SeaBIOS and boot of SATA raid card

[Rudolf Marek]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

I had some time today. I have PCI SATA RAID (fake raid) with Silicon Image chip
on it. The option rom runs fine but SeaBIOS complains that EBDA was relocated. I
fixed that in the patch. Also I fixed the stuff so now I get:

Inspecting possible rom at f1200000 (vd=31141095 bdf=00000530)
Copying option rom from f1200000 to 000ca000
Running option rom at 0000ca00:00000003
Changing serial settings was 00000003/00000002 now 00000003/00000000
Option rom at 0000ca00:00000003 attempted to move ebda from 00009fc0 to 00009f80
$PnP at 000cb220
Running BCV 000031c7
Running option rom at 0000ca00:000031c7
Option rom at 0000ca00:000031c7 attempted to move ebda from 00009fc0 to 00009f80
Adding IPL
Press F12 for boot menu.

Select boot device:

1. Floppy
2. Hard Disk
3. CD-Rom
4. Floppy [05:30-0 ST3250310AS         ]

Jump to int19
enter handle_19:
  NULL
Booting from Floppy [05:30-0 ST3250310AS         ]...
fail check_drive:384(00000080):
  a=00000201 b=00000000 c=00000001 d=00000000 si=00000000 di=00000000
  ds=00000000 es=000007c0 ip=0000e7e9 cs=0000f000 f=00000002 r=00007b5e
Boot failed: could not read the boot disk

enter handle_18:
  NULL

The WIP patch is attached. I had no time to investigate why it does not actually
boot. Maybe next time.

Rudolf
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAklIQAcACgkQ3J9wPJqZRNXPCgCePij8Nodva+Qmw8DUdBd6GosO
uskAn1PWHYmsJqDgjDYTKSH4zYd5d+27
=mBQO
-----END PGP SIGNATURE-----

[Kevin O'Connor]

Hi Ruik!

On Wed, Dec 17, 2008 at 12:55:54AM +0100, Rudolf Marek wrote:
> Hello,
>
> I had some time today. I have PCI SATA RAID (fake raid) with Silicon Image chip
> on it. The option rom runs fine but SeaBIOS complains that EBDA was relocated. I
> fixed that in the patch. Also I fixed the stuff so now I get:

Yeah, SeaBIOS needs to handle a relocatable ebda.

The simple way to do this (reload the ebda_seg from the bda) on every
access is going to lead to terrible code generation.  It may be
simpler to implement the Post Memory Manager (PMM) interface - in
theory, an option rom shouldn't relocate the ebda if it can allocate
memory via PMM.  Another possibility would be to assign a segment to
the current ebda location (eg, fs/gs) at every entry to the C code.

[...]

Hrmm.  I'm not sure what the following part of your patch does:
> @@ -341,13 +343,29 @@
>              callrom(rom, OPTION_ROM_INITVECTOR, 0);
>              continue;
>          }
>    // PnP rom.
>    if (pnp->bev)
>        // Can boot system - add to IPL list.
>        add_ipl(rom, pnp);
>    else if (pnp->bcv)
>        // Has BCV - run it now.
>        callrom(rom, pnp->bcv, 0);
> /* rebuild the PNP block, init migh have change that */
> pnp = get_pnp_rom(rom);
> /* for further details check BIOS Boot Specification */
> while (pnp) {
> 	dprintf(1, "$PnP at %p\n",pnp);
> if ((pnp->bev) && (!pnp->bcv))
>         add_ipl(rom, pnp, IPL_TYPE_BEV);
> if ((!pnp->bev) && (pnp->bcv)) {
> 	dprintf(1, "Running BCV %x\n",pnp->bcv);
>    	    callrom(rom, pnp->bcv, 0);
>         add_ipl(rom, pnp, pnp->type_lo);
> }
> if (pnp->nextoffset)
> 	pnp = (struct pnp_data *) (((u8 *) pnp) + pnp->nextoffset);
> else
> 	pnp = NULL;
> }
> }
>
> }

-Kevin

[Rudolf Marek]

> The simple way to do this (reload the ebda_seg from the bda) on every
> access is going to lead to terrible code generation.  It may be
> simpler to implement the Post Memory Manager (PMM) interface - in
> theory, an option rom shouldn't relocate the ebda if it can allocate
> memory via PMM.  Another possibility would be to assign a segment to
> the current ebda location (eg, fs/gs) at every entry to the C code.

yes perhaps, because when PPM manager fails to provide memory, the option rom 
will do the EBDA stuff anyway.
> Hrmm.  I'm not sure what the following part of your patch does:

The option rom has either BEV OR BCV not both. BEV is for network card, just a 
entrypoint. BCV will patch the int13 and add new device there.
> > @@ -341,13 +343,29 @@
> >              callrom(rom, OPTION_ROM_INITVECTOR, 0);
> >              continue;
> >          }
> >    // PnP rom.
> >    if (pnp->bev)
> >        // Can boot system - add to IPL list.
> >        add_ipl(rom, pnp);
> >    else if (pnp->bcv)
> >        // Has BCV - run it now.
> >        callrom(rom, pnp->bcv, 0);
> > /* rebuild the PNP block, init migh have change that */
> > pnp = get_pnp_rom(rom);
> > /* for further details check BIOS Boot Specification */
> > while (pnp) {
> > 	dprintf(1, "$PnP at %p\n",pnp);
> > if ((pnp->bev) && (!pnp->bcv))
> >         add_ipl(rom, pnp, IPL_TYPE_BEV);
> > if ((!pnp->bev) && (pnp->bcv)) {
> > 	dprintf(1, "Running BCV %x\n",pnp->bcv);
> >    	    callrom(rom, pnp->bcv, 0);
> >         add_ipl(rom, pnp, pnp->type_lo);

the type_lo is not the boot type, Its a PCI type it seems so it then should be 
HDD (this can fix the floppy item which is wrong)
> > }
> > if (pnp->nextoffset)

there can be more BEV/BCVs in the chain.
> > 	pnp = (struct pnp_data *) (((u8 *) pnp) + pnp->nextoffset);
> > else
> > 	pnp = NULL;
> > }
> > }
> >
> > }

Rudolf