R: R: [Fedora-ia64-list] mount: could not find filesystem '/dev/root'

[Doug Chapman]

On Tue, 2008-02-26 at 20:09 -0500, Doug Chapman wrote:
> On Tue, 2008-02-26 at 18:54 -0500, Doug Chapman wrote:
> > On Tue, 2008-02-26 at 17:48 +0100, Bettoni Fabio wrote:
> > > Hi Doug,
> > > about
> > > > > I have seen the same thing on my systems.  It has been fixed in the 
> > > > > upstream kernel so it should be fixed the next time Fedora pulls in 
> > > > > the latest kernel. 
> > > if I try to compile vanilla linux-2.6.25-rc3.tar.bz2 by myself, should
> > > it work ?
> > > 
> > > Fabio
> > 
> > Possibly but I don't know exactly when the bug upstream was fixed.  I
> > _think_ it was before rc3 but not sure.
> > 
> > I see that we have a new kernel src.rpm that is building on x86 right
> > now.  I will kick off an ia64 build of that and let you know what I
> > find.
> > 
> > - Doug
> > 
> 
> It turns out we are seeing something new here.  With the latest kernel I
> am seeing what I _think_ is the same problem as you are.  What i am
> seeing it it isn't seeing the initrd.img that is passed in.
> 
> The key bit is during bootup I don't see these messages that would
> normally be printed during bootup:
> 
> checking if image is initramfs... it is
> Freeing initrd memory: 5648kB freed
> 
> I am digging through the code to try to figure out why this is now.
> 
> - Doug
> 

OK, I have learned a few things:

1) my initrd comment above is a non-issue, it turns out the initrd is
getting loaded, it just happens at a slightly different point in the
boot when compared to older kernels.

2) I can reproduce this with the latest git pull so the problem still
exists upstream.

3) i can reproduce this on RHEL5 so that rules out a regression in
mkinitrd/nash/udev etc.

4) If I use an older config file which was based on RHEL5 as the basis
for the kernel build it works OK.  So, I am working on finding what the
difference between the configs is that breaks this.  So far I have ruled
out the SLAB vs. SLUB allocator but many more potential config
differences still to try.

5) I really shouldn't be working this late and need sleep ;)


- Doug