RFC: change default SPARC boot order to "disk:a disk"

[Mark Cave-Ayland]

Hi all,

I've had a bug report off-list indicating that r1199 broke booting 
Solaris from hard disk. Some investigation shows that the simple reason 
is because the boot list in arch/sparc32/openbios.c is missing a slice 
letter, and so the boot device in romvec is set incorrectly.

As the disk device in OpenBIOS/QEMU is effectively modelled as 
sd(0,0,0), then I believe the correct solution should be this:

diff --git a/openbios-devel/arch/sparc32/openbios.c 
b/openbios-devel/arch/sparc32/openbios.c
index abdfe7c..6d1c3dc 100644
--- a/openbios-devel/arch/sparc32/openbios.c
+++ b/openbios-devel/arch/sparc32/openbios.c
@@ -887,7 +887,7 @@ arch_init( void )
                 push_str("floppy");
                 break;
         case 'c':
-               push_str("disk");
+               push_str("disk:a disk");
                 break;
         default:
         case 'd':

Unfortunately I only have a couple of SPARC disk images as my main test 
suite is ISO based, and so while it works for me (and the original 
reporter), I'm a bit nervous about just applying the change as-is.

Can anyone confirm that my understanding of Solaris slices is correct, 
and that this is indeed the correct solution to the problem?

Many thanks,

Mark.

[Mark Cave-Ayland]

On 08/12/13 12:43, Mark Cave-Ayland wrote:
> Hi all,
>
> I've had a bug report off-list indicating that r1199 broke booting
> Solaris from hard disk. Some investigation shows that the simple reason
> is because the boot list in arch/sparc32/openbios.c is missing a slice
> letter, and so the boot device in romvec is set incorrectly.
>
> As the disk device in OpenBIOS/QEMU is effectively modelled as
> sd(0,0,0), then I believe the correct solution should be this:
>
> diff --git a/openbios-devel/arch/sparc32/openbios.c
> b/openbios-devel/arch/sparc32/openbios.c
> index abdfe7c..6d1c3dc 100644
> --- a/openbios-devel/arch/sparc32/openbios.c
> +++ b/openbios-devel/arch/sparc32/openbios.c
> @@ -887,7 +887,7 @@ arch_init( void )
> push_str("floppy");
> break;
> case 'c':
> push_str("disk");
> push_str("disk:a disk");
>
> break;
> default:
> case 'd':
>
> Unfortunately I only have a couple of SPARC disk images as my main test
> suite is ISO based, and so while it works for me (and the original
> reporter), I'm a bit nervous about just applying the change as-is.
>
> Can anyone confirm that my understanding of Solaris slices is correct,
> and that this is indeed the correct solution to the problem?

In actual fact, it seems that this regression was a combination of two 
other bugs exposed by recent changes which have now been fixed in SVN. 
So I'm currently planning to leave this as it is unless anyone feels 
strongly otherwise...

ATB,

Mark.