Re: GRUB 0.93 on Solaris 9/x86...

[Yoshinori K. Okuji]

At Thu, 3 Apr 2003 17:42:06 +0200 (MEST),
Matthias Pf�tzner - Sun Professional Services Germany - Frankfurt wrote:
> Who is the maintainer of the configure part of GRUB? Perhaps you can enhance
> that part?

I don't think such an effort is necessary. Just set CC, OBJCOPY,
etc. when running the configure script
(such as "OBJCOPY=gobjcopy ./configure").

> ==============================================================================
> *** grub/asmstub.c.orig Tue Dec  3 00:20:45 2002
> --- grub/asmstub.c      Wed Apr  2 22:42:23 2003
> ***************
> *** 1131,1137 ****
> --- 1131,1145 ----
>       goto fail;
>   
>     /* Raw mode.  */
> + #if 0
>     cfmakeraw (&termios);
> + #else
> +   termios.c_iflag &= ~(IGNBRK|BRKINT|PARMRK|ISTRIP|INLCR|IGNCR|ICRNL|IXON);
> +   termios.c_oflag &= ~OPOST;
> +   termios.c_lflag &= ~(ECHO|ECHONL|ICANON|ISIG|IEXTEN);
> +   termios.c_cflag &= ~(CSIZE|PARENB);
> +   termios.c_cflag |= CS8;
> + #endif
>   
>     /* Set the speed.  */
>     termios_speed = get_termios_speed (speed);
> ==============================================================================

I can integrate your patch, if you implement a better solution. Maybe
it would be better to detect if cfmakeraw is available in configure.

> Then, there are problems with the way, Solaris does handle disks (IDE, SCSI,
> Floppies) (I should have said: There are problems with the assumptions that
> GRUB makes on the setting up of disk devices... ;-)). Solaris uses::
> 
>         /dev/rdiskette0
> 
>         /dev/dsk/c0d0p0 (fdisk partition)
>         /dev/dsk/c0d0s0 (solaris partition in the FDISK Partition with Solaris
>                          flag (IDE))
>         /dev/dsk/c0t0d0s0 (solaris partition in the FDISK Partition with
>                         Solaris flag (SCSI)) cntr 0, target 0 disk 0 slice 0
> 
> All the numbers are free! (but most start at 0, except the t0, which is the
> SCSI id!) t*d0 is used most of the times, as I haven't seen any LUNS on disk
> (wrong, big boxes, like EMC Symmetrix, IBm Shark will have something different
> like 0 here!).

You have to write Solaris-dependent code anyway.

Thanks,
Okuji