Rene Herman napisał(a):
Is this now done in generic code? Not necessary anymore? Was no
comment
in the changelog.
The MODE_RUNNING is not used at all in the cs4231 code. I wonder what
the purpose of it.
It was used by the AD1848 code though; snd_ad1848_trigger() set/reset
it
on start/stop and it was then tested by snd_ad1848_interrupt() to
decide
whether or not to call snd_pcm_period_elapsed().
It is not used by the cs4231. The only difference is that ad1848 does
not
call the snd_pcm_period_elapsed() after calling the snd_ad1848_open() but before calling the snd_ad1848_trigger() (and similar restriction after the snd_ad1848_trigger() and snd_ad1848_close().
The cs4231 does not use such restriction. I decided it does not really
matter.
The interrupts are not enabled before and after the
snd_ad1848_trigger().
If the cs4231 driver needed this it would be already causing problems.
It seems we are talking at cross purposes. I'm not talking about cs4231. I see this MODE_RUNNING thing disappear from ad1848_lib and, unless I've missed it, not resurface in wss_lib -- that library that after these patches is used to drive AD1848 chips that used to be driven by ad1848_lib.
The MODE_RUNNING looks as something someone will have once added upon seeing spurious IRQs and as such, testing that it isn't needed on chip/card A, B and C doesn't tell us much. The problem may have been observed on type D, E or F and/or under condition foo or bar.
Looking at snd_wss_interrupt() I dom't seem to be seeing a similar guard after these patches.
If it's not needed, all for slashing it but please be convincing. Note that by now you should know a lot more about the innards of this code than I do, so please be as verbose as needed.
I cannot be convincing as I am not convinced myself. I understand your reasoning. My situation was that I saw the guard in the ad1848 which was missing in the cs4231 while they should look like twins (in the area when MODE_RUNNING existed). I had two choices:
1. Add MODE_RUNNING to the cs4231 2. Remove MODE_RUNNING
The first approach is a bloat to me as I tested on ad1848 and it was not needed. I don't know the conditions you called the "condition foo or bar" and I don't know if it ever exists. I have checked git history on the file, but this one was added earlier. Maybe there was a kind of approach in the old days then dropped/replaced later (like mutex_open which is not needed now). The second approach is more dangerous but with enough testing should be justified and provides us with cleaner code.
Regards, Krzysztof
---------------------------------------------------------------------- Te newsy nakreca Cie na caly dzien! Sprawdz >>> http://link.interia.pl/f1e94