Re: [ALSA] ISA drivers bailing on first IS_ERR

[Rene Herman <rene.herman@keyaccess.nl>]

Takashi Iwai wrote:

> These looks OK to me.  Could you regenerate patches against the latest
> git (or ALSA CVS) ?
> 
> Or, it might be better against to mm tree, since pnp registrations
> will be modified there, too.  They should go also to mainstream
> together.

Will do.

> Nevertheless, the patches (this and the previous one) are good to go
> to stable tree, too.

I'll wait a bit for a comment from Greg on the error propagation thing 
and will submit both then.

Even when simply returning the error as the patch to bus_add_device() 
did, there's a problem in that driver_probe_device() specifically 
ignores -ENODEV and -ENXIO. I suppose that's for hotpluggable stuff, 
where do you do want the driver loaded even without devices...

I guess for -stable the minimal fix would be to make sure all the probe 
methods do not return -ENODEV. I'll audit them for that. For mainline, 
it might be a better idea to have an option in the platform_driver 
struct that, no, we certainly don't want to ignore -ENODEV for this 
ancient non-hotplug non-pnp ISA stuff.

Needs a comment from Greg as well.

Rene.