Backmerge remote-tracking branch 'drm/drm-next' into drm-misc-next

Some conflicts with ttm_bo->offset removal, but drm-misc-next needs updating to v5.8.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>

1 files changed, 321 insertions, 97 deletions

diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
index c4e71abbdd53..096be97ce9f9 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gmu.c
@@ -2,14 +2,16 @@
 /* Copyright (c) 2017-2019 The Linux Foundation. All rights reserved. */
 
 #include <linux/clk.h>
-#include <linux/dma-mapping.h>
 #include <linux/interconnect.h>
 #include <linux/pm_domain.h>
 #include <linux/pm_opp.h>
 #include <soc/qcom/cmd-db.h>
+#include <drm/drm_gem.h>
 
 #include "a6xx_gpu.h"
 #include "a6xx_gmu.xml.h"
+#include "msm_gem.h"
+#include "msm_mmu.h"
 
 static void a6xx_gmu_fault(struct a6xx_gmu *gmu)
 {
@@ -127,8 +129,6 @@ static void __a6xx_gmu_set_freq(struct a6xx_gmu *gmu, int index)
 	if (ret)
 		dev_err(gmu->dev, "GMU set GPU frequency error: %d\n", ret);
 
-	gmu->freq = gmu->gpu_freqs[index];
-
 	/*
 	 * Eventually we will want to scale the path vote with the frequency but
 	 * for now leave it at max so that the performance is nominal.
@@ -151,8 +151,21 @@ void a6xx_gmu_set_freq(struct msm_gpu *gpu, unsigned long freq)
 			break;
 
 	gmu->current_perf_index = perf_index;
+	gmu->freq = gmu->gpu_freqs[perf_index];
+
+	/*
+	 * This can get called from devfreq while the hardware is idle. Don't
+	 * bring up the power if it isn't already active
+	 */
+	if (pm_runtime_get_if_in_use(gmu->dev) == 0)
+		return;
 
-	__a6xx_gmu_set_freq(gmu, perf_index);
+	if (gmu->legacy)
+		__a6xx_gmu_set_freq(gmu, perf_index);
+	else
+		a6xx_hfi_set_freq(gmu, perf_index);
+
+	pm_runtime_put(gmu->dev);
 }
 
 unsigned long a6xx_gmu_get_freq(struct msm_gpu *gpu)
@@ -196,6 +209,12 @@ static int a6xx_gmu_start(struct a6xx_gmu *gmu)
 	u32 val;
 
 	gmu_write(gmu, REG_A6XX_GMU_CM3_SYSRESET, 1);
+
+	/* Set the log wptr index
+	 * note: downstream saves the value in poweroff and restores it here
+	 */
+	gmu_write(gmu, REG_A6XX_GPU_GMU_CX_GMU_PWR_COL_CP_RESP, 0);
+
 	gmu_write(gmu, REG_A6XX_GMU_CM3_SYSRESET, 0);
 
 	ret = gmu_poll_timeout(gmu, REG_A6XX_GMU_CM3_FW_INIT_RESULT, val,
@@ -232,8 +251,13 @@ int a6xx_gmu_set_oob(struct a6xx_gmu *gmu, enum a6xx_gmu_oob_state state)
 
 	switch (state) {
 	case GMU_OOB_GPU_SET:
-		request = GMU_OOB_GPU_SET_REQUEST;
-		ack = GMU_OOB_GPU_SET_ACK;
+		if (gmu->legacy) {
+			request = GMU_OOB_GPU_SET_REQUEST;
+			ack = GMU_OOB_GPU_SET_ACK;
+		} else {
+			request = GMU_OOB_GPU_SET_REQUEST_NEW;
+			ack = GMU_OOB_GPU_SET_ACK_NEW;
+		}
 		name = "GPU_SET";
 		break;
 	case GMU_OOB_BOOT_SLUMBER:
@@ -272,6 +296,13 @@ int a6xx_gmu_set_oob(struct a6xx_gmu *gmu, enum a6xx_gmu_oob_state state)
 /* Clear a pending OOB state in the GMU */
 void a6xx_gmu_clear_oob(struct a6xx_gmu *gmu, enum a6xx_gmu_oob_state state)
 {
+	if (!gmu->legacy) {
+		WARN_ON(state != GMU_OOB_GPU_SET);
+		gmu_write(gmu, REG_A6XX_GMU_HOST2GMU_INTR_SET,
+			1 << GMU_OOB_GPU_SET_CLEAR_NEW);
+		return;
+	}
+
 	switch (state) {
 	case GMU_OOB_GPU_SET:
 		gmu_write(gmu, REG_A6XX_GMU_HOST2GMU_INTR_SET,
@@ -294,6 +325,9 @@ static int a6xx_sptprac_enable(struct a6xx_gmu *gmu)
 	int ret;
 	u32 val;
 
