From f4caf5842652f08e024741ef6d423cb0c101d863 Mon Sep 17 00:00:00 2001
From: Hawking Zhang <Hawking.Zhang@amd.com>
Date: Wed, 14 Sep 2022 16:35:50 +0800
Subject: drm/amdgpu: introduce vmhub definition for multi-partition cases (v3)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

v1: Each partition has its own gfxhub or mmhub. adjust
the num of MAX_VMHUBS and the GFXHUB/MMHUB layout (Le)

v2: re-design the AMDGPU_GFXHUB/AMDGPU_MMHUB layout (Le)

v3: apply the gfxhub/mmhub layout to new IPs (Hawking)

v4: fix up gmc11 (Alex)

v5: rebase (Alex)

Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 64dcaa2670dd..7efe7c43fffb 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1309,7 +1309,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
 
 		/* doorbell size is 2 dwords, get DWORD offset */
 		ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
-		ring->vm_hub = AMDGPU_MMHUB_0;
+		ring->vm_hub = AMDGPU_MMHUB0(0);
 
 		sprintf(ring->name, "sdma%d", i);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
@@ -1328,7 +1328,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
 			 */
 			ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
 			ring->doorbell_index += 0x400;
-			ring->vm_hub = AMDGPU_MMHUB_0;
+			ring->vm_hub = AMDGPU_MMHUB0(0);
 
 			sprintf(ring->name, "page%d", i);
 			r = amdgpu_ring_init(adev, ring, 1024,
-- 
cgit 


From 0ee20b86966cff359f51051a56f8c2d89b09aed4 Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Mon, 28 Feb 2022 19:30:10 +0800
Subject: drm/amdgpu: assign the doorbell index in 1st page to sdma page queue

Previously for vega10, the sdma_doorbell_range is only enough for sdma
gfx queue, thus the index on second doorbell page is allocated for sdma
page queue. From vega20, the sdma_doorbell_range on 1st page is enlarged.
Therefore, just leverage these index instead of allocation on 2nd page.

v2: change "(x << 1) + 2" to "(x + 1) << 1" for readability and add comments.

Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c |  3 ++-
 drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c     | 14 ++++++++++++--
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c   |  7 +++----
 3 files changed, 17 insertions(+), 7 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 54be85539ecb..38b6bbcaf186 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -1091,7 +1091,8 @@ static int amdgpu_device_doorbell_init(struct amdgpu_device *adev)
 		 * doorbells are in the first page. So with paging queue enabled,
 		 * the max num_kernel_doorbells should + 1 page (0x400 in dword)
 		 */
-		if (adev->asic_type >= CHIP_VEGA10)
+		if (adev->ip_versions[SDMA0_HWIP][0] >= IP_VERSION(4, 0, 0) &&
+		    adev->ip_versions[SDMA0_HWIP][0] < IP_VERSION(4, 2, 0))
 			adev->doorbell.num_kernel_doorbells += 0x400;
 	}
 
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index 50b6eb9bcfda..70b0d1fd9868 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -1847,8 +1847,18 @@ static int sdma_v4_0_sw_init(void *handle)
 			/* paging queue use same doorbell index/routing as gfx queue
 			 * with 0x400 (4096 dwords) offset on second doorbell page
 			 */
-			ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
-			ring->doorbell_index += 0x400;
+			if (adev->ip_versions[SDMA0_HWIP][0] >= IP_VERSION(4, 0, 0) &&
+			    adev->ip_versions[SDMA0_HWIP][0] < IP_VERSION(4, 2, 0)) {
+				ring->doorbell_index =
+					adev->doorbell_index.sdma_engine[i] << 1;
+				ring->doorbell_index += 0x400;
+			} else {
+				/* From vega20, the sdma_doorbell_range in 1st
+				 * doorbell page is reserved for page queue.
+				 */
+				ring->doorbell_index =
+					(adev->doorbell_index.sdma_engine[i] + 1) << 1;
+			}
 
 			if (adev->ip_versions[SDMA0_HWIP][0] == IP_VERSION(4, 2, 2) && i >= 5)
 				ring->vm_hub = AMDGPU_MMHUB1(0);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 7efe7c43fffb..441d6911fd20 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1323,11 +1323,10 @@ static int sdma_v4_4_2_sw_init(void *handle)
 			ring->ring_obj = NULL;
 			ring->use_doorbell = true;
 
-			/* paging queue use same doorbell index/routing as gfx queue
-			 * with 0x400 (4096 dwords) offset on second doorbell page
+			/* doorbell index of page queue is assigned right after
+			 * gfx queue on the same instance
 			 */
-			ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
-			ring->doorbell_index += 0x400;
+			ring->doorbell_index = (adev->doorbell_index.sdma_engine[i] + 1) << 1;
 			ring->vm_hub = AMDGPU_MMHUB0(0);
 
 			sprintf(ring->name, "page%d", i);
-- 
cgit 


From f786b1d4ec778a5fc23911f06a0e38c9f4953b0c Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Wed, 20 Apr 2022 17:03:00 +0800
Subject: drm/amdgpu: add support for SDMA on multiple AIDs

Initialize SDMA instances on each AID.

v2: revise coding fault in hw_fini

Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h |  8 ++++++++
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 20 +++++++++++++-------
 2 files changed, 21 insertions(+), 7 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
index 67975dcede5d..632b77138fe4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
@@ -37,6 +37,14 @@ enum amdgpu_sdma_irq {
 	AMDGPU_SDMA_IRQ_INSTANCE5,
 	AMDGPU_SDMA_IRQ_INSTANCE6,
 	AMDGPU_SDMA_IRQ_INSTANCE7,
+	AMDGPU_SDMA_IRQ_INSTANCE8,
+	AMDGPU_SDMA_IRQ_INSTANCE9,
+	AMDGPU_SDMA_IRQ_INSTANCE10,
+	AMDGPU_SDMA_IRQ_INSTANCE11,
+	AMDGPU_SDMA_IRQ_INSTANCE12,
+	AMDGPU_SDMA_IRQ_INSTANCE13,
+	AMDGPU_SDMA_IRQ_INSTANCE14,
+	AMDGPU_SDMA_IRQ_INSTANCE15,
 	AMDGPU_SDMA_IRQ_LAST
 };
 
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 441d6911fd20..7deadea03caa 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1253,9 +1253,10 @@ static int sdma_v4_4_2_sw_init(void *handle)
 	struct amdgpu_ring *ring;
 	int r, i;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	u32 aid_id;
 
 	/* SDMA trap event */
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for (i = 0; i < adev->sdma.num_inst_per_aid; i++) {
 		r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i),
 				      SDMA0_4_0__SRCID__SDMA_TRAP,
 				      &adev->sdma.trap_irq);
@@ -1264,7 +1265,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
 	}
 
 	/* SDMA SRAM ECC event */
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for (i = 0; i < adev->sdma.num_inst_per_aid; i++) {
 		r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i),
 				      SDMA0_4_0__SRCID__SDMA_SRAM_ECC,
 				      &adev->sdma.ecc_irq);
@@ -1273,7 +1274,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
 	}
 
 	/* SDMA VM_HOLE/DOORBELL_INV/POLL_TIMEOUT/SRBM_WRITE_PROTECTION event*/
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for (i = 0; i < adev->sdma.num_inst_per_aid; i++) {
 		r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i),
 				      SDMA0_4_0__SRCID__SDMA_VM_HOLE,
 				      &adev->sdma.vm_hole_irq);
@@ -1303,15 +1304,17 @@ static int sdma_v4_4_2_sw_init(void *handle)
 		ring = &adev->sdma.instance[i].ring;
 		ring->ring_obj = NULL;
 		ring->use_doorbell = true;
+		aid_id = adev->sdma.instance[i].aid_id;
 
 		DRM_DEBUG("SDMA %d use_doorbell being set to: [%s]\n", i,
 				ring->use_doorbell?"true":"false");
 
 		/* doorbell size is 2 dwords, get DWORD offset */
 		ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
-		ring->vm_hub = AMDGPU_MMHUB0(0);
+		ring->vm_hub = AMDGPU_MMHUB0(aid_id);
 
-		sprintf(ring->name, "sdma%d", i);
+		sprintf(ring->name, "sdma%d.%d", aid_id,
+				i % adev->sdma.num_inst_per_aid);
 		r = amdgpu_ring_init(adev, ring, 1024, &adev->sdma.trap_irq,
 				     AMDGPU_SDMA_IRQ_INSTANCE0 + i,
 				     AMDGPU_RING_PRIO_DEFAULT, NULL);
@@ -1327,9 +1330,10 @@ static int sdma_v4_4_2_sw_init(void *handle)
 			 * gfx queue on the same instance
 			 */
 			ring->doorbell_index = (adev->doorbell_index.sdma_engine[i] + 1) << 1;
-			ring->vm_hub = AMDGPU_MMHUB0(0);
+			ring->vm_hub = AMDGPU_MMHUB0(aid_id);
 
-			sprintf(ring->name, "page%d", i);
+			sprintf(ring->name, "page%d.%d", aid_id,
+					i % adev->sdma.num_inst_per_aid);
 			r = amdgpu_ring_init(adev, ring, 1024,
 					     &adev->sdma.trap_irq,
 					     AMDGPU_SDMA_IRQ_INSTANCE0 + i,
@@ -1811,6 +1815,8 @@ static void sdma_v4_4_2_set_ring_funcs(struct amdgpu_device *adev)
 				&sdma_v4_4_2_page_ring_funcs;
 			adev->sdma.instance[i].page.me = i;
 		}
+
+		adev->sdma.instance[i].aid_id = i / adev->sdma.num_inst_per_aid;
 	}
 }
 
-- 
cgit 


From 2a47a2d90e5cd96c24503061c8920a1e6ee248a0 Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Tue, 1 Mar 2022 19:42:29 +0800
Subject: drm/amdgpu: assign the doorbell index for sdma on non-AID0

Allocate new sdma doorbell index for the instances only on AID1 for now.

Todo: there's limitation that SDMA doorbell index on SDMA 4.4.2 needs to be
less than 0x1FF, so the tail part in _AMDGPU_VEGA20_DOORBELL_ASSIGNMENT is not
enough to store sdma doorbell range on maximum 4 AIDs if doorbell_range is 20.
So it looks better to create a new doorbell index assignment table for 4.4.2.

v2: change "(x << 1) + 2" to "(x + 1) << 1" for readability.

Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h |  6 +++++-
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c     | 18 ++++++++++++++++--
 2 files changed, 21 insertions(+), 3 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
index ffb75d23d2fc..b036d2f01930 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
@@ -85,6 +85,7 @@ struct amdgpu_doorbell_index {
 	uint32_t last_non_cp;
 	uint32_t xcc1_kiq_start;
 	uint32_t xcc1_mec_ring0_start;
+	uint32_t aid1_sdma_start;
 	uint32_t max_assignment;
 	/* Per engine SDMA doorbell size in dword */
 	uint32_t sdma_doorbell_range;
@@ -171,7 +172,10 @@ typedef enum _AMDGPU_VEGA20_DOORBELL_ASSIGNMENT
 	/* 8 compute rings per GC. Max to 0x1CE */
 	AMDGPU_VEGA20_DOORBELL_XCC1_MEC_RING0_START       = 0x197,
 
-	AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT            = 0x1CE,
+	/* AID1 SDMA: 0x1D0 ~ 0x1F7 */
+	AMDGPU_VEGA20_DOORBELL_AID1_sDMA_START           = 0x1D0,
+
+	AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT            = 0x1F7,
 	AMDGPU_VEGA20_DOORBELL_INVALID                   = 0xFFFF
 } AMDGPU_VEGA20_DOORBELL_ASSIGNMENT;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 7deadea03caa..6935a24d1e89 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1310,7 +1310,14 @@ static int sdma_v4_4_2_sw_init(void *handle)
 				ring->use_doorbell?"true":"false");
 
 		/* doorbell size is 2 dwords, get DWORD offset */
-		ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
+		if (aid_id > 0)
+			ring->doorbell_index =
+				(adev->doorbell_index.aid1_sdma_start << 1)
+				+ adev->doorbell_index.sdma_doorbell_range
+				* (i - adev->sdma.num_inst_per_aid);
+		else
+			ring->doorbell_index =
+				adev->doorbell_index.sdma_engine[i] << 1;
 		ring->vm_hub = AMDGPU_MMHUB0(aid_id);
 
 		sprintf(ring->name, "sdma%d.%d", aid_id,
@@ -1329,7 +1336,14 @@ static int sdma_v4_4_2_sw_init(void *handle)
 			/* doorbell index of page queue is assigned right after
 			 * gfx queue on the same instance
 			 */
-			ring->doorbell_index = (adev->doorbell_index.sdma_engine[i] + 1) << 1;
+			if (aid_id > 0)
+				ring->doorbell_index =
+					((adev->doorbell_index.aid1_sdma_start + 1) << 1)
+					+ adev->doorbell_index.sdma_doorbell_range
+					* (i - adev->sdma.num_inst_per_aid);
+			else
+				ring->doorbell_index =
+					(adev->doorbell_index.sdma_engine[i] + 1) << 1;
 			ring->vm_hub = AMDGPU_MMHUB0(aid_id);
 
 			sprintf(ring->name, "page%d.%d", aid_id,
-- 
cgit 


From 0d81101c190d1835f1bfba85dc3d65b9f9cacd68 Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Fri, 25 Feb 2022 15:14:19 +0800
Subject: drm/amdgpu: complement the IH node_id table for multiple AIDs

With different node_id, the SDMA interrupt from multiple AIDs can be
distinguished by sw driver.

Signed-off-by: Le Ma <le.ma@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c  | 4 ++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h  | 6 +++++-
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 3 +++
 3 files changed, 12 insertions(+), 1 deletion(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index 031610c1340a..3dafaf70c987 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -100,12 +100,16 @@ const char *soc15_ih_clientid_name[] = {
 };
 
 const int node_id_to_phys_map[NODEID_MAX] = {
+	[AID0_NODEID] = 0,
 	[XCD0_NODEID] = 0,
 	[XCD1_NODEID] = 1,
+	[AID1_NODEID] = 1,
 	[XCD2_NODEID] = 2,
 	[XCD3_NODEID] = 3,
+	[AID2_NODEID] = 2,
 	[XCD4_NODEID] = 4,
 	[XCD5_NODEID] = 5,
+	[AID3_NODEID] = 3,
 	[XCD6_NODEID] = 6,
 	[XCD7_NODEID] = 7,
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
index efe8a278cbdf..04c0b4fa17a4 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h
@@ -102,13 +102,17 @@ struct amdgpu_irq {
 	bool                            retry_cam_enabled;
 };
 
-enum interrupt_node_id_per_xcp {
+enum interrupt_node_id_per_aid {
+	AID0_NODEID = 0,
 	XCD0_NODEID = 1,
 	XCD1_NODEID = 2,
+	AID1_NODEID = 4,
 	XCD2_NODEID = 5,
 	XCD3_NODEID = 6,
+	AID2_NODEID = 8,
 	XCD4_NODEID = 9,
 	XCD5_NODEID = 10,
+	AID3_NODEID = 12,
 	XCD6_NODEID = 13,
 	XCD7_NODEID = 14,
 	NODEID_MAX,
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 6935a24d1e89..d3c7f9a43ef1 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1492,6 +1492,9 @@ static int sdma_v4_4_2_process_trap_irq(struct amdgpu_device *adev,
 
 	DRM_DEBUG("IH: SDMA trap\n");
 	instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
+	instance += node_id_to_phys_map[entry->node_id] *
+			adev->sdma.num_inst_per_aid;
+
 	switch (entry->ring_id) {
 	case 0:
 		amdgpu_fence_process(&adev->sdma.instance[instance].ring);
-- 
cgit 


From 20bedf1379b1d4b060e3f43661f35e5447d0cfed Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Mon, 25 Apr 2022 22:19:58 +0800
Subject: drm/amdgpu: introduce new doorbell assignment table for GC 9.4.3

Four basic reasons as below to do the change:
  1. number of ring expand a lot on GC 9.4.3, and adjustment on old
     assignment cannot make each ring in a continuous doorbell space.
  2. the SDMA doorbell index should not exceed 0x1FF on SDMA 4.2.2 due to
     regDOORBELLx_CTRL_ENTRY.BIF_DOORBELLx_RANGE_OFFSET_ENTRY field width.
  3. re-design the doorbell assignment and unify the calculation as
     "start + ring/inst id" will make the code much concise.
  4. only defining the START/END makes the table look simple

v2: (Lijo)
  1. replace name
  2. use num_inst_per_aid/sdma_doorbell_range instead of hardcoding

Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h | 32 ++++++++++++++++++++++++----
 drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c      |  6 +-----
 drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c      |  8 +------
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c     | 19 +++--------------
 4 files changed, 33 insertions(+), 32 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
index b036d2f01930..a29a018ec84e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h
@@ -59,7 +59,7 @@ struct amdgpu_doorbell_index {
 	uint32_t gfx_ring1;
 	uint32_t gfx_userqueue_start;
 	uint32_t gfx_userqueue_end;
-	uint32_t sdma_engine[8];
+	uint32_t sdma_engine[16];
 	uint32_t mes_ring0;
 	uint32_t mes_ring1;
 	uint32_t ih;
@@ -83,9 +83,6 @@ struct amdgpu_doorbell_index {
 	};
 	uint32_t first_non_cp;
 	uint32_t last_non_cp;
-	uint32_t xcc1_kiq_start;
-	uint32_t xcc1_mec_ring0_start;
-	uint32_t aid1_sdma_start;
 	uint32_t max_assignment;
 	/* Per engine SDMA doorbell size in dword */
 	uint32_t sdma_doorbell_range;
@@ -312,6 +309,33 @@ typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT
 	AMDGPU_DOORBELL64_INVALID                 = 0xFFFF
 } AMDGPU_DOORBELL64_ASSIGNMENT;
 
+typedef enum _AMDGPU_DOORBELL_ASSIGNMENT_LAYOUT1
+{
+	/* KIQ: 0~7 for maximum 8 XCD */
+	AMDGPU_DOORBELL_LAYOUT1_KIQ_START               = 0x000,
+	AMDGPU_DOORBELL_LAYOUT1_HIQ                     = 0x008,
+	AMDGPU_DOORBELL_LAYOUT1_DIQ                     = 0x009,
+	/* Compute: 0x0A ~ 0x49 */
+	AMDGPU_DOORBELL_LAYOUT1_MEC_RING_START          = 0x00A,
+	AMDGPU_DOORBELL_LAYOUT1_MEC_RING_END            = 0x049,
+	AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_START         = 0x04A,
+	AMDGPU_DOORBELL_LAYOUT1_USERQUEUE_END           = 0x0C9,
+	/* SDMA: 0x100 ~ 0x19F */
+	AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_START       = 0x100,
+	AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_END         = 0x19F,
+	/* IH: 0x1A0 ~ 0x1AF */
+	AMDGPU_DOORBELL_LAYOUT1_IH                      = 0x1A0,
+	/* VCN: 0x1B0 ~ 0x1C2 */
+	AMDGPU_DOORBELL_LAYOUT1_VCN_START               = 0x1B0,
+	AMDGPU_DOORBELL_LAYOUT1_VCN_END                 = 0x1C2,
+
+	AMDGPU_DOORBELL_LAYOUT1_FIRST_NON_CP            = AMDGPU_DOORBELL_LAYOUT1_sDMA_ENGINE_START,
+	AMDGPU_DOORBELL_LAYOUT1_LAST_NON_CP             = AMDGPU_DOORBELL_LAYOUT1_VCN_END,
+
+	AMDGPU_DOORBELL_LAYOUT1_MAX_ASSIGNMENT          = 0x1C2,
+	AMDGPU_DOORBELL_LAYOUT1_INVALID                 = 0xFFFF
+} AMDGPU_DOORBELL_ASSIGNMENT_LAYOUT1;
+
 u32 amdgpu_mm_rdoorbell(struct amdgpu_device *adev, u32 index);
 void amdgpu_mm_wdoorbell(struct amdgpu_device *adev, u32 index, u32 v);
 u64 amdgpu_mm_rdoorbell64(struct amdgpu_device *adev, u32 index);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
index 49b18cf987aa..76438f197de1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c
@@ -315,11 +315,7 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
 	ring->use_doorbell = true;
 	ring->xcc_id = xcc_id;
 	ring->vm_hub = AMDGPU_GFXHUB(xcc_id);
-	if (xcc_id >= 1)
-		ring->doorbell_index = (adev->doorbell_index.xcc1_kiq_start +
-					xcc_id - 1) << 1;
-	else
-		ring->doorbell_index = adev->doorbell_index.kiq << 1;
+	ring->doorbell_index = (adev->doorbell_index.kiq + xcc_id) << 1;
 
 	r = amdgpu_gfx_kiq_acquire(adev, ring, xcc_id);
 	if (r)
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
index a9fab8de29e8..52185b1d5d31 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c
@@ -817,13 +817,7 @@ static int gfx_v9_4_3_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 
 	ring->ring_obj = NULL;
 	ring->use_doorbell = true;
-	if (xcc_id >= 1)
-		ring->doorbell_index =
-				(adev->doorbell_index.xcc1_mec_ring0_start +
-				ring_id - adev->gfx.num_compute_rings) << 1;
-	else
-		ring->doorbell_index =
-				(adev->doorbell_index.mec_ring0 + ring_id) << 1;
+	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
 	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
 				+ (ring_id * GFX9_MEC_HPD_SIZE);
 	ring->vm_hub = AMDGPU_GFXHUB(xcc_id);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index d3c7f9a43ef1..9b53174925f8 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1310,14 +1310,7 @@ static int sdma_v4_4_2_sw_init(void *handle)
 				ring->use_doorbell?"true":"false");
 
