6.4 RC1 locks up solid on first reboot

[Jeremy Chadwick]

On Sat, Oct 25, 2008 at 03:42:18AM +0200, Aragon Gouveia wrote:
> | By Jeremy Chadwick <koitsu at FreeBSD.org>
> |                                          [ 2008-10-25 03:22 +0200 ]
> > > FWIW, I've had problems with 7.0's boot loader locking up too.  I'm running
> > > a -stable compile from march and during bootup, if I make too many
> > > keypresses the bootup sequence freezes.  Once loader has handed off to the
> > > kernel then everything's fine, but I can't make any use of loader's or
> > > bootX's command lines without the system freezing.  I'm careful not to touch
> > > my keyboard during bootup!
> > > 
> > > Keyboard and mouse are both USB.
> > 
> > There are known problems with some BIOSes and "USB Legacy" support.
> > Said BIOS option allows a USB keyboard and mouse to be emulated as PS/2
> > for operating systems which lack a USB stack, such as MS-DOS -- and more
> > importantly, bootloaders!  The FreeBSD bootloader only understands
> > AT/PS2 keyboards, which is why that BIOS option is needed.
> > 
> > Can you confirm this problem happens when using a PS/2 keyboard?
> 
> Unfortunately, no luck.  I've just tried:
> 
> * USB and PS/2 keyboards both plugged in, enough key input on the PS/2
> keyboard causes a freeze
> 
> * Just the PS/2 keyboard plugged in, no different.

Okay, so then the issue is probably with the bootloader in some bizarre
way.  John might have some ideas, but I have none (outside of my skill
set).

Also, it might help if you could explain "enough key input" and "too
many keypresses".  These are a little too vague.

> * Just the PS/2 keyboard plugged in, and USB legacy disabled in the BIOS. 
> Same.

Just for posterity: the USB Legacy Support BIOS option does not affect
natively-connected PS/2 keyboards; you can leave the option enabled even
in the scenario where you have a USB keyboard *and* a PS/2 keyboard
plugged in; one will not "trump" the other.  Instead, you should have
two keyboards which function in OSes/environments which lack a USB
stack.  (That is, until something resets/reassigns the BIOS-controlled
interrupt, which will then break USB->PS/2 emulation; the native PS/2
keyboard should not be affected by this)

> Is it possible that excessive build optimisations could lead to this?  Last
> time I rebuilt I did so with '-O2 -fno-strict-aliasing' and -march in
> make.conf.  I've been meaning to try with a rebuilt loader with no
> optimisations...

Anything is possible.  Can you please rebuild your system, and the
bootstraps (and don't forget to install them; bsdlabel -B <slice>),
without messing with optimisation flags?

> > P.S. -- Why are you slamming keys during the bootup sequence?  :-)
> 
> It's my workstation and I'm usually impatient to get through the bootup
> sequence quickly. :)

I'm still having trouble understanding what you're talking about or
why you're doing this.

I *think* what you're trying to say is something like: "when my machine
boots up, I don't like waiting 10 seconds at the Beastie/loader menu, so
I hit Enter to skip the counterdown".  Please clarify.  :-)

It's interesting that people are having these kinds of odd behaviours so
early in the boot stage.  I've personal experience with FreeBSD acting
wonky with USB keyboards once the OS is loaded (long delays when doing
things like switching vcons), and I solved those problems by disabling
kbdmux(4) either in loader.conf or removing it from the kernel.

Others have had the same success by removing the atkbd/atkbdc drivers
from the kernel (but keeping kbdmux).  kbdmux, obviously, does not
apply to the bootloader issue.

-- 
| Jeremy Chadwick                                jdc at parodius.com |
| Parodius Networking                       http://www.parodius.com/ |
| UNIX Systems Administrator                  Mountain View, CA, USA |
| Making life hard for others since 1977.              PGP: 4BD6C0CB |