Bug reported by Robert de Bath on the Debian BTS 8 months ago. His server locks up saying:

  f000:5054: 01 ILLEGAL EXTENDED X86 OPCODE!
  XIO:  fatal IO error 104 (Connection reset by peer) on X server ":0.0"
  after 0 requests (0 known processed) with 0 events remaining.

He's using the vesa driver with xserver 1.3 (it did not with 1.1). The full log is available at http://bugs.debian.org/cgi-bin/bugreport.cgi?msg=20;filename=Xorg.0.log;att=1;bug=404885

The same problem (maybe better reported) is also at http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=433144
and also with the i810 driver in
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=428796
which seems to prove that's the bug in the server (in the int10 code) and not vesa-specific.

Robert looked at the code in hw/xfree86/int10/xf86int10.c, in function
int1A_handler. For AX values 0xb108, 0xb109, 0xb10a, 0xb10b, 0xb10c and 0xb10d, the entire contents of EDI is used as a pci register number, only DI should be used. He fixed the problem (for his specific BIOS) by changing the following in the 0xb10a case:
< X86_ECX = pciReadLong(tag, X86_EDI);
--
> X86_ECX = pciReadLong(tag, X86_DI);

Brice