Re: [Xen-devel] [BUG] xl create -c with pygrub hangs in xen 4.4.0-rc5

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Pasi Kärkkäinen writes ("Re: [Xen-devel] [BUG] xl create -c with pygrub hangs 
in xen 4.4.0-rc5"):
> On Sun, Feb 23, 2014 at 09:36:44PM +0000, M A Young wrote:
> > I was trying xen-4.4.0-rc5 with my standard set up (booting a pv
> > guest using pygrub as a bootloader) but this no longer works. I
> > would expect to get a boot menu where I could select a kernel and
> > the guest, but nothing happens. If I examine the bootloader.1.log
> > file I find the pygrub output I would expect to see on the console,
> > and the output of xenstore-ls suggests pygrub is selecting the
> > default kernel (as it would without input) and exiting, but xentop
> > doesn't report any cpu usage on the guest. It seems xl has created a
> > child process that doesn't exit, though if I kill the child process
> > by hand the boot does continue.
> > 
> > I traced the change in behaviour to the commit 
> > http://xenbits.xenproject.org/gitweb/?p=xen.git;a=commit;h=5f0c4a78100382972b4d2a71a04b90e015e9fe87
> > "libxl: fork: Share SIGCHLD handler amongst ctxs". If I revert this
> > then I get the expected behaviour again, though I haven't worked out
> > why this patch cause the effects I am seeing.
> 
> Uh oh.. I wonder if Ian (author of the libxl changes in question) has some 
> ideas why that might be happening..
> 
> This is kind of a bug that's not very easy to detect even with automated 
> pygrub testing.. 
> interactive testing is pretty much required to notice this bug.

Thanks to Michael for reporting this.  I don't immediately see what's
wrong but will think about it.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel