On Mon, 10 Oct 2011 09:42:44 -0400, AJ (Adam) wrote:
On Sun, 2011-10-09 at 17:26 +0200, Michael Schwendt wrote:
Shortly before the Plymouth LUKS passphrase prompt, the screen turns black,
the monitor reports No signal and enters power-saving mode. The system
hasn't crashed, because instead of pressing Ctrl+Alt+Del I can also enter
the passphrase blindly for the boot procedure to continue. However,
nothing wakes up the monitor afterwards. Switching to a virtual console
causes the signal to return, and what is displayed on the screen is not
GDM but the gray background with the central animation having reached
100% or nearly that. There is no virtual console available, and one cannot
return to GDM either.
Plymouth is just a KMS client like the X server. Normally it tries to
avoid setting a video mode if it can (so boot flickers less) but it will
do in some cases. So this sounds like it's a bug in the kernel driver.
Much of the info from the Xorg debugging guide:
http://fedoraproject.org/wiki/How_to_debug_Xorg_problems
is relevant here, although obviously the X log and etc aren't. In
addition, booting with plymouth:debug on the kernel command line will
dump debugging output to /var/log/plymouth-debug.log, which should show
the mode setup that plymouth is trying for. Boot with that flag set
once, then either boot again without plymouth.debug or rhgb set, or
simply ssh into the machine and copy the log file aside.
Thanks. This sounded promising, but funnily since I've added
plymouth:debug permanently, I haven't managed to reproduce the issue.
Instead, I've encountered a few different issues. Once only:
1) The GDM background picture was divided into a few big rectangular areas
which were shifted by 1/5 screen width. Looked like an unfinished slider
puzzle. ;)
2) A completely corrupted GDM background picture. Red/white/black noise,
but a working login.
3) The boot not continueing after the LUKS passphrase prompt.
The saved plymouth-debug.log files don't contain any obviously dangerous
errors. Just lots of ply-event-loop.c debug msgs including (always) many
[ply-event-loop.c] ply_event_loop_remove_source_node:failed to
delete fd 11 from epoll watch list: Bad file descriptor
...
[ply-boot-server.c]ply_boot_connection_on_request:could not
finish writing update reply: Broken pipe
Now, this morning I've taken out the option again to find out whether the
primary issue would return. Perhaps it is influenced by interrupting the
GRUB timeout with key presses (although I think I haven't done that with
F-15 often). Or my earlier GRUB chainloading. Recently I've stored GRUB 2
in MBR. It's a tough case, but any idea might help in collecting data *if*
it will happen again or if a test-case is found.
--
Fedora release 16 (Verne) - Linux 3.1.0-0.rc9.git0.0.fc16.x86_64
loadavg: 0.13 0.06 0.05
--
test mailing list
test@lists.fedoraproject.org
To unsubscribe:
https://admin.fedoraproject.org/mailman/listinfo/test
