[coreboot-gerrit] Change in coreboot[master]: mb/pcengines/apu1/mainboard.c: fix apu1 serial number counting

[Piotr Kleinschmidt (Code Review)]

Piotr Kleinschmidt has uploaded this change for review. ( https://review.coreboot.org/c/coreboot/+/42512 )

Change subject: mb/pcengines/apu1/mainboard.c: fix apu1 serial number counting
......................................................................

mb/pcengines/apu1/mainboard.c: fix apu1 serial number counting

PC Engines apu1 platform returns serial number value as -64.
It is caused by faulty read of NIC's registers.
This patch adjust NIC's bus number to valid value, so NIC's
registers can be correctly read and hence NIC's MAC address.

TEST=`dmidecode -t 2` command in Linux Debian

Signed-off-by: Piotr Kleinschmidt 
Change-Id: I5e8690d100b38ac7889395d375c0ff32bdefda0b
---
M src/mainboard/pcengines/apu1/mainboard.c
1 file changed, 10 insertions(+), 1 deletion(-)

git pull ssh://review.coreboot.org:29418/coreboot refs/changes/12/42512/1

diff --git a/src/mainboard/pcengines/apu1/mainboard.c b/src/mainboard/pcengines/apu1/mainboard.c
index a2a78c5..b297061 100644
--- a/src/mainboard/pcengines/apu1/mainboard.c
+++ b/src/mainboard/pcengines/apu1/mainboard.c
@@ -299,9 +299,18 @@
    if (!dev)
    	return serial;

+	/* FIXME: dev->bus->secondary has 0x100 value, while it should
+	 * has 0x001 value, as it is on PCI bridge 1.
+	 * Without this workaround pci_read_config32(dev, 0x18) returns
+	 * incorrect data and hence platform's serial number is incorrect.
+	 * However, shifting dev->bus->secondary 8 bits right is not
+	 * reliable solution, as the problem probably lies earlier.
+	 */
+	dev->bus->secondary >>= 8;
+
    /* Read in the last 3 bytes of NIC's MAC address. */
    bar18 = pci_read_config32(dev, 0x18);
-	bar18 &= 0xFFFFFC00;
+	bar18 &= 0xFFFFFFF0;
    for (i = 3; i < 6; i++) {
    	mac_addr <<= 8;
    	mac_addr |= read8((u8 *)bar18 + i);