Re: [coreboot] [v2] r4698 - trunk/coreboot-v2/src/mainboard/dell/s1850

[Stefan Reinauer]

[e-mail] wrote:
> Author: rminnich
> Date: 2009-10-01 02:02:28 +0200 (Thu, 01 Oct 2009)
> New Revision: 4698
>
> Modified:
>    trunk/coreboot-v2/src/mainboard/dell/s1850/auto.c
> Log:
> This is now set up more like the real hardware likes it. 
> Some of this trickery was determined with serialice. 
> There are several lovely undocumented features to the chipset.
>
> Signed-off-by: Ronald G. Minnich 
> Acked-by: Ronald G. Minnich

:-) Nice
> Modified: trunk/coreboot-v2/src/mainboard/dell/s1850/auto.c
>
> --- trunk/coreboot-v2/src/mainboard/dell/s1850/auto.c	2009-09-30 21:36:38 UTC (rev 4697)
> +++ trunk/coreboot-v2/src/mainboard/dell/s1850/auto.c	2009-10-01 00:02:28 UTC (rev 4698)
> @@ -173,8 +173,13 @@
>  }
>
> /* end IPMI garbage */
>
> static void main(unsigned long bist)
>  {
> u8 b;
> u16 w;
> u32 l;
> int do_reset;
> /*
>
> @@ -191,6 +196,80 @@
>    	}
>    };
> /* using SerialICE, we've seen this basic reset sequence on the dell.
> * we don't understand it as it uses undocumented registers, but
> * we're going to clone it. 
> */
> /* enable a hidden device. */
> b = pci_read_config8(PCI_DEV(0, 0, 0), 0xf4);
> b |= 0x8;
> pci_write_config8(PCI_DEV(0, 0, 0), 0xf4, b);
> /* read-write lock in CMOS on LPC bridge on ICH5 */
> pci_write_config8(PCI_DEV(0, 0x1f, 0), 0xd8, 4);
> /* operate on undocumented device */

Can you dump the pci config space or read the pci id of that device?
> l = pci_read_config32(PCI_DEV(0, 0, 2), 0xa4);
> l |= 0x1000;
> pci_write_config32(PCI_DEV(0, 0, 2), 0xa4, l);
> l = pci_read_config32(PCI_DEV(0, 0, 2), 0x9c);
> l |= 0x8000;
> pci_write_config32(PCI_DEV(0, 0, 2), 0x9c, l);
> /* disable undocumented device */
> b = pci_read_config8(PCI_DEV(0, 0, 0), 0xf4);
> b &= ~0x8;
> pci_write_config8(PCI_DEV(0, 0, 0), 0xf4, b);
> /* set up LPC bridge bits, some of which reply on undocumented
> * registers
> */
> b= pci_read_config8(PCI_DEV(0, 0x1f, 0), 0xd8);
> b |= 4;
> pci_write_config8(PCI_DEV(0, 0x1f, 0), 0xd8, b);
> b= pci_read_config8(PCI_DEV(0, 0x1f, 0), 0xd4);
> b |= 2;
> pci_write_config8(PCI_DEV(0, 0x1f, 0), 0xd4, b);
> /* ACPI base address */
> pci_write_config16(PCI_DEV(0, 0x1f, 0), 0x40, 0x800);

bit 0 is hard wired to 1 so it doesn't matter much I guess.
> /* Enable specific ACPI features */
> b= pci_read_config8(PCI_DEV(0, 0x1f, 0), 0x44);
> b |= 0x10;
> pci_write_config8(PCI_DEV(0, 0x1f, 0), 0x44, b);
> /* ACPI control */
> w = inw(0x868);
> outw(w|0x800, 0x868);
> w = inw(0x866);
> outw(w|2, 0x866);

This is something in the TCO area, probably disabling the watchdog
> /* SMBUS */
> pci_write_config16(PCI_DEV(0, 0x1f, 3), 0x20, 0x08c0);
> /* unknown */
> b = inb(0x8c2);
> outb(0xdf, 0x8c2);
> /* another device enable? */
> b = pci_read_config8(PCI_DEV(0, 0, 0), 0xf4);
> b |= 2;
> pci_write_config8(PCI_DEV(0, 0, 0), 0xf4, b);
> /* ?? */
> l = pci_read_config32(PCI_DEV(0, 0, 8), 0xc0);
> do_reset = l & 0x8000000;
> l |= 0x8000000;
> pci_write_config32(PCI_DEV(0, 0, 2), 0xc0, l);
> if (! do_reset) {
> outb(2, 0xcf9);
> outb(6, 0xcf9);
> }
> if (bist == 0) {
> /* Skip this if there was a built in self test failure */
> early_mtrr_init();

