[ATrpms-users] memtest86+
Paulo Cavalcanti
promac at gmail.com
Sat Apr 21 04:52:06 CEST 2007
On 4/20/07, Phil Schaffner <P.R.Schaffner at ieee.org> wrote:
>
> On Fri, 2007-04-13 at 08:43 -0300, Paulo Cavalcanti wrote:
> > Hi,
> >
> > I prepared some rpms for the latest version of memtest86+, and they
> > are working great for me.
> >
> > The problem I had with redhat's version was
> >
> > Error 28: Selected item cannot fit into memory
> >
> > when trying to boot it via grub (same happened with memtest86-3.3).
> >
> > This problem seems to be related to newer mobos and newer grub
> > versions.
> >
> > After some research, I managed to patch it and fixed memtest-setup for
> > writing
> > the appropriate grub arguments.
> >
> > http://people.atrpms.net/~pcavalcanti/rpms/
> >
> > Suggestions and feedback are welcome.
>
> Rebuilt for Centos/EL5 from your SRPM at
>
>
> http://people.atrpms.net/~pcavalcanti/srpms/memtest86+-1.70-1.fc6.lcg.src.rpm
>
> A nit-picking suggestion (probably for grubby unless I'm missing
> something in the grubby man page) would be to have the entry in
> grub.conf be made at the bottom rather than the top.
Agreed.
A more substantive problem was the GRUB stanza created:
>
> title Memtest86+ v1.70
> kernel /boot--type=netbsd /memtest86+-1.70
>
> Should have been:
>
> kernel --type=netbsd /boot/memtest86+-1.70
>
> >From the spec file, looks like it would work on systems with a /boot
> partition but fails on those without /boot. The relevant section is:
>
> %post
> if [ -x /sbin/grubby ] ; then
> /sbin/grubby \
> --add-kernel="/boot/%{name}-%{version}" \
> --title "Memtest86+ v%{version}"
>
> sed -i -e's,/%{name},--type=netbsd /%{name},' /boot/grub/grub.conf
> fi
>
> %postun
> if [ -x /sbin/grubby ] ; then
> sed -i -e's,--type=netbsd /%{name},/%{name},' /boot/grub/grub.conf
> /sbin/grubby --remove-kernel="/boot/%{name}-%{version}"
> fi
>
> Haven't got a handle on to correct logic to handle both cases.
What complicates things is that "type=netbsd" cannot be at the end of the
line,
as grubby likes to do, according to my tests. That is why I used "sed".
Unfortunately, I do not have a system without a /boot to test.
The original redhat's rpm does not handle grub in the spec file. I include
it
because I think people do not read the comments in the rpm. But it is not
clear
for me if it was the right decision.
Anyway, I filed a bug report today in redhat's bugzilla, hoping
they fix their rpms.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=237279
The problem is that we just remember memtest86 when we have bad RAM
at hand, and it is very frustrating not being able to boot via grub and test
it quickly.
--
Paulo Roma Cavalcanti
LCG - UFRJ
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.atrpms.net/pipermail/atrpms-users/attachments/20070420/040bd970/attachment-0001.html
More information about the atrpms-users
mailing list