None of the 2.5.48- bk or mm kernels lets me mount root,
while 2.5.47 works fine. It doesn't matter wether it is
a smp machine with scsi raid-1 or a up with plain ide.
Possibly some devfs change that went into 2.5.48?
I have been using devfs for a long time now. Most
people don't, and they seems able to run 2.5.48.
Helge Hafting
On Fri, 2002-11-22 at 18:22, Helge Hafting wrote:
> None of the 2.5.48- bk or mm kernels lets me mount root,
> while 2.5.47 works fine.
You have to use the long devfs name, e.g. rather than /dev/hda1 use
/dev/ide/host0/bus0/target0/lun0/part1.
I don't know whether this is change is intentional or not or which
changeset is responsible.
On Fri, Nov 22, 2002 at 08:42:37PM +0100, Luca Barbieri wrote:
> On Fri, 2002-11-22 at 18:22, Helge Hafting wrote:
> > None of the 2.5.48- bk or mm kernels lets me mount root,
> > while 2.5.47 works fine.
>
> You have to use the long devfs name, e.g. rather than /dev/hda1 use
> /dev/ide/host0/bus0/target0/lun0/part1.
>
What do you mean? My lilo.conf obviously uses the "long"
names, because the short ones doesn't exist. It has
been like that since I started using devfs long ago.
The kernel don't deal in names when it tries to mount root though,
it tries to mount device 03:04 or 09:00
That works with 2.5.47 and all earlier 2.5 kernels, but not 2.5.48.
Helge Hafting
"Milton D. Miller II" wrote:
>
> Maybe this patch has some merit?
>
> http://www.ussg.iu.edu/hypermail/linux/kernel/0211.2/0495.html
> The patch perports to fix the case where getdirents on a directory takes
> more than 512 bytes for the result.
Maybe, but the patch claims to fix initrd problems, and I'm
not using an initrd yet. I'll mount root directly as
long as the kernel supports it.
Thanks anyway, it might be worth a try.
Helge Hafting