Merge tag 'drm-msm-next-2022-11-28' of https://gitlab.freedesktop.org/drm/msm into drm-next

msm-next for v6.2 (the gpu/gem bits)

- Remove exclusive-fence hack that caused over-synchronization
- Fix speed-bin detection vs. probe-defer
- Enable clamp_to_idle on 7c3
- Improved hangcheck detection

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rob Clark <robdclark@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/CAF6AEGvT1h_S4d=YRgphgR8i7aMaxQaNW8mru7QaoUo9uiUk2A@mail.gmail.com

1 files changed, 54 insertions, 30 deletions

diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
index e87196457b9a..36c8fb699b56 100644
--- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c
@@ -247,8 +247,7 @@ static void a6xx_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)
 	OUT_RING(ring, submit->seqno);
 
 	trace_msm_gpu_submit_flush(submit,
-		gpu_read64(gpu, REG_A6XX_CP_ALWAYS_ON_COUNTER_LO,
-			REG_A6XX_CP_ALWAYS_ON_COUNTER_HI));
+		gpu_read64(gpu, REG_A6XX_CP_ALWAYS_ON_COUNTER_LO));
 
 	a6xx_flush(gpu, ring);
 }
@@ -947,8 +946,7 @@ static int a6xx_ucode_init(struct msm_gpu *gpu)
 		}
 	}
 
-	gpu_write64(gpu, REG_A6XX_CP_SQE_INSTR_BASE,
-		REG_A6XX_CP_SQE_INSTR_BASE+1, a6xx_gpu->sqe_iova);
+	gpu_write64(gpu, REG_A6XX_CP_SQE_INSTR_BASE, a6xx_gpu->sqe_iova);
 
 	return 0;
 }
@@ -999,8 +997,7 @@ static int hw_init(struct msm_gpu *gpu)
 	 * memory rendering at this point in time and we don't want to block off
 	 * part of the virtual memory space.
 	 */
-	gpu_write64(gpu, REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO,
-		REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_HI, 0x00000000);
+	gpu_write64(gpu, REG_A6XX_RBBM_SECVID_TSB_TRUSTED_BASE_LO, 0x00000000);
 	gpu_write(gpu, REG_A6XX_RBBM_SECVID_TSB_TRUSTED_SIZE, 0x00000000);
 
 	/* Turn on 64 bit addressing for all blocks */
@@ -1049,11 +1046,9 @@ static int hw_init(struct msm_gpu *gpu)
 
 	if (!adreno_is_a650_family(adreno_gpu)) {
 		/* Set the GMEM VA range [0x100000:0x100000 + gpu->gmem - 1] */
-		gpu_write64(gpu, REG_A6XX_UCHE_GMEM_RANGE_MIN_LO,
-			REG_A6XX_UCHE_GMEM_RANGE_MIN_HI, 0x00100000);
+		gpu_write64(gpu, REG_A6XX_UCHE_GMEM_RANGE_MIN_LO, 0x00100000);
 
 		gpu_write64(gpu, REG_A6XX_UCHE_GMEM_RANGE_MAX_LO,
-			REG_A6XX_UCHE_GMEM_RANGE_MAX_HI,
 			0x00100000 + adreno_gpu->gmem - 1);
 	}
 
@@ -1145,8 +1140,7 @@ static int hw_init(struct msm_gpu *gpu)
 		goto out;
 
 	/* Set the ringbuffer address */
-	gpu_write64(gpu, REG_A6XX_CP_RB_BASE, REG_A6XX_CP_RB_BASE_HI,
-		gpu->rb[0]->iova);
+	gpu_write64(gpu, REG_A6XX_CP_RB_BASE, gpu->rb[0]->iova);
 
 	/* Targets that support extended APRIV can use the RPTR shadow from
 	 * hardware but all the other ones need to disable the feature. Targets
@@ -1178,7 +1172,6 @@ static int hw_init(struct msm_gpu *gpu)
 		}
 
 		gpu_write64(gpu, REG_A6XX_CP_RB_RPTR_ADDR_LO,
-			REG_A6XX_CP_RB_RPTR_ADDR_HI,
 			shadowptr(a6xx_gpu, gpu->rb[0]));
 	}
 
@@ -1499,9 +1492,9 @@ static void a6xx_fault_detect_irq(struct msm_gpu *gpu)
 		gpu_read(gpu, REG_A6XX_RBBM_STATUS),
 		gpu_read(gpu, REG_A6XX_CP_RB_RPTR),
 		gpu_read(gpu, REG_A6XX_CP_RB_WPTR),
-		gpu_read64(gpu, REG_A6XX_CP_IB1_BASE, REG_A6XX_CP_IB1_BASE_HI),
+		gpu_read64(gpu, REG_A6XX_CP_IB1_BASE),
 		gpu_read(gpu, REG_A6XX_CP_IB1_REM_SIZE),
-		gpu_read64(gpu, REG_A6XX_CP_IB2_BASE, REG_A6XX_CP_IB2_BASE_HI),
+		gpu_read64(gpu, REG_A6XX_CP_IB2_BASE),
 		gpu_read(gpu, REG_A6XX_CP_IB2_REM_SIZE));
 
 	/* Turn off the hangcheck timer to keep it from bothering us */
@@ -1712,8 +1705,7 @@ static int a6xx_get_timestamp(struct msm_gpu *gpu, uint64_t *value)
 	/* Force the GPU power on so we can read this register */
 	a6xx_gmu_set_oob(&a6xx_gpu->gmu, GMU_OOB_PERFCOUNTER_SET);
 