 		/* doorbell size is 2 dwords, get DWORD offset */
-		if (aid_id > 0)
-			ring->doorbell_index =
-				(adev->doorbell_index.aid1_sdma_start << 1)
-				+ adev->doorbell_index.sdma_doorbell_range
-				* (i - adev->sdma.num_inst_per_aid);
-		else
-			ring->doorbell_index =
-				adev->doorbell_index.sdma_engine[i] << 1;
+		ring->doorbell_index = adev->doorbell_index.sdma_engine[i] << 1;
 		ring->vm_hub = AMDGPU_MMHUB0(aid_id);
 
 		sprintf(ring->name, "sdma%d.%d", aid_id,
@@ -1336,14 +1329,8 @@ static int sdma_v4_4_2_sw_init(void *handle)
 			/* doorbell index of page queue is assigned right after
 			 * gfx queue on the same instance
 			 */
-			if (aid_id > 0)
-				ring->doorbell_index =
-					((adev->doorbell_index.aid1_sdma_start + 1) << 1)
-					+ adev->doorbell_index.sdma_doorbell_range
-					* (i - adev->sdma.num_inst_per_aid);
-			else
-				ring->doorbell_index =
-					(adev->doorbell_index.sdma_engine[i] + 1) << 1;
+			ring->doorbell_index =
+				(adev->doorbell_index.sdma_engine[i] + 1) << 1;
 			ring->vm_hub = AMDGPU_MMHUB0(aid_id);
 
 			sprintf(ring->name, "page%d.%d", aid_id,
-- 
cgit 


From f8b34a0518701bae8bb02fabe129f01b1dc33336 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Wed, 29 Jun 2022 21:04:39 +0530
Subject: drm/amdgpu: Use instance table for sdma 4.4.2

For ASICs with sdma IP v4.4.2, add mapping for logical to physical
instances.

v2:
	Register accesses on bare metal should be based on physical
instance. Use GET_INST() to get physical instance.

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Le Ma <Le.Ma@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 .../gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c    |  7 ++--
 drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c             |  7 ++--
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c           | 38 +++++++++++++++++-----
 3 files changed, 38 insertions(+), 14 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
index 5daec0b45545..772774615cb8 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gc_9_4_3.c
@@ -43,9 +43,10 @@ static uint32_t get_sdma_rlc_reg_offset(struct amdgpu_device *adev,
 					unsigned int engine_id,
 					unsigned int queue_id)
 {
-	uint32_t sdma_engine_reg_base = SOC15_REG_OFFSET(SDMA0, engine_id,
-					regSDMA_RLC0_RB_CNTL) -
-					regSDMA_RLC0_RB_CNTL;
+	uint32_t sdma_engine_reg_base =
+		SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, engine_id),
+				 regSDMA_RLC0_RB_CNTL) -
+		regSDMA_RLC0_RB_CNTL;
 	uint32_t retval = sdma_engine_reg_base +
 		  queue_id * (regSDMA_RLC1_RB_CNTL - regSDMA_RLC0_RB_CNTL);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c
index 5a5bab665f60..e6c1004e4e76 100644
--- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c
+++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_9.c
@@ -75,7 +75,10 @@ static void nbio_v7_9_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 			bool use_doorbell, int doorbell_index, int doorbell_size)
 {
 	u32 doorbell_range = 0, doorbell_ctrl = 0;
-	int aid_id = adev->sdma.instance[instance].aid_id;
+	int aid_id, dev_inst;
+
+	dev_inst = GET_INST(SDMA0, instance);
+	aid_id = adev->sdma.instance[instance].aid_id;
 
 	if (use_doorbell == false)
 		return;
@@ -93,7 +96,7 @@ static void nbio_v7_9_sdma_doorbell_range(struct amdgpu_device *adev, int instan
 		REG_SET_FIELD(doorbell_ctrl, S2A_DOORBELL_ENTRY_1_CTRL,
 			S2A_DOORBELL_PORT1_RANGE_SIZE, doorbell_size);
 
-	switch (instance % adev->sdma.num_inst_per_aid) {
+	switch (dev_inst % adev->sdma.num_inst_per_aid) {
 	case 0:
 		WREG32(SOC15_REG_OFFSET(NBIO, 0, regDOORBELL0_CTRL_ENTRY_1) +
 			4 * aid_id, doorbell_range);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 9b53174925f8..4350939105c5 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -57,7 +57,9 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev);
 static u32 sdma_v4_4_2_get_reg_offset(struct amdgpu_device *adev,
 		u32 instance, u32 offset)
 {
-	return (adev->reg_offset[SDMA0_HWIP][instance][0] + offset);
+	u32 dev_inst = GET_INST(SDMA0, instance);
+
+	return (adev->reg_offset[SDMA0_HWIP][dev_inst][0] + offset);
 }
 
 static unsigned sdma_v4_4_2_seq_to_irq_id(int seq_num)
@@ -1475,16 +1477,31 @@ static int sdma_v4_4_2_process_trap_irq(struct amdgpu_device *adev,
 				      struct amdgpu_irq_src *source,
 				      struct amdgpu_iv_entry *entry)
 {
-	uint32_t instance;
+	uint32_t instance, i;
 
 	DRM_DEBUG("IH: SDMA trap\n");
 	instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
-	instance += node_id_to_phys_map[entry->node_id] *
-			adev->sdma.num_inst_per_aid;
+
+	/* Client id gives the SDMA instance in AID. To know the exact SDMA
+	 * instance, interrupt entry gives the node id which corresponds to the AID instance.
+	 * Match node id with the AID id associated with the SDMA instance. */
+	for (i = instance; i < adev->sdma.num_instances;
+	     i += adev->sdma.num_inst_per_aid) {
+		if (adev->sdma.instance[i].aid_id ==
+		    node_id_to_phys_map[entry->node_id])
+			break;
+	}
+
+	if (i >= adev->sdma.num_instances) {
+		dev_WARN_ONCE(
+			adev->dev, 1,
+			"Couldn't find the right sdma instance in trap handler");
+		return 0;
+	}
 
 	switch (entry->ring_id) {
 	case 0:
-		amdgpu_fence_process(&adev->sdma.instance[instance].ring);
+		amdgpu_fence_process(&adev->sdma.instance[i].ring);
 		break;
 	default:
 		break;
@@ -1717,12 +1734,12 @@ static void sdma_v4_4_2_get_clockgating_state(void *handle, u64 *flags)
 		*flags = 0;
 
 	/* AMD_CG_SUPPORT_SDMA_MGCG */
-	data = RREG32(SOC15_REG_OFFSET(SDMA0, 0, regSDMA_CLK_CTRL));
+	data = RREG32(SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, 0), regSDMA_CLK_CTRL));
 	if (!(data & SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK))
 		*flags |= AMD_CG_SUPPORT_SDMA_MGCG;
 
 	/* AMD_CG_SUPPORT_SDMA_LS */
-	data = RREG32(SOC15_REG_OFFSET(SDMA0, 0, regSDMA_POWER_CNTL));
+	data = RREG32(SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, 0), regSDMA_POWER_CNTL));
 	if (data & SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK)
 		*flags |= AMD_CG_SUPPORT_SDMA_LS;
 }
@@ -1809,7 +1826,7 @@ static const struct amdgpu_ring_funcs sdma_v4_4_2_page_ring_funcs = {
 
 static void sdma_v4_4_2_set_ring_funcs(struct amdgpu_device *adev)
 {
-	int i;
+	int i, dev_inst;
 
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		adev->sdma.instance[i].ring.funcs = &sdma_v4_4_2_ring_funcs;
@@ -1820,7 +1837,10 @@ static void sdma_v4_4_2_set_ring_funcs(struct amdgpu_device *adev)
 			adev->sdma.instance[i].page.me = i;
 		}
 
-		adev->sdma.instance[i].aid_id = i / adev->sdma.num_inst_per_aid;
+		dev_inst = GET_INST(SDMA0, i);
+		/* AID to which SDMA belongs depends on physical instance */
+		adev->sdma.instance[i].aid_id =
+			dev_inst / adev->sdma.num_inst_per_aid;
 	}
 }
 
-- 
cgit 


From 5f09237b82e51be8a1849eb658dcb153748f8f60 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Mon, 31 Oct 2022 10:46:05 +0530
Subject: drm/amdgpu: Add SDMA v4.4.2 golden settings

Add programming of SDMA golden settings for v4.4.2

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 4350939105c5..016813b295ac 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -96,11 +96,22 @@ static int sdma_v4_4_2_irq_id_to_seq(unsigned client_id)
 
 static void sdma_v4_4_2_init_golden_registers(struct amdgpu_device *adev)
 {
-	switch (adev->ip_versions[SDMA0_HWIP][0]) {
-	case IP_VERSION(4, 4, 2):
-		break;
-	default:
-		break;
+	u32 val;
+	int i;
+
+	for (i = 0; i < adev->sdma.num_instances; i++) {
+		val = RREG32_SDMA(i, regSDMA_GB_ADDR_CONFIG);
+		val = REG_SET_FIELD(val, SDMA_GB_ADDR_CONFIG, NUM_BANKS, 4);
+		val = REG_SET_FIELD(val, SDMA_GB_ADDR_CONFIG,
+				    PIPE_INTERLEAVE_SIZE, 0);
+		WREG32_SDMA(i, regSDMA_GB_ADDR_CONFIG, val);
+
+		val = RREG32_SDMA(i, regSDMA_GB_ADDR_CONFIG_READ);
+		val = REG_SET_FIELD(val, SDMA_GB_ADDR_CONFIG_READ, NUM_BANKS,
+				    4);
+		val = REG_SET_FIELD(val, SDMA_GB_ADDR_CONFIG_READ,
+				    PIPE_INTERLEAVE_SIZE, 0);
+		WREG32_SDMA(i, regSDMA_GB_ADDR_CONFIG_READ, val);
 	}
 }
 
-- 
cgit 


From 527c670e5323414dbef8f4719dc9b348a50ac1c8 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Wed, 14 Sep 2022 12:48:08 +0530
Subject: drm/amdgpu: Add sdma instance specific functions

SDMA 4.4.2 supports multiple instances. Add functions to support
handling of each SDMA instance separately.

