Once upon a time, m.roth@5-cent.us m.roth@5-cent.us said:
Any idea if pts's are set up? Also, is this of any help http://wiki.libvirt.org/page/Unable_to_connect_to_console_of_a_running_domain?
That's the "old" way (the way virt-install defaults for RHEL6 and before). The "new and improved" way is a direct virtualized device.
Here's what virt-install would set for the old way:
<serial type='pty'> <target port='0'/> </serial> <console type='pty'> <target type='serial' port='0'/> </console>
And here's what you get when you say RHEL7:
<console type='pty'> <target type='virtio' port='0'/> </console>
That works for setting up a hvc0 device in the guest. It appears that it is implemented through a kernel module, virtio_console. Now, from past experience, the kernel is supposed to be able to handle consoles that only show up when modules are loaded (e.g. USB serial), so that _should_ work. The kernel crash is a bug I guess; the virtio_console module appears to have some "issues" as well (not refcounted properly for example; I can unload the module while I'm logged in through a getty on hvc0, which then goes away).
I need to try a Fedora guest and see what happens (have to wait on that, don't have a Fedora tree at work to install from). Could be a general kernel bug, or could just be RHEL7's kernel.
Still, no matter what, virtio console leaves out the boot loader. Maybe I'll file a bug against Fedora (or upstream) virt-manager that virtio console shouldn't be used by default.