drm/amd/display: DMCU FW loading from PSP

Update the programming sequence to allow DMCU firmware to be loaded by
PSP. This code detects whether the firmware is loaded and does a check
to verify the expected interface version and checks for correct response
from micro controller.

Added registry key method to allow force loading of firmware from kernel
mode driver for test purposes. This is old method of firmware loading
without PSP.

Moved some init sequences into dc/dmcu.

Changed loading sequence to initialize IRAM after firmware completely
loaded. Firmware will now disable features that use IRAM
until initialized.

Signed-off-by: Anthony Koo <anthony.koo@amd.com>
Reviewed-by: Tony Cheng <Tony.Cheng@amd.com>
Acked-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c

@@ -49,9 +49,18 @@
 #define PSR_EXIT 0x21
 #define PSR_SET 0x23
 #define PSR_SET_WAITLOOP 0x31
+#define MCP_INIT_DMCU 0x88
+#define MCP_INIT_IRAM 0x89
+#define MCP_DMCU_VERSION 0x90
 #define MASTER_COMM_CNTL_REG__MASTER_COMM_INTERRUPT_MASK   0x00000001L
 unsigned int cached_wait_loop_number = 0;
 
+static bool dce_dmcu_init(struct dmcu *dmcu)
+{
+	// Do nothing
+	return true;
+}
+
 bool dce_dmcu_load_iram(struct dmcu *dmcu,
 		unsigned int start_offset,
 		const char *src,
@@ -84,7 +93,7 @@ static void dce_get_dmcu_psr_state(struct dmcu *dmcu, uint32_t *psr_state)
 {
 	struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu);
 
-	uint32_t psrStateOffset = 0xf0;
+	uint32_t psr_state_offset = 0xf0;
 
 	/* Enable write access to IRAM */
 	REG_UPDATE(DMCU_RAM_ACCESS_CTRL, IRAM_HOST_ACCESS_EN, 1);
@@ -92,7 +101,7 @@ static void dce_get_dmcu_psr_state(struct dmcu *dmcu, uint32_t *psr_state)
 	REG_WAIT(DCI_MEM_PWR_STATUS, DMCU_IRAM_MEM_PWR_STATE, 0, 2, 10);
 
 	/* Write address to IRAM_RD_ADDR in DMCU_IRAM_RD_CTRL */
-	REG_WRITE(DMCU_IRAM_RD_CTRL, psrStateOffset);
+	REG_WRITE(DMCU_IRAM_RD_CTRL, psr_state_offset);
 
 	/* Read data from IRAM_RD_DATA in DMCU_IRAM_RD_DATA*/
 	*psr_state = REG_READ(DMCU_IRAM_RD_DATA);
@@ -286,7 +295,128 @@ static void dce_get_psr_wait_loop(unsigned int *psr_wait_loop_number)
 }
 
