Re: sound <-> kerneld
Itai Nahshon (nahshon@actcom.co.il)
Thu, 05 Feb 1998 03:34:21 +0200
Marc Lehmann wrote:
>
> nobody else seems to have mentioned it, but there is a conceptual problem
> with the modularized sound driver, as well as with some other modules.
>
> using kerneld for the sound driver (or ftape for example), just doesn't
> work:
>
> 1. access the sound driver.. in my case "sb" is loaded, which depends
> on sound, which is loaded and registers the device.
> 2. one minute after this '(if not using the device), kerneld tries
> to unload modules automatically, i.e. sb in this case, since
> it's unused (sound itself, though, is used)
> 3. now everybody (non-root) has to wait for another minute until
> sound itself get's unloaded, no sound operation is possible.
>
> This could be fixed when "sound" automatically loads another module
> (ftape automatically tries this), but I think there simply is a design
> problem, somehow kerneld (or the kernel) should know that 3 modules
> really are ONE, otherwise, these modularizing approaches get messy.
It is noticed and you are not the first person to write about it.
The last patch from Thoman Sailer did explicit "request_module"
so that sound.o could load the "end" drivers (namely sb.o and opl3.o
in your case) when they are needed.
Itai
--
Itai Nahshon nahshon@actcom.co.il
Also nahshon@vnet.ibm.com