On Fri, Jan 29, 2010 at 1:40 PM, Mauro Carvalho Chehab <mchehab@xxxxxxxxxx> wrote: > I doubt it would solve. IMO, having it modular is good, since you may not > need cx18 alsa on all devices. Modularity is good, but we really need to rethink about the way we are loading these modules (this applies to dvb as well). For example, on em28xx, the dvb module is often getting loaded while at the same that hald is connecting to the v4l2 device (resulting in i2c errors while attempting to talk to tvp5150). A simple initialization lock would seem like a good idea, except that doesn't really work because the em28xx submodules get loaded asynchronously. And the problem isn't specific to em28xx by any means. I've hit comparable bugs in cx88. If we didn't load the modules asynchronously, then at least we would be able to hold the lock throughout the entire device initialization (ensuring that nobody can connect to the v4l2 device while the dvb and alsa drivers are initializing). Sure, it in theory adds a second or two to the module load (depending on the device), but we would have a much simpler model that would be less prone to race conditions. We would also lose the ability to modprobe the dvb module after-the-fact (and expect it to bind to existing devices), but I don't really think that would be a big deal since everything is auto-detected anyway. In fact, it might actually be a good thing given the number of times I've had to explain to people that you cannot do "modprobe em28xx-dvb" on an unsupported device and expect it to magically start working. I didn't mean to hijack the thread, but I'm just trying to point out that this is a pretty widespread problem, and not specific to cx18. Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