+	if (!gmu->legacy)
+		return 0;
+
 	gmu_write(gmu, REG_A6XX_GMU_GX_SPTPRAC_POWER_CONTROL, 0x778000);
 
 	ret = gmu_poll_timeout(gmu, REG_A6XX_GMU_SPTPRAC_PWR_CLK_STATUS, val,
@@ -313,6 +347,9 @@ static void a6xx_sptprac_disable(struct a6xx_gmu *gmu)
 	u32 val;
 	int ret;
 
+	if (!gmu->legacy)
+		return;
+
 	/* Make sure retention is on */
 	gmu_rmw(gmu, REG_A6XX_GPU_CC_GX_GDSCR, 0, (1 << 11));
 
@@ -356,6 +393,11 @@ static int a6xx_gmu_notify_slumber(struct a6xx_gmu *gmu)
 	if (gmu->idle_level < GMU_IDLE_STATE_SPTP)
 		a6xx_sptprac_disable(gmu);
 
+	if (!gmu->legacy) {
+		ret = a6xx_hfi_send_prep_slumber(gmu);
+		goto out;
+	}
+
 	/* Tell the GMU to get ready to slumber */
 	gmu_write(gmu, REG_A6XX_GMU_BOOT_SLUMBER_OPTION, 1);
 
@@ -371,6 +413,7 @@ static int a6xx_gmu_notify_slumber(struct a6xx_gmu *gmu)
 		}
 	}
 
+out:
 	/* Put fence into allow mode */
 	gmu_write(gmu, REG_A6XX_GMU_AO_AHB_FENCE_CTRL, 0);
 	return ret;
@@ -392,7 +435,7 @@ static int a6xx_rpmh_start(struct a6xx_gmu *gmu)
 		return ret;
 	}
 
-	ret = gmu_poll_timeout(gmu, REG_A6XX_RSCC_SEQ_BUSY_DRV0, val,
+	ret = gmu_poll_timeout_rscc(gmu, REG_A6XX_RSCC_SEQ_BUSY_DRV0, val,
 		!val, 100, 10000);
 
 	if (ret) {
@@ -418,7 +461,7 @@ static void a6xx_rpmh_stop(struct a6xx_gmu *gmu)
 
 	gmu_write(gmu, REG_A6XX_GMU_RSCC_CONTROL_REQ, 1);
 
-	ret = gmu_poll_timeout(gmu, REG_A6XX_GPU_RSCC_RSC_STATUS0_DRV0,
+	ret = gmu_poll_timeout_rscc(gmu, REG_A6XX_GPU_RSCC_RSC_STATUS0_DRV0,
 		val, val & (1 << 16), 100, 10000);
 	if (ret)
 		DRM_DEV_ERROR(gmu->dev, "Unable to power off the GPU RSC\n");
@@ -441,32 +484,48 @@ static void a6xx_gmu_rpmh_init(struct a6xx_gmu *gmu)
 	struct platform_device *pdev = to_platform_device(gmu->dev);
 	void __iomem *pdcptr = a6xx_gmu_get_mmio(pdev, "gmu_pdc");
 	void __iomem *seqptr = a6xx_gmu_get_mmio(pdev, "gmu_pdc_seq");
+	uint32_t pdc_address_offset;
 
 	if (!pdcptr || !seqptr)
 		goto err;
 
+	if (adreno_is_a618(adreno_gpu) || adreno_is_a640(adreno_gpu))
+		pdc_address_offset = 0x30090;
+	else if (adreno_is_a650(adreno_gpu))
+		pdc_address_offset = 0x300a0;
+	else
+		pdc_address_offset = 0x30080;
+
 	/* Disable SDE clock gating */
-	gmu_write(gmu, REG_A6XX_GPU_RSCC_RSC_STATUS0_DRV0, BIT(24));
+	gmu_write_rscc(gmu, REG_A6XX_GPU_RSCC_RSC_STATUS0_DRV0, BIT(24));
 
 	/* Setup RSC PDC handshake for sleep and wakeup */
-	gmu_write(gmu, REG_A6XX_RSCC_PDC_SLAVE_ID_DRV0, 1);
-	gmu_write(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_DATA, 0);
-	gmu_write(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_ADDR, 0);
-	gmu_write(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_DATA + 2, 0);
-	gmu_write(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_ADDR + 2, 0);
-	gmu_write(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_DATA + 4, 0x80000000);
-	gmu_write(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_ADDR + 4, 0);
-	gmu_write(gmu, REG_A6XX_RSCC_OVERRIDE_START_ADDR, 0);
-	gmu_write(gmu, REG_A6XX_RSCC_PDC_SEQ_START_ADDR, 0x4520);
-	gmu_write(gmu, REG_A6XX_RSCC_PDC_MATCH_VALUE_LO, 0x4510);
-	gmu_write(gmu, REG_A6XX_RSCC_PDC_MATCH_VALUE_HI, 0x4514);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_PDC_SLAVE_ID_DRV0, 1);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_DATA, 0);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_ADDR, 0);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_DATA + 2, 0);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_ADDR + 2, 0);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_DATA + 4, 0x80000000);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_HIDDEN_TCS_CMD0_ADDR + 4, 0);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_OVERRIDE_START_ADDR, 0);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_PDC_SEQ_START_ADDR, 0x4520);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_PDC_MATCH_VALUE_LO, 0x4510);
+	gmu_write_rscc(gmu, REG_A6XX_RSCC_PDC_MATCH_VALUE_HI, 0x4514);
 
 	/* Load RSC sequencer uCode for sleep and wakeup */
-	gmu_write(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0, 0xa7a506a0);
-	gmu_write(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 1, 0xa1e6a6e7);
-	gmu_write(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 2, 0xa2e081e1);
-	gmu_write(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 3, 0xe9a982e2);
-	gmu_write(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 4, 0x0020e8a8);
+	if (adreno_is_a650(adreno_gpu)) {
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0, 0xeaaae5a0);
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 1, 0xe1a1ebab);
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 2, 0xa2e0a581);
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 3, 0xecac82e2);
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 4, 0x0020edad);
+	} else {
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0, 0xa7a506a0);
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 1, 0xa1e6a6e7);
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 2, 0xa2e081e1);
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 3, 0xe9a982e2);
+		gmu_write_rscc(gmu, REG_A6XX_RSCC_SEQ_MEM_0_DRV0 + 4, 0x0020e8a8);
+	}
 
 	/* Load PDC sequencer uCode for power up and power down sequence */
 	pdc_write(seqptr, REG_A6XX_PDC_GPU_SEQ_MEM_0, 0xfebea1e1);
@@ -487,10 +546,7 @@ static void a6xx_gmu_rpmh_init(struct a6xx_gmu *gmu)
 	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS1_CMD0_DATA + 4, 0x0);
 
 	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS1_CMD0_MSGID + 8, 0x10108);
-	if (adreno_is_a618(adreno_gpu))
-		pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS1_CMD0_ADDR + 8, 0x30090);
-	else
-		pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS1_CMD0_ADDR + 8, 0x30080);
+	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS1_CMD0_ADDR + 8, pdc_address_offset);
 	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS1_CMD0_DATA + 8, 0x0);
 
 	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD_ENABLE_BANK, 7);
@@ -502,17 +558,12 @@ static void a6xx_gmu_rpmh_init(struct a6xx_gmu *gmu)
 
 	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_MSGID + 4, 0x10108);
 	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_ADDR + 4, 0x30000);
-	if (adreno_is_a618(adreno_gpu))
+	if (adreno_is_a618(adreno_gpu) || adreno_is_a650(adreno_gpu))
 		pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_DATA + 4, 0x2);
 	else
 		pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_DATA + 4, 0x3);
-
-
 	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_MSGID + 8, 0x10108);
-	if (adreno_is_a618(adreno_gpu))
-		pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_ADDR + 8, 0x30090);
-	else
-		pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_ADDR + 8, 0x30080);
+	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_ADDR + 8, pdc_address_offset);
 	pdc_write(pdcptr, REG_A6XX_PDC_GPU_TCS3_CMD0_DATA + 8, 0x3);
 
