[PATCH] flashrom: SB600 SPI detection heuristic

[Carl-Daniel Hailfinger]

Create a SB600 SPI detection heuristic.

I know that the data sheets say we can't read the ROM straps, but
creative interpretation of the data sheets yielded a heuristic which
should work pretty well.

NOTE: If you test this, make sure you power down and _unplug_ the
machine for a few minutes before you boot and run flashrom with this patch.
If the machine is not unplugged for some time, the test will yield
incorrect results.
If you run a slightly older flashrom version than svn HEAD, the test
will yield incorrect results.
If you run any flashrom version (except svn HEAD plus this patch) after
poweron, the test will yield incorrect results.

Explanation:
Older flashrom versions unconditionally stomp on registers which are
used for this heuristic. These registers are in the S5 power domain, so
even powering down does not clear them, you really have to unplug the
machine.

Please run
flashrom --verbose
once and post the output to this list.

Signed-off-by: Carl-Daniel Hailfinger

Index: flashrom-sb600/chipset_enable.c
===================================================================
--- flashrom-sb600/chipset_enable.c	(Revision 464)
+++ flashrom-sb600/chipset_enable.c	(Arbeitskopie)
@@ -660,6 +660,8 @@
 {
    uint32_t tmp, prot;
    uint8_t reg;
+	struct pci_dev *smbus_dev;
+	int has_spi = 1;

/* Clear ROM protect 0-3. */
    for (reg = 0x50; reg < 0x60; reg += 4) {
@@ -688,13 +690,48 @@
    tmp &= 0xfffffff0;	/* remove low 4 bits (reserved) */
    printf_debug("SPI base address is at 0x%x\n", tmp);

+	/* If the BAR has address 0, it is unlikely SPI is used. */
+	if (!tmp)
+		has_spi = 0;
+
    /* Physical memory can only be mapped at page (4k) boundaries */
    sb600_spibar = physmap("SB600 SPI registers", tmp & 0xfffff000, 0x1000);
    /* The low bits of the SPI base address are used as offset into the mapped page */
    sb600_spibar += tmp & 0xfff;

-	flashbus = BUS_TYPE_SB600_SPI;
+	/* Look for the SMBus device. */
+	smbus_dev = pci_dev_find(0x1002, 0x4385);

+	if (!smbus_dev) {
+		fprintf(stderr, "ERROR: SMBus device not found. Not enabling SPI.\n");
+		has_spi = 0;
+	} else {
+		/* Note about the bit tests below: If a bit is zero, the GPIO is SPI. */
+		/* GPIO11/SPI_DO and GPIO12/SPI_DI status */
+		reg = pci_read_byte(smbus_dev, 0xAB);
+		reg &= 0xC0;
+		printf_debug("GPIO11 used for %s\n", (reg & (1 << 6)) ? "GPIO" : "SPI_DO");
+		printf_debug("GPIO12 used for %s\n", (reg & (1 << 7)) ? "GPIO" : "SPI_DI");
+		if (reg != 0x00)
+			has_spi = 0;
+		/* GPIO31/SPI_HOLD and GPIO32/SPI_CS status */
+		reg = pci_read_byte(smbus_dev, 0x83);
+		reg &= 0xC0;
+		printf_debug("GPIO31 used for %s\n", (reg & (1 << 6)) ? "GPIO" : "SPI_HOLD");
+		printf_debug("GPIO32 used for %s\n", (reg & (1 << 7)) ? "GPIO" : "SPI_CS");
+		if (reg != 0x00)
+			has_spi = 0;
+		/* GPIO47/SPI_CLK status */
+		reg = pci_read_byte(smbus_dev, 0xA7);
+		reg &= 0x40;
+		printf_debug("GPIO47 used for %s\n", (reg & (1 << 6)) ? "GPIO" : "SPI_CLK");
+		if (reg != 0x00)
+			has_spi = 0;
+	}
+
+	if (has_spi)
+		flashbus = BUS_TYPE_SB600_SPI;
+
    /* Force enable SPI ROM in SB600 PM register.
     * If we enable SPI ROM here, we have to disable it after we leave.
     * But how can we know which ROM we are going to handle? So we have

[Stefan Reinauer]

On 06.05.2009 2:50 Uhr, Carl-Daniel Hailfinger wrote:
> Create a SB600 SPI detection heuristic.
>
> I know that the data sheets say we can't read the ROM straps, but
> creative interpretation of the data sheets yielded a heuristic which
> should work pretty well.
>
> NOTE: If you test this, make sure you power down and _unplug_ the
> machine for a few minutes before you boot and run flashrom with this patch.
> If the machine is not unplugged for some time, the test will yield
> incorrect results.
> If you run a slightly older flashrom version than svn HEAD, the test
> will yield incorrect results.
> If you run any flashrom version (except svn HEAD plus this patch) after
> poweron, the test will yield incorrect results.
>
> Explanation:
> Older flashrom versions unconditionally stomp on registers which are
> used for this heuristic. These registers are in the S5 power domain, so
> even powering down does not clear them, you really have to unplug the
> machine.

(and remove the battery if it is a laptop)
> Signed-off-by: Carl-Daniel Hailfinger

Acked-by: Stefan Reinauer

[Carl-Daniel Hailfinger]

On 06.05.2009 03:22, Stefan Reinauer wrote:
> On 06.05.2009 2:50 Uhr, Carl-Daniel Hailfinger wrote:
> > Create a SB600 SPI detection heuristic.
> >
> > I know that the data sheets say we can't read the ROM straps, but
> > creative interpretation of the data sheets yielded a heuristic which
> > should work pretty well.
> >
> > NOTE: If you test this, make sure you power down and _unplug_ the
> > machine for a few minutes before you boot and run flashrom with this patch.
> > If the machine is not unplugged for some time, the test will yield
> > incorrect results.
> > If you run a slightly older flashrom version than svn HEAD, the test
> > will yield incorrect results.
> > If you run any flashrom version (except svn HEAD plus this patch) after
> > poweron, the test will yield incorrect results.
> >
> > Explanation:
> > Older flashrom versions unconditionally stomp on registers which are
> > used for this heuristic. These registers are in the S5 power domain, so
> > even powering down does not clear them, you really have to unplug the
> > machine.
>
> (and remove the battery if it is a laptop)
> > Signed-off-by: Carl-Daniel Hailfinger
>
> Acked-by: Stefan Reinauer

Thanks, committed in r491.

Regards,
Carl-Daniel