Testing old hardware, I used Mga7 64 bit classic install ISO on a USB stick. When i in the start menu select "Memory Test" there comes a message: "error: he kernel is too old for initrd. Press any key to continue" After a few seconds it goes away by itself, and memtest86+ shows up. It waits a few seconds so user can select i.e safe mode. Regardless if i select safe mode or not, the computer reboots after having exercised the memory for about two seconds. Strange: if i pull out one of the two memory modules, whichever, it do not crash. But there is no problem on any of the modules; running memtest86+ from the boot menu of version 6.0.3 of http://www.system-rescue-cd.org memtest go more than one pass when both modules are inserted. Main Board: ASRock P55 Pro; Intel P55 chipset BIOS defaults loaded. CPU: 13-540, RAM: Kit of two Corsair Venegance 4096MB v2.12 1600MHz 9-9-9-24
Setting it major as it is an important test tool that crash and may lead user to think RAM/Mainboard/CPU is bad, with delays and cost as result. It *may* be related to: Bug 18261 - memtest does not boot from legacy grub menu entry (immediate reboot) - but for me the reboot only happens after a couple seconds of RAM testing progress.
Assignee: bugsquad => isobuildTarget Milestone: --- => Mageia 8Severity: normal => major
Summary: memtest86 on mga7 isos reboot on some configurations. => memtest86 on mga7 isos reboot + "kernel is too old for initrd"
See Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=25235
This is two separate bugs, so I have created bug 25235 to track the "kernel is too old for initrd" error. Note that that bug only affects the CI ISOs. The reboot bug occurs on three of my systems, but not on my old Athlon X2 system or in VirtualBox. Building a test ISO where I replaced the /boot/memtest binary with the mga6 version fixes the problem. My guess is that this is a fault introduced by building with the latest GCC, as the memtest86+ source code hasn't changed since 2013.
Assignee: isobuild => bugsquadTarget Milestone: Mageia 8 => ---Summary: memtest86 on mga7 isos reboot + "kernel is too old for initrd" => memtest86+ on mga7 isos reboots almost immediately after starting test (on some hardware)Component: Release (media or process) => RPM PackagesSource RPM: (none) => memtest86+-5.01-12.mga7CC: (none) => mageia
CC: sysadmin-bugs => (none)
It looks like Fedora had the same issue using gcc-8: https://bugzilla.redhat.com/show_bug.cgi?id=1598922 They workarounded it by switching to gcc-3.4 then back to gcc-9
Source RPM: memtest86+-5.01-12.mga7 => memtest86+-5.01-12.mga7 gccCC: (none) => thierry.vignaud, tmb
See Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=18261
(In reply to Thierry Vignaud from comment #3) > It looks like Fedora had the same issue using gcc-8: > https://bugzilla.redhat.com/show_bug.cgi?id=1598922 > They workarounded it by switching to gcc-3.4 then back to gcc-9 Thanks, Thierry. I was going to assign this report to the kernel maintainers, because it is a "System/Kernel and hardware" package, but now assigning to tmb, who maintains gcc and is also our most important kernel maintainer.
Assignee: bugsquad => tmbCC: (none) => marja11
CC: (none) => javier_diaz
Summary: memtest86+ on mga7 isos reboots almost immediately after starting test (on some hardware) => memtest86+ on mga7 isos reboots after starting test (on some hardware) - gcc problem probably
Does the one recompiled with current gcc in cauldron works better?
I will try to remember to test when first mga8 iso appear.
(In reply to Thierry Vignaud from comment #5) > Does the one recompiled with current gcc in cauldron works better? No. On my affected hardware it just hangs at the point where the mga7 version reboots.
As the free version doesn't support UEFI, the only alternative is to boot the kernel with eg: "memtest=4" … Not as fancy, doesn't loop but works everywhere…
As it seems unlikely memtest86+ is ever going to be fixed, and in any case is unusable on modern hardware, is it time to remove it from the ISO boot menus?
Memtest+ was last update in 2013. See https://www.memtest.org/ That is a long time in terms of hardware development.
CC: (none) => rihoward1
I use it rather often, helping to get old computers in shape for friends. Magiea 7 xfce runs perfectly well on 10+ year young stuff :) It also helped to isolate problems in my daughters computer, and it could also detect a rouge CPU in my workstation last year: it got errors on specifically one of the cores. So it is also a partial CPU test. Fresher hardware is mostly needed only for playig games, IMO...
Cant we just keep the compiled version from mga6?
Actually, on the bug from comment 3 i see there are submissions of new versions in beginning of february, with request to test and say if it does not work, and noone have responded such. So if someone have a fedora system and can test if it works where mga7 version fail?
