diff mbox series

[FFmpeg-devel] lavu/hwcontext_vaapi: ignore nonexistent device in default DRM device selection

Message ID 20231124061736.170131-1-haihao.xiang@intel.com
State New
Headers show
Series [FFmpeg-devel] lavu/hwcontext_vaapi: ignore nonexistent device in default DRM device selection | expand

Checks

Context Check Description
andriy/make_x86 success Make finished
andriy/make_fate_x86 success Make fate finished

Commit Message

Xiang, Haihao Nov. 24, 2023, 6:17 a.m. UTC
From: Haihao Xiang <haihao.xiang@intel.com>

It is possible that renderD128 doesn't exist but renderD129 is
available in a system (see [1]). This change can make sure the default
DRM device selection works even if renderD128 doesn't exist.

[1] https://github.com/intel/intel-device-plugins-for-kubernetes/blob/main/cmd/gpu_plugin/README.md#issues-with-media-workloads-on-multi-gpu-setups

Signed-off-by: Haihao Xiang <haihao.xiang@intel.com>
---
 libavutil/hwcontext_vaapi.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

Comments

Anton Khirnov Nov. 26, 2023, 10:55 a.m. UTC | #1
Quoting Xiang, Haihao (2023-11-24 07:17:36)
> From: Haihao Xiang <haihao.xiang@intel.com>
> 
> It is possible that renderD128 doesn't exist but renderD129 is
> available in a system (see [1]). This change can make sure the default
> DRM device selection works even if renderD128 doesn't exist.
> 
> [1] https://github.com/intel/intel-device-plugins-for-kubernetes/blob/main/cmd/gpu_plugin/README.md#issues-with-media-workloads-on-multi-gpu-setups
> 
> Signed-off-by: Haihao Xiang <haihao.xiang@intel.com>
> ---
>  libavutil/hwcontext_vaapi.c | 14 ++++++++++++--
>  1 file changed, 12 insertions(+), 2 deletions(-)
> 
> diff --git a/libavutil/hwcontext_vaapi.c b/libavutil/hwcontext_vaapi.c
> index 12bc95119a..6d542a9bc9 100644
> --- a/libavutil/hwcontext_vaapi.c
> +++ b/libavutil/hwcontext_vaapi.c
> @@ -1733,8 +1733,18 @@ static int vaapi_device_create(AVHWDeviceContext *ctx, const char *device,
>                           "/dev/dri/renderD%d", 128 + n);
>                  priv->drm_fd = open(path, O_RDWR);
>                  if (priv->drm_fd < 0) {
> -                    av_log(ctx, AV_LOG_VERBOSE, "Cannot open "
> -                           "DRM render node for device %d.\n", n);
> +                    if (errno == ENOENT) {
> +                        if (n != max_devices - 1) {
> +                            av_log(ctx, AV_LOG_VERBOSE,
> +                                   "No render device %s, try next device for "
> +                                   "DRM render node.\n", path);
> +                            continue;
> +                        } else
> +                            av_log(ctx, AV_LOG_VERBOSE, "No avaialbe render device "

Typo in 'available'.