 	/* Setup GPU PDC */
@@ -542,6 +593,8 @@ static void a6xx_gmu_power_config(struct a6xx_gmu *gmu)
 {
 	/* Disable GMU WB/RB buffer */
 	gmu_write(gmu, REG_A6XX_GMU_SYS_BUS_CONFIG, 0x1);
+	gmu_write(gmu, REG_A6XX_GMU_ICACHE_CONFIG, 0x1);
+	gmu_write(gmu, REG_A6XX_GMU_DCACHE_CONFIG, 0x1);
 
 	gmu_write(gmu, REG_A6XX_GMU_PWR_COL_INTER_FRAME_CTRL, 0x9c40400);
 
@@ -571,14 +624,95 @@ static void a6xx_gmu_power_config(struct a6xx_gmu *gmu)
 		A6XX_GMU_RPMH_CTRL_GFX_VOTE_ENABLE);
 }
 
+struct block_header {
+	u32 addr;
+	u32 size;
+	u32 type;
+	u32 value;
+	u32 data[];
+};
+
+/* this should be a general kernel helper */
+static int in_range(u32 addr, u32 start, u32 size)
+{
+	return addr >= start && addr < start + size;
+}
+
+static bool fw_block_mem(struct a6xx_gmu_bo *bo, const struct block_header *blk)
+{
+	if (!in_range(blk->addr, bo->iova, bo->size))
+		return false;
+
+	memcpy(bo->virt + blk->addr - bo->iova, blk->data, blk->size);
+	return true;
+}
+
+static int a6xx_gmu_fw_load(struct a6xx_gmu *gmu)
+{
+	struct a6xx_gpu *a6xx_gpu = container_of(gmu, struct a6xx_gpu, gmu);
+	struct adreno_gpu *adreno_gpu = &a6xx_gpu->base;
+	const struct firmware *fw_image = adreno_gpu->fw[ADRENO_FW_GMU];
+	const struct block_header *blk;
+	u32 reg_offset;
+
+	u32 itcm_base = 0x00000000;
+	u32 dtcm_base = 0x00040000;
+
+	if (adreno_is_a650(adreno_gpu))
+		dtcm_base = 0x10004000;
+
+	if (gmu->legacy) {
+		/* Sanity check the size of the firmware that was loaded */
+		if (fw_image->size > 0x8000) {
+			DRM_DEV_ERROR(gmu->dev,
+				"GMU firmware is bigger than the available region\n");
+			return -EINVAL;
+		}
+
+		gmu_write_bulk(gmu, REG_A6XX_GMU_CM3_ITCM_START,
+			       (u32*) fw_image->data, fw_image->size);
+		return 0;
+	}
+
+
+	for (blk = (const struct block_header *) fw_image->data;
+	     (const u8*) blk < fw_image->data + fw_image->size;
+	     blk = (const struct block_header *) &blk->data[blk->size >> 2]) {
+		if (blk->size == 0)
+			continue;
+
+		if (in_range(blk->addr, itcm_base, SZ_16K)) {
+			reg_offset = (blk->addr - itcm_base) >> 2;
+			gmu_write_bulk(gmu,
+				REG_A6XX_GMU_CM3_ITCM_START + reg_offset,
+				blk->data, blk->size);
+		} else if (in_range(blk->addr, dtcm_base, SZ_16K)) {
+			reg_offset = (blk->addr - dtcm_base) >> 2;
+			gmu_write_bulk(gmu,
+				REG_A6XX_GMU_CM3_DTCM_START + reg_offset,
+				blk->data, blk->size);
+		} else if (!fw_block_mem(&gmu->icache, blk) &&
+			   !fw_block_mem(&gmu->dcache, blk) &&
+			   !fw_block_mem(&gmu->dummy, blk)) {
+			DRM_DEV_ERROR(gmu->dev,
+				"failed to match fw block (addr=%.8x size=%d data[0]=%.8x)\n",
+				blk->addr, blk->size, blk->data[0]);
+		}
+	}
+
+	return 0;
+}
+
 static int a6xx_gmu_fw_start(struct a6xx_gmu *gmu, unsigned int state)
 {
 	static bool rpmh_init;
 	struct a6xx_gpu *a6xx_gpu = container_of(gmu, struct a6xx_gpu, gmu);
 	struct adreno_gpu *adreno_gpu = &a6xx_gpu->base;
-	int i, ret;
+	int ret;
 	u32 chipid;
-	u32 *image;
+
+	if (adreno_is_a650(adreno_gpu))
+		gmu_write(gmu, REG_A6XX_GPU_GMU_CX_GMU_CX_FAL_INTF, 1);
 