 #if defined(CONFIG_DRM_AMD_DC_DCN1_0)
-bool dcn10_dmcu_load_iram(struct dmcu *dmcu,
+static void dcn10_get_dmcu_state(struct dmcu *dmcu)
+{
+	struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu);
+	uint32_t dmcu_state_offset = 0xf6;
+
+	/* Enable write access to IRAM */
+	REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL,
+			IRAM_HOST_ACCESS_EN, 1,
+			IRAM_RD_ADDR_AUTO_INC, 1);
+
+	REG_WAIT(DMU_MEM_PWR_CNTL, DMCU_IRAM_MEM_PWR_STATE, 0, 2, 10);
+
+	/* Write address to IRAM_RD_ADDR in DMCU_IRAM_RD_CTRL */
+	REG_WRITE(DMCU_IRAM_RD_CTRL, dmcu_state_offset);
+
+	/* Read data from IRAM_RD_DATA in DMCU_IRAM_RD_DATA*/
+	dmcu->dmcu_state = REG_READ(DMCU_IRAM_RD_DATA);
+
+	/* Disable write access to IRAM to allow dynamic sleep state */
+	REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL,
+			IRAM_HOST_ACCESS_EN, 0,
+			IRAM_RD_ADDR_AUTO_INC, 0);
+}
+
+static void dcn10_get_dmcu_version(struct dmcu *dmcu)
+{
+	struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu);
+	uint32_t dmcu_version_offset = 0xf1;
+
+	/* Clear scratch */
+	REG_WRITE(DC_DMCU_SCRATCH, 0);
+
+	/* Enable write access to IRAM */
+	REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL,
+			IRAM_HOST_ACCESS_EN, 1,
+			IRAM_RD_ADDR_AUTO_INC, 1);
+
+	REG_WAIT(DMU_MEM_PWR_CNTL, DMCU_IRAM_MEM_PWR_STATE, 0, 2, 10);
+
+	/* Write address to IRAM_RD_ADDR and read from DATA register */
+	REG_WRITE(DMCU_IRAM_RD_CTRL, dmcu_version_offset);
+	dmcu->dmcu_version.interface_version = REG_READ(DMCU_IRAM_RD_DATA);
+	dmcu->dmcu_version.year = ((REG_READ(DMCU_IRAM_RD_DATA) << 8) |
+						REG_READ(DMCU_IRAM_RD_DATA));
+	dmcu->dmcu_version.month = REG_READ(DMCU_IRAM_RD_DATA);
+	dmcu->dmcu_version.day = REG_READ(DMCU_IRAM_RD_DATA);
+
+	/* Disable write access to IRAM to allow dynamic sleep state */
+	REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL,
+			IRAM_HOST_ACCESS_EN, 0,
+			IRAM_RD_ADDR_AUTO_INC, 0);
+
+	/* Send MCP command message to DMCU to get version reply from FW.
+	 * We expect this version should match the one in IRAM, otherwise
+	 * something is wrong with DMCU and we should fail and disable UC.
+	 */
+	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800);
+
+	/* Set command to get DMCU version from microcontroller */
+	REG_UPDATE(MASTER_COMM_CMD_REG, MASTER_COMM_CMD_REG_BYTE0,
+			MCP_DMCU_VERSION);
+
+	/* Notify microcontroller of new command */
+	REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1);
+
+	/* Ensure command has been executed before continuing */
+	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800);
+
+	/* Somehow version does not match, so fail and return version 0 */
+	if (dmcu->dmcu_version.interface_version != REG_READ(DC_DMCU_SCRATCH))
+		dmcu->dmcu_version.interface_version = 0;
+}
+
+static bool dcn10_dmcu_init(struct dmcu *dmcu)
+{
+	struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu);
+
+	/* DMCU FW should populate the scratch register if running */
+	if (REG_READ(DC_DMCU_SCRATCH) == 0)
+		return false;
+
+	/* Check state is uninitialized */
+	dcn10_get_dmcu_state(dmcu);
+
+	/* If microcontroller is already initialized, do nothing */
+	if (dmcu->dmcu_state == DMCU_RUNNING)
+		return true;
+
+	/* Retrieve and cache the DMCU firmware version. */
+	dcn10_get_dmcu_version(dmcu);
+
+	/* Check interface version to confirm firmware is loaded and running */
+	if (dmcu->dmcu_version.interface_version == 0)
+		return false;
+
+	/* Wait until microcontroller is ready to process interrupt */
+	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800);
+
+	/* Set initialized ramping boundary value */
+	REG_WRITE(MASTER_COMM_DATA_REG1, 0xFFFF);
+
+	/* Set command to initialize microcontroller */
+	REG_UPDATE(MASTER_COMM_CMD_REG, MASTER_COMM_CMD_REG_BYTE0,
+			MCP_INIT_DMCU);
+
+	/* Notify microcontroller of new command */
+	REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1);
+
+	/* Ensure command has been executed before continuing */
+	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800);
+
+	// Check state is initialized
+	dcn10_get_dmcu_state(dmcu);
+
+	// If microcontroller is not in running state, fail
+	if (dmcu->dmcu_state != DMCU_RUNNING)
+		return false;
+
+	return true;
+}
+
+static bool dcn10_dmcu_load_iram(struct dmcu *dmcu,
 		unsigned int start_offset,
 		const char *src,
 		unsigned int bytes)
@@ -294,7 +424,9 @@ bool dcn10_dmcu_load_iram(struct dmcu *dmcu,
 	struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu);
 	unsigned int count = 0;
 
-	REG_UPDATE(DMCU_CTRL, DMCU_ENABLE, 1);
+	/* If microcontroller is not running, do nothing */
+	if (dmcu->dmcu_state != DMCU_RUNNING)
+		return false;
 
 	/* Enable write access to IRAM */
 	REG_UPDATE_2(DMCU_RAM_ACCESS_CTRL,
@@ -313,6 +445,19 @@ bool dcn10_dmcu_load_iram(struct dmcu *dmcu,
 			IRAM_HOST_ACCESS_EN, 0,
 			IRAM_WR_ADDR_AUTO_INC, 0);
 
+	/* Wait until microcontroller is ready to process interrupt */
+	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800);
+
+	/* Set command to signal IRAM is loaded and to initialize IRAM */
+	REG_UPDATE(MASTER_COMM_CMD_REG, MASTER_COMM_CMD_REG_BYTE0,
+			MCP_INIT_IRAM);
+
+	/* Notify microcontroller of new command */
+	REG_UPDATE(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 1);
+
+	/* Ensure command has been executed before continuing */
+	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 100, 800);
+
 	return true;
 }
 
@@ -320,7 +465,11 @@ static void dcn10_get_dmcu_psr_state(struct dmcu *dmcu, uint32_t *psr_state)
 {
 	struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu);
 
-	uint32_t psrStateOffset = 0xf0;
+	uint32_t psr_state_offset = 0xf0;
+
+	/* If microcontroller is not running, do nothing */
+	if (dmcu->dmcu_state != DMCU_RUNNING)
+		return;
 
 	/* Enable write access to IRAM */
 	REG_UPDATE(DMCU_RAM_ACCESS_CTRL, IRAM_HOST_ACCESS_EN, 1);
@@ -328,7 +477,7 @@ static void dcn10_get_dmcu_psr_state(struct dmcu *dmcu, uint32_t *psr_state)
 	REG_WAIT(DMU_MEM_PWR_CNTL, DMCU_IRAM_MEM_PWR_STATE, 0, 2, 10);
 
 	/* Write address to IRAM_RD_ADDR in DMCU_IRAM_RD_CTRL */
-	REG_WRITE(DMCU_IRAM_RD_CTRL, psrStateOffset);
+	REG_WRITE(DMCU_IRAM_RD_CTRL, psr_state_offset);
 
 	/* Read data from IRAM_RD_DATA in DMCU_IRAM_RD_DATA*/
 	*psr_state = REG_READ(DMCU_IRAM_RD_DATA);
@@ -348,6 +497,10 @@ static void dcn10_dmcu_set_psr_enable(struct dmcu *dmcu, bool enable, bool wait)
 	unsigned int retryCount;
 	uint32_t psr_state = 0;
 
+	/* If microcontroller is not running, do nothing */
+	if (dmcu->dmcu_state != DMCU_RUNNING)
+		return;
+
 	/* waitDMCUReadyForCmd */
 	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0,
 				dmcu_wait_reg_ready_interval,
@@ -399,6 +552,10 @@ static void dcn10_dmcu_setup_psr(struct dmcu *dmcu,
 	union dce_dmcu_psr_config_data_reg2 masterCmdData2;
 	union dce_dmcu_psr_config_data_reg3 masterCmdData3;
 
+	/* If microcontroller is not running, do nothing */
+	if (dmcu->dmcu_state != DMCU_RUNNING)
+		return;
+
 	link->link_enc->funcs->psr_program_dp_dphy_fast_training(link->link_enc,
 			psr_context->psrExitLinkTrainingRequired);
 
@@ -505,6 +662,11 @@ static void dcn10_psr_wait_loop(
 {
 	struct dce_dmcu *dmcu_dce = TO_DCE_DMCU(dmcu);
 	union dce_dmcu_psr_config_data_wait_loop_reg1 masterCmdData1;
+
+	/* If microcontroller is not running, do nothing */
+	if (dmcu->dmcu_state != DMCU_RUNNING)
+		return;
+
 	if (wait_loop_number != 0) {
 	/* waitDMCUReadyForCmd */
 	REG_WAIT(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, 0, 1, 10000);
@@ -531,6 +693,7 @@ static void dcn10_get_psr_wait_loop(unsigned int *psr_wait_loop_number)
 #endif
 
 static const struct dmcu_funcs dce_funcs = {
+	.dmcu_init = dce_dmcu_init,
 	.load_iram = dce_dmcu_load_iram,
 	.set_psr_enable = dce_dmcu_set_psr_enable,
 	.setup_psr = dce_dmcu_setup_psr,
@@ -541,6 +704,7 @@ static const struct dmcu_funcs dce_funcs = {
 
 #if defined(CONFIG_DRM_AMD_DC_DCN1_0)
 static const struct dmcu_funcs dcn10_funcs = {
+	.dmcu_init = dcn10_dmcu_init,
 	.load_iram = dcn10_dmcu_load_iram,
 	.set_psr_enable = dcn10_dmcu_set_psr_enable,
 	.setup_psr = dcn10_dmcu_setup_psr,

drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.h

@@ -31,6 +31,7 @@
 
 #define DMCU_COMMON_REG_LIST_DCE_BASE() \
 	SR(DMCU_CTRL), \
+	SR(DMCU_STATUS), \
 	SR(DMCU_RAM_ACCESS_CTRL), \
 	SR(DMCU_IRAM_WR_CTRL), \
 	SR(DMCU_IRAM_WR_DATA), \
@@ -42,7 +43,8 @@
 	SR(DMCU_IRAM_RD_CTRL), \
 	SR(DMCU_IRAM_RD_DATA), \
 	SR(DMCU_INTERRUPT_TO_UC_EN_MASK), \
-	SR(SMU_INTERRUPT_CONTROL)
+	SR(SMU_INTERRUPT_CONTROL), \
+	SR(DC_DMCU_SCRATCH)
 
