2.5.67-ac2 and lilo

2.5.67-ac2 and lilo

[Christian Staudenmayer]

Hello,

i'm running a machine that uses the aic7xxx driver for the old adaptec 2940 scsi
controller. i had problems with getting 2.5.67-ac2 to run, it used to end in
a kernel panic. i got told to remove the body of the function ide_xlate_1024
by just "return 0;", which fixed the problem, i could then boot the kernel.
but now, when running lilo, i get the following message:

Fatal: First boot sector doesn't have a valid LILO signature

these problems do not occur on kernels 2.4.20, 2.4.21-pre7-ac1 and 2.5.67-bk9,
if i boot another kernel, i can run/install LILO without problems.

i'd appreciate any insight on this problem.

Greetings, Christian Staudenmayer

__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com

Re: 2.5.67-ac2 and lilo

[Andries Brouwer]

On Sat, Apr 19, 2003 at 10:07:25AM -0700, Christian Staudenmayer wrote:

> i'm running a machine that uses the aic7xxx driver for the old adaptec 2940 scsi
> controller. i had problems with getting 2.5.67-ac2 to run, it used to end in
> a kernel panic. i got told to remove the body of the function ide_xlate_1024
> by just "return 0;", which fixed the problem, i could then boot the kernel.
> but now, when running lilo, i get the following message:
> 
> Fatal: First boot sector doesn't have a valid LILO signature
> 
> these problems do not occur on kernels 2.4.20, 2.4.21-pre7-ac1 and 2.5.67-bk9,
> if i boot another kernel, i can run/install LILO without problems.
> 
> i'd appreciate any insight on this problem.

Hmm. You did not forget to rerun LILO or so?
If ide_xlate_1024 does nothing but "return 0" then
handle_ide_mess() does nothing.

What other differences does -ac2 have?
It reintroduces a bug in the partition reading code - the fragment

@@ -456,7 +556,7 @@
                if (!subtypes[n].parse)
                        continue;
                subtypes[n].parse(state, bdev, START_SECT(p)*sector_size,
-                                               NR_SECTS(p)*sector_size, slot);
+                                               NR_SECTS(p)*sector_size, n);
        }
        put_dev_sector(sect);

of the -ac2 patch is bad.
In order to judge whether it is this or something else I would have
to know much more about your system. What are the kernel boot messages
for this disk under 2.5.67-bk9 that works and 2.5.67-ac2 that fails?
What is the partition table? No disk managers in sight? Does LILO
have lba32 or linear options?
Does 2.5.67-ac2 work if you replace its fs/partitions/msdos.c by
the vanilla one?

Andries

Re: 2.5.67-ac2 and lilo

[Helge Hafting]

On Sat, Apr 19, 2003 at 10:07:25AM -0700, Christian Staudenmayer wrote:
> Hello,
> 
> i'm running a machine that uses the aic7xxx driver for the old adaptec 2940 scsi
> controller. i had problems with getting 2.5.67-ac2 to run, it used to end in
> a kernel panic. i got told to remove the body of the function ide_xlate_1024
> by just "return 0;", which fixed the problem, i could then boot the kernel.
> but now, when running lilo, i get the following message:

Weird - the 2940 is scsi so how come ide is involved?
I use mine for a cdrom.  I had to use the old aic7xxx driver,
the new one didn't detect the cdrom.
It works fine with 2.5.67 and 2.5.67-mm4
 
Helge Hafting

Re: 2.5.67-ac2 and lilo

[Christian Staudenmayer]

Hey Andries,

before taking your message into account, i think you misunderstood my
question:

> Hmm. You did not forget to rerun LILO or so?

the error occurs when _running_ lilo, i.e. after typing "lilo" at the prompt.

greetings, chris

--- Andries Brouwer <aebr@win.tue.nl> wrote:
> > but now, when running lilo, i get the following message:
> > 
> > Fatal: First boot sector doesn't have a valid LILO signature
> 
> Hmm. You did not forget to rerun LILO or so?
> If ide_xlate_1024 does nothing but "return 0" then
> handle_ide_mess() does nothing.
> 
> What other differences does -ac2 have?
> It reintroduces a bug in the partition reading code - the fragment
> 
> @@ -456,7 +556,7 @@
>                 if (!subtypes[n].parse)
>                         continue;
>                 subtypes[n].parse(state, bdev, START_SECT(p)*sector_size,
> -                                               NR_SECTS(p)*sector_size, slot);
> +                                               NR_SECTS(p)*sector_size, n);
>         }
>         put_dev_sector(sect);
> 
> of the -ac2 patch is bad.
> In order to judge whether it is this or something else I would have
> to know much more about your system. What are the kernel boot messages
> for this disk under 2.5.67-bk9 that works and 2.5.67-ac2 that fails?
> What is the partition table? No disk managers in sight? Does LILO
> have lba32 or linear options?
> Does 2.5.67-ac2 work if you replace its fs/partitions/msdos.c by
> the vanilla one?
> 
> Andries
> 


__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com

Re: 2.5.67-ac2 and lilo

[Andries Brouwer]

On Sat, Apr 19, 2003 at 02:13:56PM -0700, Christian Staudenmayer wrote:

> Hey Andries,
> 
> before taking your message into account, i think you misunderstood my
> question:
> 
> > Hmm. You did not forget to rerun LILO or so?
> 
> the error occurs when _running_ lilo, i.e. after typing "lilo" at the prompt.

Yes, I misunderstood. Interesting that the running of an ordinary utility
depends on the kernel in this way.

I cry when I read the source.

    if ((size = read(l_fd,buff,sizeof(buff)+1)) < 0)
        die("read %s: %s",loader,strerror(errno));
    if (size > sizeof(buff))
        die("Chain loader %s is too big",loader);

Precisely what LILO does can be followed better by giving one or more -v
options.

The question remains: if you replace (in the kernel) the -ac2 msdos.c
by the vanilla msdos.c, does that improve things?

By the way, LILO has been a beautiful boot loader, but I think
it is slowly dying.

Andries