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 162 +++++++++++++++++--------------
 1 file changed, 91 insertions(+), 71 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 016813b295ac..6a971e15b4b0 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -94,7 +94,8 @@ static int sdma_v4_4_2_irq_id_to_seq(unsigned client_id)
 	}
 }
 
-static void sdma_v4_4_2_init_golden_registers(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_init_golden_registers(struct amdgpu_device *adev,
+						   uint32_t inst_mask)
 {
 	u32 val;
 	int i;
@@ -418,13 +419,14 @@ static void sdma_v4_4_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
  *
  * Stop the gfx async dma ring buffers.
  */
-static void sdma_v4_4_2_gfx_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_gfx_stop(struct amdgpu_device *adev,
+				      uint32_t inst_mask)
 {
 	struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES];
 	u32 rb_cntl, ib_cntl;
 	int i, unset = 0;
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		sdma[i] = &adev->sdma.instance[i].ring;
 
 		if ((adev->mman.buffer_funcs_ring == sdma[i]) && unset != 1) {
@@ -448,7 +450,8 @@ static void sdma_v4_4_2_gfx_stop(struct amdgpu_device *adev)
  *
  * Stop the compute async dma queues.
  */
-static void sdma_v4_4_2_rlc_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_rlc_stop(struct amdgpu_device *adev,
+				      uint32_t inst_mask)
 {
 	/* XXX todo */
 }
@@ -460,14 +463,15 @@ static void sdma_v4_4_2_rlc_stop(struct amdgpu_device *adev)
  *
  * Stop the page async dma ring buffers.
  */
-static void sdma_v4_4_2_page_stop(struct amdgpu_device *adev)
+static void sdma_v4_4_2_inst_page_stop(struct amdgpu_device *adev,
+				       uint32_t inst_mask)
 {
 	struct amdgpu_ring *sdma[AMDGPU_MAX_SDMA_INSTANCES];
 	u32 rb_cntl, ib_cntl;
 	int i;
 	bool unset = false;
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		sdma[i] = &adev->sdma.instance[i].page;
 
 		if ((adev->mman.buffer_funcs_ring == sdma[i]) &&
@@ -495,7 +499,8 @@ static void sdma_v4_4_2_page_stop(struct amdgpu_device *adev)
  *
  * Halt or unhalt the async dma engines context switch.
  */
-static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enable)
+static void sdma_v4_4_2_inst_ctx_switch_enable(struct amdgpu_device *adev,
+					       bool enable, uint32_t inst_mask)
 {
 	u32 f32_cntl, phase_quantum = 0;
 	int i;
@@ -524,7 +529,7 @@ static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enabl
 			unit  << SDMA_PHASE0_QUANTUM__UNIT__SHIFT;
 	}
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		f32_cntl = RREG32_SDMA(i, regSDMA_CNTL);
 		f32_cntl = REG_SET_FIELD(f32_cntl, SDMA_CNTL,
 				AUTO_CTXSW_ENABLE, enable ? 1 : 0);
@@ -538,7 +543,6 @@ static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enabl
 		/* Extend page fault timeout to avoid interrupt storm */
 		WREG32_SDMA(i, regSDMA_UTCL1_TIMEOUT, 0x00800080);
 	}
-
 }
 
 /**
@@ -546,22 +550,24 @@ static void sdma_v4_4_2_ctx_switch_enable(struct amdgpu_device *adev, bool enabl
  *
  * @adev: amdgpu_device pointer
  * @enable: enable/disable the DMA MEs.
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Halt or unhalt the async dma engines.
  */
-static void sdma_v4_4_2_enable(struct amdgpu_device *adev, bool enable)
+static void sdma_v4_4_2_inst_enable(struct amdgpu_device *adev, bool enable,
+				    uint32_t inst_mask)
 {
 	u32 f32_cntl;
 	int i;
 
 	if (!enable) {
-		sdma_v4_4_2_gfx_stop(adev);
-		sdma_v4_4_2_rlc_stop(adev);
+		sdma_v4_4_2_inst_gfx_stop(adev, inst_mask);
+		sdma_v4_4_2_inst_rlc_stop(adev, inst_mask);
 		if (adev->sdma.has_page_queue)
-			sdma_v4_4_2_page_stop(adev);
+			sdma_v4_4_2_inst_page_stop(adev, inst_mask);
 	}
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		f32_cntl = RREG32_SDMA(i, regSDMA_F32_CNTL);
 		f32_cntl = REG_SET_FIELD(f32_cntl, SDMA_F32_CNTL, HALT, enable ? 0 : 1);
 		WREG32_SDMA(i, regSDMA_F32_CNTL, f32_cntl);
@@ -780,7 +786,8 @@ static void sdma_v4_4_2_init_pg(struct amdgpu_device *adev)
  * Set up the compute DMA queues and enable them.
  * Returns 0 for success, error for failure.
  */
-static int sdma_v4_4_2_rlc_resume(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_rlc_resume(struct amdgpu_device *adev,
+				       uint32_t inst_mask)
 {
 	sdma_v4_4_2_init_pg(adev);
 
@@ -795,7 +802,8 @@ static int sdma_v4_4_2_rlc_resume(struct amdgpu_device *adev)
  * Loads the sDMA0/1 ucode.
  * Returns 0 for success, -EINVAL if the ucode is not available.
  */
-static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_load_microcode(struct amdgpu_device *adev,
+					   uint32_t inst_mask)
 {
 	const struct sdma_firmware_header_v1_0 *hdr;
 	const __le32 *fw_data;
@@ -803,9 +811,9 @@ static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
 	int i, j;
 
 	/* halt the MEs */
-	sdma_v4_4_2_enable(adev, false);
+	sdma_v4_4_2_inst_enable(adev, false, inst_mask);
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	for_each_inst(i, inst_mask) {
 		if (!adev->sdma.instance[i].fw)
 			return -EINVAL;
 
@@ -831,38 +839,41 @@ static int sdma_v4_4_2_load_microcode(struct amdgpu_device *adev)
 }
 
 /**
- * sdma_v4_4_2_start - setup and start the async dma engines
+ * sdma_v4_4_2_inst_start - setup and start the async dma engines
  *
  * @adev: amdgpu_device pointer
  *
  * Set up the DMA engines and enable them.
  * Returns 0 for success, error for failure.
  */
-static int sdma_v4_4_2_start(struct amdgpu_device *adev)
+static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev,
+				  uint32_t inst_mask)
 {
 	struct amdgpu_ring *ring;
+	uint32_t tmp_mask;
 	int i, r = 0;
 
 	if (amdgpu_sriov_vf(adev)) {
-		sdma_v4_4_2_ctx_switch_enable(adev, false);
-		sdma_v4_4_2_enable(adev, false);
+		sdma_v4_4_2_inst_ctx_switch_enable(adev, false, inst_mask);
+		sdma_v4_4_2_inst_enable(adev, false, inst_mask);
 	} else {
 		/* bypass sdma microcode loading on Gopher */
 		if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP &&
-		    !(adev->pdev->device == 0x49) && !(adev->pdev->device == 0x50)) {
-			r = sdma_v4_4_2_load_microcode(adev);
+		    adev->sdma.instance[0].fw) {
+			r = sdma_v4_4_2_inst_load_microcode(adev, inst_mask);
 			if (r)
 				return r;
 		}
 
 		/* unhalt the MEs */
-		sdma_v4_4_2_enable(adev, true);
+		sdma_v4_4_2_inst_enable(adev, true, inst_mask);
 		/* enable sdma ring preemption */
-		sdma_v4_4_2_ctx_switch_enable(adev, true);
+		sdma_v4_4_2_inst_ctx_switch_enable(adev, true, inst_mask);
 	}
 
 	/* start the gfx rings and rlc compute queues */
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	tmp_mask = inst_mask;
+	for_each_inst(i, tmp_mask) {
 		uint32_t temp;
 
 		WREG32_SDMA(i, regSDMA_SEM_WAIT_FAIL_TIMER_CNTL, 0);
@@ -889,15 +900,16 @@ static int sdma_v4_4_2_start(struct amdgpu_device *adev)
 	}
 
 	if (amdgpu_sriov_vf(adev)) {
-		sdma_v4_4_2_ctx_switch_enable(adev, true);
-		sdma_v4_4_2_enable(adev, true);
+		sdma_v4_4_2_inst_ctx_switch_enable(adev, true, inst_mask);
+		sdma_v4_4_2_inst_enable(adev, true, inst_mask);
 	} else {
-		r = sdma_v4_4_2_rlc_resume(adev);
+		r = sdma_v4_4_2_inst_rlc_resume(adev, inst_mask);
 		if (r)
 			return r;
 	}
 
-	for (i = 0; i < adev->sdma.num_instances; i++) {
+	tmp_mask = inst_mask;
+	for_each_inst(i, tmp_mask) {
 		ring = &adev->sdma.instance[i].ring;
 
 		r = amdgpu_ring_test_helper(ring);
@@ -1383,14 +1395,17 @@ static int sdma_v4_4_2_hw_init(void *handle)
 {
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	uint32_t inst_mask;
 
+	/* TODO: Check if this is needed */
 	if (adev->flags & AMD_IS_APU)
 		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, false);
 
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
 	if (!amdgpu_sriov_vf(adev))
-		sdma_v4_4_2_init_golden_registers(adev);
+		sdma_v4_4_2_inst_init_golden_registers(adev, inst_mask);
 
-	r = sdma_v4_4_2_start(adev);
+	r = sdma_v4_4_2_inst_start(adev, inst_mask);
 
 	return r;
 }
@@ -1398,22 +1413,27 @@ static int sdma_v4_4_2_hw_init(void *handle)
 static int sdma_v4_4_2_hw_fini(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	uint32_t inst_mask;
 	int i;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
 	for (i = 0; i < adev->sdma.num_instances; i++) {
 		amdgpu_irq_put(adev, &adev->sdma.ecc_irq,
 			       AMDGPU_SDMA_IRQ_INSTANCE0 + i);
 	}
 
-	sdma_v4_4_2_ctx_switch_enable(adev, false);
-	sdma_v4_4_2_enable(adev, false);
+	sdma_v4_4_2_inst_ctx_switch_enable(adev, false, inst_mask);
+	sdma_v4_4_2_inst_enable(adev, false, inst_mask);
 
 	return 0;
 }
 
+static int sdma_v4_4_2_set_clockgating_state(void *handle,
+					     enum amd_clockgating_state state);
+
 static int sdma_v4_4_2_suspend(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
@@ -1650,15 +1670,39 @@ static int sdma_v4_4_2_process_srbm_write_irq(struct amdgpu_device *adev,
 	return 0;
 }
 
