>>>>> "IM" == Ilia Mirkin <imirkin at alum.mit.edu> writes:IM> OK, well 18.1.0 is completely busted for those GPU's. Either upgrade IM> or downgrade. But don't use that release. I did [DOWNGRADE] libglapi-mesa:amd64 18.1.0-1 -> 18.0.5-1 [DOWNGRADE] libglx-mesa0:amd64 18.1.0-1 -> 18.0.5-1 and rebooted. Problem remains: $ glxinfo > /dev/null libGL error: failed to create dri screen libGL error: failed to load driver: nouveau
On Tue, Jun 12, 2018 at 9:09 PM, 積丹尼 Dan Jacobson <jidanni at jidanni.org> wrote:>>>>>> "IM" == Ilia Mirkin <imirkin at alum.mit.edu> writes: > > IM> OK, well 18.1.0 is completely busted for those GPU's. Either upgrade > IM> or downgrade. But don't use that release. > > I did > [DOWNGRADE] libglapi-mesa:amd64 18.1.0-1 -> 18.0.5-1 > [DOWNGRADE] libglx-mesa0:amd64 18.1.0-1 -> 18.0.5-1 > and rebooted. Problem remains: > $ glxinfo > /dev/null > libGL error: failed to create dri screen > libGL error: failed to load driver: nouveauInteresting. This will probably require a bunch of back-and-forth to debug. You can use gdb to work out why the load is failing. If it's in dri_fill_in_modes, then something went wrong in your downgrade. Feel free to join #nouveau on irc.freenode.net to discuss. -ilia
By the way, why does it try to access /etc/ld.so.nohwcap 49 times when
failing once should be enough? One getpid() would seem enough too.
$ strace glxinfo -B 2>&1|sort|uniq -c|sort -n|tail -n 3
54 access("/etc/ld.so.nohwcap", F_OK) = -1 ENOENT (No such
file or directory)
62 recvmsg(3, {msg_namelen=0}, 0) = -1 EAGAIN (Resource
temporarily unavailable)
116 getpid() = 1422
I'll send you the full strace.