Continuing the saga: F9alpha AMD-64 HP DC7700 SFF switch_root: no filesystems
John Summerfield
debian at herakles.homelinux.org
Wed Mar 12 17:48:31 UTC 2008
Michal Jaegermann wrote:
> On Wed, Mar 12, 2008 at 01:40:42AM -0400, Will Woods wrote:
>>
>> So, nearly all of the "cannot find /dev/root" problems we've seen have
>> been mkinitrd-related. So davej is correct: it is very likely to be a
>> mkinitrd bug.
>
> It is not entirely clear from what John writes what happens
> with what kernel and when. If you see in dmesg, like a fragment
> which he included recently (not clear from which kernel):
It's perfectly clear to me:-)
>
> ata1.00: ATA-7: ST3320620AS, 3.AAK, max UDMA/133
> ata1.00: 625142448 sectors, multi 16: LBA48 NCQ (depth 0/32)
> ata1.01: ATA-8: WDC WD5000AAKS-00YGA0, 12.01C02, max UDMA/133
> ata1.01: 976773168 sectors, multi 16: LBA48 NCQ (depth 0/32)
>
> and booting fails then most likely initrd is at fault.
That fragment is from the most recent kernel that works, 2.6.24.1-28.fc9
>
> If, OTOH, like in this fragment of dmesg for
That was captured from a failing kernel by use a crossover serial cable,
a second computer and minicom.
> 2.6.25-0.90.rc3.git5.fc9 which John added to bugzilla, you have
>
> ata1: SATA link down (SStatus 221 SControl 300)
>
> then it is hard to even consider initrd as there is no chance
> for it to be ever loaded since disks are "invisible".
>
>> It *might* actually be something in the sata driver..
_I_ think it is. The initrd _does_ load, it's loaded by grub using BIOS
calls.
I eventually got busybox in it and not being erased (another thread) and
had a look around. It contains the same sata modules as the kernel that
works.
So I figure the mkinird is working and making the same decisions as
before, but that something's changed in the driver code.
>
> That second piece points to that although it is possible that
> in this case both show up.
>
> Of course when you see a screen picture with the last few lines
> which say that / is not there, and all other information scrolled
> out a long time ago, then the most natural reaction is
> "uh-hu, that initrd is messed up" just because a kernel managed
> to get that far so in a sense it already "booted".
I posted the photo because, at that time, I couldn't locate the serial
cable. It depicts the same problem, and it does seem to have all the ata
stuff there.
>
> Michal
>
More information about the fedora-test-list
mailing list