Bug#822529: gdm3: retries to often / to fast on errors

[Laurent Bigonville]

On Mon, 25 Apr 2016 08:12:06 +0200 Tobias Frost <tobi at coldtobi.de> wrote:
 >
 > Dear gdm3 maintainers,
 >

Hello Tobias,

 > To debug another issue, I needed to boot an older kernel which has 
not nv compiled,
 > so X could not load that module and could not start.
 >
 > That it not able to start is expected behaviour, but after that 
happens gdm basically DoS the machine:
 > After the failures happens it immediatly retries and that indefintily.
 >
 > This gives a "blinking" effect switching the vts back and forth, 
making it
 > impossible to enter anything into a vt to login and actually fix the 
problem.
 >
 > To get the machine useable again, I sshed into that machine to stop 
gdm manually.
 >
 > Attached is an excerpt (~10 seconds) of /var/log/syslog.
 >
 > As a metrics, in the time I needd to stop the service, was 1minute 
23seconds, it retried 242 times.
 >
 > Please let me know if you need additonal information, let me know.
 >
 > I guess the systemd file needs some tewaking to limit retries.
 >
 > Thanks!
 >

Do you think it's a regression from 3.18? TBH, I think I already saw 
this behavior in the past.

Also, do you think you could bring this upstream?

Cheers,

Laurent Bigonville