/drivers/gpu/drm/udl/udl_connector.c b/drivers/gpu/drm/udl/udl_connector.c
> index 4709b54..a6d5e21 100644
> --- a/drivers/gpu/drm/udl/udl_connector.c
> +++ b/drivers/gpu/drm/udl/udl_connector.c
> @@ -96,8 +96,6 @@ static int udl_mode_valid(struct drm_connector *connector,
>  static enum drm_connector_status
>  udl_detect(struct drm_connector *connector, bool force)
>  {
> -       if (drm_device_is_unplugged(connector->dev))
> -               return connector_status_disconnected;
>         return connector_status_connected;

Who marks a connector as "disconnect" once the device is removed?

>  }
>
> diff --git a/drivers/gpu/drm/udl/udl_fb.c b/drivers/gpu/drm/udl/udl_fb.c
> index 200419d..29aca6c 100644
> --- a/drivers/gpu/drm/udl/udl_fb.c
> +++ b/drivers/gpu/drm/udl/udl_fb.c
> @@ -321,12 +321,6 @@ static void udl_fb_imageblit(struct fb_info *info, const struct fb_image *image)
>  static int udl_fb_open(struct fb_info *info, int user)
>  {
>         struct udl_fbdev *ufbdev = info->par;
> -       struct drm_device *dev = ufbdev->ufb.base.dev;
> -       struct udl_device *udl = dev->dev_private;
> -
> -       /* If the USB device is gone, we don't accept new opens */
> -       if (drm_device_is_unplugged(udl->ddev))
> -               return -ENODEV;
>
>         ufbdev->fb_count++;
>
> diff --git a/include/drm/drmP.h b/include/drm/drmP.h
> index d7162cf..40c6099 100644
> --- a/include/drm/drmP.h
> +++ b/include/drm/drmP.h
> @@ -748,7 +748,6 @@ struct drm_device {
>         struct drm_minor *control;              /**< Control node */
>         struct drm_minor *primary;              /**< Primary node */
>         struct drm_minor *render;               /**< Render node */
> -       atomic_t unplugged;                     /**< Flag whether dev is dead */
>         struct inode *anon_inode;               /**< inode for private address-space */
>         char *unique;                           /**< unique name of the device */
>         /*@} */
> @@ -879,19 +878,6 @@ static __inline__ int drm_core_check_feature(struct drm_device *dev,
>         return ((dev->driver->driver_features & feature) ? 1 : 0);
>  }
>
> -static inline void drm_device_set_unplugged(struct drm_device *dev)
> -{
> -       smp_wmb();
> -       atomic_set(&dev->unplugged, 1);
> -}
> -
> -static inline int drm_device_is_unplugged(struct drm_device *dev)
> -{
> -       int ret = atomic_read(&dev->unplugged);
> -       smp_rmb();
> -       return ret;
> -}
> -
>  static inline bool drm_is_render_client(const struct drm_file *file_priv)
>  {
>         return file_priv->minor->type == DRM_MINOR_RENDER;

Did you actually test this patch? How is it supposed to work? Please
explain _exactly_ in your commit message what the new semantics are. I
highly doubt that this patch does what you want. Also, please run
kmemleak on your tests to verify that all memory is actually released.

Generally, I like the approach. However, I think you still need a
state that marks a device as dropped. I'd do something like this:

1) In udl_usb_disconnect(), set udl->udev to NULL, locked by
mode_config.mutex. This should be the _first_ thing you do there. This
serves as barrier against ->get_modes(), which is the only path that
uses "udl->udev".

2) In udl_detect(), return 'connector_status_disconnected' if
"udl->udev" is NULL.

3) In udl_get_edid() add "if (WARN_ON(!udl->udev)) return NULL;"

4) Hook into f_ops.release() and call into drm_release(). But
afterwards you have to call drm_put_dev() if open_count is 0 now.

Thanks
David
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel