xf86OpenConsole: VT_WAITACTIVE failed: Interrupted system call

[Keith Packard]

On Sun, 2007-10-07 at 18:16 -0700, Linus Torvalds wrote:

> HOWEVER: I'm starting to believe that there's some other race condition 
> that is getting triggered in that path. One of my machines seems to be 
> able to trigger this almost reliably (not every time, but I've been 
> rebooting it, and I get this about 50% of the time). And the behaviour is 
> interesting: it seems to be waiting for the VT_WAITACTIVATE for something 
> like ten seconds, and but for some reason the VT switch doesn't happen 
> when the problem occurs. But if I force it manually (by switching to VT7 
> from the keyboard) X startup continues ok.

Forcing the switch will probably send a signal to the X server, causing
the VT_WAITACTIVE to return EINTR. I fear the 'fix' I suggested will now
cause that to hang and for the X server to continue sitting there. Or
maybe it will recover when you switch back this time. I don't know.

> Anyway, the problem doesn't actually happen on the first X startup: it 
> happens when the "Boot-time" X exits (with all the Fedora status stuff), 
> and the "real" X starts up to kdm. 

And you wonder why we want to get rid of the extra X server startup. I
suspect having one X server stop and another start will exercise paths
in the kernel which haven't seen a lot of use in the past, especially as
one VT_WAITACTIVE will now really have to wait for the other X server to
clean up.

While we do frequently flip between running X servers, starting an X
server while another is in the process of shutting down is fairly
uncommon.

-- 
keith.packard at intel.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.x.org/archives/xorg/attachments/20071007/02239e15/attachment.pgp>