As said in comment 3, the Fedora "fix" is to build with an old version of gcc. I don't think we have that option.
Technically, why do we need to re-compile it at all -- cant we just copy our compiled version from mga6? (or copy that compiled binary from fedora)
That would mean we cannot support it anymore as we couldn't recompile it in case of eg a security issue. The other problem being that it's not usable on UEFI systems (unless you download the non free version) (unless your firmware provides a BIOS compatible mode _and_ you switch to it -- which would break some systems such as Windows)
I've got a fix for this. I've also found a new upstream fork with some other fixes which I'm going to try: https://github.com/anphsw/memtest86
Status: NEW => ASSIGNEDAssignee: tmb => mageia
CC: (none) => rolfpedersen
I ended up forking and rewriting memtest86+. My version is available in cauldron in the pcmemtest package, and is on the CI and Live ISOs currently in QA. It has the advantage that it works with UEFI as well as with legacy BIOS (and hopefully has far fewer bugs). In the meantime Thierry has uploaded a newer version of memtest86+. I don't know if that fixes the bug described here. In case it doesn't and he still wants to maintain this version, I will attach my fix here.
Assignee: mageia => thierry.vignaudStatus: ASSIGNED => NEW
Created attachment 11684 [details] Patch to fix test hangs/reboots when built with GCC 8 (based on v5.01)
Hi, I've never got any memtest to work from the boot menu and tried this (removed what memtest I could see was installed): sudo urpme memtester sudo urpmi http://mirrors.kernel.org/mageia/distrib/cauldron/x86_64/media/core/release/pcmemtest-1.1-2.mga8.x86_64.rpm sudo update-grub Rebooted and chose the Pcmemtest entry but got an error: can't find command linux32 Thanks.
@Rolf, could you open a new bug report for that (against pcmemtest), and show the output of rpm -qa | grep grub2
(In reply to Martin Whitaker from comment #19) > Created attachment 11684 [details] > Patch to fix test hangs/reboots when built with GCC 8 (based on v5.01) It doesn't apply: 21 out of 21 hunks FAILED -- saving rejects to file test.c.rej
Maybe upstream have independently fixed the bug then.
(In reply to Rolf Pedersen from comment #20) > Pcmemtest I'm suffering from bug #18261 Just installed PCMemtest from Cauldron, works like a charm! Thanks Martin :)
CC: (none) => mageia
Just tested memtest86+ from cauldron: still fails (as in bug #18261).
(In reply to Martin Whitaker from comment #21) > @Rolf, could you open a new bug report for that (against pcmemtest), and > show the output of > > rpm -qa | grep grub2 @Martin, I have the very same problem (can't find linux32) on another machine. The command line you asked for: [root@localhost grub2]# rpm -qa|grep grub2 grub2-common-2.02.0-15.mga7 grub2-mageia-theme-2.02.0-15.mga7 grub2-efi-2.02.0-15.mga7 on the machine where pcmemtest works I have exactly the same versions - both are up to date. I think the machine where it works is booting in BIOS mode whereas the other in UEFI mode, but I'm not sure, I)m not an expert in that area.
@Jean-Christophe, see bug 26742 for explanation and solution.
(In reply to Martin Whitaker from comment #9) > As it seems unlikely memtest86+ is ever going to be fixed, and in any case > is unusable on modern hardware, is it time to remove it from the ISO boot > menus? @martin Maybe pcmtest should obsolete memtest86+ [1] and we should adjust the ISO boot menu to run pcmtest instead? [1] until someone tries to update it from one of the forks
@Thierry, I hesitated to obsolete memtest86+ as upstream had unexpectedly come back to life back (shortly after I wrote that comment!). But I guess if it's still broken, we could do that. pcmemtest has already replaced memtest86+ on the CI and Live ISOs, so it's only the netinstall ISOs that still use memtest86+. I can take a look at changing those if you want (I see you took over as maintainer from tmb).
Thanks for having a look
I've updated the netinstall ISOs to use pcmemtest (and added the Memory Test option to the UEFI boot menu). pcmemtest now obsoletes memtest86+ in cauldron.
It seems this bug is fixed? On all bootable media we serve from Mageia 8 there will be a working memory tester? ( pcmemtest )
Closing as fixed. We will not release new Mageia 7 isos, and pcmemtest is working on Mageia 8 ISOs. If any problem with pcmemtest on Mageia 8 ISOs, please file a new bug.
Status: NEW => RESOLVEDResolution: (none) => FIXED