path: root/src/northbridge/amd/amdk8/raminit_f.c

/*
 * This file is part of the coreboot project.
 *
 * Copyright (C) 2002 Linux Networx
 * (Written by Eric Biederman <ebiederman@lnxi.com> for Linux Networx)
 * Copyright (C) 2004 YingHai Lu
 * Copyright (C) 2008 Advanced Micro Devices, Inc.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; version 2 of the License.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301 USA
*/

#include <cpu/x86/cache.h>
#include <cpu/x86/mtrr.h>
#include <cpu/x86/tsc.h>

#include <stdlib.h>
#include "raminit.h"
#include "amdk8_f.h"
#include <spd_ddr2.h>

#ifndef QRANK_DIMM_SUPPORT
#define QRANK_DIMM_SUPPORT 0
#endif

#if CONFIG_DEBUG_RAM_SETUP
#define printk_raminit(fmt, arg...) printk(BIOS_DEBUG, fmt, arg)
#else
#define printk_raminit(fmt, arg...)
#endif


#if (CONFIG_RAMTOP & (CONFIG_RAMTOP -1)) != 0
# error "CONFIG_RAMTOP must be a power of 2"
#endif

#include "amdk8_f_pci.c"


	/* for PCI_ADDR(0, 0x18, 2, 0x98) index,
	 and PCI_ADDR(0x, 0x18, 2, 0x9c) data */
	/*
		index:
		[29: 0] DctOffset (Dram Controller Offset)
		[30:30] DctAccessWrite (Dram Controller Read/Write Select)
			0 = read access
			1 = write access
		[31:31] DctAccessDone (Dram Controller Access Done)
			0 = Access in progress
			1 = No access is progress

		Data:
		[31: 0] DctOffsetData (Dram Controller Offset Data)

		Read:
			- Write the register num to DctOffset with
			  DctAccessWrite = 0
			- poll the DctAccessDone until it = 1
			- Read the data from DctOffsetData
		Write:
			- Write the data to DctOffsetData
			- Write register num to DctOffset with DctAccessWrite = 1
			- poll the DctAccessDone untio it = 1
	*/


static void setup_resource_map(const unsigned int *register_values, int max)
{
	int i;
	for (i = 0; i < max; i += 3) {
		device_t dev;
		unsigned where;
		unsigned long reg;
		dev = register_values[i] & ~0xff;
		where = register_values[i] & 0xff;
		reg = pci_read_config32(dev, where);
		reg &= register_values[i+1];
		reg |= register_values[i+2];
		pci_write_config32(dev, where, reg);
	}
}

static int controller_present(const struct mem_controller *ctrl)
{
	return pci_read_config32(ctrl->f0, 0) == 0x11001022;
}

static void sdram_set_registers(const struct mem_controller *ctrl, struct sys_info *sysinfo)
{
	static const unsigned int register_values[] = {

	/* Careful set limit registers before base registers which
	   contain the enables */
	/* DRAM Limit i Registers
	 * F1:0x44 i = 0
	 * F1:0x4C i = 1
	 * F1:0x54 i = 2
	 * F1:0x5C i = 3
	 * F1:0x64 i = 4
	 * F1:0x6C i = 5
	 * F1:0x74 i = 6
	 * F1:0x7C i = 7
	 * [ 2: 0] Destination Node ID
	 *	   000 = Node 0
	 *	   001 = Node 1
	 *	   010 = Node 2
	 *	   011 = Node 3
	 *	   100 = Node 4
	 *	   101 = Node 5
	 *	   110 = Node 6
	 *	   111 = Node 7
	 * [ 7: 3] Reserved
	 * [10: 8] Interleave select
	 *	   specifies the values of A[14:12] to use with interleave enable.
	 * [15:11] Reserved
	 * [31:16] DRAM Limit Address i Bits 39-24
	 *	   This field defines the upper address bits of a 40 bit  address
	 *	   that define the end of the DRAM region.
	 */
	PCI_ADDR(0, 0x18, 1, 0x44), 0x0000f8f8, 0x00000000,
	PCI_ADDR(0, 0x18, 1, 0x4C), 0x0000f8f8, 0x00000001,
	PCI_ADDR(0, 0x18, 1, 0x54), 0x0000f8f8, 0x00000002,
	PCI_ADDR(0, 0x18, 1, 0x5C), 0x0000f8f8, 0x00000003,
	PCI_ADDR(0, 0x18, 1, 0x64), 0x0000f8f8, 0x00000004,
	PCI_ADDR(0, 0x18, 1, 0x6C), 0x0000f8f8, 0x00000005,
	PCI_ADDR(0, 0x18, 1, 0x74), 0x0000f8f8, 0x00000006,
	PCI_ADDR(0, 0x18, 1, 0x7C), 0x0000f8f8, 0x00000007,
	/* DRAM Base i Registers
	 * F1:0x40 i = 0
	 * F1:0x48 i = 1
	 * F1:0x50 i = 2
	 * F1:0x58 i = 3
	 * F1:0x60 i = 4
	 * F1:0x68 i = 5
	 * F1:0x70 i = 6
	 * F1:0x78 i = 7
	 * [ 0: 0] Read Enable
	 *	   0 = Reads Disabled
	 *	   1 = Reads Enabled
	 * [ 1: 1] Write Enable
	 *	   0 = Writes Disabled
	 *	   1 = Writes Enabled
	 * [ 7: 2] Reserved
	 * [10: 8] Interleave Enable
	 *	   000 = No interleave
	 *	   001 = Interleave on A[12] (2 nodes)
	 *	   010 = reserved
	 *	   011 = Interleave on A[12] and A[14] (4 nodes)
	 *	   100 = reserved
	 *	   101 = reserved
	 *	   110 = reserved
	 *	   111 = Interleve on A[12] and A[13] and A[14] (8 nodes)
	 * [15:11] Reserved
	 * [13:16] DRAM Base Address i Bits 39-24
	 *	   This field defines the upper address bits of a 40-bit address
	 *	   that define the start of the DRAM region.
	 */
	PCI_ADDR(0, 0x18, 1, 0x40), 0x0000f8fc, 0x00000000,
	PCI_ADDR(0, 0x18, 1, 0x48), 0x0000f8fc, 0x00000000,
	PCI_ADDR(0, 0x18, 1, 0x50), 0x0000f8fc, 0x00000000,
	PCI_ADDR(0, 0x18, 1, 0x58), 0x0000f8fc, 0x00000000,
	PCI_ADDR(0, 0x18, 1, 0x60), 0x0000f8fc, 0x00000000,
	PCI_ADDR(0, 0x18, 1, 0x68), 0x0000f8fc, 0x00000000,
	PCI_ADDR(0, 0x18, 1, 0x70), 0x0000f8fc, 0x00000000,
	PCI_ADDR(0, 0x18, 1, 0x78), 0x0000f8fc, 0x00000000,

	/* DRAM CS Base Address i Registers
	 * F2:0x40 i = 0
	 * F2:0x44 i = 1
	 * F2:0x48 i = 2
	 * F2:0x4C i = 3
	 * F2:0x50 i = 4
	 * F2:0x54 i = 5
	 * F2:0x58 i = 6
	 * F2:0x5C i = 7
	 * [ 0: 0] Chip-Select Bank Enable
	 *	   0 = Bank Disabled
	 *	   1 = Bank Enabled
	 * [ 1: 1] Spare Rank
	 * [ 2: 2] Memory Test Failed
	 * [ 4: 3] Reserved
	 * [13: 5] Base Address (21-13)
	 *	   An optimization used when all DIMM are the same size...
	 * [18:14] Reserved
	 * [28:19] Base Address (36-27)
	 *	   This field defines the top 11 addresses bit of a 40-bit
	 *	   address that define the memory address space.  These
	 *	   bits decode 32-MByte blocks of memory.
	 * [31:29] Reserved
	 */
	PCI_ADDR(0, 0x18, 2, 0x40), 0xe007c018, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x44), 0xe007c018, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x48), 0xe007c018, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x4C), 0xe007c018, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x50), 0xe007c018, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x54), 0xe007c018, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x58), 0xe007c018, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x5C), 0xe007c018, 0x00000000,
	/* DRAM CS Mask Address i Registers
	 * F2:0x60 i = 0,1
	 * F2:0x64 i = 2,3
	 * F2:0x68 i = 4,5
	 * F2:0x6C i = 6,7
	 * Select bits to exclude from comparison with the DRAM Base address register.
	 * [ 4: 0] Reserved
	 * [13: 5] Address Mask (21-13)
	 *	   Address to be excluded from the optimized case
	 * [18:14] Reserved
	 * [28:19] Address Mask (36-27)
	 *	   The bits with an address mask of 1 are excluded from address comparison
	 * [31:29] Reserved
	 *
	 */
	PCI_ADDR(0, 0x18, 2, 0x60), 0xe007c01f, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x64), 0xe007c01f, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x68), 0xe007c01f, 0x00000000,
	PCI_ADDR(0, 0x18, 2, 0x6C), 0xe007c01f, 0x00000000,

	/* DRAM Control Register
	 * F2:0x78
	 * [ 3: 0] RdPtrInit ( Read Pointer Initial Value)
	 * 	0x03-0x00: reserved
	 * [ 6: 4] RdPadRcvFifoDly (Read Delay from Pad Receive FIFO)
	 *	000 = reserved
	 *	001 = reserved
	 *	010 = 1.5 Memory Clocks
	 *	011 = 2 Memory Clocks
	 *	100 = 2.5 Memory Clocks
	 *	101 = 3 Memory Clocks
	 *	110 = 3.5 Memory Clocks
	 *	111 = Reseved
	 * [15: 7] Reserved
	 * [16:16] AltVidC3MemClkTriEn (AltVID Memory Clock Tristate Enable)
	 *	Enables the DDR memory clocks to be tristated when alternate VID
	 *	mode is enabled. This bit has no effect if the DisNbClkRamp bit
	 *	(F3, 0x88) is set
	 * [17:17] DllTempAdjTime (DLL Temperature Adjust Cycle Time)
	 *	0 = 5 ms
	 *	1 = 1 ms
	 * [18:18] DqsRcvEnTrain (DQS Receiver Enable Training Mode)
	 *	0 = Normal DQS Receiver enable operation
	 *	1 = DQS receiver enable training mode
	  * [31:19] reverved
	 */
	PCI_ADDR(0, 0x18, 2, 0x78), 0xfff80000, (6<<4)|(6<<0),

	/* DRAM Initialization Register
	 * F2:0x7C
	 * [15: 0] MrsAddress (Address for MRS/EMRS Commands)
	 *	this field specifies the dsata driven on the DRAM address pins
	 *	15-0 for MRS and EMRS commands
	 * [18:16] MrsBank (Bank Address for MRS/EMRS Commands)
	 *	this files specifies the data driven on the DRAM bank pins for
	 *	the MRS and EMRS commands
	 * [23:19] reverved
	 * [24:24] SendPchgAll (Send Precharge All Command)
	 *	Setting this bit causes the DRAM controller to send a precharge
	 *	all command. This bit is cleared by the hardware after the
	 *	command completes
	 * [25:25] SendAutoRefresh (Send Auto Refresh Command)
	 *	Setting this bit causes the DRAM controller to send an auto
	 *	refresh command. This bit is cleared by the hardware after the
	 *	command completes
	 * [26:26] SendMrsCmd (Send MRS/EMRS Command)
	 *	Setting this bit causes the DRAM controller to send the MRS or
	 *	EMRS command defined by the MrsAddress and MrsBank fields. This
	 *	bit is cleared by the hardware adter the commmand completes
	 * [27:27] DeassertMemRstX (De-assert Memory Reset)
	 * 	Setting this bit causes the DRAM controller to de-assert the
	 *	memory reset pin. This bit cannot be used to assert the memory
	 *	reset pin
	 * [28:28] AssertCke (Assert CKE)
	 *	setting this bit causes the DRAM controller to assert the CKE
	 *	pins. This bit cannot be used to de-assert the CKE pins
	 * [30:29] reverved
	 * [31:31] EnDramInit (Enable DRAM Initialization)
	 *	Setting this bit puts the DRAM controller in a BIOS controlled
	 *	DRAM initialization mode. BIOS must clear this bit aster DRAM
	 *	initialization is complete.
	 */
//	PCI_ADDR(0, 0x18, 2, 0x7C), 0x60f80000, 0,


	/* DRAM Bank Address Mapping Register
	 * F2:0x80
	 * Specify the memory module size
	 * [ 3: 0] CS1/0
	 * [ 7: 4] CS3/2
	 * [11: 8] CS5/4
	 * [15:12] CS7/6
	 * [31:16]
	      row    col   bank
	  0:  13     9      2    :128M
	  1:  13     10     2    :256M
	  2:  14     10     2    :512M
	  3:  13     11     2    :512M
	  4:  13     10     3    :512M
	  5:  14     10     3    :1G
	  6:  14     11     2    :1G
	  7:  15     10     3    :2G
	  8:  14     11     3    :2G
	  9:  15     11     3    :4G
	 10:  16     10     3    :4G
	 11:  16     11     3    :8G
	 */
	PCI_ADDR(0, 0x18, 2, 0x80), 0xffff0000, 0x00000000,
	/* DRAM Timing Low Register
	 * F2:0x88
	 * [ 2: 0] Tcl (Cas# Latency, Cas# to read-data-valid)
	 *	   000 = reserved
	 *	   001 = reserved
	 *	   010 = CL 3
	 *	   011 = CL 4
	 *	   100 = CL 5
	 *	   101 = CL 6
	 *	   110 = reserved
	 *	   111 = reserved
	 * [ 3: 3] Reserved
	 * [ 5: 4] Trcd (Ras#-active to Cas# read/write delay)
	 *	   00 = 3 clocks
	 *	   01 = 4 clocks
	 *	   10 = 5 clocks
	 *	   11 = 6 clocks
	 * [ 7: 6] Reserved
	 * [ 9: 8] Trp (Row Precharge Time, Precharge-to-Active or Auto-Refresh)
	 *	   00 = 3 clocks
	 *	   01 = 4 clocks
	 *	   10 = 5 clocks
	 *	   11 = 6 clocks
	 * [10:10] Reserved
	 * [11:11] Trtp (Read to Precharge Time, read Cas# to precharge time)
	 *	   0 = 2 clocks for Burst Length of 32 Bytes
	 *	       4 clocks for Burst Length of 64 Bytes
	 *	   1 = 3 clocks for Burst Length of 32 Bytes
	 *	       5 clocks for Burst Length of 64 Bytes
	 * [15:12] Tras (Minimum Ras# Active Time)
	 *	   0000 = reserved
	 *	   0001 = reserved
	 *	   0010 = 5 bus clocks
	 *	   ...
	 *	   1111 = 18 bus clocks
	 * [19:16] Trc (Row Cycle Time, Ras#-active to Ras#-active or auto
	 * refresh of the same bank)
	 *	   0000 = 11 bus clocks
	 *	   0010 = 12 bus clocks
	 *	   ...
	 *	   1110 = 25 bus clocks
	 *	   1111 = 26 bus clocks
	 * [21:20] Twr (Write Recovery Time, From the last data to precharge,
	 * writes can go back-to-back)
	 *	   00 = 3 bus clocks
	 *	   01 = 4 bus clocks
	 *	   10 = 5 bus clocks
	 *	   11 = 6 bus clocks
	 * [23:22] Trrd (Active-to-active(Ras#-to-Ras#) Delay of different banks)
	 *	   00 = 2 bus clocks
	 *	   01 = 3 bus clocks
	 *	   10 = 4 bus clocks
	 *	   11 = 5 bus clocks
	 * [31:24] MemClkDis ( Disable the MEMCLK outputs for DRAM channel A,
	 * BIOS should set it to reduce the power consumption)
	 *	  Bit		F(1207)		M2 Package	S1g1 Package
	 *	    0		N/A		MA1_CLK1	N/A
	 *	    1		N/A		MA0_CLK1	MA0_CLK1
	 *	    2		MA3_CLK		N/A		N/A
	 *	    3		MA2_CLK		N/A		N/A
	 *	    4		MA1_CLK		MA1_CLK0	N/A
	 *	    5		MA0_CLK		MA0_CLK0	MA0_CLK0
	 *	    6		N/A		MA1_CLK2	N/A
	 *	    7		N/A		MA0_CLK2	MA0_CLK2
	 */
	PCI_ADDR(0, 0x18, 2, 0x88), 0x000004c8, 0xff000002 /* 0x03623125 */ ,
	/* DRAM Timing High Register
	 * F2:0x8C
	 * [ 3: 0] Reserved
	 * [ 6: 4] TrwtTO (Read-to-Write Turnaround for Data, DQS Contention)
	 *	   000 = 2 bus clocks
	 *	   001 = 3 bus clocks
	 *	   010 = 4 bus clocks
	 *	   011 = 5 bus clocks
	 *	   100 = 6 bus clocks
	 *	   101 = 7 bus clocks
	 *	   110 = 8 bus clocks
	 *	   111 = 9 bus clocks
	 * [ 7: 7] Reserved
	 * [ 9: 8] Twtr (Internal DRAM Write-to-Read Command Delay,
	 * minium write-to-read delay when both access the same chip select)
	 *	   00 = Reserved
	 *	   01 = 1 bus clocks
	 *	   10 = 2 bus clocks
	 *	   11 = 3 bus clocks
	 * [11:10] Twrrd (Write to Read DIMM Termination Turnaround, minimum
	 * write-to-read delay when accessing two different DIMMs)
	 *	   00 = 0 bus clocks
	 *	   01 = 1 bus clocks
	 *	   10 = 2 bus clocks
	 *	   11 = 3 bus clocks
	 * [13:12] Twrwr (Write to Write Timing)
	 *	   00 = 1 bus clocks ( 0 idle cycle on the bus)
	 *	   01 = 2 bus clocks ( 1 idle cycle on the bus)
	 *	   10 = 3 bus clocks ( 2 idle cycles on the bus)
	 *	   11 = Reserved
	 * [15:14] Trdrd ( Read to Read Timing)
	 *	   00 = 2 bus clocks ( 1 idle cycle on the bus)
	 *	   01 = 3 bus clocks ( 2 idle cycles on the bus)
	 *	   10 = 4 bus clocks ( 3 idle cycles on the bus)
	 *	   11 = 5 bus clocks ( 4 idel cycles on the bus)
	 * [17:16] Tref (Refresh Rate)
	 *	   00 = Undefined behavior
	 *	   01 = Reserved
	 *	   10 = Refresh interval of 7.8 microseconds
	 *	   11 = Refresh interval of 3.9 microseconds
	 * [19:18] Reserved
	 * [22:20] Trfc0 ( Auto-Refresh Row Cycle Time for the Logical DIMM0,
	 *	based on DRAM density and speed)
	 *	   000 = 75 ns (all speeds, 256Mbit)
	 *	   001 = 105 ns (all speeds, 512Mbit)
	 *	   010 = 127.5 ns (all speeds, 1Gbit)
	 *	   011 = 195 ns (all speeds, 2Gbit)
	 *	   100 = 327.5 ns (all speeds, 4Gbit)
	 *	   101 = reserved
	 *	   110 = reserved
	 *	   111 = reserved
	 * [25:23] Trfc1 ( Auto-Refresh Row Cycle Time for the Logical DIMM1,
	 *	based on DRAM density and speed)
	 * [28:26] Trfc2 ( Auto-Refresh Row Cycle Time for the Logical DIMM2,
	 *	based on DRAM density and speed)
	 * [31:29] Trfc3 ( Auto-Refresh Row Cycle Time for the Logical DIMM3,
	 *	based on DRAM density and speed)
	 */
	PCI_ADDR(0, 0x18, 2, 0x8c), 0x000c008f, (2 << 16)|(1 << 8),
	/* DRAM Config Low Register
	 * F2:0x90
	 * [ 0: 0] InitDram (Initialize DRAM)
	 *	   1 = write 1 cause DRAM controller to execute the DRAM
	 *	       initialization, when done it read to 0
	 * [ 1: 1] ExitSelfRef ( Exit Self Refresh Command )
	 *	   1 = write 1 causes the DRAM controller to bring the DRAMs out
	 *	       for self refresh mode
	 * [ 3: 2] Reserved
	 * [ 5: 4] DramTerm (DRAM Termination)
	 *	   00 = On die termination disabled
	 *	   01 = 75 ohms
	 *	   10 = 150 ohms
	 *	   11 = 50 ohms
	 * [ 6: 6] Reserved
	 * [ 7: 7] DramDrvWeak ( DRAM Drivers Weak Mode)
	 *	   0 = Normal drive strength mode.
	 *	   1 = Weak drive strength mode
	 * [ 8: 8] ParEn (Parity Enable)
	 *	   1 = Enable address parity computation output, PAR,
	 *	       and enables the parity error input, ERR
	 * [ 9: 9] SelfRefRateEn (Faster Self Refresh Rate Enable)
	 * 	  1 = Enable high temperature ( two times normal )
	 *	      self refresh rate
	 * [10:10] BurstLength32 ( DRAM Burst Length Set for 32 Bytes)
	 *	   0 = 64-byte mode
	 *	   1 = 32-byte mode
	 * [11:11] Width128 ( Width of DRAM interface)
	 *	   0 = the controller DRAM interface is 64-bits wide
	 *	   1 = the controller DRAM interface is 128-bits wide
	 * [12:12] X4Dimm (DIMM 0 is x4)
	 * [13:13] X4Dimm (DIMM 1 is x4)
	 * [14:14] X4Dimm (DIMM 2 is x4)
	 * [15:15] X4Dimm (DIMM 3 is x4)
	 *         0 = DIMM is not x4
	 *         1 = x4 DIMM present
	 * [16:16] UnBuffDimm ( Unbuffered DIMMs)
	 *	   0 = Buffered DIMMs
	 *	   1 = Unbuffered DIMMs
	 * [18:17] Reserved
	 * [19:19] DimmEccEn ( DIMM ECC Enable )
	 *	   1 =  ECC checking is being enabled for all DIMMs on the DRAM
	 *	        controller ( Through F3 0x44[EccEn])
	 * [31:20] Reserved
	 */
	PCI_ADDR(0, 0x18, 2, 0x90), 0xfff6004c, 0x00000010,
	/* DRAM Config High Register
	 * F2:0x94
	 * [ 0: 2] MemClkFreq ( Memory Clock Frequency)
	 *	   000 = 200MHz
	 *	   001 = 266MHz
	 *	   010 = 333MHz
	 *	   011 = reserved
	 *	   1xx = reserved
	 * [ 3: 3] MemClkFreqVal (Memory Clock Freqency Valid)
	 *	   1 = BIOS need to set the bit when setting up MemClkFreq to
	 *	       the proper value
	 * [ 7: 4] MaxAsyncLat ( Maximum Asynchronous Latency)
	 *         0000 = 0 ns
	 *         ...
	 *         1111 = 15 ns
	 * [11: 8] Reserved
	 * [12:12] RDqsEn ( Read DQS Enable) This bit is only be set if x8
	 *	   registered DIMMs are present in the system
	 *	   0 = DM pins function as data mask pins
	 *	   1 = DM pins function as read DQS pins
	 * [13:13] Reserved
	 * [14:14] DisDramInterface ( Disable the DRAM interface ) When this bit
	 * is set, the DRAM controller is disabled, and interface in low power
	 * state
	 *	   0 = Enabled (default)
	 *	   1 = Disabled
	 * [15:15] PowerDownEn ( Power Down Mode Enable )
	 *	   0 = Disabled (default)
	 *	   1 = Enabled
	 * [16:16] PowerDown ( Power Down Mode )
	 *	   0 = Channel CKE Control
	 *	   1 = Chip Select CKE Control
	 * [17:17] FourRankSODimm (Four Rank SO-DIMM)
	 *	   1 = this bit is set by BIOS to indicate that a four rank
	 *	       SO-DIMM is present
	 * [18:18] FourRankRDimm (Four Rank Registered DIMM)
	 *         1 = this bit is set by BIOS to indicate that a four rank
	 *	       registered DIMM is present
	 * [19:19] Reserved
	 * [20:20] SlowAccessMode (Slow Access Mode (2T Mode))
	 *	   0 = DRAM address and control signals are driven for one
	 *	       MEMCLK cycle
	 *	   1 = One additional MEMCLK of setup time is provided on all
	 *	       DRAM address and control signals except CS, CKE, and ODT;
	 *	       i.e., these signals are drivern for two MEMCLK cycles
	 *	       rather than one
	 * [21:21] Reserved
	 * [22:22] BankSwizzleMode ( Bank Swizzle Mode),
	 *         0 = Disabled (default)
	 *         1 = Enabled
	 * [23:23] Reserved
	 * [27:24] DcqBypassMax ( DRAM Controller Queue Bypass Maximum)
	 *	   0000 = No bypass; the oldest request is never bypassed
	 *	   0001 = The oldest request may be bypassed no more than 1 time
	 *	   ...
	 *	   1111 = The oldest request may be bypassed no more than 15\
	 *	          times
	 * [31:28] FourActWindow ( Four Bank Activate Window) , not more than
	 *	   4 banks in a 8 bank device are activated
	 *	   0000 = No tFAW window restriction
	 *	   0001 = 8 MEMCLK cycles
	 *	   0010 = 9 MEMCLK cycles
	 *	   ...
	 *	   1101 = 20 MEMCLK cycles
	 *	   111x = reserved
	 */
	PCI_ADDR(0, 0x18, 2, 0x94), 0x00a82f00,0x00008000,
	/* DRAM Delay Line Register
	 * F2:0xa0
	 * [ 0: 0] MemClrStatus (Memory Clear Status) : Readonly
	 *	   when set, this bit indicates that the memory clear function
	 *	   is complete. Only clear by reset. BIOS should not write or
	 *	   read the DRAM until this bit is set by hardware
	 * [ 1: 1] DisableJitter ( Disable Jitter)
	 *	   When set the DDR compensation circuit will not change the
	 *	   values unless the change is more than one step from the
	 *	   current value
	 * [ 3: 2] RdWrQByp ( Read/Write Queue Bypass Count)
	 *	   00 = 2
	 *	   01 = 4
	 *	   10 = 8
	 *	   11 = 16
	 * [ 4: 4] Mode64BitMux (Mismatched DIMM Support Enable)
	 *	   1 When bit enables support for mismatched DIMMs when using
	 *	   128-bit DRAM interface, the Width128 no effect, only for
	 *	   AM2 and s1g1
	 * [ 5: 5] DCC_EN ( Dynamica Idle Cycle Counter Enable)
	 *	   When set to 1, indicates that each entry in the page tables
	 *	   dynamically adjusts the idle cycle limit based on page
	 *	    Conflict/Page Miss (PC/PM) traffic
	 * [ 8: 6] ILD_lmt ( Idle Cycle Limit)
	 *	   000 = 0 cycles
	 *	   001 = 4 cycles
	 *	   010 = 8 cycles
	 *	   011 = 16 cycles
	 *	   100 = 32 cycles
	 *	   101 = 64 cycles
	 *	   110 = 128 cycles
	 *	   111 = 256 cycles
	 * [ 9: 9] DramEnabled ( DRAM Enabled)
	 *	   When Set, this bit indicates that the DRAM is enabled, this
	 *	   bit is set by hardware after DRAM initialization or on an exit
	 *	   from self refresh. The DRAM controller is intialized after the
	 *	   hardware-controlled initialization process ( initiated by the
	 *	   F2 0x90[DramInit]) completes or when the BIOS-controlled
	 *	   initialization process completes (F2 0x7c(EnDramInit] is
	 *	   written from 1 to 0)
	 * [23:10] Reserved
	 * [31:24] MemClkDis ( Disable the MEMCLK outputs for DRAM channel B,
	 *	   BIOS should set it to reduce the power consumption)
	 *         Bit          F(1207)         M2 Package      S1g1 Package
	 *          0           N/A             MA1_CLK1        N/A
	 *          1           N/A             MA0_CLK1        MA0_CLK1
	 *          2           MA3_CLK         N/A             N/A
	 *          3           MA2_CLK         N/A             N/A
	 *          4           MA1_CLK         MA1_CLK0        N/A
	 *          5           MA0_CLK         MA0_CLK0        MA0_CLK0
	 *          6           N/A             MA1_CLK2        N/A
	 *          7           N/A             MA0_CLK2        MA0_CLK2
	 */
	PCI_ADDR(0, 0x18, 2, 0xa0), 0x00fffc00, 0xff000000,

	/* DRAM Scrub Control Register
	 * F3:0x58
	 * [ 4: 0] DRAM Scrube Rate
	 * [ 7: 5] reserved
	 * [12: 8] L2 Scrub Rate
	 * [15:13] reserved
	 * [20:16] Dcache Scrub
	 * [31:21] reserved
	 *	   Scrub Rates
	 *	   00000 = Do not scrub
	 *	   00001 =  40.00 ns
	 *	   00010 =  80.00 ns
	 *	   00011 = 160.00 ns
	 *	   00100 = 320.00 ns
	 *	   00101 = 640.00 ns
	 *	   00110 =   1.28 us
	 *	   00111 =   2.56 us
	 *	   01000 =   5.12 us
	 *	   01001 =  10.20 us
	 *	   01011 =  41.00 us
	 *	   01100 =  81.90 us
	 *	   01101 = 163.80 us
	 *	   01110 = 327.70 us
	 *	   01111 = 655.40 us
	 *	   10000 =   1.31 ms
	 *	   10001 =   2.62 ms
	 *	   10010 =   5.24 ms
	 *	   10011 =  10.49 ms
	 *	   10100 =  20.97 ms
	 *	   10101 =  42.00 ms
	 *	   10110 =  84.00 ms
	 *	   All Others = Reserved
	 */
	PCI_ADDR(0, 0x18, 3, 0x58), 0xffe0e0e0, 0x00000000,
	/* DRAM Scrub Address Low Register
	 * F3:0x5C
	 * [ 0: 0] DRAM Scrubber Redirect Enable
	 *	   0 = Do nothing
	 *	   1 = Scrubber Corrects errors found in normal operation
	 * [ 5: 1] Reserved
	 * [31: 6] DRAM Scrub Address 31-6
	 */
	PCI_ADDR(0, 0x18, 3, 0x5C), 0x0000003e, 0x00000000,
	/* DRAM Scrub Address High Register
	 * F3:0x60
	 * [ 7: 0] DRAM Scrubb Address 39-32
	 * [31: 8] Reserved
	 */
	PCI_ADDR(0, 0x18, 3, 0x60), 0xffffff00, 0x00000000,
	};
	/* for PCI_ADDR(0, 0x18, 2, 0x98) index,
	 and PCI_ADDR(0x, 0x18, 2, 0x9c) data */
	/*
		index:
		[29: 0] DctOffset (Dram Controller Offset)
		[30:30] DctAccessWrite (Dram Controller Read/Write Select)
			0 = read access
			1 = write access
		[31:31] DctAccessDone (Dram Controller Access Done)
			0 = Access in progress
			1 = No access is progress

		Data:
		[31: 0] DctOffsetData (Dram Controller Offset Data)

		Read:
			- Write the register num to DctOffset with DctAccessWrite = 0
			- poll the DctAccessDone until it = 1
			- Read the data from DctOffsetData
		Write:
			- Write the data to DctOffsetData
			- Write register num to DctOffset with DctAccessWrite = 1
			- poll the DctAccessDone untio it = 1