-static void sdma_v4_4_2_update_medium_grain_clock_gating(
-		struct amdgpu_device *adev,
-		bool enable)
+static void sdma_v4_4_2_inst_update_medium_grain_light_sleep(
+	struct amdgpu_device *adev, bool enable, uint32_t inst_mask)
+{
+	uint32_t data, def;
+	int i;
+
+	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) {
+		for_each_inst(i, inst_mask) {
+			/* 1-not override: enable sdma mem light sleep */
+			def = data = RREG32_SDMA(i, regSDMA_POWER_CNTL);
+			data |= SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
+			if (def != data)
+				WREG32_SDMA(i, regSDMA_POWER_CNTL, data);
+		}
+	} else {
+		for_each_inst(i, inst_mask) {
+			/* 0-override:disable sdma mem light sleep */
+			def = data = RREG32_SDMA(i, regSDMA_POWER_CNTL);
+			data &= ~SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
+			if (def != data)
+				WREG32_SDMA(i, regSDMA_POWER_CNTL, data);
+		}
+	}
+}
+
+static void sdma_v4_4_2_inst_update_medium_grain_clock_gating(
+	struct amdgpu_device *adev, bool enable, uint32_t inst_mask)
 {
 	uint32_t data, def;
 	int i;
 
 	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG)) {
-		for (i = 0; i < adev->sdma.num_instances; i++) {
+		for_each_inst(i, inst_mask) {
 			def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
 			data &= ~(SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
@@ -1672,7 +1716,7 @@ static void sdma_v4_4_2_update_medium_grain_clock_gating(
 				WREG32_SDMA(i, regSDMA_CLK_CTRL, data);
 		}
 	} else {
-		for (i = 0; i < adev->sdma.num_instances; i++) {
+		for_each_inst(i, inst_mask) {
 			def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
 			data |= (SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
@@ -1688,45 +1732,21 @@ static void sdma_v4_4_2_update_medium_grain_clock_gating(
 	}
 }
 
-
-static void sdma_v4_4_2_update_medium_grain_light_sleep(
-		struct amdgpu_device *adev,
-		bool enable)
-{
-	uint32_t data, def;
-	int i;
-
-	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) {
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-			/* 1-not override: enable sdma mem light sleep */
-			def = data = RREG32_SDMA(0, regSDMA_POWER_CNTL);
-			data |= SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
-			if (def != data)
-				WREG32_SDMA(0, regSDMA_POWER_CNTL, data);
-		}
-	} else {
-		for (i = 0; i < adev->sdma.num_instances; i++) {
-		/* 0-override:disable sdma mem light sleep */
-			def = data = RREG32_SDMA(0, regSDMA_POWER_CNTL);
-			data &= ~SDMA_POWER_CNTL__MEM_POWER_OVERRIDE_MASK;
-			if (def != data)
-				WREG32_SDMA(0, regSDMA_POWER_CNTL, data);
-		}
-	}
-}
-
 static int sdma_v4_4_2_set_clockgating_state(void *handle,
 					  enum amd_clockgating_state state)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	uint32_t inst_mask;
 
 	if (amdgpu_sriov_vf(adev))
 		return 0;
 
-	sdma_v4_4_2_update_medium_grain_clock_gating(adev,
-			state == AMD_CG_STATE_GATE);
-	sdma_v4_4_2_update_medium_grain_light_sleep(adev,
-			state == AMD_CG_STATE_GATE);
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
+
+	sdma_v4_4_2_inst_update_medium_grain_clock_gating(
+		adev, state == AMD_CG_STATE_GATE, inst_mask);
+	sdma_v4_4_2_inst_update_medium_grain_light_sleep(
+		adev, state == AMD_CG_STATE_GATE, inst_mask);
 	return 0;
 }
 
-- 
cgit 


From 3446cb78f3a8033fda416825282e7cafecc83129 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Fri, 23 Sep 2022 15:40:15 +0530
Subject: drm/amdgpu: Add SDMA v4.4.2 XCP funcs

Add functions required to suspend/resume instances of SDMA which
are part of an XCP.

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 36 ++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 6a971e15b4b0..184eb7902722 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -27,6 +27,7 @@
 #include <linux/pci.h>
 
 #include "amdgpu.h"
+#include "amdgpu_xcp.h"
 #include "amdgpu_ucode.h"
 #include "amdgpu_trace.h"
 
@@ -2025,3 +2026,38 @@ const struct amdgpu_ip_block_version sdma_v4_4_2_ip_block = {
 	.rev = 0,
 	.funcs = &sdma_v4_4_2_ip_funcs,
 };
+
+static int sdma_v4_4_2_xcp_resume(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	int r;
+
+	if (!amdgpu_sriov_vf(adev))
+		sdma_v4_4_2_inst_init_golden_registers(adev, inst_mask);
+
+	r = sdma_v4_4_2_inst_start(adev, inst_mask);
+
+	return r;
+}
+
+static int sdma_v4_4_2_xcp_suspend(void *handle, uint32_t inst_mask)
+{
+	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
+	uint32_t tmp_mask = inst_mask;
+	int i;
+
+	for_each_inst(i, tmp_mask) {
+		amdgpu_irq_put(adev, &adev->sdma.ecc_irq,
+			       AMDGPU_SDMA_IRQ_INSTANCE0 + i);
+	}
+
+	sdma_v4_4_2_inst_ctx_switch_enable(adev, false, inst_mask);
+	sdma_v4_4_2_inst_enable(adev, false, inst_mask);
+
+	return 0;
+}
+
+struct amdgpu_xcp_ip_funcs sdma_v4_4_2_xcp_funcs = {
+	.suspend = &sdma_v4_4_2_xcp_suspend,
+	.resume = &sdma_v4_4_2_xcp_resume
+};
-- 
cgit 


From 30b52995c294b3c9a002ff2f90f773cbf8850714 Mon Sep 17 00:00:00 2001
From: Asad kamal <asad.kamal@amd.com>
Date: Tue, 3 Jan 2023 13:14:58 +0800
Subject: drm/amdgpu: Remove SMU powergate message call for SDMA

SDMA v4.4.2 doesn't need explicit power gating control through PMFW

Signed-off-by: Asad kamal <asad.kamal@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 4 ----
 1 file changed, 4 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 184eb7902722..7c91cbd91ac8 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1398,10 +1398,6 @@ static int sdma_v4_4_2_hw_init(void *handle)
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 	uint32_t inst_mask;
 
-	/* TODO: Check if this is needed */
-	if (adev->flags & AMD_IS_APU)
-		amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, false);
-
 	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
 	if (!amdgpu_sriov_vf(adev))
 		sdma_v4_4_2_inst_init_golden_registers(adev, inst_mask);
-- 
cgit 


From 7389c75114c53b061d686f19dff5833adaf96cb8 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Tue, 17 Jan 2023 16:54:49 +0530
Subject: drm/amdgpu: Keep SDMAv4.4.2 active during reset

During ASIC wide reset, SDMA shouldn't be clockgated and be ready to
accept freeze requests from PMFW. For that, don't stop SDMA engine
during reset and keep the clocks active.

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 8 ++++++++
 1 file changed, 8 insertions(+)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 7c91cbd91ac8..729e26a4a2e7 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -566,6 +566,11 @@ static void sdma_v4_4_2_inst_enable(struct amdgpu_device *adev, bool enable,
 		sdma_v4_4_2_inst_rlc_stop(adev, inst_mask);
 		if (adev->sdma.has_page_queue)
 			sdma_v4_4_2_inst_page_stop(adev, inst_mask);
+
+		/* SDMA FW needs to respond to FREEZE requests during reset.
+		 * Keep it running during reset */
+		if (!amdgpu_sriov_vf(adev) && amdgpu_in_reset(adev))
+			return;
 	}
 
 	for_each_inst(i, inst_mask) {
@@ -1435,6 +1440,9 @@ static int sdma_v4_4_2_suspend(void *handle)
 {
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	if (amdgpu_in_reset(adev))
+		sdma_v4_4_2_set_clockgating_state(adev, AMD_CG_STATE_UNGATE);
+
 	return sdma_v4_4_2_hw_fini(adev);
 }
 
-- 
cgit 


From 35ff4301ebc37bd45c18edae08afd2983dc9a338 Mon Sep 17 00:00:00 2001
From: Le Ma <le.ma@amd.com>
Date: Fri, 3 Feb 2023 14:38:33 +0800
Subject: drm/amdgpu: enable context empty interrupt on sdma v4.4.2

With SDMA_CTNL.CTXEMPTY_INT_ENABLE set, the F32 clock can be gated when
SDMA finishes all job and goes to idle.

And no specific interrupt handling is required in driver.

Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 2 ++
 1 file changed, 2 insertions(+)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 729e26a4a2e7..7aa26e716a2d 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -890,6 +890,8 @@ static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev,
 		/* set utc l1 enable flag always to 1 */
 		temp = RREG32_SDMA(i, regSDMA_CNTL);
 		temp = REG_SET_FIELD(temp, SDMA_CNTL, UTC_L1_ENABLE, 1);
+		/* enable context empty interrupt during initialization */
+		temp = REG_SET_FIELD(temp, SDMA_CNTL, CTXEMPTY_INT_ENABLE, 1);
 		WREG32_SDMA(i, regSDMA_CNTL, temp);
 
 		if (!amdgpu_sriov_vf(adev)) {
-- 
cgit 


From cbf9e46ab0fec29ac39d05d9a87fa66122bc9783 Mon Sep 17 00:00:00 2001
From: Lijo Lazar <lijo.lazar@amd.com>
Date: Fri, 3 Feb 2023 13:17:51 +0530
Subject: drm/amdgpu: Enable MGCG on SDMAv4.4.2

Enable clock gating on SDMAv4.4.2 versions. Leave memory light sleep to
default.

