On 12/10/16 15:37, Felix Miata wrote:
> John Pilkington composed on 2016-10-12 11:23 (UTC+0100):
>
>> On 12/10/16 10:50, John Pilkington wrote:
>
>>> My video card has worked fine with recent builds of the f23 nvidia304xx
>>> driver from rpmfusion, but an update two days ago left me with black
>>> screens (two - VGA and HDMI). I have a thread on the rpmfusion list,
>>> but here may now be more appropriate.
>
>>> I have removed the nvidia packages, and nouveau gives me the full
>>> plasma
>>> window with taskbar etc, but after a few seconds that is replaced by
>>> featureless black screens with a mouse arrow that moves - from one
>>> screen to the other - but does nothing else.
>
>>> Ctrl/Alt/F2 lets me view Xorg.0.log - on both screens - and all looks
>>> harmless until, about 80 seconds after the last NOUVEAU message, about
>>> HDMI max frequency, I get
>
>>> AIGLX: Suspending AIGLX clients for VT switch
>>> NOUVEAU(0): NVLeaveVT is called
>
>>> It now looks as if this is just recording the switch to tty2
>
>>> Switching between Ctrl/Alt/F2 and Ctrl/Alt/F1 adds more instances of it
>>> but I still don't get a working GUI
>
>>> Ideas?
>
>> The final two lines of dmesg are
>
>> show_signal_msg: 9 callbacks suppressed
>
>> plasmashell[1901]: segfault at 0 ip 00007fcabe9244a7 sp 00007ffd195bf9c0
>> error 4 in nouveau_dri.so[7feabe49a000+8ce000]
>
>> These occur at 205s. The HDMI max TDMS frequency message in Xorg.0.log
>> was at 165s.
>
>> All messages copied manually from the screens.
>
> What happens if you purge all of nouveau and NVidia drivers, leaving
> Xorg to
> use its integral modesetting driver?
>
> cf.
>
http://www.phoronix.com/scan.php?page=news_item&px=Ubuntu-Debian-Aban...
>
>
Thanks for the suggestion. I wasn't hopeful, having tried removing the
intel driver (with an old i915 chip) in *this* box running SL7, but
tried it:
dmesg | grep segfault gives 7 lines:
sddm[xx]: segfault at 0 ip xxxx sp xxxx error 4 in libQT5Core.so.5.6.1[xxx]
or similar but with error 14 in sddm.
dnf history shows that 58 packages were updated on 10 Oct, but that
would be with yumex-dnf and I can't find where it stores its history for
non-GUI access.
I had previously tried the nvidia build from scrpms, which didn't
install cleanly (rpm -i --force) but also pointed to libQT5Core.so.5.6.1
John
I have to update. There were still nvidia*-libs packages installed.
Now the first 4 segfaults are in sddm, error 14, 15, 15, 14. The ones
after that are in sddm .... libQT5Core.so.5.6.1