memtest doesn't support current platforms/processors

[Daniel J Blueman]

memtest-1.70 in the current 8.04LTS Hardy Heron doesn't support and/or recognise quite a few chipsets and processors which have been around for some time (1-2 years?), such as the Intel 965.

As we rely on this tool to establish if we can trust the platform or not (especially if a kernel crash looks dubious), it is in our interest to use the most mature version:

http://www.memtest.org/#change

I have found the SPD display in v2.0 to be particularly useful to identify which serial # DIMM is faulty, so it is a boon. The impact of updating this is near-zero, as it doesn't run on a live system, and it's a tiny package.

[Daniel J Blueman (danielblueman)]

The updated memtest86+ 2.01 is in Debian lenny, so just needs uploading:

http://packages.debian.org/lenny/memtest86+

Is it possible to consider SRU approval for this and upload to hardy-proposed?

[Dereck Wonnacott (dereck)]

Thank you for your suggestion. However, the changes you are requesting aren't really a bug and require more discussion, which should be done on an appropriate mailing list or forum. [WWW] http://www.ubuntu.com/support/community/mailinglists might be a good start for determining which mailing list to use.

If you wish to try a go:
https://wiki.ubuntu.com/StableReleaseUpdates?action=show&redirect=SRU

[Daniel J Blueman (danielblueman)]

I should have mentioned also that the current HH memtest-1.70 has issues with X38 chipsets and mis-initialises them, which has been shown to give false positives in certain configurations; that's a fairly drastic bug.

Let me know if you'd like more information on this.

[Daniel J Blueman (danielblueman)]

TESTCASE:
 1. use Intel X38/X48/S5400 chipset system or eg 12MB L2$ Intel 45nm processors
 2. boot info memtest from GRUB menu
 3. observe incorrect chipset identification
 4. if really unlucky with X38 chipset and particular DIMMs, observe false positive memory errors

[Matt Zimmerman (mdz)]

In general, we are very cautious about bringing in new code from upstream into a stable release, particularly an LTS. However, as you rightly point out, memtest86+ is something of a special case, as it can't affect the normal operation of the system, only its own operation. It still requires care, but may be a candidate for an exception here. The release team will make the decision about whether it would be appropriate to update it.

However, even if they approve, someone would need to package the new version and prepare an update before anything further can happen. Probably the best thing to do would be to try to get this updated in Intrepid (possibly via Debian) first, then consider backporting (some changes from) it.

[Daniel J Blueman (danielblueman)]

Thanks for your feedback Matt. The updated memtest (ie v2.01) package is already in intrepid:

http://packages.ubuntu.com/intrepid/memtest86+

This is then gated by the release team's decision I guess. Thanks again!

[Brian Murray (brian-murray)]

I approved the nomination for Hardy and Intrepid. I also have marked Intrepid task as Fix Released since Intrepid has a current version of memtest.

[Rolf Leggewie (r0lf)]

Hardy has seen the end of its life and is no longer receiving any updates. Marking the Hardy task for this ticket as "Won't Fix".