Bug 11661

Still happens with memtest86+-5.01-3.mga4 :(

CC: (none) => pterjan

Just a question: Does this happens with a UEFI boot?
Because memtest is bootable, it must be allowed to boot...

CC: (none) => jarillon

(In reply to Pierre Jarillon from comment #2)
> Just a question: Does this happens with a UEFI boot?
> Because memtest is bootable, it must be allowed to boot...

I don't have a UEFI system to test, however memtest boots OK, the regular memtest screen appears with a couple of function key options displayed:

F1 (Select failsafe mode) causes instant re-boot.

F2 (force multi-threading) changes to the same screen as when no key is pressed, prior to re-boot.

The initial screen is displayed for maybe 3 seconds before the re-boot if no key is pressed.

I have managed to get images of both screens by extracting frames from a video - these are attached below.

Created attachment 4611 [details]
Initial screen on booting into memtest

Created attachment 4612 [details]
Screen just before it re-boots

This sounds like something machine specific, like an instruction not implemented on your cpu, or some call the BIOS doesn't like, etc
What is your processor?

Created attachment 4614 [details]
cpu info

Does you bug already reported on: http://forum.canardpc.com/forums/73-Memtest86-Official-forum (http://forum.canardpc.com/threads/83126-5.01-freezes for example)?

Reverting to 4.2 is not a good idea for me as 4.2 does not support all new Intel and AMD CPUs as Sandy Bridge, Ivy Bridge, Haswell, Kabini, Bulldozer, Trinity, Bobcat, Pineview, Cedar Trail.

So better not having memtest86+ rather than having 4.2 in 2013/2014 I think.

WIP.

So are you implying that 5.01 works for you?
There seem to be many reports of issues with it and the general advice I see seems to be to use stable 4.2 when it fails.
I don't see any Mageia bug reports related to 4.2, so if it ain't broke...

It is not machine specific. During M4B2 prerelease testing several users saw it.

Easy to reproduce: qemu-kvm -kernel memtest.bin -no-reboot -no-shutdown

(also fails without kvm but takes more time to boot)

Binary built by upstream works fine.

Building with an old gcc also works fine.

Crash seems to be:
                ticks = (ch + ch/act_cpus + c*ch);

act_cpus is an extern from smp.c

I tried changing smp.c so that it can not be 0 and it no longer crashes but sees 0 active cores.

Fixed in memtest86+-5.01-5.mga4

Closing.

Resolution: (none) => FIXED
Status: ASSIGNED => RESOLVED

Thanks a lot for working on it

CC: (none) => ennael1