Signed-off-by: Lijo Lazar <lijo.lazar@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 22 +++++++++++++---------
 drivers/gpu/drm/amd/amdgpu/soc15.c       |  4 ++--
 2 files changed, 15 insertions(+), 11 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 7aa26e716a2d..dca0beec0252 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -1683,7 +1683,11 @@ static void sdma_v4_4_2_inst_update_medium_grain_light_sleep(
 	uint32_t data, def;
 	int i;
 
-	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS)) {
+	/* leave as default if it is not driver controlled */
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS))
+		return;
+
+	if (enable) {
 		for_each_inst(i, inst_mask) {
 			/* 1-not override: enable sdma mem light sleep */
 			def = data = RREG32_SDMA(i, regSDMA_POWER_CNTL);
@@ -1708,12 +1712,14 @@ static void sdma_v4_4_2_inst_update_medium_grain_clock_gating(
 	uint32_t data, def;
 	int i;
 
-	if (enable && (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG)) {
+	/* leave as default if it is not driver controlled */
+	if (!(adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG))
+		return;
+
+	if (enable) {
 		for_each_inst(i, inst_mask) {
 			def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
-			data &= ~(SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
-				  SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
-				  SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
+			data &= ~(SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE4_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE3_MASK |
 				  SDMA_CLK_CTRL__SOFT_OVERRIDE2_MASK |
@@ -1725,9 +1731,7 @@ static void sdma_v4_4_2_inst_update_medium_grain_clock_gating(
 	} else {
 		for_each_inst(i, inst_mask) {
 			def = data = RREG32_SDMA(i, regSDMA_CLK_CTRL);
-			data |= (SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK |
-				 SDMA_CLK_CTRL__SOFT_OVERRIDE6_MASK |
-				 SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
+			data |= (SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE4_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE3_MASK |
 				 SDMA_CLK_CTRL__SOFT_OVERRIDE2_MASK |
@@ -1773,7 +1777,7 @@ static void sdma_v4_4_2_get_clockgating_state(void *handle, u64 *flags)
 
 	/* AMD_CG_SUPPORT_SDMA_MGCG */
 	data = RREG32(SOC15_REG_OFFSET(SDMA0, GET_INST(SDMA0, 0), regSDMA_CLK_CTRL));
-	if (!(data & SDMA_CLK_CTRL__SOFT_OVERRIDE7_MASK))
+	if (!(data & SDMA_CLK_CTRL__SOFT_OVERRIDE5_MASK))
 		*flags |= AMD_CG_SUPPORT_SDMA_MGCG;
 
 	/* AMD_CG_SUPPORT_SDMA_LS */
diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c
index 5a63ec441a5b..100fdf5074b4 100644
--- a/drivers/gpu/drm/amd/amdgpu/soc15.c
+++ b/drivers/gpu/drm/amd/amdgpu/soc15.c
@@ -1146,8 +1146,8 @@ static int soc15_common_early_init(void *handle)
 		adev->asic_funcs = &aqua_vanjaram_asic_funcs;
 		adev->cg_flags =
 			AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_CGCG |
-			AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_VCN_MGCG |
-			AMD_CG_SUPPORT_JPEG_MGCG;
+			AMD_CG_SUPPORT_GFX_CGLS | AMD_CG_SUPPORT_SDMA_MGCG |
+			AMD_CG_SUPPORT_VCN_MGCG | AMD_CG_SUPPORT_JPEG_MGCG;
 		adev->pg_flags =
 			AMD_PG_SUPPORT_VCN |
 			AMD_PG_SUPPORT_VCN_DPG |
-- 
cgit 


From dc37a9194ad20c4f09f22ec79cc2b5e5eb57c5a2 Mon Sep 17 00:00:00 2001
From: Hawking Zhang <Hawking.Zhang@amd.com>
Date: Sun, 5 Feb 2023 22:54:50 +0800
Subject: drm/amdgpu: Add query_ras_error_count for sdma v4_4_2

Add query_ras_error_count callback for sdma
v4_4_2. It will be used to query and log sdma
uncorrectable error count and memory block.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h | 28 ++++++++++++++
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 64 ++++++++++++++++++++++++++++++++
 2 files changed, 92 insertions(+)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
index 62afb282a3ff..513ac22120c1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h
@@ -62,6 +62,34 @@ struct amdgpu_sdma_instance {
 	uint32_t		aid_id;
 };
 
+enum amdgpu_sdma_ras_memory_id {
+	AMDGPU_SDMA_MBANK_DATA_BUF0 = 1,
+	AMDGPU_SDMA_MBANK_DATA_BUF1 = 2,
+	AMDGPU_SDMA_MBANK_DATA_BUF2 = 3,
+	AMDGPU_SDMA_MBANK_DATA_BUF3 = 4,
+	AMDGPU_SDMA_MBANK_DATA_BUF4 = 5,
+	AMDGPU_SDMA_MBANK_DATA_BUF5 = 6,
+	AMDGPU_SDMA_MBANK_DATA_BUF6 = 7,
+	AMDGPU_SDMA_MBANK_DATA_BUF7 = 8,
+	AMDGPU_SDMA_MBANK_DATA_BUF8 = 9,
+	AMDGPU_SDMA_MBANK_DATA_BUF9 = 10,
+	AMDGPU_SDMA_MBANK_DATA_BUF10 = 11,
+	AMDGPU_SDMA_MBANK_DATA_BUF11 = 12,
+	AMDGPU_SDMA_MBANK_DATA_BUF12 = 13,
+	AMDGPU_SDMA_MBANK_DATA_BUF13 = 14,
+	AMDGPU_SDMA_MBANK_DATA_BUF14 = 15,
+	AMDGPU_SDMA_MBANK_DATA_BUF15 = 16,
+	AMDGPU_SDMA_UCODE_BUF = 17,
+	AMDGPU_SDMA_RB_CMD_BUF = 18,
+	AMDGPU_SDMA_IB_CMD_BUF = 19,
+	AMDGPU_SDMA_UTCL1_RD_FIFO = 20,
+	AMDGPU_SDMA_UTCL1_RDBST_FIFO = 21,
+	AMDGPU_SDMA_UTCL1_WR_FIFO = 22,
+	AMDGPU_SDMA_DATA_LUT_FIFO = 23,
+	AMDGPU_SDMA_SPLIT_DAT_BUF = 24,
+	AMDGPU_SDMA_MEMORY_BLOCK_LAST,
+};
+
 struct amdgpu_sdma_ras {
 	struct amdgpu_ras_block_object ras_block;
 };
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index dca0beec0252..925ca6218a5e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -2071,3 +2071,67 @@ struct amdgpu_xcp_ip_funcs sdma_v4_4_2_xcp_funcs = {
 	.suspend = &sdma_v4_4_2_xcp_suspend,
 	.resume = &sdma_v4_4_2_xcp_resume
 };
+
+static const struct amdgpu_ras_err_status_reg_entry sdma_v4_2_2_ue_reg_list[] = {
+	{AMDGPU_RAS_REG_ENTRY(SDMA0, 0, regSDMA_UE_ERR_STATUS_LO, regSDMA_UE_ERR_STATUS_HI),
+	1, (AMDGPU_RAS_ERR_INFO_VALID | AMDGPU_RAS_ERR_STATUS_VALID), "SDMA"},
+};
+
+static const struct amdgpu_ras_memory_id_entry sdma_v4_4_2_ras_memory_list[] = {
+	{AMDGPU_SDMA_MBANK_DATA_BUF0, "SDMA_MBANK_DATA_BUF0"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF1, "SDMA_MBANK_DATA_BUF1"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF2, "SDMA_MBANK_DATA_BUF2"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF3, "SDMA_MBANK_DATA_BUF3"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF4, "SDMA_MBANK_DATA_BUF4"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF5, "SDMA_MBANK_DATA_BUF5"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF6, "SDMA_MBANK_DATA_BUF6"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF7, "SDMA_MBANK_DATA_BUF7"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF8, "SDMA_MBANK_DATA_BUF8"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF9, "SDMA_MBANK_DATA_BUF9"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF10, "SDMA_MBANK_DATA_BUF10"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF11, "SDMA_MBANK_DATA_BUF11"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF12, "SDMA_MBANK_DATA_BUF12"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF13, "SDMA_MBANK_DATA_BUF13"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF14, "SDMA_MBANK_DATA_BUF14"},
+	{AMDGPU_SDMA_MBANK_DATA_BUF15, "SDMA_MBANK_DATA_BUF15"},
+	{AMDGPU_SDMA_UCODE_BUF, "SDMA_UCODE_BUF"},
+	{AMDGPU_SDMA_RB_CMD_BUF, "SDMA_RB_CMD_BUF"},
+	{AMDGPU_SDMA_IB_CMD_BUF, "SDMA_IB_CMD_BUF"},
+	{AMDGPU_SDMA_UTCL1_RD_FIFO, "SDMA_UTCL1_RD_FIFO"},
+	{AMDGPU_SDMA_UTCL1_RDBST_FIFO, "SDMA_UTCL1_RDBST_FIFO"},
+	{AMDGPU_SDMA_UTCL1_WR_FIFO, "SDMA_UTCL1_WR_FIFO"},
+	{AMDGPU_SDMA_DATA_LUT_FIFO, "SDMA_DATA_LUT_FIFO"},
+	{AMDGPU_SDMA_SPLIT_DAT_BUF, "SDMA_SPLIT_DAT_BUF"},
+};
+
+static void sdma_v4_4_2_inst_query_ras_error_count(struct amdgpu_device *adev,
+						   uint32_t sdma_inst,
+						   void *ras_err_status)
+{
+	struct ras_err_data *err_data = (struct ras_err_data *)ras_err_status;
+
+	/* sdma v4_4_2 doesn't support query ce counts */
+	amdgpu_ras_inst_query_ras_error_count(adev,
+					sdma_v4_2_2_ue_reg_list,
+					ARRAY_SIZE(sdma_v4_2_2_ue_reg_list),
+					sdma_v4_4_2_ras_memory_list,
+					ARRAY_SIZE(sdma_v4_4_2_ras_memory_list),
+					sdma_inst,
+					AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
+					&err_data->ue_count);
+}
+
+static void sdma_v4_4_2_query_ras_error_count(struct amdgpu_device *adev,
+					      void *ras_err_status)
+{
+	uint32_t inst_mask;
+	int i = 0;
+
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
+		for_each_inst(i, inst_mask)
+			sdma_v4_4_2_inst_query_ras_error_count(adev, i, ras_err_status);
+	} else {
+		dev_warn(adev->dev, "SDMA RAS is not supported\n");
+	}
+}
-- 
cgit 


From a64b15520cc3a19bc6ca807870b51b37774a4d3d Mon Sep 17 00:00:00 2001
From: Hawking Zhang <Hawking.Zhang@amd.com>
Date: Sun, 22 Jan 2023 12:19:57 +0800
Subject: drm/amdgpu: Add reset_ras_error_count for sdma v4_4_2

Add reset_ras_error_count callback for sdma
v4_4_2. It will be used to reset sdma ras error
count.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 925ca6218a5e..f0333822df78 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -2135,3 +2135,26 @@ static void sdma_v4_4_2_query_ras_error_count(struct amdgpu_device *adev,
 		dev_warn(adev->dev, "SDMA RAS is not supported\n");
 	}
 }
+
+static void sdma_v4_4_2_inst_reset_ras_error_count(struct amdgpu_device *adev,
+						   uint32_t sdma_inst)
+{
+	amdgpu_ras_inst_reset_ras_error_count(adev,
+					sdma_v4_2_2_ue_reg_list,
+					ARRAY_SIZE(sdma_v4_2_2_ue_reg_list),
+					sdma_inst);
+}
+
+static void sdma_v4_4_2_reset_ras_error_count(struct amdgpu_device *adev)
+{
+	uint32_t inst_mask;
+	int i = 0;
+
+	inst_mask = GENMASK(adev->sdma.num_instances - 1, 0);
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) {
+		for_each_inst(i, inst_mask)
+			sdma_v4_4_2_inst_reset_ras_error_count(adev, i);
+	} else {
+		dev_warn(adev->dev, "SDMA RAS is not supported\n");
+	}
+}
-- 
cgit 


From 1e69fde70030e2fd2b729de5998d3fd6c94be238 Mon Sep 17 00:00:00 2001
From: Hawking Zhang <Hawking.Zhang@amd.com>
Date: Sun, 22 Jan 2023 23:29:28 +0800
Subject: drm/amdgpu: Initialize sdma v4_4_2 ras function

Initialize sdma v4_4_2 ras function and interrupt
handler.