	*/
	int i;
	int max;

	if (!controller_present(ctrl)) {
		sysinfo->ctrl_present[ctrl->node_id] = 0;
		return;
	}
	sysinfo->ctrl_present[ctrl->node_id] = 1;

	printk(BIOS_SPEW, "setting up CPU %02x northbridge registers\n", ctrl->node_id);
	max = ARRAY_SIZE(register_values);
	for (i = 0; i < max; i += 3) {
		device_t dev;
		unsigned where;
		unsigned long reg;
		dev = (register_values[i] & ~0xff) - PCI_DEV(0, 0x18, 0) + ctrl->f0;
		where = register_values[i] & 0xff;
		reg = pci_read_config32(dev, where);
		reg &= register_values[i+1];
		reg |= register_values[i+2];
		pci_write_config32(dev, where, reg);
	}
	printk(BIOS_SPEW, "done.\n");
}

#if 0
static int is_dual_channel(const struct mem_controller *ctrl)
{
	uint32_t dcl;
	dcl = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
	return dcl & DCL_Width128;
}
#endif

static int is_opteron(const struct mem_controller *ctrl)
{
	/* Test to see if I am an Opteron. M2 and S1G1 support dual
	 * channel, too, but only support unbuffered DIMMs so we need a
	 * better test for Opterons.
	 * However, all code uses is_opteron() to find out whether to
	 * use dual channel, so if we really check for opteron here, we
	 * need to fix up all code using this function, too.
	 */

	uint32_t nbcap;
	nbcap = pci_read_config32(ctrl->f3, NORTHBRIDGE_CAP);
	return !!(nbcap & NBCAP_128Bit);
}

#if 0
static int is_registered(const struct mem_controller *ctrl)
{
	/* Test to see if we are dealing with registered SDRAM.
	 * If we are not registered we are unbuffered.
	 * This function must be called after spd_handle_unbuffered_dimms.
	 */
	uint32_t dcl;
	dcl = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
	return !(dcl & DCL_UnBuffDimm);
}
#endif

static void spd_get_dimm_size(unsigned device, struct dimm_size *sz)
{
	/* Calculate the log base 2 size of a DIMM in bits */
	int value;
	sz->per_rank = 0;
	sz->rows = 0;
	sz->col = 0;
	sz->rank = 0;

	value = spd_read_byte(device, SPD_ROW_NUM);	/* rows */
	if (value < 0) goto hw_err;
	if ((value & 0xff) == 0) goto val_err; /* max is 16 ? */
	sz->per_rank += value & 0xff;
	sz->rows = value & 0xff;

	value = spd_read_byte(device, SPD_COL_NUM);	/* columns */
	if (value < 0) goto hw_err;
	if ((value & 0xff) == 0) goto val_err;  /* max is 11 */
	sz->per_rank += value & 0xff;
	sz->col = value & 0xff;

	value = spd_read_byte(device, SPD_BANK_NUM);	/* banks */
	if (value < 0) goto hw_err;
	if ((value & 0xff) == 0) goto val_err;
	sz->bank = log2(value & 0xff);  // convert 4 to 2, and 8 to 3
	sz->per_rank += sz->bank;

	/* Get the module data width and convert it to a power of two */
	value = spd_read_byte(device, SPD_DATA_WIDTH);
	if (value < 0) goto hw_err;
	value &= 0xff;
	if ((value != 72) && (value != 64)) goto val_err;
	sz->per_rank += log2(value) - 3; //64 bit So another 3 lines

	/* How many ranks? */
	/* number of physical banks */
	value = spd_read_byte(device, SPD_MOD_ATTRIB_RANK);
	if (value < 0) goto hw_err;
/*	value >>= SPD_MOD_ATTRIB_RANK_NUM_SHIFT; */
	value &= SPD_MOD_ATTRIB_RANK_NUM_MASK;
	value += SPD_MOD_ATTRIB_RANK_NUM_BASE; // 0-->1, 1-->2, 3-->4
	/*
	  rank == 1 only one rank or say one side
	  rank == 2 two side , and two ranks
	  rank == 4 two side , and four ranks total
	  Some one side two ranks, because of stacked
	*/
	if ((value != 1) && (value != 2) && (value != 4 )) {
		goto val_err;
	}
	sz->rank = value;

	/* verify if per_rank is equal byte 31
	  it has the DIMM size as a multiple of 128MB.
	  */
	value = spd_read_byte(device, SPD_RANK_SIZE);
	if (value < 0) goto hw_err;
	value &= 0xff;
	value = log2(value);
	if (value <=4 ) value += 8; // add back to 1G to high
	value += (27-5); // make 128MB to the real lines
	if ( value != (sz->per_rank)) {
		printk(BIOS_ERR, "Bad RANK Size --\n");
		goto val_err;
	}

	goto out;

 val_err:
	die("Bad SPD value\n");
	/* If an hw_error occurs report that I have no memory */
 hw_err:
	sz->per_rank = 0;
	sz->rows = 0;
	sz->col = 0;
	sz->bank = 0;
	sz->rank = 0;
 out:
	return;
}


static void set_dimm_size(const struct mem_controller *ctrl,
			  struct dimm_size *sz, unsigned index,
			  struct mem_info *meminfo)
{
	uint32_t base0, base1;

	/* For each base register.
	 * Place the dimm size in 32 MB quantities in the bits 31 - 21.
	 * The initialize dimm size is in bits.
	 * Set the base enable bit0.
	 */

	base0 = base1 = 0;

	/* Make certain side1 of the dimm is at least 128MB */
	if (sz->per_rank >= 27) {
		base0 = (1 << ((sz->per_rank - 27 ) + 19)) | 1;
	}

	/* Make certain side2 of the dimm is at least 128MB */
	if (sz->rank > 1) { // 2 ranks or 4 ranks
		base1 = (1 << ((sz->per_rank - 27 ) + 19)) | 1;
	}

	/* Double the size if we are using dual channel memory */
	if (meminfo->is_Width128) {
		base0 = (base0 << 1) | (base0 & 1);
		base1 = (base1 << 1) | (base1 & 1);
	}

	/* Clear the reserved bits */
	base0 &= ~0xe007fffe;
	base1 &= ~0xe007fffe;

	if (!(meminfo->dimm_mask & 0x0F) && (meminfo->dimm_mask & 0xF0)) { /* channelB only? */
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 4) << 2), base0);
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 5) << 2), base1);
	} else {
		/* Set the appropriate DIMM base address register */
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 0) << 2), base0);
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 1) << 2), base1);
#if QRANK_DIMM_SUPPORT == 1
		if (sz->rank == 4) {
			pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 4) << 2), base0);
			pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 5) << 2), base1);
		}
#endif
	}

	/* Enable the memory clocks for this DIMM by Clear the MemClkDis bit*/
	if (base0) {
		uint32_t dword;
		uint32_t ClkDis0;
#if CONFIG_CPU_SOCKET_TYPE == 0x10 /* L1 */
		ClkDis0 = DTL_MemClkDis0;
#elif CONFIG_CPU_SOCKET_TYPE == 0x11 /* AM2 */
		ClkDis0 = DTL_MemClkDis0_AM2;
#elif CONFIG_CPU_SOCKET_TYPE == 0x12	/* S1G1 */
		ClkDis0 = DTL_MemClkDis0_S1g1;
#endif

		if (!(meminfo->dimm_mask & 0x0F) && (meminfo->dimm_mask & 0xF0)) { /* channelB only? */
			dword = pci_read_config32(ctrl->f2, DRAM_CTRL_MISC);
			dword &= ~(ClkDis0 >> index);
			pci_write_config32(ctrl->f2, DRAM_CTRL_MISC, dword);

		} else {
			dword = pci_read_config32(ctrl->f2, DRAM_TIMING_LOW); //Channel A
			dword &= ~(ClkDis0 >> index);
#if QRANK_DIMM_SUPPORT == 1
			if (sz->rank == 4) {
				dword &= ~(ClkDis0 >> (index+2));
			}
#endif
			pci_write_config32(ctrl->f2, DRAM_TIMING_LOW, dword);

			if (meminfo->is_Width128) { // ChannelA+B
				dword = pci_read_config32(ctrl->f2, DRAM_CTRL_MISC);
				dword &= ~(ClkDis0 >> index);
#if QRANK_DIMM_SUPPORT == 1
				if (sz->rank == 4) {
					dword &= ~(ClkDis0 >> (index+2));
				}
#endif
				pci_write_config32(ctrl->f2, DRAM_CTRL_MISC, dword);
			}
		}

	}
}

/*    row    col   bank  for 64 bit
  0:  13     9      2    :128M
  1:  13     10     2    :256M
  2:  14     10     2    :512M
  3:  13     11     2    :512M
  4:  13     10     3    :512M
  5:  14     10     3    :1G
  6:  14     11     2    :1G
  7:  15     10     3    :2G
  8:  14     11     3    :2G
  9:  15     11     3    :4G
 10:  16     10     3    :4G
 11:  16     11     3    :8G
*/


static void set_dimm_cs_map(const struct mem_controller *ctrl,
			     struct dimm_size *sz, unsigned index,
			     struct mem_info *meminfo)
{
	static const uint8_t cs_map_aaa[24] = {
		/* (bank=2, row=13, col=9)(3, 16, 11) ---> (0, 0, 0) (1, 3, 2) */
	//Bank2
		0, 1, 3,
		0, 2, 6,
		0, 0, 0,
		0, 0, 0,
	//Bank3
		0, 4, 0,
		0, 5, 8,
		0, 7, 9,
		0,10,11,
	};

	uint32_t map;

	if (!(meminfo->dimm_mask & 0x0F) && (meminfo->dimm_mask & 0xF0)) { /* channelB only? */
		index += 2;
	}
	map = pci_read_config32(ctrl->f2, DRAM_BANK_ADDR_MAP);
	map &= ~(0xf << (index * 4));
#if QRANK_DIMM_SUPPORT == 1
	if (sz->rank == 4) {
		map &= ~(0xf << ( (index + 2) * 4));
	}
#endif

	/* Make certain side1 of the dimm is at least 128MB */
	if (sz->per_rank >= 27) {
		unsigned temp_map;
		temp_map = cs_map_aaa[(sz->bank-2)*3*4 + (sz->rows - 13)*3 + (sz->col - 9) ];
		map |= temp_map << (index*4);
#if QRANK_DIMM_SUPPORT == 1
		if (sz->rank == 4) {
			map |=  temp_map << ( (index + 2) * 4);
		}
#endif
	}

	pci_write_config32(ctrl->f2, DRAM_BANK_ADDR_MAP, map);

}


static long spd_set_ram_size(const struct mem_controller *ctrl,
			      struct mem_info *meminfo)
{
	int i;

	for (i = 0; i < DIMM_SOCKETS; i++) {
		struct dimm_size *sz = &(meminfo->sz[i]);
		u32 spd_device = ctrl->channel0[i];

		if (!(meminfo->dimm_mask & (1 << i))) {
			if (meminfo->dimm_mask & (1 << (DIMM_SOCKETS + i))) { /* channelB only? */
				spd_device = ctrl->channel1[i];
			} else {
				continue;
			}
		}

		spd_get_dimm_size(spd_device, sz);
		if (sz->per_rank == 0) {
			return -1; /* Report SPD error */
		}
		set_dimm_size(ctrl, sz, i, meminfo);
		set_dimm_cs_map(ctrl, sz, i, meminfo);
	}
	return meminfo->dimm_mask;
}

static void route_dram_accesses(const struct mem_controller *ctrl,
				unsigned long base_k, unsigned long limit_k)
{
	/* Route the addresses to the controller node */
	unsigned node_id;
	unsigned limit;
	unsigned base;
	unsigned index;
	unsigned limit_reg, base_reg;
	device_t device;

	node_id = ctrl->node_id;
	index = (node_id << 3);
	limit = (limit_k << 2);
	limit &= 0xffff0000;
	limit -= 0x00010000;
	limit |= ( 0 << 8) | (node_id << 0);
	base = (base_k << 2);
	base &= 0xffff0000;
	base |= (0 << 8) | (1<<1) | (1<<0);

	limit_reg = 0x44 + index;
	base_reg = 0x40 + index;
	for (device = PCI_DEV(0, 0x18, 1); device <= PCI_DEV(0, 0x1f, 1);
	     device += PCI_DEV(0, 1, 0)) {
		pci_write_config32(device, limit_reg, limit);
		pci_write_config32(device, base_reg, base);
	}
}

static void set_top_mem(unsigned tom_k, unsigned hole_startk)
{
	/* Error if I don't have memory */
	if (!tom_k) {
		die("No memory?");
	}

	/* Report the amount of memory. */
	printk(BIOS_DEBUG, "RAM end at 0x%08x kB\n", tom_k);

	/* Now set top of memory */
	msr_t msr;
	if (tom_k > (4*1024*1024)) {
		printk_raminit("Handling memory mapped above 4 GB\n");
		printk_raminit("Upper RAM end at 0x%08x kB\n", tom_k);
		msr.lo = (tom_k & 0x003fffff) << 10;
		msr.hi = (tom_k & 0xffc00000) >> 22;
		wrmsr(TOP_MEM2, msr);
		printk_raminit("Correcting memory amount mapped below 4 GB\n");
	}

	/* Leave a 64M hole between TOP_MEM and TOP_MEM2
	 * so I can see my rom chip and other I/O devices.
	 */
	if (tom_k >= 0x003f0000) {
#if CONFIG_HW_MEM_HOLE_SIZEK != 0
		if (hole_startk != 0) {
			tom_k = hole_startk;
		} else
#endif
		tom_k = 0x3f0000;
		printk_raminit("Adjusting lower RAM end\n");
	}
	printk_raminit("Lower RAM end at 0x%08x kB\n", tom_k);
	msr.lo = (tom_k & 0x003fffff) << 10;
	msr.hi = (tom_k & 0xffc00000) >> 22;
	wrmsr(TOP_MEM, msr);
}

static unsigned long interleave_chip_selects(const struct mem_controller *ctrl, int is_Width128)
{
	/* 35 - 27 */

	static const uint8_t csbase_low_f0_shift[] = {
	 /* 128MB */       (14 - (13-5)),
	 /* 256MB */       (15 - (13-5)),
	 /* 512MB */       (15 - (13-5)),
	 /* 512MB */       (16 - (13-5)),
	 /* 512MB */       (16 - (13-5)),
	 /* 1GB   */       (16 - (13-5)),
	 /* 1GB   */       (16 - (13-5)),
	 /* 2GB   */       (16 - (13-5)),
	 /* 2GB   */       (17 - (13-5)),
	 /* 4GB   */       (17 - (13-5)),
	 /* 4GB   */       (16 - (13-5)),
	 /* 8GB   */       (17 - (13-5)),
	};

	/* cs_base_high is not changed */

	uint32_t csbase_inc;
	int chip_selects, index;
	int bits;
	unsigned common_size;
	unsigned common_cs_mode;
	uint32_t csbase, csmask;

	/* See if all of the memory chip selects are the same size
	 * and if so count them.
	 */
	chip_selects = 0;
	common_size = 0;
	common_cs_mode = 0xff;
	for (index = 0; index < 8; index++) {
		unsigned size;
		unsigned cs_mode;
		uint32_t value;

		value = pci_read_config32(ctrl->f2, DRAM_CSBASE + (index << 2));

		/* Is it enabled? */
		if (!(value & 1)) {
			continue;
		}
		chip_selects++;
		size = (value >> 19) & 0x3ff;
		if (common_size == 0) {
			common_size = size;
		}
		/* The size differed fail */
		if (common_size != size) {
			return 0;
		}

		value = pci_read_config32(ctrl->f2, DRAM_BANK_ADDR_MAP);
		cs_mode =( value >> ((index>>1)*4)) & 0xf;
		if (common_cs_mode == 0xff) {
			common_cs_mode = cs_mode;
		}
		/* The cs_mode differed fail */
		if (common_cs_mode != cs_mode) {
			return 0;
		}
	}

	/* Chip selects can only be interleaved when there is
	 * more than one and their is a power of two of them.
	 */
	bits = log2(chip_selects);
	if (((1 << bits) != chip_selects) || (bits < 1) || (bits > 3)) {
		//chip_selects max = 8
		return 0;
	}

