Two times I got an UML crashing when I left it active on my box. The first
time I didn't report it since I wasn't able to get backtraces(i.e. no
debugging info) and it was a patched version.
Now I've left 2.4.22-5um running and idle(with two consoles open): the UML had
been started, then I Ctrl+z it and bg'ed it(and it worked fine). But now
those consoles are open but I can't type anything(as previous time). There
are only 3 threads(the fourth is died: the shell under which I ran it
reported a segmentation fault about it) and I've got with gdb) their
backtraces(they seem to be good: I've used gdb 6.0).
This was the command line:
./linux umid=slack90 ubd0=slack90.rootfs root=/dev/ubd0 ubd7=slack90.swapfs
mem=64m eth0=tuntap,,,192.168.0.1 UML_XDISPLAY=192.168.0.1:0
UML_IP=192.168.0.3 dsp=/dev/dsp mixer=/dev/mixer con=xterm con0=fd:3
They are currently all in S state.
I'm attaching backtraces, the content of uml-slack90 and the startup output.
The host is a 2.4.22-ck2 + skas(ck2 patchset is from Con Kolivas, it's O(1) +
preemptible + low latency backports basically, very stable and no conflicts).
I'm willing to help: I can gdb it and set breakpoints, to get the point of
failure, if you say me how (Or better, set a trap for SIGSEGV probably, if
you say me the exact "handle" syntax for this, since the shell reported a
segfault). What do you need to investigate on this?
Paolo Giarrusso, aka Blaisorblade
Linux Kernel 2.4.21/2.6.0-test on an i686; Linux registered user n. 292729