 	if (state == GMU_WARM_BOOT) {
 		ret = a6xx_rpmh_start(gmu);
@@ -589,13 +723,6 @@ static int a6xx_gmu_fw_start(struct a6xx_gmu *gmu, unsigned int state)
 			"GMU firmware is not loaded\n"))
 			return -ENOENT;
 
-		/* Sanity check the size of the firmware that was loaded */
-		if (adreno_gpu->fw[ADRENO_FW_GMU]->size > 0x8000) {
-			DRM_DEV_ERROR(gmu->dev,
-				"GMU firmware is bigger than the available region\n");
-			return -EINVAL;
-		}
-
 		/* Turn on register retention */
 		gmu_write(gmu, REG_A6XX_GMU_GENERAL_7, 1);
 
@@ -609,18 +736,16 @@ static int a6xx_gmu_fw_start(struct a6xx_gmu *gmu, unsigned int state)
 				return ret;
 		}
 
-		image = (u32 *) adreno_gpu->fw[ADRENO_FW_GMU]->data;
-
-		for (i = 0; i < adreno_gpu->fw[ADRENO_FW_GMU]->size >> 2; i++)
-			gmu_write(gmu, REG_A6XX_GMU_CM3_ITCM_START + i,
-				image[i]);
+		ret = a6xx_gmu_fw_load(gmu);
+		if (ret)
+			return ret;
 	}
 
 	gmu_write(gmu, REG_A6XX_GMU_CM3_FW_INIT_RESULT, 0);
 	gmu_write(gmu, REG_A6XX_GMU_CM3_BOOT_CONFIG, 0x02);
 
 	/* Write the iova of the HFI table */
-	gmu_write(gmu, REG_A6XX_GMU_HFI_QTBL_ADDR, gmu->hfi->iova);
+	gmu_write(gmu, REG_A6XX_GMU_HFI_QTBL_ADDR, gmu->hfi.iova);
 	gmu_write(gmu, REG_A6XX_GMU_HFI_QTBL_INFO, 1);
 
 	gmu_write(gmu, REG_A6XX_GMU_AHB_FENCE_RANGE_0,
@@ -633,6 +758,9 @@ static int a6xx_gmu_fw_start(struct a6xx_gmu *gmu, unsigned int state)
 
 	gmu_write(gmu, REG_A6XX_GMU_HFI_SFR_ADDR, chipid);
 
+	gmu_write(gmu, REG_A6XX_GPU_GMU_CX_GMU_PWR_COL_CP_MSG,
+		  gmu->log.iova | (gmu->log.size / SZ_4K - 1));
+
 	/* Set up the lowest idle level on the GMU */
 	a6xx_gmu_power_config(gmu);
 
@@ -640,9 +768,11 @@ static int a6xx_gmu_fw_start(struct a6xx_gmu *gmu, unsigned int state)
 	if (ret)
 		return ret;
 
-	ret = a6xx_gmu_gfx_rail_on(gmu);
-	if (ret)
-		return ret;
+	if (gmu->legacy) {
+		ret = a6xx_gmu_gfx_rail_on(gmu);
+		if (ret)
+			return ret;
+	}
 
 	/* Enable SPTP_PC if the CPU is responsible for it */
 	if (gmu->idle_level < GMU_IDLE_STATE_SPTP) {
@@ -683,13 +813,13 @@ static void a6xx_gmu_rpmh_off(struct a6xx_gmu *gmu)
 	u32 val;
 
