On 03/02/2016 01:41 PM, Dan Carpenter wrote:
> On Wed, Mar 02, 2016 at 09:50:31AM -0700, Shuah Khan wrote:
>> + mctl = kzalloc(sizeof(*mctl), GFP_KERNEL);
>> + if (!mctl)
>> + return -ENOMEM;
>> +
>> + mctl->media_dev = mdev;
>> + if (stream == SNDRV_PCM_STREAM_PLAYBACK) {
>> + intf_type = MEDIA_INTF_T_ALSA_PCM_PLAYBACK;
>> + mctl->media_entity.function = MEDIA_ENT_F_AUDIO_PLAYBACK;
>> + mctl->media_pad.flags = MEDIA_PAD_FL_SOURCE;
>> + mixer_pad = 1;
>> + } else {
>> + intf_type = MEDIA_INTF_T_ALSA_PCM_CAPTURE;
>> + mctl->media_entity.function = MEDIA_ENT_F_AUDIO_CAPTURE;
>> + mctl->media_pad.flags = MEDIA_PAD_FL_SINK;
>> + mixer_pad = 2;
>> + }
>> + mctl->media_entity.name = pcm->name;
>> + media_entity_pads_init(&mctl->media_entity, 1, &mctl->media_pad);
>> + ret = media_device_register_entity(mctl->media_dev,
>> + &mctl->media_entity);
>> + if (ret)
>> + goto err1;
>
> Could we give this label a meaningful name instead of a number?
> goto free_mctl;
I do see other places where numbered labels are used.
Names might help with code readability.
register_entity_fail probably makes more sense as a
label than free_mctl. In any case, I can address the
labels in a follow-on patch.
>
>> +
>> + mctl->intf_devnode = media_devnode_create(mdev, intf_type, 0,
>> + MAJOR(pcm_dev->devt),
>> + MINOR(pcm_dev->devt));
>> + if (!mctl->intf_devnode) {
>> + ret = -ENOMEM;
>> + goto err2;
>
> goto unregister_device;
>
>> + }
>> + mctl->intf_link = media_create_intf_link(&mctl->media_entity,
>> + &mctl->intf_devnode->intf,
>> + MEDIA_LNK_FL_ENABLED);
>> + if (!mctl->intf_link) {
>> + ret = -ENOMEM;
>> + goto err3;
>
> goto delete_devnode;
>
>> + }
>> +
>> + /* create link between mixer and audio */
>> + media_device_for_each_entity(entity, mdev) {
>> + switch (entity->function) {
>> + case MEDIA_ENT_F_AUDIO_MIXER:
>> + ret = media_create_pad_link(entity, mixer_pad,
>> + &mctl->media_entity, 0,
>> + MEDIA_LNK_FL_ENABLED);
>> + if (ret)
>> + goto err4;
>
> This is a bit weird because we're inside a loop. Shouldn't we call
> media_entity_remove_links() or something if this is the second time
> through the loop?
Links are removed from media_device_unregister_entity()
which is called in the error path.
>
> I don't understand this. The kernel has the media_entity_cleanup() stub
> function which is supposed to do this but it hasn't been implemented
> yet?
>
Please see above. Links are removed when entity is
unregistered. media_entity_cleanup() is a stub. It
isn't intended for removing links.
thanks,
-- Shuah
--
Shuah Khan
Sr. Linux Kernel Developer
Open Source Innovation Group
Samsung Research America (Silicon Valley)
shua...@osg.samsung.com | (970) 217-8978