	/* Find the bits of csbase that we need to interleave on */
	csbase_inc = 1 << (csbase_low_f0_shift[common_cs_mode]);
	if (is_Width128) {
		csbase_inc <<=1;
	}

	/* Compute the initial values for csbase and csbask.
	 * In csbase just set the enable bit and the base to zero.
	 * In csmask set the mask bits for the size and page level interleave.
	 */
	csbase = 0 | 1;
	csmask = (((common_size  << bits) - 1) << 19);
	csmask |= 0x3fe0 & ~((csbase_inc << bits) - csbase_inc);
	for (index = 0; index < 8; index++) {
		uint32_t value;

		value = pci_read_config32(ctrl->f2, DRAM_CSBASE + (index << 2));
		/* Is it enabled? */
		if (!(value & 1)) {
			continue;
		}
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (index << 2), csbase);
		if ((index & 1) == 0) {  //only have 4 CSMASK
			pci_write_config32(ctrl->f2, DRAM_CSMASK + ((index>>1) << 2), csmask);
		}
		csbase += csbase_inc;
	}

	printk(BIOS_DEBUG, "Interleaved\n");

	/* Return the memory size in K */
	return common_size << ((27-10) + bits);
}

static unsigned long order_chip_selects(const struct mem_controller *ctrl)
{
	unsigned long tom;

	/* Remember which registers we have used in the high 8 bits of tom */
	tom = 0;
	for (;;) {
		/* Find the largest remaining canidate */
		unsigned index, canidate;
		uint32_t csbase, csmask;
		unsigned size;
		csbase = 0;
		canidate = 0;
		for (index = 0; index < 8; index++) {
			uint32_t value;
			value = pci_read_config32(ctrl->f2, DRAM_CSBASE + (index << 2));

			/* Is it enabled? */
			if (!(value & 1)) {
				continue;
			}

			/* Is it greater? */
			if (value <= csbase) {
				continue;
			}

			/* Has it already been selected */
			if (tom & (1 << (index + 24))) {
				continue;
			}
			/* I have a new canidate */
			csbase = value;
			canidate = index;
		}

		/* See if I have found a new canidate */
		if (csbase == 0) {
			break;
		}

		/* Remember the dimm size */
		size = csbase >> 19;

		/* Remember I have used this register */
		tom |= (1 << (canidate + 24));

		/* Recompute the cs base register value */
		csbase = (tom << 19) | 1;

		/* Increment the top of memory */
		tom += size;

		/* Compute the memory mask */
		csmask = ((size -1) << 19);
		csmask |= 0x3fe0;		/* For now don't optimize */

		/* Write the new base register */
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (canidate << 2), csbase);
		/* Write the new mask register */
		if ((canidate & 1) == 0) {  //only have 4 CSMASK
			pci_write_config32(ctrl->f2, DRAM_CSMASK + ((canidate >> 1) << 2), csmask);
		}

	}
	/* Return the memory size in K */
	return (tom & ~0xff000000) << (27-10);
}

static unsigned long memory_end_k(const struct mem_controller *ctrl, int max_node_id)
{
	unsigned node_id;
	unsigned end_k;
	/* Find the last memory address used */
	end_k = 0;
	for (node_id = 0; node_id < max_node_id; node_id++) {
		uint32_t limit, base;
		unsigned index;
		index = node_id << 3;
		base = pci_read_config32(ctrl->f1, 0x40 + index);
		/* Only look at the limit if the base is enabled */
		if ((base & 3) == 3) {
			limit = pci_read_config32(ctrl->f1, 0x44 + index);
			end_k = ((limit + 0x00010000) & 0xffff0000) >> 2;
		}
	}
	return end_k;
}

static void order_dimms(const struct mem_controller *ctrl,
			 struct mem_info *meminfo)
{
	unsigned long tom_k, base_k;

	if (read_option(CMOS_VSTART_interleave_chip_selects,
	    CMOS_VLEN_interleave_chip_selects, 1) != 0) {
		tom_k = interleave_chip_selects(ctrl, meminfo->is_Width128);
	} else {
		printk(BIOS_DEBUG, "Interleaving disabled\n");
		tom_k = 0;
	}

	if (!tom_k) {
		tom_k = order_chip_selects(ctrl);
	}

	/* Compute the memory base address */
	base_k = memory_end_k(ctrl, ctrl->node_id);
	tom_k += base_k;
	route_dram_accesses(ctrl, base_k, tom_k);
	set_top_mem(tom_k, 0);
}

static long disable_dimm(const struct mem_controller *ctrl, unsigned index,
			  struct mem_info *meminfo)
{
	printk(BIOS_DEBUG, "disabling dimm %02x\n", index);
	if (!(meminfo->dimm_mask & 0x0F) && (meminfo->dimm_mask & 0xF0)) { /* channelB only? */
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 4) << 2), 0);
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 5) << 2), 0);
	} else {
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 0) << 2), 0);
		pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 1) << 2), 0);
#if QRANK_DIMM_SUPPORT == 1
		if (meminfo->sz[index].rank == 4) {
			pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 4) << 2), 0);
			pci_write_config32(ctrl->f2, DRAM_CSBASE + (((index << 1) + 5) << 2), 0);
		}
#endif
	}

	meminfo->dimm_mask &= ~(1 << index);
	return meminfo->dimm_mask;
}

static long spd_handle_unbuffered_dimms(const struct mem_controller *ctrl,
					 struct mem_info *meminfo)
{
	int i;
	uint32_t registered;
	uint32_t dcl;
	registered = 0;
	for (i = 0; (i < DIMM_SOCKETS); i++) {
		int value;
		u32 spd_device = ctrl->channel0[i];
		if (!(meminfo->dimm_mask & (1 << i))) {
			if (meminfo->dimm_mask & (1 << (DIMM_SOCKETS + i))) { /* channelB only? */
				spd_device = ctrl->channel1[i];
			} else {
				continue;
			}
		}
		value = spd_read_byte(spd_device, SPD_DIMM_TYPE);
		if (value < 0) {
			return -1;
		}

		/* Registered dimm ? */
		value &= 0x3f;
		if ((value == SPD_DIMM_TYPE_RDIMM) || (value == SPD_DIMM_TYPE_mRDIMM)) {
			//check SPD_MOD_ATTRIB to verify it is SPD_MOD_ATTRIB_REGADC (0x11)?
			registered |= (1<<i);
		}
	}

	if (is_opteron(ctrl)) {
#if 0
		if ( registered != (meminfo->dimm_mask & ((1<<DIMM_SOCKETS)-1)) ) {
			meminfo->dimm_mask &= (registered | (registered << DIMM_SOCKETS) ); //disable unbuffed dimm
//			die("Mixed buffered and registered dimms not supported");
		}
		//By yhlu for debug M2, s1g1 can do dual channel, but it use unbuffer DIMM
		if (!registered) {
			die("Unbuffered Dimms not supported on Opteron");
		}
#endif
	}


	dcl = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
	dcl &= ~DCL_UnBuffDimm;
	meminfo->is_registered = 1;
	if (!registered) {
		dcl |= DCL_UnBuffDimm;
		meminfo->is_registered = 0;
	}
	pci_write_config32(ctrl->f2, DRAM_CONFIG_LOW, dcl);

	if (meminfo->is_registered) {
		printk(BIOS_SPEW, "Registered\n");
	} else {
		printk(BIOS_SPEW, "Unbuffered\n");
	}
	return meminfo->dimm_mask;
}

static unsigned int spd_detect_dimms(const struct mem_controller *ctrl)
{
	unsigned dimm_mask;
	int i;
	dimm_mask = 0;
	for (i = 0; i < DIMM_SOCKETS; i++) {
		int byte;
		unsigned device;
		device = ctrl->channel0[i];
		printk_raminit("DIMM socket %i, channel 0 SPD device is 0x%02x\n", i, device);
		if (device) {
			byte = spd_read_byte(ctrl->channel0[i], SPD_MEM_TYPE);  /* Type */
			if (byte == SPD_MEM_TYPE_SDRAM_DDR2) {
				dimm_mask |= (1 << i);
			}
		}
		device = ctrl->channel1[i];
		printk_raminit("DIMM socket %i, channel 1 SPD device is 0x%02x\n", i, device);
		if (device) {
			byte = spd_read_byte(ctrl->channel1[i], SPD_MEM_TYPE);
			if (byte == SPD_MEM_TYPE_SDRAM_DDR2) {
				dimm_mask |= (1 << (i + DIMM_SOCKETS));
			}
		}
	}
	return dimm_mask;
}

static long spd_enable_2channels(const struct mem_controller *ctrl, struct mem_info *meminfo)
{
	int i;
	uint32_t nbcap;
	/* SPD addresses to verify are identical */
	static const uint8_t addresses[] = {
		2,	/* Type should be DDR2 SDRAM */
		3,	/* *Row addresses */
		4,	/* *Column addresses */
		5,	/* *Number of DIMM Ranks */
		6,	/* *Module Data Width*/
		11,	/* *DIMM Conf Type */
		13,	/* *Pri SDRAM Width */
		17,	/* *Logical Banks */
		20,	/* *DIMM Type Info */
		21,	/* *SDRAM Module Attributes */
		27,	/* *tRP Row precharge time */
		28,	/* *Minimum Row Active to Row Active Delay (tRRD) */
		29,	/* *tRCD RAS to CAS */
		30,	/* *tRAS Activate to Precharge */
		36,	/* *Write recovery time (tWR) */
		37,	/* *Internal write to read command delay (tRDP) */
		38,	/* *Internal read to precharge command delay (tRTP) */
		40,	/* *Extension of Byte 41 tRC and Byte 42 tRFC */
		41,	/* *Minimum Active to Active/Auto Refresh Time(Trc) */
		42,	/* *Minimum Auto Refresh Command Time(Trfc) */
		/* The SPD addresses 18, 9, 23, 26 need special treatment like
		 * in spd_set_memclk. Right now they cause many false negatives.
		 * Keep them at the end to see other mismatches (if any).
		 */
		18,	/* *Supported CAS Latencies */
		9,	/* *Cycle time at highest CAS Latency CL=X */
		23,	/* *Cycle time at CAS Latency (CLX - 1) */
		26,	/* *Cycle time at CAS Latency (CLX - 2) */
	};
	u32 dcl, dcm;
	u8 common_cl;

/* S1G1 and AM2 sockets are Mod64BitMux capable. */
#if CONFIG_CPU_SOCKET_TYPE == 0x11 || CONFIG_CPU_SOCKET_TYPE == 0x12
	u8 mux_cap = 1;
#else
	u8 mux_cap = 0;
#endif

	/* If the dimms are not in pairs do not do dual channels */
	if ((meminfo->dimm_mask & ((1 << DIMM_SOCKETS) - 1)) !=
		((meminfo->dimm_mask >> DIMM_SOCKETS) & ((1 << DIMM_SOCKETS) - 1))) {
		goto single_channel;
	}
	/* If the cpu is not capable of doing dual channels don't do dual channels */
	nbcap = pci_read_config32(ctrl->f3, NORTHBRIDGE_CAP);
	if (!(nbcap & NBCAP_128Bit)) {
		goto single_channel;
	}
	for (i = 0; (i < 4) && (ctrl->channel0[i]); i++) {
		unsigned device0, device1;
		int value0, value1;
		int j;
		/* If I don't have a dimm skip this one */
		if (!(meminfo->dimm_mask & (1 << i))) {
			continue;
		}
		device0 = ctrl->channel0[i];
		device1 = ctrl->channel1[i];
		/* Abort if the chips don't support a common CAS latency. */
		common_cl = spd_read_byte(device0, 18) & spd_read_byte(device1, 18);
		if (!common_cl) {
			printk(BIOS_DEBUG, "No common CAS latency supported\n");
			goto single_channel;
		} else {
			printk_raminit("Common CAS latency bitfield: 0x%02x\n", common_cl);
		}
		for (j = 0; j < ARRAY_SIZE(addresses); j++) {
			unsigned addr;
			addr = addresses[j];
			value0 = spd_read_byte(device0, addr);
			if (value0 < 0) {
				return -1;
			}
			value1 = spd_read_byte(device1, addr);
			if (value1 < 0) {
				return -1;
			}
			if (value0 != value1) {
				printk_raminit("SPD values differ between channel 0/1 for byte %i\n", addr);
				goto single_channel;
			}
		}
	}
	printk(BIOS_SPEW, "Enabling dual channel memory\n");
	dcl = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
	dcl &= ~DCL_BurstLength32;  /*	32byte mode may be preferred in platforms that include graphics controllers that generate a lot of 32-bytes system memory accesses
					32byte mode is not supported when the DRAM interface is 128 bits wides, even 32byte mode is set, system still use 64 byte mode	*/
	dcl |= DCL_Width128;
	pci_write_config32(ctrl->f2, DRAM_CONFIG_LOW, dcl);
	meminfo->is_Width128 = 1;
	return meminfo->dimm_mask;

 single_channel:
	meminfo->is_Width128 = 0;
	meminfo->is_64MuxMode = 0;

	/* single dimm */
	if ((meminfo->dimm_mask & ((1 << DIMM_SOCKETS) - 1)) !=
	   ((meminfo->dimm_mask >> DIMM_SOCKETS) & ((1 << DIMM_SOCKETS) - 1))) {
		if (((meminfo->dimm_mask >> DIMM_SOCKETS) & ((1 << DIMM_SOCKETS) - 1))) {
			/* mux capable and single dimm in channelB */
			if (mux_cap) {
				printk(BIOS_SPEW, "Enable 64MuxMode & BurstLength32\n");
				dcm = pci_read_config32(ctrl->f2, DRAM_CTRL_MISC);
				dcm |= DCM_Mode64BitMux;
				pci_write_config32(ctrl->f2, DRAM_CTRL_MISC, dcm);
				dcl = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
				//dcl |= DCL_BurstLength32; /* 32byte mode for channelB only */
				pci_write_config32(ctrl->f2, DRAM_CONFIG_LOW, dcl);
				meminfo->is_64MuxMode = 1;
			} else {
				meminfo->dimm_mask &= ~((1 << (DIMM_SOCKETS * 2)) - (1 << DIMM_SOCKETS));
			}
		}
	} else { /* unmatched dual dimms ? */
		/* unmatched dual dimms not supported by meminit code. Use single channelA dimm. */
		meminfo->dimm_mask &= ~((1 << (DIMM_SOCKETS * 2)) - (1 << DIMM_SOCKETS));
		printk(BIOS_SPEW, "Unmatched dual dimms. Use single channelA dimm.\n");
	}
	return meminfo->dimm_mask;
}

struct mem_param {
	uint16_t cycle_time;
	uint8_t divisor; /* In 1/40 ns increments */
	uint8_t TrwtTO;
	uint8_t Twrrd;
	uint8_t Twrwr;
	uint8_t Trdrd;
	uint8_t DcqByPassMax;
	uint32_t dch_memclk;
	char name[9];
};

	static const struct mem_param speed[] = {
		{
			.name	    = "200MHz",
			.cycle_time = 0x500,
			.divisor    = 200, // how many 1/40ns per clock
			.dch_memclk = DCH_MemClkFreq_200MHz, //0
			.TrwtTO	    = 7,
			.Twrrd	    = 2,
			.Twrwr	    = 2,
			.Trdrd	    = 3,
			.DcqByPassMax = 4,

		},
		{
			.name	    = "266MHz",
			.cycle_time = 0x375,
			.divisor    = 150, //????
			.dch_memclk = DCH_MemClkFreq_266MHz, //1
			.TrwtTO     = 7,
			.Twrrd      = 2,
			.Twrwr      = 2,
			.Trdrd      = 3,
			.DcqByPassMax = 4,
		},
		 {
			.name       = "333MHz",
			.cycle_time = 0x300,
			.divisor    = 120,
			.dch_memclk = DCH_MemClkFreq_333MHz, //2
			.TrwtTO     = 7,
			.Twrrd      = 2,
			.Twrwr      = 2,
			.Trdrd      = 3,
			.DcqByPassMax = 4,

