On 12/03/2013 10:52 AM, Michal Privoznik wrote:
https://bugzilla.redhat.com/show_bug.cgi?id=1035955
There's a window when starting a qemu process between fork() and exec()
during which we are doing things that may fail but not tunnelling the
error to the daemon. This is basically all within qemuProcessHook().
Or more precisely, where things fail so early that the child decided not
to exec() a qemu process in the first place.
So whenever we fail in something, e.g. placing a process onto numa
node,
users ale left with:
s/ale/are/
error: Child quit during startup handshake: Input/output error
while the original error is thrown into the domain log:
libvirt: error : internal error: NUMA memory tuning in 'preferred'
mode only supports single node
Hence, we should read the log file and search for the error message and
report it to users.
An alternative might be to use a pipe back to the parent, marked
close-on-exec, where we dump any untranslated error messages (the parent
can then read until EOF, and if it gets anything on the pipe, then we
know the child raised an error). Doing so is probably a good idea
anyways, as printing into the log file in the child process is dangerous
(if the child uses _("...") to translate an error message, then it is
calling a function that is not async-signal-safe, and could deadlock).
But it's more invasive, and is not written, so your approach is find to
use in the meantime.
/*
+ * Read domain log and probably overwrite error if there's one in
+ * the domain log file. This functions exists to cover the small
s/functions/function/
+ * window between fork() and exec() during which child may fail
+ * by libvirt hand, e.g. placing onto a NUMA node failed.
s/libvirt/libvirt's/
+ if (off > 0) {
+ /* Found an error in the log. Report it */
+ virResetLastError();
+ virReportError(VIR_ERR_INTERNAL_ERROR,
+ _("Process exited prior exec: %s"),
s/prior/prior to/
--
Eric Blake eblake redhat com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org