drm/amdgpu: make pad_ib a ring function v3

The padding depends on the firmware version and we need that for BO moves as
well, not only for VM updates.

v2: new approach of making pad_ib a ring function
v3: fix typo in macro name
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucer@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -289,8 +289,6 @@ struct amdgpu_vm_pte_funcs {
 			    uint64_t pe,
 			    uint64_t addr, unsigned count,
 			    uint32_t incr, uint32_t flags);
-	/* pad the indirect buffer to the necessary number of dw */
-	void (*pad_ib)(struct amdgpu_ib *ib);
 };
 
 /* provided by the gmc block */
@@ -340,6 +338,8 @@ struct amdgpu_ring_funcs {
 	int (*test_ib)(struct amdgpu_ring *ring);
 	/* insert NOP packets */
 	void (*insert_nop)(struct amdgpu_ring *ring, uint32_t count);
+	/* pad the indirect buffer to the necessary number of dw */
+	void (*pad_ib)(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
 };
 
 /*
@@ -1188,6 +1188,7 @@ void amdgpu_ib_pool_fini(struct amdgpu_device *adev);
 int amdgpu_ib_ring_tests(struct amdgpu_device *adev);
 int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw);
 void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count);
+void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
 void amdgpu_ring_commit(struct amdgpu_ring *ring);
 void amdgpu_ring_undo(struct amdgpu_ring *ring);
 unsigned amdgpu_ring_backup(struct amdgpu_ring *ring,
@@ -2224,7 +2225,6 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count)))
 #define amdgpu_vm_write_pte(adev, ib, pa, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pa), (pe), (addr), (count), (incr), (flags)))
 #define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags)))
-#define amdgpu_vm_pad_ib(adev, ib) ((adev)->vm_manager.vm_pte_funcs->pad_ib((ib)))
 #define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib)))
 #define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r))
 #define amdgpu_ring_test_ib(r) (r)->funcs->test_ib((r))
@@ -2236,6 +2236,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring)
 #define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags))
 #define amdgpu_ring_emit_gds_switch(r, v, db, ds, wb, ws, ab, as) (r)->funcs->emit_gds_switch((r), (v), (db), (ds), (wb), (ws), (ab), (as))
 #define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r))
+#define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib)))
 #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev))
 #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv))
 #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev))

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c

@@ -90,6 +90,19 @@ void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count)
 		amdgpu_ring_write(ring, ring->nop);
 }
 
+/** amdgpu_ring_generic_pad_ib - pad IB with NOP packets
+ *
+ * @ring: amdgpu_ring structure holding ring information
+ * @ib: IB to add NOP packets to
+ *
+ * This is the generic pad_ib function for rings except SDMA
+ */
+void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
+{
+	while (ib->length_dw & ring->align_mask)
+		ib->ptr[ib->length_dw++] = ring->nop;
+}
+
 /**
  * amdgpu_ring_commit - tell the GPU to execute the new
  * commands on the ring buffer

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -1058,7 +1058,7 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring,
 		byte_count -= cur_size_in_bytes;
 	}
 
-	amdgpu_vm_pad_ib(adev, ib);
+	amdgpu_ring_pad_ib(ring, ib);
 	WARN_ON(ib->length_dw > num_dw);
 	r = amdgpu_sched_ib_submit_kernel_helper(adev, ring, ib, 1,
 						 &amdgpu_vm_free_job,

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

@@ -362,7 +362,7 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev,
 	ib->length_dw = 0;
 
 	amdgpu_vm_update_pages(adev, NULL, 0, ib, addr, 0, entries, 0, 0);
-	amdgpu_vm_pad_ib(adev, ib);
+	amdgpu_ring_pad_ib(ring, ib);
 
 	WARN_ON(ib->length_dw > 64);
 	r = amdgpu_sched_ib_submit_kernel_helper(adev, ring, ib, 1,
@@ -492,7 +492,7 @@ int amdgpu_vm_update_page_directory(struct amdgpu_device *adev,
 				       count, incr, AMDGPU_PTE_VALID);
 
 	if (ib->length_dw != 0) {
-		amdgpu_vm_pad_ib(adev, ib);
+		amdgpu_ring_pad_ib(ring, ib);
 		amdgpu_sync_resv(adev, &ib->sync, pd->tbo.resv, AMDGPU_FENCE_OWNER_VM);
 		WARN_ON(ib->length_dw > ndw);
 		r = amdgpu_sched_ib_submit_kernel_helper(adev, ring, ib, 1,
@@ -755,7 +755,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	amdgpu_vm_update_ptes(adev, gtt, gtt_flags, vm, ib, start, last + 1,
 			      addr, flags);
 
-	amdgpu_vm_pad_ib(adev, ib);
+	amdgpu_ring_pad_ib(ring, ib);
 	WARN_ON(ib->length_dw > ndw);
 	r = amdgpu_sched_ib_submit_kernel_helper(adev, ring, ib, 1,
 						 &amdgpu_vm_free_job,

drivers/gpu/drm/amd/amdgpu/cik_sdma.c

@@ -796,9 +796,9 @@ static void cik_sdma_vm_set_pte_pde(struct amdgpu_ib *ib,
  * @ib: indirect buffer to fill with padding
  *
  */
-static void cik_sdma_vm_pad_ib(struct amdgpu_ib *ib)
+static void cik_sdma_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
 {
-	struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ib->ring);
+	struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
 	u32 pad_count;
 	int i;
 
