Following patch adds resume (exit from self refresh) support for AMD K8 revF

CPUs. It handles both type of erratas on those CPUs.

Signed-off-by: Rudolf Marek <r.marek@assembler.cz>
Acked-by: Peter Stuge <peter@stuge.se>



git-svn-id: svn://svn.coreboot.org/coreboot/trunk@4102 2b7e53f0-3cfb-0310-b3e9-8179ed1497e1

3 files changed, 290 insertions, 2 deletions

diff --git a/src/northbridge/amd/amdk8/exit_from_self.c b/src/northbridge/amd/amdk8/exit_from_self.c
new file mode 100644
index 0000000000..f8c6744ed1
--- /dev/null
+++ b/src/northbridge/amd/amdk8/exit_from_self.c
@@ -0,0 +1,189 @@
+/*
+ * This file is part of the coreboot project.
+ *
+ * Copyright (C) 2009 Rudolf Marek <r.marek@assembler.cz>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License v2 as published by
+ * the Free Software Foundation.
+ *
+ * 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
+ */
+
+void exit_from_self(int controllers, const struct mem_controller *ctrl,
+		    struct sys_info *sysinfo)
+{
+	int i;
+	u32 dcl, dch;
+	u32 pcidev;
+	u8 bitmask;
+	u8 is_post_rev_g;
+	u32 cpuid;
+
+	for (i = 0; i < controllers; i++) {
+		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;
+		}
+
+		cpuid = pci_read_config32(ctrl[i].f3, 0xfc);
+		is_post_rev_g = ((cpuid & 0xfff00) > 0x50f00);
+
+		/* ChipKill */
+		dcl = pci_read_config32(ctrl[i].f2, DRAM_CONFIG_LOW);
+		if (dcl & DCL_DimmEccEn) {
+			u32 mnc;
+			printk_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);
+		}
+
+		printk_debug("before resume errata #%d\n",
+			     (is_post_rev_g) ? 270 : 125);
+		/* 
+		   1. Restore memory controller registers as normal.
+		   2. Set the DisAutoRefresh bit (Dev:2x8C[18]). (270 only)
+		   3. Set the EnDramInit bit (Dev:2x7C[31]), clear all other bits in the same register).
+		   4. Wait at least 750 us.
+		   5. Clear the EnDramInit bit.
+		   6. Clear the DisAutoRefresh bit. (270 only)
+		   7. Read the value of Dev:2x80 and write that value back to Dev:2x80.
+		   8. Set the exit from the self refresh bit (Dev:2x90[1]).
+		   9. Clear the exit from self refresh bit immediately.
+		   Note: Steps 8 and 9 must be executed in a single 64-byte aligned uninterrupted instruction stream.
+		 */
+
+		enable_lapic();
+		init_timer();
+
+		printk_debug("before exit errata - timer enabled\n");
+
+		if (is_post_rev_g) {
+			dcl =
+			    pci_read_config32(ctrl[i].f2,
+					      DRAM_TIMING_HIGH);
+			dcl |= (1 << 18);
+			pci_write_config32(ctrl[i].f2, DRAM_TIMING_HIGH,
+					   dcl);
+		}
+
+		dcl = DI_EnDramInit;
+		pci_write_config32(ctrl[i].f2, DRAM_INIT, dcl);
+
+		udelay(800);
+
+		printk_debug("before exit errata - after mdelay\n");
+
+		dcl = pci_read_config32(ctrl[i].f2, DRAM_INIT);
+		dcl &= ~DI_EnDramInit;
+		pci_write_config32(ctrl[i].f2, DRAM_INIT, dcl);
+
+		if (is_post_rev_g) {
+			dcl =
+			    pci_read_config32(ctrl[i].f2,
+					      DRAM_TIMING_HIGH);
+			dcl &= ~(1 << 18);
+			pci_write_config32(ctrl[i].f2, DRAM_TIMING_HIGH,
+					   dcl);
+		}
+
+		dcl = pci_read_config32(ctrl[i].f2, DRAM_BANK_ADDR_MAP);
+		pci_write_config32(ctrl[i].f2, DRAM_BANK_ADDR_MAP, dcl);
+
+		/* I was unable to do that like: ctrl[i].f2->path.pci.devfn << 8 */
+		pcidev =
+		    0x80000000 | ((((ctrl[i].node_id + 0x18) << 3) | 0x2)
+				  << 8) | 0x90;
+		printk_debug("pcidev is %x\n", pcidev);
+		bitmask = 2;
+		__asm__ __volatile__("pushl %0\n\t"
+				     "movw $0xcf8, %%dx\n\t"
+				     "out %%eax, (%%dx)\n\t"
+				     "movw $0xcfc, %%dx\n\t"
+				     "inl %%dx, %%eax\n\t"
+				     "orb %1, %%al\n\t"
+				     "not %1\n\t"
+				     ".align 64\n\t"
+				     "outl  %%eax, (%%dx) \n\t"
+				     "andb %1, %%al\n\t"
+				     "outl %%eax, (%%dx)\n\t"
+				     "popl %0\n\t"::"a"(pcidev),
+				     "q"(bitmask):"edx");
+	}
+
+	printk_debug("after exit errata\n");
+
+
+	for (i = 0; i < controllers; i++) {
+		u32 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_debug("Exiting memory from self refresh: ");
+		int loops = 0;
+		do {
+			loops++;
+			if ((loops & 1023) == 0) {
+				printk_debug(".");
+			}
+			dcm =
+			    pci_read_config32(ctrl[i].f2, DRAM_CTRL_MISC);
+		} while (((dcm & DCM_MemClrStatus) ==
+			  0) /* || ((dcm & DCM_DramEnabled) == 0) */ );
+
+		if (loops >= TIMEOUT_LOOPS) {
+			printk_debug("timeout with with cntrl[%d]\n", i);
+			continue;
+		}
+
+		printk_debug(" done\n");
+	}
+
+#if 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++) {
+
+		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;
+
+		dqs_restore_MC_NVRAM((ctrl + i)->f2);
+		sysinfo->mem_trained[i] = 1;	// mem was trained
+	}
+}
diff --git a/src/northbridge/amd/amdk8/raminit_f.c b/src/northbridge/amd/amdk8/raminit_f.c
index 2a9cc43c29..097e3b53a5 100644
--- a/src/northbridge/amd/amdk8/raminit_f.c
+++ b/src/northbridge/amd/amdk8/raminit_f.c
@@ -3009,12 +3009,18 @@ static void set_hw_mem_hole(int controllers, const struct mem_controller *ctrl)
 }
 #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? */
@@ -3060,6 +3066,14 @@ static void sdram_enable(int controllers, const struct mem_controller *ctrl,
 	printk_debug("\n");
 #endif
 
