[ipxe-devel] [1c838002] memtest86+.bin vs. memtest86.bin?

Ivan Shmakov oneingray at gmail.com
Fri Aug 17 21:44:41 BST 2012

>>>>> Michael Brown writes:
>>>>> On Friday 17 Aug 2012 11:25:05 Ivan Shmakov wrote:

 >>> The advantage of using memtest.0 is that you can exit back to iPXE
 >>> without rebooting,

 >> This didn't work for me — hitting ESC in memtest86+.0 (f9d3679f3770,
 >> running under QEMU/KVM) results in:

 >> Could not boot: Error 0x00000002 (http://ipxe.org/00000002)

 > That's expected behaviour.  Exit codes from memtest.0 are:

 >   0 : success
 >   1 : at least one failure was observed
 >   2 : a complete test was not performed

	Indeed, a silly mistake on my part:

chain /diskless/memtest86+.0
boot || exit 1

	While it should've read something like:

chain /diskless/memtest86+.0
# deal with the exit code here somehow?
goto menu


 >> While Memtest86+/PXE binary is a worthwhile addition, I'm still
 >> interested in what exactly are the iPXE criteria the “stock”
 >> Memtest86+ binary fails to meet?  I'd like to try iPXE to bootstrap
 >> other systems (such as FreeBSD), and it'd probably save me some time
 >> if I'd know where to look should it fail.

 > The motivation for creating memtest.0 was that the other memtest
 > formats often don't work, for reasons that can't easily be fixed.
 > For example, the default memtest.bin is an old-format zImage, which
 > requires part of its code to be loaded at 0x90000.  On many systems,
 > this area of memory will already be occupied when loading via PXE.

	ACK, thanks for the explanation.

	… At the closer view, I don't seem to see any build options for
	Memtest86+ other than the .bin and ELF formats (or what was
	meant by “default memtest.bin” above?) and neither seems to boot
	under iPXE (though the error codes are different — the one for
	the ELF image is 0x2e018001, which corresponds to ENOEXEC in the
	code, AIUI.)

	Also, it seems that the “plain” Memtest86 uses the same format
	as Memtest86+, but is perfectly bootable under iPXE.

	Apparently, I miss some fine point here…

FSF associate member #7257	http://sf-day.org/

More information about the ipxe-devel mailing list