On 11/07/2012 01:05 PM, Alexandre Oliva wrote: > On Nov 7, 2012, Alexey Smirnoff <[email protected]> wrote: > > > Do you have the radeon module built into the kernel or as a loadable > module? > > > Aah! Good to know it's not something I messed up! > Anyway, let me see if I can help you out. > > > I tend to agree ;-) > > > Do you happen to know what hotplug script is in effect at the time of > the request? > > E.g., if this module is built into the kernel, probably none whatsoever > (in which case it would make sense for the kernel to just abort the > request right away); if it's loaded early by initrd's init, then you > gotta make sure the init script configures a hotplug script for the > kernel to use, and that the hotplug script will let the kernel know if > it can't find the requested file (looks like it doesn't). > > Even if the hotplug script is configured to wait for some filesystem to > be mounted or somesuch, you might want to adjust it to recognize > “/*(DEBLOBBED)*/” as the blob name and fail fast. Depending on how it's > implemented, even creating a directory named '*(DEBLOBBED)*/' in > initrd's /lib/firmware might suffice. > > > I can't really fix this in GNU Linux-libre, because the whole point of > keeping the /*(DEBLOBBED)*/ request in place is to signal userland that > some (non-Free) firmware was requested by a kernel module. I guess I > could shorten the timeout in reject_firmware, but the timeout is not > exposed on a per-request basis, so this would require some significant > improvements to the deblobbing machinery and changes to the firmware > loading machinery that I'd rather not get into, for I try to keep > changes WRT upstream to a minimum. > > > I hope this helps, >
For those of us non-kernel-programmer types, who just want the really long wait to go away: does someone have a quick and dirty patch we could apply to our own kernel source? I use libre kernel through Gentoo's "deblob" USE flag. Gentoo allows local editing of the source before compiling through dropped in patches (the epatch_user functionality). And so I was thinking I might be able to leverage that if I had some kind of patch to work with. -- frigidcode.com
signature.asc
Description: OpenPGP digital signature
_______________________________________________ linux-libre mailing list [email protected] http://www.fsfla.org/cgi-bin/mailman/listinfo/linux-libre