 	/* Make sure there are no outstanding RPMh votes */
-	gmu_poll_timeout(gmu, REG_A6XX_RSCC_TCS0_DRV0_STATUS, val,
+	gmu_poll_timeout_rscc(gmu, REG_A6XX_RSCC_TCS0_DRV0_STATUS, val,
 		(val & 1), 100, 10000);
-	gmu_poll_timeout(gmu, REG_A6XX_RSCC_TCS1_DRV0_STATUS, val,
+	gmu_poll_timeout_rscc(gmu, REG_A6XX_RSCC_TCS1_DRV0_STATUS, val,
 		(val & 1), 100, 10000);
-	gmu_poll_timeout(gmu, REG_A6XX_RSCC_TCS2_DRV0_STATUS, val,
+	gmu_poll_timeout_rscc(gmu, REG_A6XX_RSCC_TCS2_DRV0_STATUS, val,
 		(val & 1), 100, 10000);
-	gmu_poll_timeout(gmu, REG_A6XX_RSCC_TCS3_DRV0_STATUS, val,
+	gmu_poll_timeout_rscc(gmu, REG_A6XX_RSCC_TCS3_DRV0_STATUS, val,
 		(val & 1), 100, 1000);
 }
 
@@ -744,6 +874,13 @@ int a6xx_gmu_resume(struct a6xx_gpu *a6xx_gpu)
 	status = gmu_read(gmu, REG_A6XX_GMU_GENERAL_7) == 1 ?
 		GMU_WARM_BOOT : GMU_COLD_BOOT;
 
+	/*
+	 * Warm boot path does not work on newer GPUs
+	 * Presumably this is because icache/dcache regions must be restored
+	 */
+	if (!gmu->legacy)
+		status = GMU_COLD_BOOT;
+
 	ret = a6xx_gmu_fw_start(gmu, status);
 	if (ret)
 		goto out;
@@ -761,7 +898,10 @@ int a6xx_gmu_resume(struct a6xx_gpu *a6xx_gpu)
 	enable_irq(gmu->hfi_irq);
 
 	/* Set the GPU to the current freq */
-	__a6xx_gmu_set_freq(gmu, gmu->current_perf_index);
+	if (gmu->legacy)
+		__a6xx_gmu_set_freq(gmu, gmu->current_perf_index);
+	else
+		a6xx_hfi_set_freq(gmu, gmu->current_perf_index);
 
 	/*
 	 * "enable" the GX power domain which won't actually do anything but it
@@ -919,34 +1059,75 @@ int a6xx_gmu_stop(struct a6xx_gpu *a6xx_gpu)
 	return 0;
 }
 
-static void a6xx_gmu_memory_free(struct a6xx_gmu *gmu, struct a6xx_gmu_bo *bo)
+static void a6xx_gmu_memory_free(struct a6xx_gmu *gmu)
 {
-	if (IS_ERR_OR_NULL(bo))
-		return;
-
-	dma_free_wc(gmu->dev, bo->size, bo->virt, bo->iova);
-	kfree(bo);
+	msm_gem_kernel_put(gmu->hfi.obj, gmu->aspace, false);
+	msm_gem_kernel_put(gmu->debug.obj, gmu->aspace, false);
+	msm_gem_kernel_put(gmu->icache.obj, gmu->aspace, false);
+	msm_gem_kernel_put(gmu->dcache.obj, gmu->aspace, false);
+	msm_gem_kernel_put(gmu->dummy.obj, gmu->aspace, false);
+	msm_gem_kernel_put(gmu->log.obj, gmu->aspace, false);
+
+	gmu->aspace->mmu->funcs->detach(gmu->aspace->mmu);
+	msm_gem_address_space_put(gmu->aspace);
 }
 
-static struct a6xx_gmu_bo *a6xx_gmu_memory_alloc(struct a6xx_gmu *gmu,
-		size_t size)
+static int a6xx_gmu_memory_alloc(struct a6xx_gmu *gmu, struct a6xx_gmu_bo *bo,
+		size_t size, u64 iova)
 {
-	struct a6xx_gmu_bo *bo;
+	struct a6xx_gpu *a6xx_gpu = container_of(gmu, struct a6xx_gpu, gmu);
+	struct drm_device *dev = a6xx_gpu->base.base.dev;
+	uint32_t flags = MSM_BO_WC;
+	u64 range_start, range_end;
+	int ret;
 
-	bo = kzalloc(sizeof(*bo), GFP_KERNEL);
-	if (!bo)
-		return ERR_PTR(-ENOMEM);
+	size = PAGE_ALIGN(size);
+	if (!iova) {
+		/* no fixed address - use GMU's uncached range */
+		range_start = 0x60000000 + PAGE_SIZE; /* skip dummy page */
+		range_end = 0x80000000;
+	} else {
+		/* range for fixed address */
+		range_start = iova;
+		range_end = iova + size;
+		/* use IOMMU_PRIV for icache/dcache */
+		flags |= MSM_BO_MAP_PRIV;
+	}
 