		 },
		{
			.name	    = "400MHz",
			.cycle_time = 0x250,
			.divisor    = 100,
			.dch_memclk = DCH_MemClkFreq_400MHz,//3
			.TrwtTO     = 7,
			.Twrrd      = 2,
			.Twrwr      = 2,
			.Trdrd      = 3,
			.DcqByPassMax = 4,
		},
		{
			.cycle_time = 0x000,
		},
	};

static const struct mem_param *get_mem_param(unsigned min_cycle_time)
{

	const struct mem_param *param;
	for (param = &speed[0]; param->cycle_time ; param++) {
		if (min_cycle_time > (param+1)->cycle_time) {
			break;
		}
	}
	if (!param->cycle_time) {
		die("min_cycle_time to low");
	}
	printk(BIOS_SPEW, "%s\n", param->name);
	return param;
}

static uint8_t get_exact_divisor(int i, uint8_t divisor)
{
	//input divisor could be 200(200), 150(266), 120(333), 100 (400)
	static const uint8_t dv_a[] = {
	       /* 200  266  333  400 */
	 /*4 */	  250, 250, 250, 250,
	 /*5 */   200, 200, 200, 100,
	 /*6 */   200, 166, 166, 100,
	 /*7 */   200, 171, 142, 100,

	  /*8 */   200, 150, 125, 100,
	  /*9 */   200, 156, 133, 100,
	  /*10*/   200, 160, 120, 100,
	  /*11*/   200, 163, 127, 100,

	  /*12*/   200, 150, 133, 100,
	  /*13*/   200, 153, 123, 100,
	  /*14*/   200, 157, 128, 100,
	  /*15*/   200, 160, 120, 100,
	};


	int index;
	msr_t msr;

	/* Check for FID control support */
	struct cpuid_result cpuid1;
	cpuid1 = cpuid(0x80000007);
	if( cpuid1.edx & 0x02 ) {
		/* Use current FID */
		unsigned fid_cur;
		msr = rdmsr(0xc0010042);
		fid_cur = msr.lo & 0x3f;

		index = fid_cur>>1;
	} else {
		/* Use startup FID */
		unsigned fid_start;
		msr = rdmsr(0xc0010015);
		fid_start = (msr.lo & (0x3f << 24));

		index = fid_start>>25;
	}

	if (index>12) return divisor;

	if (i>3) return divisor;

	return dv_a[index * 4+i];

}


struct spd_set_memclk_result {
	const struct mem_param *param;
	long dimm_mask;
};


static unsigned convert_to_linear(unsigned value)
{
	static const unsigned fraction[] = { 0x25, 0x33, 0x66, 0x75 };
	unsigned valuex;

	/* We need to convert value to more readable */
	if ((value & 0xf) < 10) { //no .25, .33, .66, .75
		value <<= 4;
	} else {
		valuex = ((value & 0xf0) << 4) | fraction [(value & 0xf)-10];
		value = valuex;
	}
	return value;
}

static const uint8_t latency_indicies[] = { 25, 23, 9 };

static int find_optimum_spd_latency(u32 spd_device, unsigned *min_latency, unsigned *min_cycle_time)
{
	int new_cycle_time, new_latency;
	int index;
	int latencies;
	int latency;

	/* First find the supported CAS latencies
	 * Byte 18 for DDR SDRAM is interpreted:
	 * bit 3 == CAS Latency = 3
	 * bit 4 == CAS Latency = 4
	 * bit 5 == CAS Latency = 5
	 * bit 6 == CAS Latency = 6
	 */
	new_cycle_time = 0x500;
	new_latency = 6;

	latencies = spd_read_byte(spd_device, SPD_CAS_LAT);
	if (latencies <= 0)
		return 1;

	printk_raminit("\tlatencies: %08x\n", latencies);
	/* Compute the lowest cas latency which can be expressed in this
	 * particular SPD EEPROM. You can store at most settings for 3
	 * contiguous CAS latencies, so by taking the highest CAS
	 * latency maked as supported in the SPD and subtracting 2 you
	 * get the lowest expressable CAS latency. That latency is not
	 * necessarily supported, but a (maybe invalid) entry exists
	 * for it.
	 */
	latency = log2(latencies) - 2;

	/* Loop through and find a fast clock with a low latency */
	for (index = 0; index < 3; index++, latency++) {
		int value;
		if ((latency < 3) || (latency > 6) ||
			(!(latencies & (1 << latency)))) {
			continue;
		}
		value = spd_read_byte(spd_device, latency_indicies[index]);
		if (value < 0) {
			return -1;
		}

		printk_raminit("\tindex: %08x\n", index);
		printk_raminit("\t\tlatency: %08x\n", latency);
		printk_raminit("\t\tvalue1: %08x\n", value);

		value = convert_to_linear(value);

		printk_raminit("\t\tvalue2: %08x\n", value);

		/* Only increase the latency if we decrease the clock */
		if (value >= *min_cycle_time ) {
			if (value < new_cycle_time) {
				new_cycle_time = value;
				new_latency = latency;
			} else if (value == new_cycle_time) {
				if (new_latency > latency) {
					new_latency = latency;
				}
			}
		}
		printk_raminit("\t\tnew_cycle_time: %08x\n", new_cycle_time);
		printk_raminit("\t\tnew_latency: %08x\n", new_latency);

	}

	if (new_latency > 6){
		return 1;
	}

	/* Does min_latency need to be increased? */
	if (new_cycle_time > *min_cycle_time) {
		*min_cycle_time = new_cycle_time;
	}

	/* Does min_cycle_time need to be increased? */
	if (new_latency > *min_latency) {
		*min_latency = new_latency;
	}

	printk_raminit("2 min_cycle_time: %08x\n", *min_cycle_time);
	printk_raminit("2 min_latency: %08x\n", *min_latency);

	return 0;
}

static struct spd_set_memclk_result spd_set_memclk(const struct mem_controller *ctrl, struct mem_info *meminfo)
{
	/* Compute the minimum cycle time for these dimms */
	struct spd_set_memclk_result result;
	unsigned min_cycle_time, min_latency, bios_cycle_time;
	int i;
	uint32_t value;

	static const uint16_t min_cycle_times[] = { // use full speed to compare
		[NBCAP_MEMCLK_NOLIMIT] = 0x250, /*2.5ns */
		[NBCAP_MEMCLK_333MHZ] = 0x300, /* 3.0ns */
		[NBCAP_MEMCLK_266MHZ] = 0x375, /* 3.75ns */
		[NBCAP_MEMCLK_200MHZ] = 0x500, /* 5.0s */
	};


	value = pci_read_config32(ctrl->f3, NORTHBRIDGE_CAP);
	min_cycle_time = min_cycle_times[(value >> NBCAP_MEMCLK_SHIFT) & NBCAP_MEMCLK_MASK];
	bios_cycle_time = min_cycle_times[
		read_option(CMOS_VSTART_max_mem_clock, CMOS_VLEN_max_mem_clock, 0)];
	if (bios_cycle_time > min_cycle_time) {
		min_cycle_time = bios_cycle_time;
	}
	min_latency = 3;

	printk_raminit("1 min_cycle_time: %08x\n", min_cycle_time);

	/* Compute the least latency with the fastest clock supported
	 * by both the memory controller and the dimms.
	 */
	for (i = 0; i < DIMM_SOCKETS; i++) {
		u32 spd_device;

		printk_raminit("1.1 dimm_mask: %08x\n", meminfo->dimm_mask);
		printk_raminit("i: %08x\n",i);

		if (meminfo->dimm_mask & (1 << i)) {
			spd_device = ctrl->channel0[i];
			printk_raminit("Channel 0 settings:\n");

			switch (find_optimum_spd_latency(spd_device, &min_latency, &min_cycle_time)) {
			case -1:
				goto hw_error;
				break;
			case 1:
				continue;
			}
		}
		if (meminfo->dimm_mask & (1 << (DIMM_SOCKETS + i))) {
			spd_device = ctrl->channel1[i];
			printk_raminit("Channel 1 settings:\n");

			switch (find_optimum_spd_latency(spd_device, &min_latency, &min_cycle_time)) {
			case -1:
				goto hw_error;
				break;
			case 1:
				continue;
			}
		}

	}
	/* Make a second pass through the dimms and disable
	 * any that cannot support the selected memclk and cas latency.
	 */

	printk_raminit("3 min_cycle_time: %08x\n", min_cycle_time);
	printk_raminit("3 min_latency: %08x\n", min_latency);

	for (i = 0; (i < DIMM_SOCKETS); i++) {
		int latencies;
		int latency;
		int index;
		int val;
		u32 spd_device = ctrl->channel0[i];

		if (!(meminfo->dimm_mask & (1 << i))) {
			if (meminfo->dimm_mask & (1 << (DIMM_SOCKETS + i))) { /* channelB only? */
				spd_device = ctrl->channel1[i];
			} else {
				continue;
			}
		}

		latencies = spd_read_byte(spd_device, SPD_CAS_LAT);
		if (latencies < 0) goto hw_error;
		if (latencies == 0) {
			continue;
		}

		/* Compute the lowest cas latency supported */
		latency = log2(latencies) -2;

		/* Walk through searching for the selected latency */
		for (index = 0; index < 3; index++, latency++) {
			if (!(latencies & (1 << latency))) {
				continue;
			}
			if (latency == min_latency)
				break;
		}
		/* If I can't find the latency or my index is bad error */
		if ((latency != min_latency) || (index >= 3)) {
			goto dimm_err;
		}

		/* Read the min_cycle_time for this latency */
		val = spd_read_byte(spd_device, latency_indicies[index]);
		if (val < 0) goto hw_error;

		val = convert_to_linear(val);
		/* All is good if the selected clock speed
		 * is what I need or slower.
		 */
		if (val <= min_cycle_time) {
			continue;
		}
		/* Otherwise I have an error, disable the dimm */
	dimm_err:
		meminfo->dimm_mask = disable_dimm(ctrl, i, meminfo);
	}

	printk_raminit("4 min_cycle_time: %08x\n", min_cycle_time);

	/* Now that I know the minimum cycle time lookup the memory parameters */
	result.param = get_mem_param(min_cycle_time);

	/* Update DRAM Config High with our selected memory speed */
	value = pci_read_config32(ctrl->f2, DRAM_CONFIG_HIGH);
	value &= ~(DCH_MemClkFreq_MASK << DCH_MemClkFreq_SHIFT);

	value |= result.param->dch_memclk << DCH_MemClkFreq_SHIFT;
	pci_write_config32(ctrl->f2, DRAM_CONFIG_HIGH, value);

	printk(BIOS_DEBUG, "%s\n", result.param->name);

	/* Update DRAM Timing Low with our selected cas latency */
	value = pci_read_config32(ctrl->f2, DRAM_TIMING_LOW);
	value &= ~(DTL_TCL_MASK << DTL_TCL_SHIFT);
	value |= (min_latency - DTL_TCL_BASE)  << DTL_TCL_SHIFT;
	pci_write_config32(ctrl->f2, DRAM_TIMING_LOW, value);

	result.dimm_mask = meminfo->dimm_mask;
	return result;
 hw_error:
	result.param = (const struct mem_param *)0;
	result.dimm_mask = -1;
	return result;
}

static unsigned convert_to_1_4(unsigned value)
{
	static const uint8_t fraction[] = { 0, 1, 2, 2, 3, 3, 0 };
	unsigned valuex;

	/* We need to convert value to more readable */
	valuex =  fraction [value & 0x7];
	return valuex;
}

static int get_dimm_Trc_clocks(u32 spd_device, const struct mem_param *param)
{
	int value;
	int value2;
	int clocks;
	value = spd_read_byte(spd_device, SPD_TRC);
	if (value < 0)
		return -1;
	printk_raminit("update_dimm_Trc: tRC (41) = %08x\n", value);

	value2 = spd_read_byte(spd_device, SPD_TRC -1);
	value <<= 2;
	value += convert_to_1_4(value2>>4);

	value *= 10;
	printk_raminit("update_dimm_Trc: tRC final value = %i\n", value);

	clocks = (value + param->divisor - 1)/param->divisor;
	printk_raminit("update_dimm_Trc: clocks = %i\n", clocks);

	if (clocks < DTL_TRC_MIN) {
		// We might want to die here instead or (at least|better) disable this bank.
		printk(BIOS_NOTICE, "update_dimm_Trc: Can't refresh fast enough, "
			"want %i clocks, minimum is %i clocks.\n", clocks, DTL_TRC_MIN);
		clocks = DTL_TRC_MIN;
	}
	return clocks;
}

static int update_dimm_Trc(const struct mem_controller *ctrl,
			    const struct mem_param *param,
			    int i, long dimm_mask)
{
	int clocks, old_clocks;
	uint32_t dtl;
	u32 spd_device = ctrl->channel0[i];

	if (!(dimm_mask & (1 << i)) && (dimm_mask & (1 << (DIMM_SOCKETS + i)))) { /* channelB only? */
		spd_device = ctrl->channel1[i];
	}

	clocks = get_dimm_Trc_clocks(spd_device, param);
	if (clocks == -1)
		return clocks;
	if (clocks > DTL_TRC_MAX) {
		return 0;
	}
	printk_raminit("update_dimm_Trc: clocks after adjustment = %i\n", clocks);

	dtl = pci_read_config32(ctrl->f2, DRAM_TIMING_LOW);
	old_clocks = ((dtl >> DTL_TRC_SHIFT) & DTL_TRC_MASK) + DTL_TRC_BASE;
	if (old_clocks >= clocks) {  //?? someone did it
		// clocks = old_clocks;
		return 1;
	}
	dtl &= ~(DTL_TRC_MASK << DTL_TRC_SHIFT);
	dtl |=	((clocks - DTL_TRC_BASE) << DTL_TRC_SHIFT);
	pci_write_config32(ctrl->f2, DRAM_TIMING_LOW, dtl);
	return 1;
}

static int update_dimm_Trfc(const struct mem_controller *ctrl, const struct mem_param *param, int i, struct mem_info *meminfo)
{
	unsigned clocks, old_clocks;
	uint32_t dth;
	int value;
	u8 ch_b = 0;
	u32 spd_device = ctrl->channel0[i];

	if (!(meminfo->dimm_mask & (1 << i)) && (meminfo->dimm_mask & (1 << (DIMM_SOCKETS + i)))) { /* channelB only? */
		spd_device = ctrl->channel1[i];
		ch_b = 2; /* offset to channelB trfc setting */
	}

	//get the cs_size --> logic dimm size
	value = spd_read_byte(spd_device, SPD_PRI_WIDTH);
	if (value < 0) {
		return -1;
	}

	value = 6 - log2(value); //4-->4, 8-->3, 16-->2

	clocks = meminfo->sz[i].per_rank - 27 + 2 - value;

	dth = pci_read_config32(ctrl->f2, DRAM_TIMING_HIGH);

	old_clocks = ((dth >> (DTH_TRFC0_SHIFT + ((i + ch_b) * 3))) & DTH_TRFC_MASK);

	if (old_clocks >= clocks) { // some one did it?
		return 1;
	}
	dth &= ~(DTH_TRFC_MASK << (DTH_TRFC0_SHIFT + ((i + ch_b) * 3)));
	dth |= clocks  << (DTH_TRFC0_SHIFT + ((i + ch_b) * 3));
	pci_write_config32(ctrl->f2, DRAM_TIMING_HIGH, dth);
	return 1;
}

static int update_dimm_TT_1_4(const struct mem_controller *ctrl, const struct mem_param *param, int i, long dimm_mask,
					unsigned TT_REG,
					unsigned SPD_TT, unsigned TT_SHIFT, unsigned TT_MASK, unsigned TT_BASE, unsigned TT_MIN, unsigned TT_MAX )
{
	unsigned clocks, old_clocks;
	uint32_t dtl;
	int value;
	u32 spd_device = ctrl->channel0[i];

	if (!(dimm_mask & (1 << i)) && (dimm_mask & (1 << (DIMM_SOCKETS + i)))) { /* channelB only? */
		spd_device = ctrl->channel1[i];
	}

	value = spd_read_byte(spd_device, SPD_TT); //already in 1/4 ns
	if (value < 0) return -1;
	value *=10;
	clocks = (value + param->divisor -1)/param->divisor;
	if (clocks < TT_MIN) {
		clocks = TT_MIN;
	}

	if (clocks > TT_MAX) {
		printk(BIOS_INFO, "warning spd byte : %x = %x > TT_MAX: %x, setting TT_MAX", SPD_TT, value, TT_MAX);
		clocks = TT_MAX;
	}

	dtl = pci_read_config32(ctrl->f2, TT_REG);

	old_clocks = ((dtl >> TT_SHIFT) & TT_MASK) + TT_BASE;
	if (old_clocks >= clocks) { //some one did it?
//		clocks = old_clocks;
		return 1;
	}
	dtl &= ~(TT_MASK << TT_SHIFT);
	dtl |= ((clocks - TT_BASE) << TT_SHIFT);
	pci_write_config32(ctrl->f2, TT_REG, dtl);
	return 1;
}

static int update_dimm_Trcd(const struct mem_controller *ctrl,
			     const struct mem_param *param, int i, long dimm_mask)
{
	return update_dimm_TT_1_4(ctrl, param, i, dimm_mask, DRAM_TIMING_LOW, SPD_TRCD, DTL_TRCD_SHIFT, DTL_TRCD_MASK, DTL_TRCD_BASE, DTL_TRCD_MIN, DTL_TRCD_MAX);
}

static int update_dimm_Trrd(const struct mem_controller *ctrl, const struct mem_param *param, int i, long dimm_mask)
{
	return update_dimm_TT_1_4(ctrl, param, i, dimm_mask, DRAM_TIMING_LOW, SPD_TRRD, DTL_TRRD_SHIFT, DTL_TRRD_MASK, DTL_TRRD_BASE, DTL_TRRD_MIN, DTL_TRRD_MAX);
}

static int update_dimm_Tras(const struct mem_controller *ctrl, const struct mem_param *param, int i, long dimm_mask)
{
	unsigned clocks, old_clocks;
	uint32_t dtl;
	int value;
	u32 spd_device = ctrl->channel0[i];

	if (!(dimm_mask & (1 << i)) && (dimm_mask & (1 << (DIMM_SOCKETS + i)))) { /* channelB only? */
		spd_device = ctrl->channel1[i];
	}

	value = spd_read_byte(spd_device, SPD_TRAS); //in 1 ns
	if (value < 0) return -1;
	printk_raminit("update_dimm_Tras: 0 value= %08x\n", value);

	value <<= 2; //convert it to in 1/4ns

	value *= 10;
	printk_raminit("update_dimm_Tras:  1 value= %08x\n", value);

	clocks = (value  + param->divisor - 1)/param->divisor;
	printk_raminit("update_dimm_Tras: divisor= %08x\n", param->divisor);
	printk_raminit("update_dimm_Tras: clocks= %08x\n", clocks);
	if (clocks < DTL_TRAS_MIN) {
		clocks = DTL_TRAS_MIN;
	}
	if (clocks > DTL_TRAS_MAX) {
		return 0;
	}
	dtl = pci_read_config32(ctrl->f2, DRAM_TIMING_LOW);
	old_clocks = ((dtl >> DTL_TRAS_SHIFT) & DTL_TRAS_MASK) + DTL_TRAS_BASE;
	if (old_clocks >= clocks) { // someone did it?
		return 1;
	}
	dtl &= ~(DTL_TRAS_MASK << DTL_TRAS_SHIFT);
	dtl |= ((clocks - DTL_TRAS_BASE) << DTL_TRAS_SHIFT);
	pci_write_config32(ctrl->f2, DRAM_TIMING_LOW, dtl);
	return 1;
}

static int update_dimm_Trp(const struct mem_controller *ctrl,
			    const struct mem_param *param, int i, long dimm_mask)
{
	return update_dimm_TT_1_4(ctrl, param, i, dimm_mask, DRAM_TIMING_LOW, SPD_TRP, DTL_TRP_SHIFT, DTL_TRP_MASK, DTL_TRP_BASE, DTL_TRP_MIN, DTL_TRP_MAX);
}


static int update_dimm_Trtp(const struct mem_controller *ctrl,
		const struct mem_param *param, int i, struct mem_info *meminfo)
{
	/* need to figure if it is 32 byte burst or 64 bytes burst */
	int offset = 2;
	if (!meminfo->is_Width128) {
		uint32_t dword;
		dword = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
		if ((dword &  DCL_BurstLength32)) offset = 0;
	}
	return update_dimm_TT_1_4(ctrl, param, i, meminfo->dimm_mask, DRAM_TIMING_LOW, SPD_TRTP, DTL_TRTP_SHIFT, DTL_TRTP_MASK, DTL_TRTP_BASE+offset, DTL_TRTP_MIN+offset, DTL_TRTP_MAX+offset);
}


static int update_dimm_Twr(const struct mem_controller *ctrl, const struct mem_param *param, int i, long dimm_mask)
{
	return update_dimm_TT_1_4(ctrl, param, i, dimm_mask, DRAM_TIMING_LOW, SPD_TWR, DTL_TWR_SHIFT, DTL_TWR_MASK, DTL_TWR_BASE, DTL_TWR_MIN, DTL_TWR_MAX);
}


static int update_dimm_Tref(const struct mem_controller *ctrl,
			     const struct mem_param *param, int i, long dimm_mask)
{
	uint32_t dth, dth_old;
	int value;
	u32 spd_device = ctrl->channel0[i];

	if (!(dimm_mask & (1 << i)) && (dimm_mask & (1 << (DIMM_SOCKETS + i)))) { /* channelB only? */
		spd_device = ctrl->channel1[i];
	}

	value = spd_read_byte(spd_device, SPD_TREF); // 0: 15.625us, 1: 3.9us 2: 7.8 us....
	if (value < 0) return -1;

	if (value == 1 ) {
		value = 3;
	} else {
		value = 2;
	}

	dth = pci_read_config32(ctrl->f2, DRAM_TIMING_HIGH);

	dth_old = dth;
	dth &= ~(DTH_TREF_MASK << DTH_TREF_SHIFT);
	dth |= (value << DTH_TREF_SHIFT);
	if (dth_old != dth) {
		pci_write_config32(ctrl->f2, DRAM_TIMING_HIGH, dth);
	}
	return 1;
}

static void set_4RankRDimm(const struct mem_controller *ctrl,
			const struct mem_param *param, struct mem_info *meminfo)
{
#if QRANK_DIMM_SUPPORT == 1
	int value;
	int i;
	long dimm_mask = meminfo->dimm_mask;


	if (!(meminfo->is_registered)) return;

	value = 0;

	for (i = 0; i < DIMM_SOCKETS; i++) {
		if (!(dimm_mask & (1 << i))) {
			continue;
		}

		if (meminfo->sz[i].rank == 4) {
			value = 1;
			break;
		}
	}

	if (value == 1) {
		uint32_t dch;
		dch = pci_read_config32(ctrl->f2, DRAM_CONFIG_HIGH);
		dch |= DCH_FourRankRDimm;
		pci_write_config32(ctrl->f2, DRAM_CONFIG_HIGH, dch);
	}
#endif
}

static uint32_t get_extra_dimm_mask(const struct mem_controller *ctrl,
				     struct mem_info *meminfo)
{
	int i;

	uint32_t mask_x4;
	uint32_t mask_x16;
	uint32_t mask_single_rank;
	uint32_t mask_page_1k;
	int value;
#if QRANK_DIMM_SUPPORT == 1
	int rank;
#endif

	long dimm_mask = meminfo->dimm_mask;


	mask_x4 = 0;
	mask_x16 = 0;
	mask_single_rank = 0;
	mask_page_1k = 0;

	for (i = 0; i < DIMM_SOCKETS; i++) {
		u32 spd_device = ctrl->channel0[i];
		if (!(dimm_mask & (1 << i))) {
			if (dimm_mask & (1 << (DIMM_SOCKETS + i))) { /* channelB only? */
				spd_device = ctrl->channel1[i];
			} else {
				continue;
			}
		}

		if (meminfo->sz[i].rank == 1) {
			mask_single_rank |= 1<<i;
		}

		if (meminfo->sz[i].col==10) {
			mask_page_1k |= 1<<i;
		}


		value = spd_read_byte(spd_device, SPD_PRI_WIDTH);

		#if QRANK_DIMM_SUPPORT == 1
			rank = meminfo->sz[i].rank;
		#endif

		if (value==4) {
			mask_x4 |= (1<<i);
			#if QRANK_DIMM_SUPPORT == 1
			if (rank==4) {
				mask_x4 |= 1<<(i+2);
			}
			#endif
		} else if (value==16) {
			mask_x16 |= (1<<i);
			#if QRANK_DIMM_SUPPORT == 1
			 if (rank==4) {
				 mask_x16 |= 1<<(i+2);
			 }
			#endif
		}

	}

	meminfo->x4_mask= mask_x4;
	meminfo->x16_mask = mask_x16;

	meminfo->single_rank_mask = mask_single_rank;
	meminfo->page_1k_mask = mask_page_1k;

	return mask_x4;

}


static void set_dimm_x4(const struct mem_controller *ctrl, const struct mem_param *param, struct mem_info *meminfo)
{
	uint32_t dcl;
	dcl = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
	dcl &= ~(DCL_X4Dimm_MASK<<DCL_X4Dimm_SHIFT);
	dcl |= ((meminfo->x4_mask) & 0xf) << (DCL_X4Dimm_SHIFT);
	pci_write_config32(ctrl->f2, DRAM_CONFIG_LOW, dcl);
}


static int count_ones(uint32_t dimm_mask)
{
	int dimms;
	unsigned index;
	dimms = 0;
	for (index = 0; index < (2 * DIMM_SOCKETS); index++, dimm_mask >>= 1) {
		if (dimm_mask & 1) {
			dimms++;
		}
	}
	return dimms;
}


static void set_DramTerm(const struct mem_controller *ctrl,
			const struct mem_param *param, struct mem_info *meminfo)
{
	uint32_t dcl;
	unsigned odt;
	odt = 1; // 75 ohms

	if (param->divisor == 100) { //DDR2 800
		if (meminfo->is_Width128) {
			if (count_ones(meminfo->dimm_mask & 0x0f)==2) {
				odt = 3;  //50 ohms
			}
		}

	}


#if CONFIG_DIMM_SUPPORT == 0x0204
	odt = 0x2;		/* 150 ohms */
#endif

	dcl = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
	dcl &= ~(DCL_DramTerm_MASK<<DCL_DramTerm_SHIFT);
	dcl |= (odt & DCL_DramTerm_MASK) << (DCL_DramTerm_SHIFT);
	pci_write_config32(ctrl->f2, DRAM_CONFIG_LOW, dcl);
}

static void set_ecc(const struct mem_controller *ctrl,
	const struct mem_param *param, struct mem_info *meminfo)
{
	int i;
	int value;

	uint32_t dcl, nbcap;
	nbcap = pci_read_config32(ctrl->f3, NORTHBRIDGE_CAP);
	dcl = pci_read_config32(ctrl->f2, DRAM_CONFIG_LOW);
	dcl &= ~DCL_DimmEccEn;
	if (nbcap & NBCAP_ECC) {
		dcl |= DCL_DimmEccEn;
	}
	if (read_option(CMOS_VSTART_ECC_memory, CMOS_VLEN_ECC_memory, 1) == 0) {
		dcl &= ~DCL_DimmEccEn;
	}
	pci_write_config32(ctrl->f2, DRAM_CONFIG_LOW, dcl);

	meminfo->is_ecc = 1;
	if (!(dcl & DCL_DimmEccEn)) {
		meminfo->is_ecc = 0;
		return; // already disabled the ECC, so don't need to read SPD any more
	}

	for (i = 0; i < DIMM_SOCKETS; i++) {
		u32 spd_device = ctrl->channel0[i];
		if (!(meminfo->dimm_mask & (1 << i))) {
			if (meminfo->dimm_mask & (1 << (DIMM_SOCKETS + i))) { /* channelB only? */
				spd_device = ctrl->channel1[i];
				printk(BIOS_DEBUG, "set_ecc spd_device: 0x%x\n", spd_device);
			} else {
				continue;
			}
		}

		value = spd_read_byte(ctrl->channel0[i], SPD_DIMM_CONF_TYPE);

		if (!(value & SPD_DIMM_CONF_TYPE_ECC)) {
			dcl &= ~DCL_DimmEccEn;
			pci_write_config32(ctrl->f2, DRAM_CONFIG_LOW, dcl);
			meminfo->is_ecc = 0;
			return;
		}

	}
}


static int update_dimm_Twtr(const struct mem_controller *ctrl,
			     const struct mem_param *param, int i, long dimm_mask)
{
	return update_dimm_TT_1_4(ctrl, param, i, dimm_mask, DRAM_TIMING_HIGH, SPD_TWTR, DTH_TWTR_SHIFT, DTH_TWTR_MASK, DTH_TWTR_BASE, DTH_TWTR_MIN, DTH_TWTR_MAX);
}

static void set_TT(const struct mem_controller *ctrl,
	const struct mem_param *param, unsigned TT_REG, unsigned TT_SHIFT,
	unsigned TT_MASK, unsigned TT_BASE, unsigned TT_MIN, unsigned TT_MAX,
	unsigned val, const char *str)
{
	uint32_t reg;

	if ((val < TT_MIN) || (val > TT_MAX)) {
		printk(BIOS_ERR, str);
		die(" Unknown\n");
	}

	reg = pci_read_config32(ctrl->f2, TT_REG);
	reg &= ~(TT_MASK << TT_SHIFT);
	reg |= ((val - TT_BASE) << TT_SHIFT);
	pci_write_config32(ctrl->f2, TT_REG, reg);
	return;
}


static void set_TrwtTO(const struct mem_controller *ctrl,
			const struct mem_param *param)
{
	set_TT(ctrl, param, DRAM_TIMING_HIGH, DTH_TRWTTO_SHIFT, DTH_TRWTTO_MASK,DTH_TRWTTO_BASE, DTH_TRWTTO_MIN, DTH_TRWTTO_MAX, param->TrwtTO, "TrwtTO");
}


static void set_Twrrd(const struct mem_controller *ctrl, const struct mem_param *param)
{
	set_TT(ctrl, param, DRAM_TIMING_HIGH, DTH_TWRRD_SHIFT, DTH_TWRRD_MASK,DTH_TWRRD_BASE, DTH_TWRRD_MIN, DTH_TWRRD_MAX, param->Twrrd, "Twrrd");
}


static void set_Twrwr(const struct mem_controller *ctrl, const struct mem_param *param)
{
	set_TT(ctrl, param, DRAM_TIMING_HIGH, DTH_TWRWR_SHIFT, DTH_TWRWR_MASK,DTH_TWRWR_BASE, DTH_TWRWR_MIN, DTH_TWRWR_MAX, param->Twrwr, "Twrwr");
}

static void set_Trdrd(const struct mem_controller *ctrl, const struct mem_param *param)
{
	set_TT(ctrl, param, DRAM_TIMING_HIGH, DTH_TRDRD_SHIFT, DTH_TRDRD_MASK,DTH_TRDRD_BASE, DTH_TRDRD_MIN, DTH_TRDRD_MAX, param->Trdrd, "Trdrd");
}

static void set_DcqBypassMax(const struct mem_controller *ctrl, const struct mem_param *param)
{
	set_TT(ctrl, param, DRAM_CONFIG_HIGH, DCH_DcqBypassMax_SHIFT, DCH_DcqBypassMax_MASK,DCH_DcqBypassMax_BASE, DCH_DcqBypassMax_MIN, DCH_DcqBypassMax_MAX, param->DcqByPassMax, "DcqBypassMax"); // value need to be in CMOS
}

static void set_Tfaw(const struct mem_controller *ctrl, const struct mem_param *param, struct mem_info *meminfo)
{
	static const uint8_t faw_1k[] = {8, 10, 13, 14};
	static const uint8_t faw_2k[] = {10, 14, 17, 18};
	unsigned memclkfreq_index;
	unsigned faw;


	memclkfreq_index = param->dch_memclk;

	if (meminfo->page_1k_mask != 0) { //1k page
		faw = faw_1k[memclkfreq_index];
	} else {
		faw = faw_2k[memclkfreq_index];
	}

	set_TT(ctrl, param, DRAM_CONFIG_HIGH, DCH_FourActWindow_SHIFT, DCH_FourActWindow_MASK, DCH_FourActWindow_BASE, DCH_FourActWindow_MIN, DCH_FourActWindow_MAX, faw, "FourActWindow");
}

static void set_max_async_latency(const struct mem_controller *ctrl, const struct mem_param *param)
{
	uint32_t dch;
	unsigned async_lat;


	dch = pci_read_config32(ctrl->f2, DRAM_CONFIG_HIGH);
	dch &= ~(DCH_MaxAsyncLat_MASK << DCH_MaxAsyncLat_SHIFT);