Signed-off-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 48 ++++++++++++++++++++++++++------
 1 file changed, 39 insertions(+), 9 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index f0333822df78..570ea68c521f 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -54,6 +54,7 @@ static void sdma_v4_4_2_set_ring_funcs(struct amdgpu_device *adev);
 static void sdma_v4_4_2_set_buffer_funcs(struct amdgpu_device *adev);
 static void sdma_v4_4_2_set_vm_pte_funcs(struct amdgpu_device *adev);
 static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev);
+static void sdma_v4_4_2_set_ras_funcs(struct amdgpu_device *adev);
 
 static u32 sdma_v4_4_2_get_reg_offset(struct amdgpu_device *adev,
 		u32 instance, u32 offset)
@@ -1254,6 +1255,7 @@ static int sdma_v4_4_2_early_init(void *handle)
 	sdma_v4_4_2_set_buffer_funcs(adev);
 	sdma_v4_4_2_set_vm_pte_funcs(adev);
 	sdma_v4_4_2_set_irq_funcs(adev);
+	sdma_v4_4_2_set_ras_funcs(adev);
 
 	return 0;
 }
@@ -1377,6 +1379,11 @@ static int sdma_v4_4_2_sw_init(void *handle)
 		}
 	}
 
+	if (amdgpu_sdma_ras_sw_init(adev)) {
+		dev_err(adev->dev, "fail to initialize sdma ras block\n");
+		return -EINVAL;
+	}
+
 	return r;
 }
 
@@ -1558,7 +1565,7 @@ static int sdma_v4_4_2_process_ras_data_cb(struct amdgpu_device *adev,
 	 * be disabled and the driver should only look for the aggregated
 	 * interrupt via sync flood
 	 */
-	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__GFX))
+	if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA))
 		goto out;
 
 	instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id);
@@ -1597,15 +1604,22 @@ static int sdma_v4_4_2_set_ecc_irq_state(struct amdgpu_device *adev,
 					unsigned type,
 					enum amdgpu_interrupt_state state)
 {
-	u32 sdma_edc_config;
+	u32 sdma_cntl;
 
-	sdma_edc_config = RREG32_SDMA(type, regCC_SDMA_EDC_CONFIG);
-	/*
-	 * FIXME: This was inherited from Aldebaran, but no this field
-	 * definition in the regspec of both Aldebaran and SDMA 4.4.2
-	 */
-	sdma_edc_config |= (state == AMDGPU_IRQ_STATE_ENABLE) ? (1 << 2) : 0;
-	WREG32_SDMA(type, regCC_SDMA_EDC_CONFIG, sdma_edc_config);
+	sdma_cntl = RREG32_SDMA(type, regSDMA_CNTL);
+	switch (state) {
+	case AMDGPU_IRQ_STATE_DISABLE:
+		sdma_cntl = REG_SET_FIELD(sdma_cntl, SDMA_CNTL,
+					  DRAM_ECC_INT_ENABLE, 0);
+		WREG32_SDMA(type, regSDMA_CNTL, sdma_cntl);
+		break;
+	/* sdma ecc interrupt is enabled by default
+	 * driver doesn't need to do anything to
+	 * enable the interrupt */
+	case AMDGPU_IRQ_STATE_ENABLE:
+	default:
+		break;
+	}
 
 	return 0;
 }
@@ -2158,3 +2172,19 @@ static void sdma_v4_4_2_reset_ras_error_count(struct amdgpu_device *adev)
 		dev_warn(adev->dev, "SDMA RAS is not supported\n");
 	}
 }
+
+static const struct amdgpu_ras_block_hw_ops sdma_v4_4_2_ras_hw_ops = {
+	.query_ras_error_count = sdma_v4_4_2_query_ras_error_count,
+	.reset_ras_error_count = sdma_v4_4_2_reset_ras_error_count,
+};
+
+static struct amdgpu_sdma_ras sdma_v4_4_2_ras = {
+	.ras_block = {
+		.hw_ops = &sdma_v4_4_2_ras_hw_ops,
+	},
+};
+
+static void sdma_v4_4_2_set_ras_funcs(struct amdgpu_device *adev)
+{
+	adev->sdma.ras = &sdma_v4_4_2_ras;
+}
-- 
cgit 


From 1ad29cb3438175a9517f3b6cfeb0e331be213d8c Mon Sep 17 00:00:00 2001
From: "Stanley.Yang" <Stanley.Yang@amd.com>
Date: Wed, 22 Mar 2023 11:16:53 +0800
Subject: drm/amdgpu: fix sdma instance

It should change logical instance to device instance
to query ras info

Signed-off-by: Stanley.Yang <Stanley.Yang@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 570ea68c521f..bf47eb33c12e 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -2123,6 +2123,7 @@ static void sdma_v4_4_2_inst_query_ras_error_count(struct amdgpu_device *adev,
 						   void *ras_err_status)
 {
 	struct ras_err_data *err_data = (struct ras_err_data *)ras_err_status;
+	uint32_t sdma_dev_inst = GET_INST(SDMA0, sdma_inst);
 
 	/* sdma v4_4_2 doesn't support query ce counts */
 	amdgpu_ras_inst_query_ras_error_count(adev,
@@ -2130,7 +2131,7 @@ static void sdma_v4_4_2_inst_query_ras_error_count(struct amdgpu_device *adev,
 					ARRAY_SIZE(sdma_v4_2_2_ue_reg_list),
 					sdma_v4_4_2_ras_memory_list,
 					ARRAY_SIZE(sdma_v4_4_2_ras_memory_list),
-					sdma_inst,
+					sdma_dev_inst,
 					AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE,
 					&err_data->ue_count);
 }
@@ -2153,10 +2154,12 @@ static void sdma_v4_4_2_query_ras_error_count(struct amdgpu_device *adev,
 static void sdma_v4_4_2_inst_reset_ras_error_count(struct amdgpu_device *adev,
 						   uint32_t sdma_inst)
 {
+	uint32_t sdma_dev_inst = GET_INST(SDMA0, sdma_inst);
+
 	amdgpu_ras_inst_reset_ras_error_count(adev,
 					sdma_v4_2_2_ue_reg_list,
 					ARRAY_SIZE(sdma_v4_2_2_ue_reg_list),
-					sdma_inst);
+					sdma_dev_inst);
 }
 
 static void sdma_v4_4_2_reset_ras_error_count(struct amdgpu_device *adev)
-- 
cgit 


From aaff9c089947cdb366ffddf4d4fb9747b6469d3e Mon Sep 17 00:00:00 2001
From: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Date: Tue, 23 May 2023 11:17:09 +0800
Subject: drm/amdgpu: Modify mismatched function name

No functional modification involved.

drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:426: warning: expecting prototype for sdma_v4_4_2_gfx_stop(). Prototype was for sdma_v4_4_2_inst_gfx_stop() instead.
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:457: warning: expecting prototype for sdma_v4_4_2_rlc_stop(). Prototype was for sdma_v4_4_2_inst_rlc_stop() instead.
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:470: warning: expecting prototype for sdma_v4_4_2_page_stop(). Prototype was for sdma_v4_4_2_inst_page_stop() instead.
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:506: warning: expecting prototype for sdma_v4_4_2_ctx_switch_enable(). Prototype was for sdma_v4_4_2_inst_ctx_switch_enable() instead.
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:561: warning: expecting prototype for sdma_v4_4_2_enable(). Prototype was for sdma_v4_4_2_inst_enable() instead.
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:798: warning: expecting prototype for sdma_v4_4_2_rlc_resume(). Prototype was for sdma_v4_4_2_inst_rlc_resume() instead.
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:814: warning: expecting prototype for sdma_v4_4_2_load_microcode(). Prototype was for sdma_v4_4_2_inst_load_microcode() instead.

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Link: https://bugzilla.openanolis.cn/show_bug.cgi?id=5283
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index bf47eb33c12e..590b08585901 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -415,7 +415,7 @@ static void sdma_v4_4_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
 
 
 /**
- * sdma_v4_4_2_gfx_stop - stop the gfx async dma engines
+ * sdma_v4_4_2_inst_gfx_stop - stop the gfx async dma engines
  *
  * @adev: amdgpu_device pointer
  *
@@ -446,7 +446,7 @@ static void sdma_v4_4_2_inst_gfx_stop(struct amdgpu_device *adev,
 }
 
 /**
- * sdma_v4_4_2_rlc_stop - stop the compute async dma engines
+ * sdma_v4_4_2_inst_rlc_stop - stop the compute async dma engines
  *
  * @adev: amdgpu_device pointer
  *
@@ -459,7 +459,7 @@ static void sdma_v4_4_2_inst_rlc_stop(struct amdgpu_device *adev,
 }
 
 /**
- * sdma_v4_4_2_page_stop - stop the page async dma engines
+ * sdma_v4_4_2_inst_page_stop - stop the page async dma engines
  *
  * @adev: amdgpu_device pointer
  *
@@ -494,7 +494,7 @@ static void sdma_v4_4_2_inst_page_stop(struct amdgpu_device *adev,
 }
 
 /**
- * sdma_v4_4_2_ctx_switch_enable - stop the async dma engines context switch
+ * sdma_v4_4_2_inst_ctx_switch_enable - stop the async dma engines context switch
  *
  * @adev: amdgpu_device pointer
  * @enable: enable/disable the DMA MEs context switch.
@@ -548,7 +548,7 @@ static void sdma_v4_4_2_inst_ctx_switch_enable(struct amdgpu_device *adev,
 }
 
 /**
- * sdma_v4_4_2_enable - stop the async dma engines
+ * sdma_v4_4_2_inst_enable - stop the async dma engines
  *
  * @adev: amdgpu_device pointer
  * @enable: enable/disable the DMA MEs.
@@ -786,7 +786,7 @@ static void sdma_v4_4_2_init_pg(struct amdgpu_device *adev)
 }
 
 /**
- * sdma_v4_4_2_rlc_resume - setup and start the async dma engines
+ * sdma_v4_4_2_inst_rlc_resume - setup and start the async dma engines
  *
  * @adev: amdgpu_device pointer
  *
@@ -802,7 +802,7 @@ static int sdma_v4_4_2_inst_rlc_resume(struct amdgpu_device *adev,
 }
 
 /**
- * sdma_v4_4_2_load_microcode - load the sDMA ME ucode
+ * sdma_v4_4_2_inst_load_microcode - load the sDMA ME ucode
  *
  * @adev: amdgpu_device pointer
  *
-- 
cgit 


From 61c31b8b6c5e386a9c2ddc4e2cf9d8ae46c8a1f9 Mon Sep 17 00:00:00 2001
From: Guchun Chen <guchun.chen@amd.com>
Date: Fri, 12 May 2023 16:04:56 +0800
Subject: drm/amdgpu/sdma: set sched.ready status after ring/IB test in sdma

sched.ready is nothing with ring initialization, it needs to set
to be true after ring/IB test in amdgpu_ring_test_helper to tell
the ring is ready for submission.

Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/cik_sdma.c    | 2 --
 drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c   | 2 --
 drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c   | 2 --
 drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c   | 4 ----
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 4 ----
 drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c   | 2 --
 drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c   | 4 ++--
 drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c   | 6 +-----
 drivers/gpu/drm/amd/amdgpu/si_dma.c      | 2 --
 9 files changed, 3 insertions(+), 25 deletions(-)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
index 67d16236b216..52598fbc9b39 100644
--- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
+++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c
@@ -489,8 +489,6 @@ static int cik_sdma_gfx_resume(struct amdgpu_device *adev)
 #endif
 		/* enable DMA IBs */
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
-
-		ring->sched.ready = true;
 	}
 
 	cik_sdma_enable(adev, true);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
index fd2a7b66ac56..51afc92994a8 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c
@@ -466,8 +466,6 @@ static int sdma_v2_4_gfx_resume(struct amdgpu_device *adev)
 #endif
 		/* enable DMA IBs */
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
-
-		ring->sched.ready = true;
 	}
 
 	sdma_v2_4_enable(adev, true);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
index e572389089d2..344202870aeb 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c
@@ -734,8 +734,6 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev)
 #endif
 		/* enable DMA IBs */
 		WREG32(mmSDMA0_GFX_IB_CNTL + sdma_offsets[i], ib_cntl);
-
-		ring->sched.ready = true;
 	}
 
 	/* unhalt the MEs */
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
index 70b0d1fd9868..1f83eebfc8a7 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c
@@ -1114,8 +1114,6 @@ static void sdma_v4_0_gfx_resume(struct amdgpu_device *adev, unsigned int i)
 #endif
 	/* enable DMA IBs */
 	WREG32_SDMA(i, mmSDMA0_GFX_IB_CNTL, ib_cntl);
-
-	ring->sched.ready = true;
 }
 
 /**
@@ -1202,8 +1200,6 @@ static void sdma_v4_0_page_resume(struct amdgpu_device *adev, unsigned int i)
 #endif
 	/* enable DMA IBs */
 	WREG32_SDMA(i, mmSDMA0_PAGE_IB_CNTL, ib_cntl);
-
-	ring->sched.ready = true;
 }
 
 static void
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index 590b08585901..ff41fb577cdd 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -685,8 +685,6 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i)
 #endif
 	/* enable DMA IBs */
 	WREG32_SDMA(i, regSDMA_GFX_IB_CNTL, ib_cntl);
-
-	ring->sched.ready = true;
 }
 
 /**
@@ -776,8 +774,6 @@ static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i)
 #endif
 	/* enable DMA IBs */
 	WREG32_SDMA(i, regSDMA_PAGE_IB_CNTL, ib_cntl);
-
-	ring->sched.ready = true;
 }
 
 static void sdma_v4_4_2_init_pg(struct amdgpu_device *adev)
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
index a0077cf41295..5c4d4df9cf94 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_0.c
@@ -819,8 +819,6 @@ static int sdma_v5_0_gfx_resume(struct amdgpu_device *adev)
 		/* enable DMA IBs */
 		WREG32_SOC15_IP(GC, sdma_v5_0_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl);
 
-		ring->sched.ready = true;
-
 		if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
 			sdma_v5_0_ctx_switch_enable(adev, true);
 			sdma_v5_0_enable(adev, true);
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
index efa2c84ee78e..6aae62b68f32 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c
@@ -617,8 +617,6 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
 		/* enable DMA IBs */
 		WREG32_SOC15_IP(GC, sdma_v5_2_get_reg_offset(adev, i, mmSDMA0_GFX_IB_CNTL), ib_cntl);
 
-		ring->sched.ready = true;
-
 		if (amdgpu_sriov_vf(adev)) { /* bare-metal sequence doesn't need below to lines */
 			sdma_v5_2_ctx_switch_enable(adev, true);
 			sdma_v5_2_enable(adev, true);
@@ -630,6 +628,8 @@ static int sdma_v5_2_gfx_resume(struct amdgpu_device *adev)
 			return r;
 		}
 
+		ring->sched.ready = true;
+
 		if (adev->mman.buffer_funcs_ring == ring)
 			amdgpu_ttm_set_buffer_funcs_status(adev, true);
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
index 79d09792d2ce..1c90b5c661fb 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v6_0.c
@@ -585,16 +585,12 @@ static int sdma_v6_0_gfx_resume(struct amdgpu_device *adev)
 		/* enable DMA IBs */
 		WREG32_SOC15_IP(GC, sdma_v6_0_get_reg_offset(adev, i, regSDMA0_QUEUE0_IB_CNTL), ib_cntl);
 
-		ring->sched.ready = true;
-
 		if (amdgpu_sriov_vf(adev))
 			sdma_v6_0_enable(adev, true);
 
 		r = amdgpu_ring_test_helper(ring);
-		if (r) {
-			ring->sched.ready = false;
+		if (r)
 			return r;
-		}
 
 		if (adev->mman.buffer_funcs_ring == ring)
 			amdgpu_ttm_set_buffer_funcs_status(adev, true);
diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c
index abca8b529721..42c4547f32ec 100644
--- a/drivers/gpu/drm/amd/amdgpu/si_dma.c
+++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c
@@ -174,8 +174,6 @@ static int si_dma_start(struct amdgpu_device *adev)
 		WREG32(DMA_RB_WPTR + sdma_offsets[i], ring->wptr << 2);
 		WREG32(DMA_RB_CNTL + sdma_offsets[i], rb_cntl | DMA_RB_ENABLE);
 
-		ring->sched.ready = true;
-
 		r = amdgpu_ring_test_helper(ring);
 		if (r)
 			return r;
-- 
cgit 


From 23616d1ff31d6e8ffd4f1e12b6b1e2e783fa8280 Mon Sep 17 00:00:00 2001
From: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Date: Thu, 25 May 2023 22:32:35 +0530
Subject: drm/amdgpu: Fix up kdoc in sdma_v4_4_2.c
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Address a bunch of kdoc warnings:

gcc with W=1
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:426: warning: Function parameter or member 'inst_mask' not described in 'sdma_v4_4_2_inst_gfx_stop'
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:457: warning: Function parameter or member 'inst_mask' not described in 'sdma_v4_4_2_inst_rlc_stop'
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:470: warning: Function parameter or member 'inst_mask' not described in 'sdma_v4_4_2_inst_page_stop'
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:506: warning: Function parameter or member 'inst_mask' not described in 'sdma_v4_4_2_inst_ctx_switch_enable'
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:794: warning: Function parameter or member 'inst_mask' not described in 'sdma_v4_4_2_inst_rlc_resume'
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:810: warning: Function parameter or member 'inst_mask' not described in 'sdma_v4_4_2_inst_load_microcode'
drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c:854: warning: Function parameter or member 'inst_mask' not described in 'sdma_v4_4_2_inst_start'

Cc: Christian König <christian.koenig@amd.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Srinivasan Shanmugam <srinivasan.shanmugam@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 7 +++++++
 1 file changed, 7 insertions(+)

(limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')

diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
index ff41fb577cdd..8eebf9c2bbcd 100644
--- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c
@@ -418,6 +418,7 @@ static void sdma_v4_4_2_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64
  * sdma_v4_4_2_inst_gfx_stop - stop the gfx async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be disabled
  *
  * Stop the gfx async dma ring buffers.
  */
@@ -449,6 +450,7 @@ static void sdma_v4_4_2_inst_gfx_stop(struct amdgpu_device *adev,
  * sdma_v4_4_2_inst_rlc_stop - stop the compute async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be disabled
  *
  * Stop the compute async dma queues.
  */
@@ -462,6 +464,7 @@ static void sdma_v4_4_2_inst_rlc_stop(struct amdgpu_device *adev,
  * sdma_v4_4_2_inst_page_stop - stop the page async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be disabled
  *
  * Stop the page async dma ring buffers.
  */
@@ -498,6 +501,7 @@ static void sdma_v4_4_2_inst_page_stop(struct amdgpu_device *adev,
  *
  * @adev: amdgpu_device pointer
  * @enable: enable/disable the DMA MEs context switch.
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Halt or unhalt the async dma engines context switch.
  */
@@ -785,6 +789,7 @@ static void sdma_v4_4_2_init_pg(struct amdgpu_device *adev)
  * sdma_v4_4_2_inst_rlc_resume - setup and start the async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Set up the compute DMA queues and enable them.
  * Returns 0 for success, error for failure.
@@ -801,6 +806,7 @@ static int sdma_v4_4_2_inst_rlc_resume(struct amdgpu_device *adev,
  * sdma_v4_4_2_inst_load_microcode - load the sDMA ME ucode
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Loads the sDMA0/1 ucode.
  * Returns 0 for success, -EINVAL if the ucode is not available.
@@ -845,6 +851,7 @@ static int sdma_v4_4_2_inst_load_microcode(struct amdgpu_device *adev,
  * sdma_v4_4_2_inst_start - setup and start the async dma engines
  *
  * @adev: amdgpu_device pointer
+ * @inst_mask: mask of dma engine instances to be enabled
  *
  * Set up the DMA engines and enable them.
  * Returns 0 for success, error for failure.
-- 
cgit