T124: perform ram_repair when CPU rail is powered on in coldboot

This patch is to perform software triggered RAM re-repair in
the cold boot path.

"RAM" actually refers to the CPU cache here (yeah, I know, but that's
how the manuals call it). This is some magic hardware thing that must be
done every time after applying power to the main CPU cores or their
cache may have random failures in some very rare cases.

BUG=chrome-os-partner:30430
BRANCH=nyan
TEST=run cold reboot test on nyan.

Original-Signed-off-by: Yen Lin <yelin@nvidia.com>
Original-Change-Id: I87869431e80e7bc66948a7f67f35e5b907993765
Original-Reviewed-on: https://chromium-review.googlesource.com/207362
Original-Tested-by: Yen Lin <yelin@nvidia.com>
Original-Reviewed-by: Tom Warren <twarren@nvidia.com>
Original-Reviewed-by: Julius Werner <jwerner@chromium.org>
Original-Commit-Queue: Yen Lin <yelin@nvidia.com>
(cherry picked from commit d999f5ecc31d90c8dce1dd91533bc34ffd3c03f2)
Signed-off-by: Marc Jones <marc.jones@se-eng.com>

Change-Id: Iaee1d7f9fa8856f26d7ead70eaeeff9d80dbb181
Reviewed-on: http://review.coreboot.org/8415
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>

4 files changed, 35 insertions, 1 deletions

diff --git a/src/soc/nvidia/tegra124/bootblock.c b/src/soc/nvidia/tegra124/bootblock.c
index 59001491ff..aec914532e 100644
--- a/src/soc/nvidia/tegra124/bootblock.c
+++ b/src/soc/nvidia/tegra124/bootblock.c
@@ -79,6 +79,10 @@ void main(void)
 	clock_cpu0_config(entry);
 
 	power_enable_and_ungate_cpu();
+
+	/* Repair ram on cluster0 and cluster1 after CPU is powered on. */
+	ram_repair();
+
 	clock_cpu0_remove_reset();
 
 	clock_halt_avp();
diff --git a/src/soc/nvidia/tegra124/flow.h b/src/soc/nvidia/tegra124/flow.h
index 531278c402..416681ac8f 100644
--- a/src/soc/nvidia/tegra124/flow.h
+++ b/src/soc/nvidia/tegra124/flow.h
@@ -35,8 +35,14 @@ struct flow_ctlr {
 	u32 cpu_pwr_csr;	/* offset 0x38 */
 	u32 mpid;		/* offset 0x3c */
 	u32 ram_repair;		/* offset 0x40 */
+	u32 flow_dbg_sel;	/* offset 0x44 */
+	u32 flow_dbg_cnt0;	/* offset 0x48 */
+	u32 flow_dbg_cnt1;	/* offset 0x4c */
+	u32 flow_dbg_qual;	/* offset 0x50 */
+	u32 flow_ctlr_spare;	/* offset 0x54 */
+	u32 ram_repair_cluster1;/* offset 0x58 */
 };
-check_member(flow_ctlr, ram_repair, 0x40);
+check_member(flow_ctlr, ram_repair_cluster1, 0x58);
 
 enum {
 	FLOW_MODE_SHIFT = 29,
@@ -76,4 +82,10 @@ enum {
 	FLOW_EVENT_JTAG = 1 << 28
 };
 
+/* RAM_REPAIR, 0x40, 0x58 */
+enum {
+	RAM_REPAIR_REQ = 0x1 << 0,
+	RAM_REPAIR_STS = 0x1 << 1,
+};
+
 #endif	/*  _TEGRA124_FLOW_H_ */
diff --git a/src/soc/nvidia/tegra124/power.c b/src/soc/nvidia/tegra124/power.c
index ec44a0fb02..75f04942b9 100644
--- a/src/soc/nvidia/tegra124/power.c
+++ b/src/soc/nvidia/tegra124/power.c
@@ -25,8 +25,10 @@
 
 #include "pmc.h"
 #include "power.h"
+#include "flow.h"
 
 static struct tegra_pmc_regs * const pmc = (void *)TEGRA_PMC_BASE;
+static struct flow_ctlr * const flow = (void *)TEGRA_FLOW_BASE;
 
 static int partition_powered(int id)
 {
@@ -92,3 +94,17 @@ int power_reset_status(void)
 {
 	return read32(&pmc->rst_status) & 0x7;
 }
+
+void ram_repair(void)
+{
+	// Request RAM repair for cluster 0
+	setbits_le32(&flow->ram_repair, RAM_REPAIR_REQ);
+	// Poll for completion
+	while (!(read32(&flow->ram_repair) & RAM_REPAIR_STS))
+		;
+	// Request RAM repair for cluster 1
+	setbits_le32(&flow->ram_repair_cluster1, RAM_REPAIR_REQ);
+	// Poll for completion
+	while (!(read32(&flow->ram_repair_cluster1) & RAM_REPAIR_STS))
+		;
+}
diff --git a/src/soc/nvidia/tegra124/power.h b/src/soc/nvidia/tegra124/power.h
index bce6faf9ec..889dede2dd 100644
--- a/src/soc/nvidia/tegra124/power.h
+++ b/src/soc/nvidia/tegra124/power.h
@@ -35,4 +35,6 @@ enum {
 };
 int power_reset_status(void);
 
+void ram_repair(void);
+
 #endif	/* __SOC_NVIDIA_TEGRA124_POWER_H__ */