Re: [PATCH v3] drm/connector: hdmi: Fix writing Dynamic Range Mastering infoframes

[Derek Foreman <derek.foreman@xxxxxxxxxxxxx>]

On 2024-08-27 05:19, Jani Nikula wrote:
> On Tue, 27 Aug 2024, Maxime Ripard <mripard@xxxxxxxxxx> wrote:
> > Hi,
> >
> > On Mon, Aug 26, 2024 at 07:10:11AM GMT, Derek Foreman wrote:
> > > The largest infoframe we create is the DRM (Dynamic Range Mastering)
> > > infoframe which is 26 bytes + a 4 byte header, for a total of 30
> > > bytes.
> > >
> > > With HDMI_MAX_INFOFRAME_SIZE set to 29 bytes, as it is now, we
> > > allocate too little space to pack a DRM infoframe in
> > > write_device_infoframe(), leading to an ENOSPC return from
> > > hdmi_infoframe_pack(), and never calling the connector's
> > > write_infoframe() vfunc.
> > >
> > > Instead of having HDMI_MAX_INFOFRAME_SIZE defined in two places,
> > > replace HDMI_MAX_INFOFRAME_SIZE with HDMI_INFOFRAME_SIZE(MAX) and make
> > > MAX the same size as the DRM infoframe.
> > >
> > > Signed-off-by: Derek Foreman <derek.foreman@xxxxxxxxxxxxx>
> > > ---
> > >   drivers/gpu/drm/display/drm_hdmi_state_helper.c | 4 +---
> > >   drivers/gpu/drm/drm_debugfs.c                   | 4 +---
> > >   include/linux/hdmi.h                            | 3 +++
> > >   3 files changed, 5 insertions(+), 6 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/display/drm_hdmi_state_helper.c b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
> > > index 7854820089ec..feb7a3a75981 100644
> > > --- a/drivers/gpu/drm/display/drm_hdmi_state_helper.c
> > > +++ b/drivers/gpu/drm/display/drm_hdmi_state_helper.c
> > > @@ -521,8 +521,6 @@ int drm_atomic_helper_connector_hdmi_check(struct drm_connector *connector,
> > >   }
> > >   EXPORT_SYMBOL(drm_atomic_helper_connector_hdmi_check);
> > >   
> > > -#define HDMI_MAX_INFOFRAME_SIZE		29
> > > -
> > >   static int clear_device_infoframe(struct drm_connector *connector,
> > >   				  enum hdmi_infoframe_type type)
> > >   {
> > > @@ -563,7 +561,7 @@ static int write_device_infoframe(struct drm_connector *connector,
> > >   {
> > >   	const struct drm_connector_hdmi_funcs *funcs = connector->hdmi.funcs;
> > >   	struct drm_device *dev = connector->dev;
> > > -	u8 buffer[HDMI_MAX_INFOFRAME_SIZE];
> > > +	u8 buffer[HDMI_INFOFRAME_SIZE(MAX)];
> > >   	int ret;
> > >   	int len;
> > >   
> > > diff --git a/drivers/gpu/drm/drm_debugfs.c b/drivers/gpu/drm/drm_debugfs.c
> > > index 6b239a24f1df..9d3e6dd68810 100644
> > > --- a/drivers/gpu/drm/drm_debugfs.c
> > > +++ b/drivers/gpu/drm/drm_debugfs.c
> > > @@ -520,8 +520,6 @@ static const struct file_operations drm_connector_fops = {
> > >   	.write = connector_write
> > >   };
> > >   
> > > -#define HDMI_MAX_INFOFRAME_SIZE		29
> > > -
> > >   static ssize_t
> > >   audio_infoframe_read(struct file *filp, char __user *ubuf, size_t count, loff_t *ppos)
> > >   {
> > > @@ -579,7 +577,7 @@ static ssize_t _f##_read_infoframe(struct file *filp, \
> > >   	struct drm_connector *connector; \
> > >   	union hdmi_infoframe *frame; \
> > >   	struct drm_device *dev; \
> > > -	u8 buf[HDMI_MAX_INFOFRAME_SIZE]; \
> > > +	u8 buf[HDMI_INFOFRAME_SIZE(MAX)]; \
> > >   	ssize_t len = 0; \
> > >   	\
> > >   	connector = filp->private_data; \
> > > diff --git a/include/linux/hdmi.h b/include/linux/hdmi.h
> > > index 3bb87bf6bc65..3a442a59919e 100644
> > > --- a/include/linux/hdmi.h
> > > +++ b/include/linux/hdmi.h
> > > @@ -59,6 +59,9 @@ enum hdmi_infoframe_type {
> > >   #define HDMI_DRM_INFOFRAME_SIZE    26
> > >   #define HDMI_VENDOR_INFOFRAME_SIZE  4
> > >   
> > > +/* The biggest infoframe size */
> > > +#define HDMI_MAX_INFOFRAME_SIZE		HDMI_DRM_INFOFRAME_SIZE
> > Thanks for that patch, it was definitely an oversight on my part. The
> > spec defines the max size of an infoframe to be 30 bytes, so we should
> > probably use that here?

Hmm, I think the spec (I'm looking at 1.3a, which I think is the latest 
freely available version from hdmi.org?) says a maximum length of 27 
bytes, not including the 3 byte header + the checksum byte. So I think 31?

I'm seeing this in section 5.3.5 / Table 5-14.

> Defining it like this allows the use of HDMI_INFOFRAME_SIZE(MAX) which
> adds the header size. Having some macros include the header size and
> some others not would be confusing, I think.
>
> Or then the whole thing needs to be renamed.

How about:

#define HDMI_MAX_INFOFRAME_SIZE 27

Then we can use HDMI_DRM_INFOFRAME_SIZE(MAX) to get 31

Would that be ok?

Thanks,
Derek

> BR,
> Jani.
>
> > Maxime