[v2] r4698 - trunk/coreboot-v2/src/mainboard/dell/s1850

[ron minnich]

On Wed, Sep 30, 2009 at 5:11 PM, Stefan Reinauer  wrote:
> :-) Nice

Sorry. I just assumed nobody would even care about this antique. Hope
no one is too annoyed by the self-ack.
> > /* operate on undocumented device */
>
> Can you dump the pci config space or read the pci id of that device?

no, because the Dell BIOS turns it completely off.
> > /* ACPI base address */
> > pci_write_config16(PCI_DEV(0, 0x1f, 0), 0x40, 0x800);
>
> bit 0 is hard wired to 1 so it doesn't matter much I guess.

yes, I'm cloning what they do by watching IO with serialice. They did
not read it and or it in in this one case.

lotsa magic bits here. This is a really old part so I'm a little
surprised. Not too much, any more.

Going to have a long delay now while I convert to CAR. Also having a
meeting in 2 weeks that may kill the whole project.

ron

[v2] r4698 - trunk/coreboot-v2/src/mainboard/dell/s1850

[Stefan Reinauer]

On Oct 1, 2009, at 2:44, ron minnich  wrote:
> On Wed, Sep 30, 2009 at 5:11 PM, Stefan Reinauer <[e-mail]
> > wrote:
> > :-) Nice
>
> Sorry. I just assumed nobody would even care about this antique. Hope
> no one is too annoyed by the self-ack.
> > > /* operate on undocumented device */
> >
> > Can you dump the pci config space or read the pci id of that device?
>
> no, because the Dell BIOS turns it completely off.

There's a pci config space dumper in one or two of the debug.c versions
> > > /* ACPI base address */
> > > pci_write_config16(PCI_DEV(0, 0x1f, 0), 0x40, 0x800);
> >
> > bit 0 is hard wired to 1 so it doesn't matter much I guess.
>
> yes, I'm cloning what they do by watching IO with serialice. They did
> not read it and or it in in this one case.
>
> lotsa magic bits here. This is a really old part so I'm a little
> surprised. Not too much, any more.
>
> Going to have a long delay now while I convert to CAR. Also having a
> meeting in 2 weeks that may kill the whole project.

Sure you want to add new points of failure? ;-)
> ron

Stefan

[v2] r4698 - trunk/coreboot-v2/src/mainboard/dell/s1850

[Ed Swierk]

On Wed, Sep 30, 2009 at 5:44 PM, ron minnich  wrote:
> On Wed, Sep 30, 2009 at 5:11 PM, Stefan Reinauer  wrote:
> > > /* operate on undocumented device */
> >
> > Can you dump the pci config space or read the pci id of that device?
>
> no, because the Dell BIOS turns it completely off.

It shouldn't harm anything to leave the undocumented device enabled.
Then you can boot Linux and sniff around with lspci.

--Ed

[v2] r4698 - trunk/coreboot-v2/src/mainboard/dell/s1850

[Joseph Smith]

On 09/30/2009 08:44 PM, ron minnich wrote:
> On Wed, Sep 30, 2009 at 5:11 PM, Stefan Reinauer  wrote:
> > :-) Nice
>
> Sorry. I just assumed nobody would even care about this antique. Hope
> no one is too annoyed by the self-ack.
> > > /* operate on undocumented device */
> >
> > Can you dump the pci config space or read the pci id of that device?
>
> no, because the Dell BIOS turns it completely off.
> > > /* ACPI base address */
> > > pci_write_config16(PCI_DEV(0, 0x1f, 0), 0x40, 0x800);
> >
> > bit 0 is hard wired to 1 so it doesn't matter much I guess.
>
> yes, I'm cloning what they do by watching IO with serialice. They did
> not read it and or it in in this one case.

Very cool, I have yet to play with serialice. If your watching IO then 
how do you know what the PCI regs are doing? If I could use serialice to 
capture PCI regs, it would solve alot of mysteries....

Or are you just using serialice to capture ACPI read/writes?
> lotsa magic bits here. This is a really old part so I'm a little
> surprised. Not too much, any more.
>
> Going to have a long delay now while I convert to CAR. Also having a
> meeting in 2 weeks that may kill the whole project.

bummer....would you continue it for personal interest?

---------------------------
> /* ACPI control */
> > w = inw(0x868);
> > outw(w|0x800, 0x868);
> > w = inw(0x866);
> > outw(w|2, 0x866);
>
> This is something in the TCO area, probably disabling the watchdog

Nice! If the TCO timer is not disabled you will reboot for ever....I 
beleive the OS re-enables it?