Bug#646987: /usr/bin/xinit: xinit fails with xkbcomp could not be invoked
The thing that returns ENOMEM in fork.c when this failure occurs is:
dup_mm
I still have to see why.
To try distinguishing between the ENOMEM returns, I tried adding
a flag to the return value (eg:
retval = -(ENOMEM | 0x8000)
retval = -(ENOMEM | 0x4000)
retval = -(ENOMEM | 0x2000)
etc
but that just caused the X server to disappear. At least it was a
way to distinguish whether that return was being used, and I used
that to tell which return ENOMEM was the problem, but a better method
would be handy (without filling the logs too much reporting on
every process rather than just the Xorg process).
Anyway, I will look further at this to see why it's failing -
suggestions on directions to try and methods to use are welcome.
Also, I noticed during the several reboots for these experiments
that when /etc/init.d/xend runs (under xen hypervisor), a bunch
of errors show up on the console - and that's probably when this
shows up in kern.log:
Nov 24 08:52:08 blueeyes kernel: [ 30.913604] XENBUS: Unable to read cpu state
Nov 24 08:52:08 blueeyes kernel: [ 30.913774] XENBUS: Unable to read cpu state
Nov 24 08:52:08 blueeyes kernel: [ 30.913904] XENBUS: Unable to read cpu state
Nov 24 08:52:08 blueeyes kernel: [ 30.914017] XENBUS: Unable to read cpu state
Nov 24 08:52:08 blueeyes kernel: [ 30.914129] XENBUS: Unable to read cpu state
Nov 24 08:52:08 blueeyes kernel: [ 30.914409] XENBUS: Unable to read cpu state
Nov 24 08:52:08 blueeyes kernel: [ 30.914521] XENBUS: Unable to read cpu state
Nov 24 08:52:08 blueeyes kernel: [ 30.914631] XENBUS: Unable to read cpu state
This is before the system tries to run Xorg, as far as I can tell, because
the console is still visible and spewing bootup text.
Then later in kern.log, this appears (when booting under xen, not without xen):
Nov 24 08:52:21 blueeyes kernel: [ 43.881792] Xorg:2940 map pfn expected mapping type write-back for d0000000-e0000000, got write-combining
Nov 24 08:52:21 blueeyes kernel: [ 43.889937] BUG: unable to handle kernel paging request at fffffffffffff254
I am attaching a full kern.log for two boots. The first one is under xen,
the second is not under xen. Same kernel and Xorg binaries.
Maybe I will attach that in a second email. It's kind of big and might not
get through.
bjb
Reply to: