From: Doyle, Patrick <WPD@dtccom.com> - 2006-04-13 15:08:54
> -----Original Message-----
> From: Brad Midgley [mailto:bmidgley@...]
> Sent: Thursday, April 13, 2006 10:51 AM
> To: gumstix-users@...
> Subject: Re: [Gumstix-users] ALSA vs. OSS my $.02
> thanks for digging in! it sounds like you're finding a lot of
> noteworthy quirks.
> > But I expect that, going forward, the OSS emulation will be better
> > supported, better tested, etc...
> yes... Craig could fill us in on how "actively" the kernel
> maintainer is trying to kill off the old oss driver. It will
> have some bearing on maintenance.
Yes, I am struggling with which way to go on this one. On the one hand, my
background and training is in getting low level device drivers to talk to
actual hardware. (Hence the references to specific bits and register in my
previous email). Thus, the path of least resistance for me, believe it or
not, is to modify the ALSA driver -- because I can see exactly what software
is modifying what bits where and when.
To bring the OSS driver up to date requires an understanding and
appreciation of the history and evolution of the 2.6 kernel. Somewhere
along the line, __MUTEX_INITIALIZER(x) was removed from the kernel. It
seems likely that it should be replaced with __SEMAPHORE_INIT(x, 1), but
without knowledge of the history of the change, I can't be sure. In a
similar vein, io_remap_page_range() has been changed to
io_remap_pfn_range(), with a change to (at least) one of the arguments.
There are folks out there who are much more cognizant of the evolution of
the 2.6 kernel. Perhaps they would look at those sorts of things and say,
"Those are easy". I look at them and say, "Why did they change? What are
the implications of how it used to be done? What are the implications of
doing it the new way?", etc...
And yet, in my heart of hearts, I think that the simpler, less memory
intensive, OSS driver is a better fit for typical Gumstix applications.