Comment # 25 on bug 1020283 from Thiago Macieira

(In reply to Thiago Macieira from comment #24)
> (In reply to Andrei Borzenkov from comment #23)
> > If it fails to mount root, add rd.shell kernel parameter, you should be
> > dropped into dracut shell. In which you can mount USB stick and copy dmesg
> > there. You may need to regenerate initrd to include usb-storage and stick
> > filesystem kernel modules and to pass log_buf_len=16M (or larger) to ensure
> > kernel printk buffer is large enough.
> 
> Thanks, I will try. I don't think I will need USB storage, as I can save to
> /boot which is ext4.
> 
> It will take a few days. Hopefully by Friday.

BTW, don't get your hopes up. I don't think I'll get anything useful. Possible
scenarios:

* with rd.shell, it not ask for the password, in which case the problem will
not have (yet) happened and there will be nothing useful to be logged and
inspected
* with rd.shell, it will ask for the password like "single" does, where the
problem is bypassed
* the problem will happen and, like right now, leave the keyboard inoperative
and therefore I won't be able to even run dmesg (when it happens, CapsLock
doesn't turn the LED on; I have to do Alt+SysRq,B to reboot)

Are you sure that rd.shell will drop me to a shell *after* it tries to mount
the root filesystem?