On Mon, Oct 11, 2010 at 10:47:18AM +0100, Mark Brown wrote:
On Sun, Oct 10, 2010 at 01:54:10PM +0300, Mika Westerberg wrote:
- mutex_lock(&info->lock);
- INIT_COMPLETION(info->done);
Do you really need to do this on every single register I/O?
Sorry, didn't quite understand you. Do you mean that should I take the mutex and init the completion or wait for an interrupt?
- ep93xx_ac97_write_reg(info, AC97S1DATA, reg);
- ep93xx_ac97_write_reg(info, AC97IM, AC97_SLOT2RXVALID);
- if (!wait_for_completion_timeout(&info->done, AC97_TIMEOUT)) {
dev_warn(info->dev, "timeout reading register %x\n", reg);mutex_unlock(&info->lock);return -1;Return a real error code.
Ok, will do.
+module_init(ep93xx_ac97_init); +module_exit(ep93xx_ac97_exit);
Put these next to the functions they're referencing.
Ok.
+#ifndef _EP93XX_SND_SOC_AC97_H +#define _EP93XX_SND_SOC_AC97_H
+extern struct snd_soc_dai ep93xx_ac97_dai;
+#endif /* _EP93XX_SND_SOC_AC97_H */
This is not needed with current ASoC. Are you sure you've tested with current code? You should always submit against the development version of the subsystem you're working with for any new drivers - for ASoC Jassi already pointed you at the tree, and you can in the general case also look at -next.
Yeah, sorry about that. I had the impression that using the latest mainline would be right thing to do :( I will base the next version against that tree pointed by Jassi.
Thanks for the comments, MW