	//FIXME: We need to use Max of DqsRcvEnDelay + 6ns here: After trainning and get that from index reg 0x10, 0x13, 0x16, 0x19, 0x30, 0x33, 0x36, 0x39
	async_lat = 6 + 6;


	dch |= ((async_lat - DCH_MaxAsyncLat_BASE) << DCH_MaxAsyncLat_SHIFT);
	pci_write_config32(ctrl->f2, DRAM_CONFIG_HIGH, dch);
}

#if (CONFIG_DIMM_SUPPORT & 0x0100)==0x0000 /* 2T mode only used for unbuffered DIMM */
static void set_SlowAccessMode(const struct mem_controller *ctrl)
{
	uint32_t dch;

	dch = pci_read_config32(ctrl->f2, DRAM_CONFIG_HIGH);

	dch |= (1<<20);

	pci_write_config32(ctrl->f2, DRAM_CONFIG_HIGH, dch);
}
#endif

/*
	DRAM_OUTPUT_DRV_COMP_CTRL 0, 0x20
	DRAM_ADDR_TIMING_CTRL 04, 0x24
*/
static void set_misc_timing(const struct mem_controller *ctrl, struct mem_info *meminfo)
{
	uint32_t dword;
	uint32_t dwordx;
#if (CONFIG_DIMM_SUPPORT & 0x0100)==0x0000 /* 2T mode only used for unbuffered DIMM */
	unsigned SlowAccessMode = 0;
#endif

#if CONFIG_DIMM_SUPPORT==0x0104   /* DDR2 and REG */
	long dimm_mask = meminfo->dimm_mask & 0x0f;
	/* for REG DIMM */
	dword = 0x00111222;
	dwordx = 0x002f0000;
	switch (meminfo->memclk_set) {
	case DCH_MemClkFreq_266MHz:
		if ( (dimm_mask == 0x03) || (dimm_mask == 0x02) || (dimm_mask == 0x01)) {
			dwordx = 0x002f2700;
		}
		break;
	case DCH_MemClkFreq_333MHz:
		if ( (dimm_mask == 0x03) || (dimm_mask == 0x02) || (dimm_mask == 0x01)) {
			if ((meminfo->single_rank_mask & 0x03)!=0x03) { //any double rank there?
				dwordx = 0x002f2f00;
			}
		}
		break;
	case DCH_MemClkFreq_400MHz:
		dwordx = 0x002f3300;
		break;
	}

#endif

#if CONFIG_DIMM_SUPPORT==0x0204	/* DDR2 and SO-DIMM, S1G1 */
	dword = 0x00111222;
	dwordx = 0x002F2F00;

	switch (meminfo->memclk_set) {
	case DCH_MemClkFreq_200MHz:	/* nothing to be set here */
		break;
	case DCH_MemClkFreq_266MHz:
		if ((meminfo->single_rank_mask == 0)
		    && (meminfo->x4_mask == 0) && (meminfo->x16_mask))
			dwordx = 0x002C2C00;	/* Double rank x8 */
		/* else SRx16, SRx8, DRx16 == 0x002F2F00 */
		break;
	case DCH_MemClkFreq_333MHz:
		if ((meminfo->single_rank_mask == 1)
		   && (meminfo->x16_mask == 1))	/* SR x16 */
			dwordx = 0x00272700;
		else if ((meminfo->x4_mask == 0) && (meminfo->x16_mask == 0)
			 && (meminfo->single_rank_mask == 0)) {	/* DR x8 */
			SlowAccessMode = 1;
			dwordx = 0x00002800;
		} else {	/* SR x8, DR x16 */
			dwordx = 0x002A2A00;
		}
		break;
	case DCH_MemClkFreq_400MHz:
		if ((meminfo->single_rank_mask == 1)
		   && (meminfo->x16_mask == 1))	/* SR x16 */
			dwordx = 0x00292900;
		else if ((meminfo->x4_mask == 0) && (meminfo->x16_mask == 0)
			 && (meminfo->single_rank_mask == 0)) {	/* DR x8 */
			SlowAccessMode = 1;
			dwordx = 0x00002A00;
		} else {	/* SR x8, DR x16 */
			dwordx = 0x002A2A00;
		}
		break;
	}
#endif

#if CONFIG_DIMM_SUPPORT==0x0004  /* DDR2 and unbuffered */
	long dimm_mask = meminfo->dimm_mask & 0x0f;
	/* for UNBUF DIMM */
	dword = 0x00111222;
	dwordx = 0x002f2f00;
	switch (meminfo->memclk_set) {
	case DCH_MemClkFreq_200MHz:
		if (dimm_mask == 0x03) {
			SlowAccessMode = 1;
			dword = 0x00111322;
		}
		break;
	case DCH_MemClkFreq_266MHz:
		if (dimm_mask == 0x03) {
			SlowAccessMode = 1;
			dword = 0x00111322;
			if ((meminfo->x4_mask == 0 ) && (meminfo->x16_mask == 0)) {
				switch (meminfo->single_rank_mask) {
				case 0x03:
					dwordx = 0x00002f00; //x8 single Rank
					break;
				case 0x00:
					dwordx = 0x00342f00; //x8 double Rank
					break;
				default:
					dwordx = 0x00372f00; //x8 single Rank and double Rank mixed
				}
			} else if ((meminfo->x4_mask == 0 ) && (meminfo->x16_mask == 0x01) && (meminfo->single_rank_mask == 0x01)) {
					 dwordx = 0x00382f00; //x8 Double Rank and x16 single Rank mixed
			 } else if ((meminfo->x4_mask == 0 ) && (meminfo->x16_mask == 0x02) && (meminfo->single_rank_mask == 0x02)) {
					 dwordx = 0x00382f00; //x16 single Rank and x8 double Rank mixed
			}

		} else {
			if ((meminfo->x4_mask == 0 ) && (meminfo->x16_mask == 0x00) && ((meminfo->single_rank_mask == 0x01)||(meminfo->single_rank_mask == 0x02)))  { //x8 single rank
				dwordx = 0x002f2f00;
			} else {
				dwordx = 0x002b2f00;
			}
		}
		break;
	case DCH_MemClkFreq_333MHz:
		dwordx = 0x00202220;
		if (dimm_mask == 0x03) {
			SlowAccessMode = 1;
			dword = 0x00111322;
			if ((meminfo->x4_mask == 0 ) && (meminfo->x16_mask == 0)) {
				switch (meminfo->single_rank_mask) {
				case 0x03:
					dwordx = 0x00302220; //x8 single Rank
					break;
				case 0x00:
					dwordx = 0x002b2220; //x8 double Rank
					break;
				default:
					dwordx = 0x002a2220; //x8 single Rank and double Rank mixed
				}
			} else if ((meminfo->x4_mask == 0) && (meminfo->x16_mask == 0x01) && (meminfo->single_rank_mask == 0x01)) {
					dwordx = 0x002c2220; //x8 Double Rank and x16 single Rank mixed
			} else if ((meminfo->x4_mask == 0) && (meminfo->x16_mask == 0x02) && (meminfo->single_rank_mask == 0x02)) {
					dwordx = 0x002c2220; //x16 single Rank and x8 double Rank mixed
			}
		}
		break;
	case DCH_MemClkFreq_400MHz:
		dwordx = 0x00202520;
		SlowAccessMode = 1;
		if (dimm_mask == 0x03) {
			dword = 0x00113322;
		} else {
			dword = 0x00113222;
		}
		break;
	}

	printk_raminit("\tdimm_mask = %08x\n", meminfo->dimm_mask);
	printk_raminit("\tx4_mask = %08x\n", meminfo->x4_mask);
	printk_raminit("\tx16_mask = %08x\n", meminfo->x16_mask);
	printk_raminit("\tsingle_rank_mask = %08x\n", meminfo->single_rank_mask);
	printk_raminit("\tODC = %08x\n", dword);
	printk_raminit("\tAddr Timing= %08x\n", dwordx);
#endif

#if (CONFIG_DIMM_SUPPORT & 0x0100)==0x0000 /* 2T mode only used for unbuffered DIMM */
	if (SlowAccessMode) {
		set_SlowAccessMode(ctrl);
	}
#endif

	if (!(meminfo->dimm_mask & 0x0F) && (meminfo->dimm_mask & 0xF0)) { /* channelB only? */
		/* Program the Output Driver Compensation Control Registers (Function 2:Offset 0x9c, index 0, 0x20) */
		pci_write_config32_index_wait(ctrl->f2, 0x98, 0x20, dword);

		/* Program the Address Timing Control Registers (Function 2:Offset 0x9c, index 4, 0x24) */
		pci_write_config32_index_wait(ctrl->f2, 0x98, 0x24, dwordx);
	} else {
		/* Program the Output Driver Compensation Control Registers (Function 2:Offset 0x9c, index 0, 0x20) */
		pci_write_config32_index_wait(ctrl->f2, 0x98, 0, dword);
		if (meminfo->is_Width128) {
			pci_write_config32_index_wait(ctrl->f2, 0x98, 0x20, dword);
		}

		/* Program the Address Timing Control Registers (Function 2:Offset 0x9c, index 4, 0x24) */
		pci_write_config32_index_wait(ctrl->f2, 0x98, 4, dwordx);
		if (meminfo->is_Width128) {
			pci_write_config32_index_wait(ctrl->f2, 0x98, 0x24, dwordx);
		}
	}
}


static void set_RDqsEn(const struct mem_controller *ctrl,
			const struct mem_param *param, struct mem_info *meminfo)
{
#if CONFIG_CPU_SOCKET_TYPE==0x10
	//only need to set for reg and x8
	uint32_t dch;

	dch = pci_read_config32(ctrl->f2, DRAM_CONFIG_HIGH);

	dch &= ~DCH_RDqsEn;
	if ((!meminfo->x4_mask) && (!meminfo->x16_mask)) {
		dch |= DCH_RDqsEn;
	}

	pci_write_config32(ctrl->f2, DRAM_CONFIG_HIGH, dch);
#endif
}

static void set_idle_cycle_limit(const struct mem_controller *ctrl,
				  const struct mem_param *param)
{
	uint32_t dcm;
	/* AMD says to Hardcode this */
	dcm = pci_read_config32(ctrl->f2, DRAM_CTRL_MISC);
	dcm &= ~(DCM_ILD_lmt_MASK << DCM_ILD_lmt_SHIFT);
	dcm |= DCM_ILD_lmt_16 << DCM_ILD_lmt_SHIFT;
	dcm |= DCM_DCC_EN;
	pci_write_config32(ctrl->f2, DRAM_CTRL_MISC, dcm);
}

static void set_RdWrQByp(const struct mem_controller *ctrl,
			  const struct mem_param *param)
{
	set_TT(ctrl, param, DRAM_CTRL_MISC, DCM_RdWrQByp_SHIFT, DCM_RdWrQByp_MASK,0, 0, 3, 2, "RdWrQByp");
}

static long spd_set_dram_timing(const struct mem_controller *ctrl,
				 const struct mem_param *param,
				 struct mem_info *meminfo)
{
	int i;

	for (i = 0; i < DIMM_SOCKETS; i++) {
		int rc;
		if (!(meminfo->dimm_mask & (1 << i)) &&
		    !(meminfo->dimm_mask & (1 << (DIMM_SOCKETS + i))) ) {
			continue;
		}
		printk_raminit("spd_set_dram_timing dimm socket:  %08x\n", i);
		/* DRAM Timing Low Register */
		printk_raminit("\ttrc\n");
		if ((rc = update_dimm_Trc (ctrl, param, i, meminfo->dimm_mask)) <= 0) goto dimm_err;

		printk_raminit("\ttrcd\n");
		if ((rc = update_dimm_Trcd(ctrl, param, i, meminfo->dimm_mask)) <= 0) goto dimm_err;

		printk_raminit("\ttrrd\n");
		if ((rc = update_dimm_Trrd(ctrl, param, i, meminfo->dimm_mask)) <= 0) goto dimm_err;

		printk_raminit("\ttras\n");
		if ((rc = update_dimm_Tras(ctrl, param, i, meminfo->dimm_mask)) <= 0) goto dimm_err;

		printk_raminit("\ttrp\n");
		if ((rc = update_dimm_Trp (ctrl, param, i, meminfo->dimm_mask)) <= 0) goto dimm_err;

		printk_raminit("\ttrtp\n");
		if ((rc = update_dimm_Trtp(ctrl, param, i, meminfo)) <= 0) goto dimm_err;

		printk_raminit("\ttwr\n");
		if ((rc = update_dimm_Twr (ctrl, param, i, meminfo->dimm_mask)) <= 0) goto dimm_err;

		/* DRAM Timing High Register */
		printk_raminit("\ttref\n");
		if ((rc = update_dimm_Tref(ctrl, param, i, meminfo->dimm_mask)) <= 0) goto dimm_err;

		printk_raminit("\ttwtr\n");
		if ((rc = update_dimm_Twtr(ctrl, param, i, meminfo->dimm_mask)) <= 0) goto dimm_err;

		printk_raminit("\ttrfc\n");
		if ((rc = update_dimm_Trfc(ctrl, param, i, meminfo)) <= 0) goto dimm_err;

		/* DRAM Config Low */

		continue;
	dimm_err:
		printk(BIOS_DEBUG, "spd_set_dram_timing dimm_err!\n");
		if (rc < 0) {
			return -1;
		}
		meminfo->dimm_mask = disable_dimm(ctrl, i, meminfo);
	}

	get_extra_dimm_mask(ctrl, meminfo); // will be used by RDqsEn and dimm_x4
	/* DRAM Timing Low Register */

	/* DRAM Timing High Register */
	set_TrwtTO(ctrl, param);
	set_Twrrd (ctrl, param);
	set_Twrwr (ctrl, param);
	set_Trdrd (ctrl, param);

	set_4RankRDimm(ctrl, param, meminfo);

	/* DRAM Config High */
	set_Tfaw(ctrl, param, meminfo);
	set_DcqBypassMax(ctrl, param);
	set_max_async_latency(ctrl, param);
	set_RDqsEn(ctrl, param, meminfo);

	/* DRAM Config Low */
	set_ecc(ctrl, param, meminfo);
	set_dimm_x4(ctrl, param, meminfo);
	set_DramTerm(ctrl, param, meminfo);

