Re: [flashrom] [PATCH] Check class for nvidia southbridge

[Carl-Daniel Hailfinger]

Hi Yinghai,

maybe you remember what the CK804 A01 comment mentioned in the mail
below refers to.

The chipset_enable.c code we're referring to is available at
http://www.flashrom.org/trac/flashrom/browser/trunk/chipset_enable.c

Michael, I would mask with the PCI class with 0xff00 and allow all bridges.

Regards,
Carl-Daniel

Am 13.03.2012 22:29 schrieb Michael Karcher:
> On dual-socket mainboards, there are two nvidia southbridge chips (one
> per socket).  The south bridge really used as south bridge has the
> device class of the ISA/LPC bridge device set to 0x0601 [ISA Bridge],
> while the other one has a device class in the 0x05xx range [RAM stuff].
>
> This will select the correct south bridge for the chipset enable, and
> removes the hack that the 0x361 device needed to be ignored to get rid
> of the secondary south bridge on the Tyan S2915 / HP xw9400.
>
> The CK804 entry with the ID 10de:00d3 (which triggers on the secondary
> south bridge of the L1N64) has the comment
>   /* Slave, should not be here, to fix known bug for A01. */
> Does anyone still remember what this 7 year old comment is trying to tell
> us? What is "A01"? A chipset revision? What class does this device
> have in "A01"? Will the class-must-be-0601-check break "it"?
> Signed-off-by: Michael Karcher
>
> chipset_enable.c |   22 +++++++++++++---------
>  1 files changed, 13 insertions(+), 9 deletions(-)
>
> diff --git a/chipset_enable.c b/chipset_enable.c
> index 0aba1e0..adf2e0d 100644
> --- a/chipset_enable.c
> +++ b/chipset_enable.c
> @@ -906,6 +906,12 @@ static int enable_flash_ck804(struct pci_dev *dev, const char *name)
>  {
>    uint8_t old, new;
> if (pci_read_word(dev, PCI_CLASS_DEVICE) != 0x0601) {
> /* output is followed by "OK - searching for ...\n" */
> msg_pinfo(" Seems to be a secondary south bridge, ");
> return NOT_DONE_YET;
> }
> pci_write_byte(dev, 0x92, 0x00);
>  if (pci_read_byte(dev, 0x92) != 0x00) {
>   msg_pinfo("Setting register 0x%x to 0x%x on %s failed "
>
> @@ -995,6 +1001,12 @@ static int enable_flash_mcp55(struct pci_dev *dev, const char *name)
>    uint8_t old, new, val;
>    uint16_t wordval;
> if (pci_read_word(dev, PCI_CLASS_DEVICE) != 0x0601) {
> /* output is followed by "OK - searching for ...\n" */
> msg_pinfo(" Seems to be a secondary south bridge, ");
> return NOT_DONE_YET;
> }
> /* Set the 0-16 MB enable bits. */
>  val = pci_read_byte(dev, 0x88);
>  val |= 0xff;		/* 256K */
>
> @@ -1207,15 +1219,7 @@ const struct penable chipset_enables[] = {
>    {0x10de, 0x0262, NT, "NVIDIA", "MCP51",		enable_flash_ck804},
>    {0x10de, 0x0263, NT, "NVIDIA", "MCP51",		enable_flash_ck804},
>    {0x10de, 0x0360, OK, "NVIDIA", "MCP55",		enable_flash_mcp55}, /* M57SLI*/
> /* 10de:0361 is present in Tyan S2915 OEM systems, but not connected to
> * the flash chip. Instead, 10de:0364 is connected to the flash chip.
> * Until we have PCI device class matching or some fallback mechanism,
> * this is needed to get flashrom working on Tyan S2915 and maybe other
> * dual-MCP55 boards.
> */
>
> -#if 0
> {0x10de, 0x0361, NT, "NVIDIA", "MCP55",		enable_flash_mcp55}, /* LPC */
>
> -#endif
> {0x10de, 0x0361, OK, "NVIDIA", "MCP55",		enable_flash_mcp55}, /* LPC */
> {0x10de, 0x0362, OK, "NVIDIA", "MCP55",		enable_flash_mcp55}, /* LPC */
> {0x10de, 0x0363, OK, "NVIDIA", "MCP55",		enable_flash_mcp55}, /* LPC */
> {0x10de, 0x0364, OK, "NVIDIA", "MCP55",		enable_flash_mcp55}, /* LPC */

[PATCH] Check class for nvidia southbridge

[Yinghai Lu]

On Tue, Mar 13, 2012 at 2:47 PM, Carl-Daniel Hailfinger
 wrote:
> Hi Yinghai,
>
> maybe you remember what the CK804 A01 comment mentioned in the mail
> below refers to.
>
> The chipset_enable.c code we're referring to is available at
> http://www.flashrom.org/trac/flashrom/browser/trunk/chipset_enable.c

I only remember that old chipset would have same device id even it is
on HT chain without legacy support.

new rev would change the device id if it is not legacy ...

Thanks

Yinghai