+	/* lets override the rest of the routine */
+	if (suspend) {
+		printk_debug("Wakeup!\n");
+		exit_from_self(controllers, ctrl, sysinfo);
+		printk_debug("Mem running !\n");
+		return;
+	}
+
 	for (i = 0; i < controllers; i++) {
 		uint32_t dcl, dch;
 		if (!sysinfo->ctrl_present[ i ])
diff --git a/src/northbridge/amd/amdk8/raminit_f_dqs.c b/src/northbridge/amd/amdk8/raminit_f_dqs.c
index b110480e7c..8154f41423 100644
--- a/src/northbridge/amd/amdk8/raminit_f_dqs.c
+++ b/src/northbridge/amd/amdk8/raminit_f_dqs.c
@@ -1821,10 +1821,94 @@ static void set_sysinfo_in_ram(unsigned val)
 	set_htic_bit(0, val, 9);
 }
 
+#ifdef S3_NVRAM_EARLY
+int s3_save_nvram_early(u32 dword, int size, int  nvram_pos);
+int s3_load_nvram_early(int size, u32 *old_dword, int nvram_pos);
+#else
+int s3_save_nvram_early(u32 dword, int size, int  nvram_pos) {
+}
 
-#if MEM_TRAIN_SEQ == 0
+int s3_load_nvram_early(int size, u32 *old_dword, int nvram_pos) {
+die("No memory NVRAM loader for DQS data! Unable to restore memory state\n");
+}
+#endif
 
