10 Jan
2011
10 Jan
'11
7:03 p.m.
Mark Brown wrote:
This is all code that's evolved over time, and originally there was no readback facility at all (everything came from cache) so there were no errors to report in the first place. Besides, in general there's nothing constructive we can do about errors anyway except log them.
Well, I'm not sure I agree. For one thing, the current code doesn't just fail to report I/O errors to the callers, it re-interprets an error code as valid date and plows right on ahead.
In addition, there are places where getting a return code would be useful. I call snd_soc_update_bits() in my probe function, so that's a great opportunity to tell ASoC that this device is broken and it shouldn't instantiate a sound card.
--
Timur Tabi
Linux kernel developer at Freescale