-	bo->size = PAGE_ALIGN(size);
+	bo->obj = msm_gem_new(dev, size, flags);
+	if (IS_ERR(bo->obj))
+		return PTR_ERR(bo->obj);
 
-	bo->virt = dma_alloc_wc(gmu->dev, bo->size, &bo->iova, GFP_KERNEL);
+	ret = msm_gem_get_and_pin_iova_range(bo->obj, gmu->aspace, &bo->iova,
+		range_start >> PAGE_SHIFT, range_end >> PAGE_SHIFT);
+	if (ret) {
+		drm_gem_object_put(bo->obj);
+		return ret;
+	}
+
+	bo->virt = msm_gem_get_vaddr(bo->obj);
+	bo->size = size;
+
+	return 0;
+}
+
+static int a6xx_gmu_memory_probe(struct a6xx_gmu *gmu)
+{
+	struct iommu_domain *domain;
+	struct msm_mmu *mmu;
 
-	if (!bo->virt) {
-		kfree(bo);
-		return ERR_PTR(-ENOMEM);
+	domain = iommu_domain_alloc(&platform_bus_type);
+	if (!domain)
+		return -ENODEV;
+
+	mmu = msm_iommu_new(gmu->dev, domain);
+	gmu->aspace = msm_gem_address_space_create(mmu, "gmu", 0x0, 0x7fffffff);
+	if (IS_ERR(gmu->aspace)) {
+		iommu_domain_free(domain);
+		return PTR_ERR(gmu->aspace);
 	}
 
-	return bo;
+	return 0;
 }
 
 /* Return the 'arc-level' for the given frequency */
@@ -1011,8 +1192,8 @@ static int a6xx_gmu_rpmh_arc_votes_init(struct device *dev, u32 *votes,
 
 		if (j == pri_count) {
 			DRM_DEV_ERROR(dev,
-				"Level %u not found in in the RPMh list\n",
-					level);
+				      "Level %u not found in the RPMh list\n",
+				      level);
 			DRM_DEV_ERROR(dev, "Available levels:\n");
 			for (j = 0; j < pri_count; j++)
 				DRM_DEV_ERROR(dev, "  %u\n", pri[j]);
@@ -1190,6 +1371,7 @@ static int a6xx_gmu_get_irq(struct a6xx_gmu *gmu, struct platform_device *pdev,
 void a6xx_gmu_remove(struct a6xx_gpu *a6xx_gpu)
 {
 	struct a6xx_gmu *gmu = &a6xx_gpu->gmu;
+	struct platform_device *pdev = to_platform_device(gmu->dev);
 
 	if (!gmu->initialized)
 		return;
@@ -1202,9 +1384,12 @@ void a6xx_gmu_remove(struct a6xx_gpu *a6xx_gpu)
 	}
 
 	iounmap(gmu->mmio);
+	if (platform_get_resource_byname(pdev, IORESOURCE_MEM, "rscc"))
+		iounmap(gmu->rscc);
 	gmu->mmio = NULL;
+	gmu->rscc = NULL;
 
-	a6xx_gmu_memory_free(gmu, gmu->hfi);
+	a6xx_gmu_memory_free(gmu);
 
 	free_irq(gmu->gmu_irq, gmu);
 	free_irq(gmu->hfi_irq, gmu);
@@ -1217,6 +1402,7 @@ void a6xx_gmu_remove(struct a6xx_gpu *a6xx_gpu)
 
 int a6xx_gmu_init(struct a6xx_gpu *a6xx_gpu, struct device_node *node)
 {
+	struct adreno_gpu *adreno_gpu = &a6xx_gpu->base;
 	struct a6xx_gmu *gmu = &a6xx_gpu->gmu;
 	struct platform_device *pdev = of_find_device_by_node(node);
 	int ret;
@@ -1226,15 +1412,7 @@ int a6xx_gmu_init(struct a6xx_gpu *a6xx_gpu, struct device_node *node)
 
 	gmu->dev = &pdev->dev;
 