@@ -1271,6 +1271,7 @@ static const struct amdgpu_ring_funcs cik_sdma_ring_funcs = {
 	.test_ring = cik_sdma_ring_test_ring,
 	.test_ib = cik_sdma_ring_test_ib,
 	.insert_nop = cik_sdma_ring_insert_nop,
+	.pad_ib = cik_sdma_ring_pad_ib,
 };
 
 static void cik_sdma_set_ring_funcs(struct amdgpu_device *adev)
@@ -1367,7 +1368,6 @@ static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = {
 	.copy_pte = cik_sdma_vm_copy_pte,
 	.write_pte = cik_sdma_vm_write_pte,
 	.set_pte_pde = cik_sdma_vm_set_pte_pde,
-	.pad_ib = cik_sdma_vm_pad_ib,
 };
 
 static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c

@@ -5526,6 +5526,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = {
 	.test_ring = gfx_v7_0_ring_test_ring,
 	.test_ib = gfx_v7_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
@@ -5541,6 +5542,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = {
 	.test_ring = gfx_v7_0_ring_test_ring,
 	.test_ib = gfx_v7_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

@@ -5104,6 +5104,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = {
 	.test_ring = gfx_v8_0_ring_test_ring,
 	.test_ib = gfx_v8_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
@@ -5119,6 +5120,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = {
 	.test_ring = gfx_v8_0_ring_test_ring,
 	.test_ib = gfx_v8_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static void gfx_v8_0_set_ring_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c

@@ -849,14 +849,14 @@ static void sdma_v2_4_vm_set_pte_pde(struct amdgpu_ib *ib,
 }
 
 /**
- * sdma_v2_4_vm_pad_ib - pad the IB to the required number of dw
+ * sdma_v2_4_ring_pad_ib - pad the IB to the required number of dw
  *
  * @ib: indirect buffer to fill with padding
  *
  */
-static void sdma_v2_4_vm_pad_ib(struct amdgpu_ib *ib)
+static void sdma_v2_4_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
 {
-	struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ib->ring);
+	struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
 	u32 pad_count;
 	int i;
 
@@ -1275,6 +1275,7 @@ static const struct amdgpu_ring_funcs sdma_v2_4_ring_funcs = {
 	.test_ring = sdma_v2_4_ring_test_ring,
 	.test_ib = sdma_v2_4_ring_test_ib,
 	.insert_nop = sdma_v2_4_ring_insert_nop,
+	.pad_ib = sdma_v2_4_ring_pad_ib,
 };
 
 static void sdma_v2_4_set_ring_funcs(struct amdgpu_device *adev)
@@ -1372,7 +1373,6 @@ static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = {
 	.copy_pte = sdma_v2_4_vm_copy_pte,
 	.write_pte = sdma_v2_4_vm_write_pte,
 	.set_pte_pde = sdma_v2_4_vm_set_pte_pde,
-	.pad_ib = sdma_v2_4_vm_pad_ib,
 };
 
 static void sdma_v2_4_set_vm_pte_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c

@@ -999,14 +999,14 @@ static void sdma_v3_0_vm_set_pte_pde(struct amdgpu_ib *ib,
 }
 
 /**
- * sdma_v3_0_vm_pad_ib - pad the IB to the required number of dw
+ * sdma_v3_0_ring_pad_ib - pad the IB to the required number of dw
  *
  * @ib: indirect buffer to fill with padding
  *
  */
-static void sdma_v3_0_vm_pad_ib(struct amdgpu_ib *ib)
+static void sdma_v3_0_ring_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib)
 {
-	struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ib->ring);
+	struct amdgpu_sdma_instance *sdma = amdgpu_get_sdma_instance(ring);
 	u32 pad_count;
 	int i;
 
@@ -1542,6 +1542,7 @@ static const struct amdgpu_ring_funcs sdma_v3_0_ring_funcs = {
 	.test_ring = sdma_v3_0_ring_test_ring,
 	.test_ib = sdma_v3_0_ring_test_ib,
 	.insert_nop = sdma_v3_0_ring_insert_nop,
+	.pad_ib = sdma_v3_0_ring_pad_ib,
 };
 
 static void sdma_v3_0_set_ring_funcs(struct amdgpu_device *adev)
@@ -1639,7 +1640,6 @@ static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = {
 	.copy_pte = sdma_v3_0_vm_copy_pte,
 	.write_pte = sdma_v3_0_vm_write_pte,
 	.set_pte_pde = sdma_v3_0_vm_set_pte_pde,
-	.pad_ib = sdma_v3_0_vm_pad_ib,
 };
 
 static void sdma_v3_0_set_vm_pte_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c

@@ -858,6 +858,7 @@ static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = {
 	.test_ring = uvd_v4_2_ring_test_ring,
 	.test_ib = uvd_v4_2_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static void uvd_v4_2_set_ring_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c

@@ -797,6 +797,7 @@ static const struct amdgpu_ring_funcs uvd_v5_0_ring_funcs = {
 	.test_ring = uvd_v5_0_ring_test_ring,
 	.test_ib = uvd_v5_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static void uvd_v5_0_set_ring_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c

@@ -1038,6 +1038,7 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_funcs = {
 	.test_ring = uvd_v6_0_ring_test_ring,
 	.test_ib = uvd_v6_0_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/vce_v2_0.c

@@ -642,6 +642,7 @@ static const struct amdgpu_ring_funcs vce_v2_0_ring_funcs = {
 	.test_ring = amdgpu_vce_ring_test_ring,
 	.test_ib = amdgpu_vce_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev)

drivers/gpu/drm/amd/amdgpu/vce_v3_0.c

@@ -762,6 +762,7 @@ static const struct amdgpu_ring_funcs vce_v3_0_ring_funcs = {
 	.test_ring = amdgpu_vce_ring_test_ring,
 	.test_ib = amdgpu_vce_ring_test_ib,
 	.insert_nop = amdgpu_ring_insert_nop,
+	.pad_ib = amdgpu_ring_generic_pad_ib,
 };
 
 static void vce_v3_0_set_ring_funcs(struct amdgpu_device *adev)