 #define DMCU_DCE110_COMMON_REG_LIST() \
 	DMCU_COMMON_REG_LIST_DCE_BASE(), \
@@ -58,10 +60,14 @@
 #define DMCU_COMMON_MASK_SH_LIST_DCE_COMMON_BASE(mask_sh) \
 	DMCU_SF(DMCU_CTRL, \
 			DMCU_ENABLE, mask_sh), \
+	DMCU_SF(DMCU_STATUS, \
+			UC_IN_STOP_MODE, mask_sh), \
 	DMCU_SF(DMCU_RAM_ACCESS_CTRL, \
 			IRAM_HOST_ACCESS_EN, mask_sh), \
 	DMCU_SF(DMCU_RAM_ACCESS_CTRL, \
 			IRAM_WR_ADDR_AUTO_INC, mask_sh), \
+	DMCU_SF(DMCU_RAM_ACCESS_CTRL, \
+			IRAM_RD_ADDR_AUTO_INC, mask_sh), \
 	DMCU_SF(MASTER_COMM_CMD_REG, \
 			MASTER_COMM_CMD_REG_BYTE0, mask_sh), \
 	DMCU_SF(MASTER_COMM_CNTL_REG, MASTER_COMM_INTERRUPT, mask_sh), \
@@ -89,7 +95,9 @@
 	type DMCU_IRAM_MEM_PWR_STATE; \
 	type IRAM_HOST_ACCESS_EN; \
 	type IRAM_WR_ADDR_AUTO_INC; \
+	type IRAM_RD_ADDR_AUTO_INC; \
 	type DMCU_ENABLE; \
+	type UC_IN_STOP_MODE; \
 	type MASTER_COMM_CMD_REG_BYTE0; \
 	type MASTER_COMM_INTERRUPT; \
 	type DPHY_RX_FAST_TRAINING_CAPABLE; \
@@ -112,6 +120,7 @@ struct dce_dmcu_mask {
 
 struct dce_dmcu_registers {
 	uint32_t DMCU_CTRL;
+	uint32_t DMCU_STATUS;
 	uint32_t DMCU_RAM_ACCESS_CTRL;
 	uint32_t DCI_MEM_PWR_STATUS;
 	uint32_t DMU_MEM_PWR_CNTL;
@@ -127,6 +136,7 @@ struct dce_dmcu_registers {
 	uint32_t DMCU_IRAM_RD_DATA;
 	uint32_t DMCU_INTERRUPT_TO_UC_EN_MASK;
 	uint32_t SMU_INTERRUPT_CONTROL;
+	uint32_t DC_DMCU_SCRATCH;
 };
 
 struct dce_dmcu {

drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c

@@ -31,6 +31,7 @@
 #include "dce110/dce110_hw_sequencer.h"
 #include "dce/dce_hwseq.h"
 #include "abm.h"
+#include "dmcu.h"
 #include "dcn10/dcn10_timing_generator.h"
 #include "dcn10/dcn10_dpp.h"
 #include "dcn10/dcn10_mpc.h"
@@ -863,6 +864,7 @@ static void dcn10_init_hw(struct dc *dc)
 {
 	int i;
 	struct abm *abm = dc->res_pool->abm;
+	struct dmcu *dmcu = dc->res_pool->dmcu;
 	struct dce_hwseq *hws = dc->hwseq;
 
 	if (IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) {
@@ -925,6 +927,9 @@ static void dcn10_init_hw(struct dc *dc)
 		abm->funcs->abm_init(abm);
 	}
 
+	if (dmcu != NULL)
+		dmcu->funcs->dmcu_init(dmcu);
+
 	/* power AFMT HDMI memory TODO: may move to dis/en output save power*/
 	REG_WRITE(DIO_MEM_PWR_CTRL, 0);
 

drivers/gpu/drm/amd/display/dc/inc/hw/dmcu.h

@@ -27,12 +27,28 @@
 
 #include "dm_services_types.h"
 
+enum dmcu_state {
+	DMCU_NOT_INITIALIZED = 0,
+	DMCU_RUNNING = 1
+};
+
+struct dmcu_version {
+	unsigned int day;
+	unsigned int month;
+	unsigned int year;
+	unsigned int interface_version;
+};
+
 struct dmcu {
 	struct dc_context *ctx;
 	const struct dmcu_funcs *funcs;
+
+	enum dmcu_state dmcu_state;
+	struct dmcu_version dmcu_version;
 };
 
 struct dmcu_funcs {
+	bool (*dmcu_init)(struct dmcu *dmcu);
 	bool (*load_iram)(struct dmcu *dmcu,
 			unsigned int start_offset,
 			const char *src,