On Sat, 11 Feb 2017 00:06:32 +0100 Andreas Beckmann <[email protected]> wrote: > Control: tag -1 - moreinfo > > On Fri, 10 Oct 2014 09:14:17 +0200 Andreas Beckmann <[email protected]> wrote: > > The only point I could imagine to "block" is when the script tries to > > read /proc/driver/nvidia/version - which should only exist if the driver > > got loaded properly. > > Today I ran into this problem, too. :-) > > Problem is that the 340xx driver (didn't check 304xx) does not > unregister the proc entries if it aborts loading (due to nouveau being > loaded already). The error unwinding is a mess! 375.26 is correct in > that respect. So the script tries to access stale proc entries resulting in > BUG: unable to handle kernel paging request at ... > I have tested a small patch that adds a call to nv_unregister_proc(), > which was sufficient to fix my problem. It is probably not correct for > all unwinding paths ... > > > Andreas
Nice catch! Never came across this issue myself. Should we try to get this fix in Stretch if no issues are found in sid for a while? Kind regards, Luca Boccassi