+static int save_index_to_pos(unsigned int dev, int size, int index, int nvram_pos) {
+	u32 dword = pci_read_config32_index_wait(dev, 0x98, index);
+
+	return s3_save_nvram_early(dword, size, nvram_pos);
+}
+
+static int load_index_to_pos(unsigned int dev, int size, int index, int nvram_pos) {
+
+	u32 old_dword = pci_read_config32_index_wait(dev, 0x98, index);
+	nvram_pos = s3_load_nvram_early(size, &old_dword, nvram_pos);
+	pci_write_config32_index_wait(dev, 0x98, index, old_dword);
+	return nvram_pos;
+}
+
+static int dqs_load_MC_NVRAM_ch(unsigned int dev, int ch, int pos) {
+	/* 30 bytes per channel */
+	ch *= 0x20;
+	pos = load_index_to_pos(dev, 4, 0x00 + ch, pos);
+	pos = load_index_to_pos(dev, 4, 0x01 + ch, pos);
+	pos = load_index_to_pos(dev, 4, 0x02 + ch, pos);
+	pos = load_index_to_pos(dev, 1, 0x03 + ch, pos);
+	pos = load_index_to_pos(dev, 4, 0x04 + ch, pos);
+	pos = load_index_to_pos(dev, 4, 0x05 + ch, pos);
+	pos = load_index_to_pos(dev, 4, 0x06 + ch, pos);
+	pos = load_index_to_pos(dev, 1, 0x07 + ch, pos);
+	pos = load_index_to_pos(dev, 1, 0x10 + ch, pos);
+	pos = load_index_to_pos(dev, 1, 0x13 + ch, pos);
+	pos = load_index_to_pos(dev, 1, 0x16 + ch, pos);
+	pos = load_index_to_pos(dev, 1, 0x19 + ch, pos);
+	return pos;
+}
+
+static int dqs_save_MC_NVRAM_ch(unsigned int dev, int ch, int pos) {
+	/* 30 bytes per channel */
+	ch *= 0x20;
+	pos = save_index_to_pos(dev, 4, 0x00 + ch, pos);
+	pos = save_index_to_pos(dev, 4, 0x01 + ch, pos);
+	pos = save_index_to_pos(dev, 4, 0x02 + ch, pos);
+	pos = save_index_to_pos(dev, 1, 0x03 + ch, pos);
+	pos = save_index_to_pos(dev, 4, 0x04 + ch, pos);
+	pos = save_index_to_pos(dev, 4, 0x05 + ch, pos);
+	pos = save_index_to_pos(dev, 4, 0x06 + ch, pos);
+	pos = save_index_to_pos(dev, 1, 0x07 + ch, pos);
+	pos = save_index_to_pos(dev, 1, 0x10 + ch, pos);
+	pos = save_index_to_pos(dev, 1, 0x13 + ch, pos);
+	pos = save_index_to_pos(dev, 1, 0x16 + ch, pos);
+	pos = save_index_to_pos(dev, 1, 0x19 + ch, pos);
+	return pos;
+}
+
+static void dqs_save_MC_NVRAM(unsigned int dev) {
+	int pos = 0;
+	u32 reg;
+	printk_debug("DQS SAVE NVRAM: %x\n", dev);
+	pos = dqs_save_MC_NVRAM_ch(dev, 0, pos);
+	pos = dqs_save_MC_NVRAM_ch(dev, 1, pos);
+	/* save the maxasync lat here */
+	reg = pci_read_config32(dev, DRAM_CONFIG_HIGH);
+	pos = s3_save_nvram_early(reg, 4, pos);
+}
+
+static void dqs_restore_MC_NVRAM(unsigned int dev) {
+	int pos = 0;
+	u32 reg;
+
+	printk_debug("DQS RESTORE FROM NVRAM: %x\n", dev);
+	pos = dqs_load_MC_NVRAM_ch(dev, 0, pos);
+	pos = dqs_load_MC_NVRAM_ch(dev, 1, pos);
+	/* load the maxasync lat here */
+	pos = s3_load_nvram_early(4, &reg, pos);
+	reg &= (DCH_MaxAsyncLat_MASK <<DCH_MaxAsyncLat_SHIFT);
+	reg |= pci_read_config32(dev, DRAM_CONFIG_HIGH);
+	pci_write_config32(dev, DRAM_CONFIG_HIGH, reg);
+}
 
+#if MEM_TRAIN_SEQ == 0
 #if K8_REV_F_SUPPORT_F0_F1_WORKAROUND == 1
 static void dqs_timing(int controllers, const struct mem_controller *ctrl, tsc_t *tsc0, struct sys_info *sysinfo)
 #else
@@ -1891,6 +1975,7 @@ static void dqs_timing(int controllers, const struct mem_controller *ctrl, struc
 		if(train_DqsRcvrEn(ctrl+i, 2, sysinfo)) goto out;
 		printk_debug(" done\r\n");
 		sysinfo->mem_trained[i]=1;
+		dqs_save_MC_NVRAM((ctrl+i)->f2);
 	}
 
 out: