Description of problem: Launching memtest86+ from a grub2 menu in UEFI fails to display the application's screen although the usual tones from the motherboard sounder are heard, so it's maybe running without display. There are no errors in the grub2 terminal window after the launch message. Version-Release number of selected component (if applicable): How reproducible: Steps to Reproduce: 1. 2. 3.
Hmm.. I just found this in 20_memtest86+ script in Fedora's package: if [ -d /sys/firmware/efi ]; then echo "ERROR: memtest86+ does not support EFI platforms." exit 254 fi
Summary: memtest86+ fails to run => memtest86+ fails to run on UEFI system
(In reply to Barry Jackson from comment #1) > Hmm.. > > I just found this in 20_memtest86+ script in Fedora's package: > > if [ -d /sys/firmware/efi ]; then > echo "ERROR: memtest86+ does not support EFI platforms." > exit 254 > fi Assigning to all packagers collectively, since there is no maintainer for this package.
CC: (none) => marja11Assignee: bugsquad => pkg-bugs
OK I think I have a solution for this issue. 1. If we remove the "Recommends: memtest86+" from the grub2-efi sub-package in grub2 spec and move it to the main grub2 package (PC-BIOS only). 2. Add the scriptlet in #2 above to %pre in memtest86+ package then all should be well. 3. Add a note in memtest86+ description to say that it is incompatible with UEFI. Grub2-efi will no longer request memtest86+ so it will not try to install it. If an attempt is made to install memtest86+ on a UEFI system manually then it will fail. e.g. Preparing... ######################################################################################################## ERROR: memtest86+ does not support EFI platforms. error: %prein(memtest86+-5.01-10.1.mga6.x86_64) scriptlet failed, exit status 1 ERROR: 'script' failed for memtest86+-5.01-10.1.mga6.x86_64 error: memtest86+-5.01-10.1.mga6.x86_64: install failed Unfortunately due to https://bugs.mageia.org/show_bug.cgi?id=1670#c10 this second part may not be acceptable? Barry
Although this should not cause a transaction to fail as memtest+ would no longer be recommended in a situation where it would fail, so I think that a silent fail for an attempted manual install should be OK
Hmm.. another issue: drakx-installer-images BuildRequires: memtest86+ on all x86_64 --------snip-------- %ifarch %ix86 x86_64 BuildRequires: memtest86+ BuildRequires: grub BuildRequires: syslinux >= 6.03-5 %endif ---------snip------ I'm out of ideas as I don't know where/when this is used.
(In reply to Barry Jackson from comment #5) > Hmm.. another issue: > > drakx-installer-images BuildRequires: memtest86+ on all x86_64 > --------snip-------- > %ifarch %ix86 x86_64 > BuildRequires: memtest86+ > BuildRequires: grub > BuildRequires: syslinux >= 6.03-5 > %endif > ---------snip------ > > I'm out of ideas as I don't know where/when this is used. CC'ing tv and tmb, they'll know :-D (In reply to Barry Jackson from comment #4) > Although this should not cause a transaction to fail as memtest+ would no > longer be recommended in a situation where it would fail, so I think that a > silent fail for an attempted manual install should be OK Yeah, I had to re-read bug 1670, but going by the information there, silently failing is the best option (at the moment!... I do hope it won't stay like this for ever, though ;-) )
CC: (none) => thierry.vignaud, tmb
(In reply to Marja van Waes from comment #6) > (In reply to Barry Jackson from comment #5) > > Hmm.. another issue: > > > > drakx-installer-images BuildRequires: memtest86+ on all x86_64 > > --------snip-------- > > %ifarch %ix86 x86_64 > > BuildRequires: memtest86+ > > BuildRequires: grub > > BuildRequires: syslinux >= 6.03-5 > > %endif > > ---------snip------ > > > > I'm out of ideas as I don't know where/when this is used. > > CC'ing tv and tmb, they'll know :-D > It's done to provide possibility to run memtest direcly from iso when people have trouble installing and faulty memory is suspected. Note that memtest96+ "works on uefi hw" as long as you boot with csm enabled so it can boot in legacy mode. I'm thinking of packaging the nonfree memtest (http://www.memtest86.com/) as that one provides uefi support (the free edition has no licensing restrictions what so ever)
From a quick test, it no longer works on legacy HW too... Maybe a gcc5 issue...
(In reply to Thierry Vignaud from comment #8) > From a quick test, it no longer works on legacy HW too... > Maybe a gcc5 issue... It's still working here on an old pIII i586 (fully updated Cauldron) with grub2.
I tested with grub legacy, both the regular build and the ELF one, both ends in reboot.
See Also: (none) => https://bugs.mageia.org/show_bug.cgi?id=18261
Summary: memtest86+ fails to run on UEFI system => memtest86+ fails to run on UEFI system (unsupported)
CC: (none) => bequimao.de
I have 2 PC, one with uefi and grub2 the other with legacy grub memtest86+ (installed with urpmi) works fine with grub2 and just reboot with grub. I have also tried the last Mageia-6-sta2-x86_64-DVD on another PC (Grub, no UEFI) it can only reboot the PC.
CC: (none) => jarillon
(In reply to Thierry Vignaud from comment #10) > I tested with grub legacy, both the regular build and the ELF one, both ends > in reboot. Yes, Memtest86+ 5 just reboots with most systems I have tried. The old 4 version still works, you can try it from the MGA3 medias. Anyway, I have found that we have much better now : simply boot the normal system adding "memtest" as kernel parameter : it disables automaticaly bad ram, allowing to boot directly. This is at the expense of boot time, something like 2 minutes for a slow system with a lot of RAM. Maybe we could put a boot line using it instead of memtest86+?
CC: (none) => lists.jjorge
CC: (none) => javier_diaz
memtest86+ can't definitely run in UEFI mode on x86_64. As a result we are setting this bug to RESOLVED:INVALID
Status: NEW => RESOLVEDResolution: (none) => INVALID