	/* DRAM Control Misc */
	set_idle_cycle_limit(ctrl, param);
	set_RdWrQByp(ctrl, param);

	return meminfo->dimm_mask;
}

static void sdram_set_spd_registers(const struct mem_controller *ctrl,
				     struct sys_info *sysinfo)
{
	struct spd_set_memclk_result result;
	const struct mem_param *param;
	struct mem_param paramx;
	struct mem_info *meminfo;
#if 1
	if (!sysinfo->ctrl_present[ctrl->node_id]) {
		return;
	}
#endif
	meminfo = &sysinfo->meminfo[ctrl->node_id];

	printk(BIOS_DEBUG, "sdram_set_spd_registers: paramx :%p\n", &paramx);

	activate_spd_rom(ctrl);
	meminfo->dimm_mask = spd_detect_dimms(ctrl);

	printk_raminit("sdram_set_spd_registers: dimm_mask=0x%x\n", meminfo->dimm_mask);

	if (!(meminfo->dimm_mask & ((1 << 2*DIMM_SOCKETS) - 1)))
	{
		printk(BIOS_DEBUG, "No memory for this cpu\n");
		return;
	}
	meminfo->dimm_mask = spd_enable_2channels(ctrl, meminfo);
	printk_raminit("spd_enable_2channels: dimm_mask=0x%x\n", meminfo->dimm_mask);
	if (meminfo->dimm_mask == -1)
		goto hw_spd_err;

	meminfo->dimm_mask = spd_set_ram_size(ctrl, meminfo);
	printk_raminit("spd_set_ram_size: dimm_mask=0x%x\n", meminfo->dimm_mask);
	if (meminfo->dimm_mask == -1)
		goto hw_spd_err;

	meminfo->dimm_mask = spd_handle_unbuffered_dimms(ctrl, meminfo);
	printk_raminit("spd_handle_unbuffered_dimms: dimm_mask=0x%x\n", meminfo->dimm_mask);
	if (meminfo->dimm_mask == -1)
		goto hw_spd_err;

	result = spd_set_memclk(ctrl, meminfo);
	param     = result.param;
	meminfo->dimm_mask = result.dimm_mask;
	printk_raminit("spd_set_memclk: dimm_mask=0x%x\n", meminfo->dimm_mask);
	if (meminfo->dimm_mask == -1)
		goto hw_spd_err;

	//store memclk set to sysinfo, incase we need rebuilt param again
	meminfo->memclk_set = param->dch_memclk;

	memcpy(&paramx, param, sizeof(paramx));

	paramx.divisor = get_exact_divisor(param->dch_memclk, paramx.divisor);

	meminfo->dimm_mask = spd_set_dram_timing(ctrl, &paramx, meminfo);
	printk_raminit("spd_set_dram_timing: dimm_mask=0x%x\n", meminfo->dimm_mask);
	if (meminfo->dimm_mask == -1)
		goto hw_spd_err;

	order_dimms(ctrl, meminfo);

	return;
 hw_spd_err:
	/* Unrecoverable error reading SPD data */
	die("Unrecoverable error reading SPD data. No qualified DIMMs?");
	return;
}

#define TIMEOUT_LOOPS 300000

#include "raminit_f_dqs.c"

#if CONFIG_HW_MEM_HOLE_SIZEK != 0
static uint32_t hoist_memory(int controllers, const struct mem_controller *ctrl,unsigned hole_startk, int i)
{
	int ii;
	uint32_t carry_over;
	device_t dev;
	uint32_t base, limit;
	uint32_t basek;
	uint32_t hoist;
	int j;

	carry_over = (4*1024*1024) - hole_startk;

	for (ii=controllers - 1;ii>i;ii--) {
		base  = pci_read_config32(ctrl[0].f1, 0x40 + (ii << 3));
		if ((base & ((1<<1)|(1<<0))) != ((1<<1)|(1<<0))) {
			continue;
		}
		limit = pci_read_config32(ctrl[0].f1, 0x44 + (ii << 3));
		limit += (carry_over << 2 );
		base  += (carry_over << 2 );
		for (j = 0; j < controllers; j++) {
			pci_write_config32(ctrl[j].f1, 0x44 + (ii << 3), limit);
			pci_write_config32(ctrl[j].f1, 0x40 + (ii << 3), base );
		}
	}
	limit = pci_read_config32(ctrl[0].f1, 0x44 + (i << 3));
	limit += (carry_over << 2);
	for (j = 0; j < controllers; j++) {
		pci_write_config32(ctrl[j].f1, 0x44 + (i << 3), limit);
	}
	dev = ctrl[i].f1;
	base  = pci_read_config32(dev, 0x40 + (i << 3));
	basek  = (base & 0xffff0000) >> 2;
	if (basek == hole_startk) {
		//don't need set memhole here, because hole off set will be 0, overflow
		//so need to change base reg instead, new basek will be 4*1024*1024
		base &= 0x0000ffff;
		base |= (4*1024*1024)<<2;
		for (j = 0; j < controllers; j++) {
			pci_write_config32(ctrl[j].f1, 0x40 + (i<<3), base);
		}
	}  else  {
		hoist = /* hole start address */
			((hole_startk << 10) & 0xff000000) +
			/* hole address to memory controller address */
			(((basek + carry_over) >> 6) & 0x0000ff00) +
			/* enable */
			1;
		pci_write_config32(dev, 0xf0, hoist);
	}

	return carry_over;
}

static void set_hw_mem_hole(int controllers, const struct mem_controller *ctrl)
{

	uint32_t hole_startk;
	int i;

	hole_startk = 4*1024*1024 - CONFIG_HW_MEM_HOLE_SIZEK;

	printk_raminit("Handling memory hole at 0x%08x (default)\n", hole_startk);
#if CONFIG_HW_MEM_HOLE_SIZE_AUTO_INC == 1
	/* We need to double check if the hole_startk is valid, if it is equal
	   to basek, we need to decrease it some */
	uint32_t basek_pri;
	for (i=0; i<controllers; i++) {
			uint32_t base;
			unsigned base_k;
			base  = pci_read_config32(ctrl[0].f1, 0x40 + (i << 3));
			if ((base & ((1<<1)|(1<<0))) != ((1<<1)|(1<<0))) {
				continue;
			}
			base_k = (base & 0xffff0000) >> 2;
			if (base_k == hole_startk) {
				/* decrease mem hole startk to make sure it is
				   on middle of previous node */
				hole_startk -= (base_k - basek_pri) >> 1;
				break; //only one hole
			}
			basek_pri = base_k;
	}
	printk_raminit("Handling memory hole at 0x%08x (adjusted)\n", hole_startk);
#endif
	/* find node index that need do set hole */
	for (i=0; i < controllers; i++) {
		uint32_t base, limit;
		unsigned base_k, limit_k;
		base  = pci_read_config32(ctrl[0].f1, 0x40 + (i << 3));
		if ((base & ((1 << 1) | (1 << 0))) != ((1 << 1) | (1 << 0))) {
			continue;
		}
		limit = pci_read_config32(ctrl[0].f1, 0x44 + (i << 3));
		base_k = (base & 0xffff0000) >> 2;
		limit_k = ((limit + 0x00010000) & 0xffff0000) >> 2;
		if ((base_k <= hole_startk) && (limit_k > hole_startk)) {
			unsigned end_k;
			hoist_memory(controllers, ctrl, hole_startk, i);
			end_k = memory_end_k(ctrl, controllers);
			set_top_mem(end_k, hole_startk);
			break; //only one hole
		}
	}

}
#endif

#include "exit_from_self.c"

static void sdram_enable(int controllers, const struct mem_controller *ctrl,
			  struct sys_info *sysinfo)
{
	int i;
#ifdef ACPI_IS_WAKEUP_EARLY
	int suspend = acpi_is_wakeup_early();
#else
	int suspend = 0;
#endif

#if K8_REV_F_SUPPORT_F0_F1_WORKAROUND == 1
	 unsigned cpu_f0_f1[8];
	/* FIXME: How about 32 node machine later? */
	tsc_t tsc, tsc0[8];

	printk(BIOS_DEBUG, "sdram_enable: tsc0[8]: %p", &tsc0[0]);
	uint32_t dword;
#endif

	/* Error if I don't have memory */
	if (memory_end_k(ctrl, controllers) == 0) {
		die("No memory\n");
	}

	/* Before enabling memory start the memory clocks */
	for (i = 0; i < controllers; i++) {
		uint32_t dch;
		if (!sysinfo->ctrl_present[ i ])
			continue;
		dch = pci_read_config32(ctrl[i].f2, DRAM_CONFIG_HIGH);

		/* if no memory installed, disabled the interface */
		if (sysinfo->meminfo[i].dimm_mask==0x00){
			dch |= DCH_DisDramInterface;
			pci_write_config32(ctrl[i].f2, DRAM_CONFIG_HIGH, dch);

		} else {
			dch |= DCH_MemClkFreqVal;
			pci_write_config32(ctrl[i].f2, DRAM_CONFIG_HIGH, dch);
			/* address timing and Output driver comp Control */
			set_misc_timing(ctrl+i, sysinfo->meminfo+i );
		}
	}

	/* We need to wait a minimum of 20 MEMCLKS to enable the InitDram */
	memreset(controllers, ctrl);

	/* lets override the rest of the routine */
	if (suspend) {
		printk(BIOS_DEBUG, "Wakeup!\n");
		exit_from_self(controllers, ctrl, sysinfo);
		printk(BIOS_DEBUG, "Mem running !\n");
		return;
	}

	for (i = 0; i < controllers; i++) {
		uint32_t dcl, dch;
		if (!sysinfo->ctrl_present[ i ])
			continue;
		/* Skip everything if I don't have any memory on this controller */
		dch = pci_read_config32(ctrl[i].f2, DRAM_CONFIG_HIGH);
		if (!(dch & DCH_MemClkFreqVal)) {
			continue;
		}

		/* ChipKill */
		dcl = pci_read_config32(ctrl[i].f2, DRAM_CONFIG_LOW);
		if (dcl & DCL_DimmEccEn) {
			uint32_t mnc;
			printk(BIOS_SPEW, "ECC enabled\n");
			mnc = pci_read_config32(ctrl[i].f3, MCA_NB_CONFIG);
			mnc |= MNC_ECC_EN;
			if (dcl & DCL_Width128) {
				mnc |= MNC_CHIPKILL_EN;
			}
			pci_write_config32(ctrl[i].f3, MCA_NB_CONFIG, mnc);
		}

#if K8_REV_F_SUPPORT_F0_F1_WORKAROUND == 1
		cpu_f0_f1[i] = is_cpu_pre_f2_in_bsp(i);
		if (cpu_f0_f1[i]) {
			//Rev F0/F1 workaround
#if 1
				/* Set the DqsRcvEnTrain bit */
			dword = pci_read_config32(ctrl[i].f2, DRAM_CTRL);
			dword |= DC_DqsRcvEnTrain;
			pci_write_config32(ctrl[i].f2, DRAM_CTRL, dword);
#endif
			tsc0[i] = rdtsc();
		}
#endif

		pci_write_config32(ctrl[i].f2, DRAM_CONFIG_LOW, dcl);
		dcl |= DCL_InitDram;
		pci_write_config32(ctrl[i].f2, DRAM_CONFIG_LOW, dcl);
	}

	for (i = 0; i < controllers; i++) {
		uint32_t dcl, dcm;
		if (!sysinfo->ctrl_present[ i ])
			continue;
		/* Skip everything if I don't have any memory on this controller */
		if (sysinfo->meminfo[i].dimm_mask==0x00) continue;

		printk(BIOS_DEBUG, "Initializing memory: ");
		int loops = 0;
		do {
			dcl = pci_read_config32(ctrl[i].f2, DRAM_CONFIG_LOW);
			loops++;
			if ((loops & 1023) == 0) {
				printk(BIOS_DEBUG, ".");
			}
		} while(((dcl & DCL_InitDram) != 0) && (loops < TIMEOUT_LOOPS));
		if (loops >= TIMEOUT_LOOPS) {
			printk(BIOS_DEBUG, " failed\n");
			continue;
		}

		/* Wait until it is safe to touch memory */
		do {
			dcm = pci_read_config32(ctrl[i].f2, DRAM_CTRL_MISC);
		} while(((dcm & DCM_MemClrStatus) == 0) /* || ((dcm & DCM_DramEnabled) == 0)*/ );

#if K8_REV_F_SUPPORT_F0_F1_WORKAROUND == 1
		if (cpu_f0_f1[i]) {
			tsc= rdtsc();

			print_debug_dqs_tsc("\nbegin tsc0", i, tsc0[i].hi, tsc0[i].lo, 2);
			print_debug_dqs_tsc("end   tsc ", i, tsc.hi, tsc.lo, 2);

			if (tsc.lo<tsc0[i].lo) {
				tsc.hi--;
			}
			tsc.lo -= tsc0[i].lo;
			tsc.hi -= tsc0[i].hi;

			tsc0[i].lo = tsc.lo;
			tsc0[i].hi = tsc.hi;

			print_debug_dqs_tsc("     dtsc0", i, tsc0[i].hi, tsc0[i].lo, 2);
		}
#endif
		printk(BIOS_DEBUG, " done\n");
	}

#if CONFIG_HW_MEM_HOLE_SIZEK != 0
	/* init hw mem hole here */
	/* DramHoleValid bit only can be set after MemClrStatus is set by Hardware */
	set_hw_mem_hole(controllers, ctrl);
#endif

	/* store tom to sysinfo, and it will be used by dqs_timing */
	{
		msr_t msr;
		//[1M, TOM)
		msr = rdmsr(TOP_MEM);
		sysinfo->tom_k = ((msr.hi<<24) | (msr.lo>>8))>>2;

		//[4G, TOM2)
		msr = rdmsr(TOP_MEM2);
		sysinfo->tom2_k = ((msr.hi<<24)| (msr.lo>>8))>>2;
	}

	for (i = 0; i < controllers; i++) {
		sysinfo->mem_trained[i] = 0;

		if (!sysinfo->ctrl_present[ i ])
			continue;

		/* Skip everything if I don't have any memory on this controller */
		if (sysinfo->meminfo[i].dimm_mask==0x00)
			continue;

		sysinfo->mem_trained[i] = 0x80; // mem need to be trained
	}


#if CONFIG_MEM_TRAIN_SEQ ==  0
   #if K8_REV_F_SUPPORT_F0_F1_WORKAROUND == 1
	dqs_timing(controllers, ctrl, tsc0, sysinfo);
   #else
	dqs_timing(controllers, ctrl, sysinfo);
   #endif
#else

#if CONFIG_MEM_TRAIN_SEQ == 2
	/* need to enable mtrr, so dqs training could access the test address  */
	setup_mtrr_dqs(sysinfo->tom_k, sysinfo->tom2_k);
#endif

	for (i = 0; i < controllers; i++) {
		/* Skip everything if I don't have any memory on this controller */
		if (sysinfo->mem_trained[i]!=0x80)
			continue;

		dqs_timing(i, &ctrl[i], sysinfo, 1);

#if CONFIG_MEM_TRAIN_SEQ == 1
		break; // only train the first node with ram
#endif
	}

#if CONFIG_MEM_TRAIN_SEQ == 2
	clear_mtrr_dqs(sysinfo->tom2_k);
#endif

#endif

#if CONFIG_MEM_TRAIN_SEQ != 1
	wait_all_core0_mem_trained(sysinfo);
#endif

}

static void fill_mem_ctrl(int controllers, struct mem_controller *ctrl_a,
			  const uint16_t *spd_addr)
{
	int i;
	int j;
	struct mem_controller *ctrl;
	for (i=0;i<controllers; i++) {
		ctrl = &ctrl_a[i];
		ctrl->node_id = i;
		ctrl->f0 = PCI_DEV(0, 0x18+i, 0);
		ctrl->f1 = PCI_DEV(0, 0x18+i, 1);
		ctrl->f2 = PCI_DEV(0, 0x18+i, 2);
		ctrl->f3 = PCI_DEV(0, 0x18+i, 3);

		if (spd_addr == (void *)0) continue;

		for (j=0;j<DIMM_SOCKETS;j++) {
			ctrl->channel0[j] = spd_addr[(i*2+0)*DIMM_SOCKETS + j];
			ctrl->channel1[j] = spd_addr[(i*2+1)*DIMM_SOCKETS + j];
		}
	}
}