On 06/27/2014 02:39 PM, Tomi Valkeinen wrote:
On 26/06/14 22:20, Jyri Sarha wrote:
...
+#if defined(CONFIG_OMAP4_DSS_HDMI_AUDIO)
- /* drvdata is only used by HDMI audio */
- hdmi.audio.hdmi_lock = &hdmi.lock;
- hdmi.audio.wp = &hdmi.wp;
- hdmi.audio.core = &hdmi.core;
- hdmi.audio.cfg = &hdmi.cfg;
- hdmi.audio.hdmi_dai_drv = &omap_hdmi_dai;
- hdmi.audio.audio_start = hdmi4_audio_start;
- hdmi.audio.audio_stop = hdmi4_audio_stop;
- hdmi.audio.audio_config = hdmi4_audio_config;
- dev_set_drvdata(&pdev->dev, &hdmi.audio);
- r = hdmi_audio_register(pdev);
- if (r) {
DSSERR("Registering HDMI audio failed\n");hdmi_uninit_output(pdev);pm_runtime_disable(&pdev->dev);return r;- }
+#endif
I don't think it's a good idea to reserve the driver data for audio. We need to change the hdmi drivers sooner or later to support multiple instances, and at that point having drvdata for audio will cause problems.
If you don't want to make the hdmi struct public yet, maybe just add a function to this file which takes a pointer to the hdmi struct and returns the audio part. That way you can set the whole hdmi struct as driver data, which should be future safe.
Adding a function is not as simple as that because the OMAP4 and OMAP5 structs are subtly different and hdmi_audio.c is common to both. I'll make the minimal changes to make the struct public.
Cheers, Jyri