-	/* Pass force_dma false to require the DT to set the dma region */
-	ret = of_dma_configure(gmu->dev, node, false);
-	if (ret)
-		return ret;
-
-	/* Set the mask after the of_dma_configure() */
-	ret = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(31));
-	if (ret)
-		return ret;
+	of_dma_configure(gmu->dev, node, true);
 
 	/* Fow now, don't do anything fancy until we get our feet under us */
 	gmu->idle_level = GMU_IDLE_STATE_ACTIVE;
@@ -1246,20 +1424,64 @@ int a6xx_gmu_init(struct a6xx_gpu *a6xx_gpu, struct device_node *node)
 	if (ret)
 		goto err_put_device;
 
+	ret = a6xx_gmu_memory_probe(gmu);
+	if (ret)
+		goto err_put_device;
+
+	/* Allocate memory for the GMU dummy page */
+	ret = a6xx_gmu_memory_alloc(gmu, &gmu->dummy, SZ_4K, 0x60000000);
+	if (ret)
+		goto err_memory;
+
+	if (adreno_is_a650(adreno_gpu)) {
+		ret = a6xx_gmu_memory_alloc(gmu, &gmu->icache,
+			SZ_16M - SZ_16K, 0x04000);
+		if (ret)
+			goto err_memory;
+	} else if (adreno_is_a640(adreno_gpu)) {
+		ret = a6xx_gmu_memory_alloc(gmu, &gmu->icache,
+			SZ_256K - SZ_16K, 0x04000);
+		if (ret)
+			goto err_memory;
+
+		ret = a6xx_gmu_memory_alloc(gmu, &gmu->dcache,
+			SZ_256K - SZ_16K, 0x44000);
+		if (ret)
+			goto err_memory;
+	} else {
+		/* HFI v1, has sptprac */
+		gmu->legacy = true;
+
+		/* Allocate memory for the GMU debug region */
+		ret = a6xx_gmu_memory_alloc(gmu, &gmu->debug, SZ_16K, 0);
+		if (ret)
+			goto err_memory;
+	}
+
 	/* Allocate memory for for the HFI queues */
-	gmu->hfi = a6xx_gmu_memory_alloc(gmu, SZ_16K);
-	if (IS_ERR(gmu->hfi))
+	ret = a6xx_gmu_memory_alloc(gmu, &gmu->hfi, SZ_16K, 0);
+	if (ret)
 		goto err_memory;
 
-	/* Allocate memory for the GMU debug region */
-	gmu->debug = a6xx_gmu_memory_alloc(gmu, SZ_16K);
-	if (IS_ERR(gmu->debug))
+	/* Allocate memory for the GMU log region */
+	ret = a6xx_gmu_memory_alloc(gmu, &gmu->log, SZ_4K, 0);
+	if (ret)
 		goto err_memory;
 
 	/* Map the GMU registers */
 	gmu->mmio = a6xx_gmu_get_mmio(pdev, "gmu");
-	if (IS_ERR(gmu->mmio))
+	if (IS_ERR(gmu->mmio)) {
+		ret = PTR_ERR(gmu->mmio);
 		goto err_memory;
+	}
+
+	if (adreno_is_a650(adreno_gpu)) {
+		gmu->rscc = a6xx_gmu_get_mmio(pdev, "rscc");
+		if (IS_ERR(gmu->rscc))
+			goto err_mmio;
+	} else {
+		gmu->rscc = gmu->mmio + 0x23000;
+	}
 
 	/* Get the HFI and GMU interrupts */
 	gmu->hfi_irq = a6xx_gmu_get_irq(gmu, pdev, "hfi", a6xx_hfi_irq);
@@ -1286,13 +1508,15 @@ int a6xx_gmu_init(struct a6xx_gpu *a6xx_gpu, struct device_node *node)
 
 err_mmio:
 	iounmap(gmu->mmio);
+	if (platform_get_resource_byname(pdev, IORESOURCE_MEM, "rscc"))
+		iounmap(gmu->rscc);
 	free_irq(gmu->gmu_irq, gmu);
 	free_irq(gmu->hfi_irq, gmu);
-err_memory:
-	a6xx_gmu_memory_free(gmu, gmu->hfi);
 
 	ret = -ENODEV;
 
+err_memory:
+	a6xx_gmu_memory_free(gmu);
 err_put_device:
 	/* Drop reference taken in of_find_device_by_node */
 	put_device(gmu->dev);