2003-05-11 12:57:24

by Gregoire Favre

[permalink] [raw]
Subject: lilo and 2.5.69?

Hello,

Normally I have in my lili.conf:

append = "root=/dev/scsi/host0/bus0/target15/lun0/part2 video=matrox:1600x1200-16@75"

But I got:

Fatal: open /dev/ide/host0/bus0/target0/lun0/par: No such file or
directory
Exit 1

Same if I put:
root=/dev/scsi/host0/bus0/target15/lun0/part2

Or root=/dev/sdb2

Is there something special to do with this new kernel?

Thank you very much,

Gr?goire
________________________________________________________________
http://ulima.unil.ch/greg ICQ:16624071 mailto:[email protected]


2003-05-11 13:43:49

by Christoph Hellwig

[permalink] [raw]
Subject: Re: lilo and 2.5.69?

On Sun, May 11, 2003 at 03:09:45PM +0200, Gregoire Favre wrote:
> Is there something special to do with this new kernel?

You need to fix lilo to not assume the names listet in /proc/partitions are
actual device files. In 2.5.69 there's a bug that it prints truncated devfs
names instead of traditional device names as it should, but relying on the
names to mean anything is broken - that kernel can't enforce the device names
used.

The following patch that is in Linus BK tree should get it working for you
again for now..


--- 1.108/fs/partitions/check.c Tue Apr 29 17:42:50 2003
+++ edited/fs/partitions/check.c Fri May 2 09:41:04 2003
@@ -96,19 +96,6 @@

char *disk_name(struct gendisk *hd, int part, char *buf)
{
-#ifdef CONFIG_DEVFS_FS
- if (hd->devfs_name[0] != '\0') {
- if (part)
- snprintf(buf, BDEVNAME_SIZE, "%s/part%d",
- hd->devfs_name, part);
- else if (hd->minors != 1)
- snprintf(buf, BDEVNAME_SIZE, "%s/disc", hd->devfs_name);
- else
- snprintf(buf, BDEVNAME_SIZE, "%s", hd->devfs_name);
- return buf;
- }
-#endif
-
if (!part)
snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name);
else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1]))

2003-05-11 14:11:45

by Gregoire Favre

[permalink] [raw]
Subject: Re: lilo and 2.5.69?

On Sun, May 11, 2003 at 04:22:24PM +0200, Helge Hafting wrote:

> Looks like a bug that truncates long device names.
> Looks like your'e using devfs, using
> /dev/discs/discX/part2
> is a fine workaround - because it is short enough.
> Replace the X with whatever number your
> host0-target15 disk has.

I have put:

append = "root=/dev/discs/disc2/part2 video=matrox:1600x1200-16@75"

And it produces the same error:

Fatal: open /dev/ide/host0/bus0/target0/lun0/par: No such file or directory 8,1 Top
Exit 1

which was previsible as dev/sdb2 produced the same problem.

Thank you very much,

Gr?goire
________________________________________________________________
http://ulima.unil.ch/greg ICQ:16624071 mailto:[email protected]

2003-05-11 14:06:32

by Helge Hafting

[permalink] [raw]
Subject: Re: lilo and 2.5.69?

On Sun, May 11, 2003 at 03:09:45PM +0200, Gregoire Favre wrote:
> Hello,
>
> Normally I have in my lili.conf:
>
> append = "root=/dev/scsi/host0/bus0/target15/lun0/part2 video=matrox:1600x1200-16@75"
>
> But I got:
>
> Fatal: open /dev/ide/host0/bus0/target0/lun0/par: No such file or
> directory
> Exit 1
>
> Same if I put:
> root=/dev/scsi/host0/bus0/target15/lun0/part2
>
> Or root=/dev/sdb2
>
> Is there something special to do with this new kernel?

Looks like a bug that truncates long device names.
Looks like your'e using devfs, using
/dev/discs/discX/part2
is a fine workaround - because it is short enough.
Replace the X with whatever number your
host0-target15 disk has.

Helge Hafting

2003-05-11 14:20:49

by Gregoire Favre

[permalink] [raw]
Subject: Re: lilo and 2.5.69?

In fa.linux.kernel, you wrote:

> You need to fix lilo to not assume the names listet in /proc/partitions are
> actual device files. In 2.5.69 there's a bug that it prints truncated devfs
> names instead of traditional device names as it should, but relying on the
> names to mean anything is broken - that kernel can't enforce the device names
> used.
>
> The following patch that is in Linus BK tree should get it working for you
> again for now..

Great ;-)

I'll compil with your patch: thank you very much,

Gr?goire
________________________________________________________________
http://ulima.unil.ch/greg ICQ:16624071 mailto:[email protected]