Re: [LinuxBIOS] r2546 - trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804

[Lu, Yinghai]

YES

-----Original Message-----
From: linuxbios-bounces+yinghai.lu=[e-mail]
[mailto:linuxbios-bounces+yinghai.lu=[e-mail]] On Behalf Of
[e-mail]
Sent: Friday, February 02, 2007 2:40 PM
To: [e-mail]
Subject: [LinuxBIOS] r2546 -
trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804

Author: rminnich
Date: 2007-02-02 23:40:10 +0100 (Fri, 02 Feb 2007)
New Revision: 2546

Modified:
   trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804/chip.h
   trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804/ck804_usb.c
Log:
I have Sun Ultra40 workstation. Southbridge is nVidia CrushK8-04/nforce
2200 (too many names, sounds like a criminal).

1) Linuxbios loads kernel A; kernel A loads kernel B. Everything works
fine.

2) Then I push the reset button.

3) Linuxbios loads kernel A; kernel A loads kernel B. Kernel B complains
about wrong checksum of the mptable and crushes later.

An investigation showed that in 3), short after kernel A (v2.6.19.2)
sets
the Bus Master Enable bit of the nVidia's USB1 controller
(pci_set_master()),
the mptable gets two bytes at physical address 0x80 damaged.

Nothing is plugged to the USB ports. Other two Sun workstations had the
same
behavior. This does not make sense to me unless the controller has a HW
bug.

I believe, this should better be fixed in the kernel USB driver.

For now this patch offers a possibility for linuxbios to reset the USB
controller by setting HostControllerReset bit in HcCommandStatus
Register.
It is enablead by using 'register "usb1_hc_reset"="1"' in 'chip
southbridge/nvidia/ck804' section of the mainboard's Config.lb.

Signed-off-by: Roman Kononov 
Acked-by: Ronald G. Minnich <[e-mail]

Modified: trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804/chip.h
===================================================================
--- trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804/chip.h
2007-02-02 17:08:04 UTC (rev 2545)
+++ trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804/chip.h
2007-02-02 22:40:10 UTC (rev 2546)
@@ -1,10 +1,11 @@
 #ifndef CK804_CHIP_H
 #define CK804_CHIP_H

-struct southbridge_nvidia_ck804_config 
+struct southbridge_nvidia_ck804_config
 {
-        unsigned int ide0_enable : 1;
-        unsigned int ide1_enable : 1;
+	unsigned int usb1_hc_reset : 1;
+	unsigned int ide0_enable : 1;
+	unsigned int ide1_enable : 1;
    unsigned int sata0_enable : 1;
    unsigned int sata1_enable : 1;
    unsigned long nic_rom_address;

Modified: trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804/ck804_usb.c
===================================================================
--- trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804/ck804_usb.c
2007-02-02 17:08:04 UTC (rev 2545)
+++ trunk/LinuxBIOSv2/src/southbridge/nvidia/ck804/ck804_usb.c
2007-02-02 22:40:10 UTC (rev 2546)
@@ -9,9 +9,22 @@
 #include <device/pci_ops.h>
 #include "ck804.h"

+static void usb1_init(struct device *dev) {
+	struct southbridge_nvidia_ck804_config const *
conf=dev->chip_info;
+	if (conf->usb1_hc_reset) {
+		//Somehow the warm reset does not really resets the USB
controller.
+		//Later, during boot, when the Bus Master bit is set,
the USB
+		//controller trashes the memory, causing weird
misbehavior.
+		//Was detected on Sun Ultra40, where mptable was
damaged.
+		uint32_t bar0=pci_read_config32(dev,0x10);
+		uint32_t* regs=(uint32_t*)(bar0&~0xfff);
+		regs[2]|=1;	//OHCI USB HCCommandStatus Register,
HostControllerReset bit
+	}
+}
+
 static void lpci_set_subsystem(device_t dev, unsigned vendor, unsigned
device)
-{       
-        pci_write_config32(dev, 0x40, 
+{
+        pci_write_config32(dev, 0x40,
                 ((device & 0xffff) << 16) | (vendor & 0xffff));
 }
 static struct pci_operations lops_pci = {
@@ -22,7 +35,7 @@
    .read_resources   = pci_dev_read_resources,
    .set_resources    = pci_dev_set_resources,
    .enable_resources = pci_dev_enable_resources,
-	.init             = 0,
+	.init             = usb1_init,
 //	.enable           = ck804_enable,
    .scan_bus         = 0,
    .ops_pci          = &lops_pci,