-	*value = gpu_read64(gpu, REG_A6XX_CP_ALWAYS_ON_COUNTER_LO,
-			    REG_A6XX_CP_ALWAYS_ON_COUNTER_HI);
+	*value = gpu_read64(gpu, REG_A6XX_CP_ALWAYS_ON_COUNTER_LO);
 
 	a6xx_gmu_clear_oob(&a6xx_gpu->gmu, GMU_OOB_PERFCOUNTER_SET);
 
@@ -1824,6 +1816,39 @@ static uint32_t a6xx_get_rptr(struct msm_gpu *gpu, struct msm_ringbuffer *ring)
 	return ring->memptrs->rptr = gpu_read(gpu, REG_A6XX_CP_RB_RPTR);
 }
 
+static bool a6xx_progress(struct msm_gpu *gpu, struct msm_ringbuffer *ring)
+{
+	struct msm_cp_state cp_state = {
+		.ib1_base = gpu_read64(gpu, REG_A6XX_CP_IB1_BASE),
+		.ib2_base = gpu_read64(gpu, REG_A6XX_CP_IB2_BASE),
+		.ib1_rem  = gpu_read(gpu, REG_A6XX_CP_IB1_REM_SIZE),
+		.ib2_rem  = gpu_read(gpu, REG_A6XX_CP_IB2_REM_SIZE),
+	};
+	bool progress;
+
+	/*
+	 * Adjust the remaining data to account for what has already been
+	 * fetched from memory, but not yet consumed by the SQE.
+	 *
+	 * This is not *technically* correct, the amount buffered could
+	 * exceed the IB size due to hw prefetching ahead, but:
+	 *
+	 * (1) We aren't trying to find the exact position, just whether
+	 *     progress has been made
+	 * (2) The CP_REG_TO_MEM at the end of a submit should be enough
+	 *     to prevent prefetching into an unrelated submit.  (And
+	 *     either way, at some point the ROQ will be full.)
+	 */
+	cp_state.ib1_rem += gpu_read(gpu, REG_A6XX_CP_CSQ_IB1_STAT) >> 16;
+	cp_state.ib2_rem += gpu_read(gpu, REG_A6XX_CP_CSQ_IB2_STAT) >> 16;
+
+	progress = !!memcmp(&cp_state, &ring->last_cp_state, sizeof(cp_state));
+
+	ring->last_cp_state = cp_state;
+
+	return progress;
+}
+
 static u32 a618_get_speed_bin(u32 fuse)
 {
 	if (fuse == 0)
@@ -1879,7 +1904,7 @@ static u32 fuse_to_supp_hw(struct device *dev, struct adreno_rev rev, u32 fuse)
 
 	if (val == UINT_MAX) {
 		DRM_DEV_ERROR(dev,
-			"missing support for speed-bin: %u. Some OPPs may not be supported by hardware",
+			"missing support for speed-bin: %u. Some OPPs may not be supported by hardware\n",
 			fuse);
 		return UINT_MAX;
 	}
@@ -1889,7 +1914,7 @@ static u32 fuse_to_supp_hw(struct device *dev, struct adreno_rev rev, u32 fuse)
 
 static int a6xx_set_supported_hw(struct device *dev, struct adreno_rev rev)
 {
-	u32 supp_hw = UINT_MAX;
+	u32 supp_hw;
 	u32 speedbin;
 	int ret;
 
@@ -1901,15 +1926,13 @@ static int a6xx_set_supported_hw(struct device *dev, struct adreno_rev rev)
 	if (ret == -ENOENT) {
 		return 0;
 	} else if (ret) {
-		DRM_DEV_ERROR(dev,
-			      "failed to read speed-bin (%d). Some OPPs may not be supported by hardware",
-			      ret);
-		goto done;
+		dev_err_probe(dev, ret,
+			      "failed to read speed-bin. Some OPPs may not be supported by hardware\n");
+		return ret;
 	}
 
 	supp_hw = fuse_to_supp_hw(dev, rev, speedbin);
 
-done:
 	ret = devm_pm_opp_set_supported_hw(dev, &supp_hw, 1);
 	if (ret)
 		return ret;
@@ -1942,6 +1965,7 @@ static const struct adreno_gpu_funcs funcs = {
 		.create_address_space = a6xx_create_address_space,
 		.create_private_address_space = a6xx_create_private_address_space,
 		.get_rptr = a6xx_get_rptr,
+		.progress = a6xx_progress,
 	},
 	.get_timestamp = a6xx_get_timestamp,
 };
@@ -1978,13 +2002,6 @@ struct msm_gpu *a6xx_gpu_init(struct drm_device *dev)
 			adreno_cmp_rev(ADRENO_REV(6, 3, 5, ANY_ID), info->rev)))
 		adreno_gpu->base.hw_apriv = true;
 
-	/*
-	 * For now only clamp to idle freq for devices where this is known not
-	 * to cause power supply issues:
-	 */
-	if (info && (info->revn == 618))
-		gpu->clamp_to_idle = true;
-
 	a6xx_llc_slices_init(pdev, a6xx_gpu);
 
 	ret = a6xx_set_supported_hw(&pdev->dev, config->rev);
@@ -1999,6 +2016,13 @@ struct msm_gpu *a6xx_gpu_init(struct drm_device *dev)
 		return ERR_PTR(ret);
 	}
 
+	/*
+	 * For now only clamp to idle freq for devices where this is known not
+	 * to cause power supply issues:
+	 */
+	if (adreno_is_a618(adreno_gpu) || adreno_is_7c3(adreno_gpu))
+		gpu->clamp_to_idle = true;
+
 	/* Check if there is a GMU phandle and set it up */
 	node = of_parse_phandle(pdev->dev.of_node, "qcom,gmu", 0);