Also, the else is unnecessary since the if() block ends with continue.
Xiang, Haihao Nov. 27, 2023, 4:42 a.m. UTC | #2
On So, 2023-11-26 at 11:55 +0100, Anton Khirnov wrote:
> Quoting Xiang, Haihao (2023-11-24 07:17:36)
> > From: Haihao Xiang <haihao.xiang@intel.com>
> > 
> > It is possible that renderD128 doesn't exist but renderD129 is
> > available in a system (see [1]). This change can make sure the default
> > DRM device selection works even if renderD128 doesn't exist.
> > 
> > [1]
> > https://github.com/intel/intel-device-plugins-for-kubernetes/blob/main/cmd/gpu_plugin/README.md#issues-with-media-workloads-on-multi-gpu-setups
> > 
> > Signed-off-by: Haihao Xiang <haihao.xiang@intel.com>
> > ---
> >  libavutil/hwcontext_vaapi.c | 14 ++++++++++++--
> >  1 file changed, 12 insertions(+), 2 deletions(-)
> > 
> > diff --git a/libavutil/hwcontext_vaapi.c b/libavutil/hwcontext_vaapi.c
> > index 12bc95119a..6d542a9bc9 100644
> > --- a/libavutil/hwcontext_vaapi.c
> > +++ b/libavutil/hwcontext_vaapi.c
> > @@ -1733,8 +1733,18 @@ static int vaapi_device_create(AVHWDeviceContext
> > *ctx, const char *device,
> >                           "/dev/dri/renderD%d", 128 + n);
> >                  priv->drm_fd = open(path, O_RDWR);
> >                  if (priv->drm_fd < 0) {
> > -                    av_log(ctx, AV_LOG_VERBOSE, "Cannot open "
> > -                           "DRM render node for device %d.\n", n);
> > +                    if (errno == ENOENT) {
> > +                        if (n != max_devices - 1) {
> > +                            av_log(ctx, AV_LOG_VERBOSE,
> > +                                   "No render device %s, try next device
> > for "
> > +                                   "DRM render node.\n", path);
> > +                            continue;
> > +                        } else
> > +                            av_log(ctx, AV_LOG_VERBOSE, "No avaialbe render
> > device "
> 
> Typo in 'available'.

Thanks for catching this typo

> 
> Also, the else is unnecessary since the if() block ends with continue.
> 

I wanted to print a message if all devices don't exist. I'll remove it in the
new version if you think such info is unnecessary. 

BRs
Haihao
Anton Khirnov Nov. 27, 2023, 8:37 a.m. UTC | #3
Quoting Xiang, Haihao (2023-11-27 05:42:20)
> > 
> > Also, the else is unnecessary since the if() block ends with continue.
> > 
> 
> I wanted to print a message if all devices don't exist. I'll remove it in the
> new version if you think such info is unnecessary. 

I don't object to the message, I mean just that

if (foo) {
  ....
  continue;
} else
  bar;

is equivalent to

if (foo) {
  ....
  continue;
}

bar;

The latter saves you one level of indentation and I find it more
readable.
Xiang, Haihao Nov. 28, 2023, 2:45 a.m. UTC | #4
> Quoting Xiang, Haihao (2023-11-27 05:42:20)
> > > 
> > > Also, the else is unnecessary since the if() block ends with continue.
> > > 
> > 
> > I wanted to print a message if all devices don't exist. I'll remove it in
> > the
> > new version if you think such info is unnecessary. 
> 
> I don't object to the message, I mean just that
> 
> if (foo) {
>   ....
>   continue;
> } else
>   bar;
> 
> is equivalent to
> 
> if (foo) {
>   ....
>   continue;
> }
> 
> bar;
> 
> The latter saves you one level of indentation and I find it more
> readable.

I got your point, many thanks for your explanations. 

BRs
Haihao

>
diff mbox series

Patch

diff --git a/libavutil/hwcontext_vaapi.c b/libavutil/hwcontext_vaapi.c
index 12bc95119a..6d542a9bc9 100644
--- a/libavutil/hwcontext_vaapi.c
+++ b/libavutil/hwcontext_vaapi.c
@@ -1733,8 +1733,18 @@  static int vaapi_device_create(AVHWDeviceContext *ctx, const char *device,
                          "/dev/dri/renderD%d", 128 + n);
                 priv->drm_fd = open(path, O_RDWR);
                 if (priv->drm_fd < 0) {
-                    av_log(ctx, AV_LOG_VERBOSE, "Cannot open "
-                           "DRM render node for device %d.\n", n);
+                    if (errno == ENOENT) {
+                        if (n != max_devices - 1) {
+                            av_log(ctx, AV_LOG_VERBOSE,
+                                   "No render device %s, try next device for "
+                                   "DRM render node.\n", path);
+                            continue;
+                        } else
+                            av_log(ctx, AV_LOG_VERBOSE, "No avaialbe render device "
+                                   "for DRM render node.\n");
+                    } else
+                        av_log(ctx, AV_LOG_VERBOSE, "Cannot open "
+                               "DRM render node for device %d.\n", n);
                     break;
                 }
 #if CONFIG_LIBDRM