amd-drm-fixes-6.14-2025-02-13:

amdgpu:
 - Fix shutdown regression on old APUs
 - Fix compute queue hang on gfx9 APUs
 - Fix possible invalid access in PSP failure path
 - Avoid possible buffer overflow in pptable override
 
 amdkfd:
 - Properly free gang bo in failure path
 - GFX12 trap handler fix
 -----BEGIN PGP SIGNATURE-----
 
 iHUEABYKAB0WIQQgO5Idg2tXNTSZAr293/aFa7yZ2AUCZ64QYgAKCRC93/aFa7yZ
 2E4uAPoCIug8zq4eyn2FIke05YUsKIleaWqnyfx5lOxE0liX6AEAqVjST+WNWmjx
 6vC1MAOUitiVjOjLcsGIHG+xXK2T0Ak=
 =H/SZ
 -----END PGP SIGNATURE-----

Merge tag 'amd-drm-fixes-6.14-2025-02-13' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-6.14-2025-02-13:

amdgpu:
- Fix shutdown regression on old APUs
- Fix compute queue hang on gfx9 APUs
- Fix possible invalid access in PSP failure path
- Avoid possible buffer overflow in pptable override

amdkfd:
- Properly free gang bo in failure path
- GFX12 trap handler fix

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250213153843.242640-1-alexander.deucher@amd.com

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

@@ -120,9 +120,10 @@
  * - 3.58.0 - Add GFX12 DCC support
  * - 3.59.0 - Cleared VRAM
  * - 3.60.0 - Add AMDGPU_TILING_GFX12_DCC_WRITE_COMPRESS_DISABLE (Vulkan requirement)
+ * - 3.61.0 - Contains fix for RV/PCO compute queues
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	60
+#define KMS_DRIVER_MINOR	61
 #define KMS_DRIVER_PATCHLEVEL	0
 
 /*

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

@@ -3815,9 +3815,10 @@ int psp_init_cap_microcode(struct psp_context *psp, const char *chip_name)
 		if (err == -ENODEV) {
 			dev_warn(adev->dev, "cap microcode does not exist, skip\n");
 			err = 0;
-			goto out;
+		} else {
+			dev_err(adev->dev, "fail to initialize cap microcode\n");
 		}
-		dev_err(adev->dev, "fail to initialize cap microcode\n");
+		goto out;
 	}
 
 	info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CAP];

drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c

@@ -7437,6 +7437,38 @@ static void gfx_v9_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring)
 	amdgpu_ring_write(ring, 0);  /* RESERVED field, programmed to zero */
 }
 
+static void gfx_v9_0_ring_begin_use_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ip_block *gfx_block =
+		amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+
+	amdgpu_gfx_enforce_isolation_ring_begin_use(ring);
+
+	/* Raven and PCO APUs seem to have stability issues
+	 * with compute and gfxoff and gfx pg.  Disable gfx pg during
+	 * submission and allow again afterwards.
+	 */
+	if (gfx_block && amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0))
+		gfx_v9_0_set_powergating_state(gfx_block, AMD_PG_STATE_UNGATE);
+}
+
+static void gfx_v9_0_ring_end_use_compute(struct amdgpu_ring *ring)
+{
+	struct amdgpu_device *adev = ring->adev;
+	struct amdgpu_ip_block *gfx_block =
+		amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_GFX);
+
+	/* Raven and PCO APUs seem to have stability issues
+	 * with compute and gfxoff and gfx pg.  Disable gfx pg during
+	 * submission and allow again afterwards.
+	 */
+	if (gfx_block && amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 1, 0))
+		gfx_v9_0_set_powergating_state(gfx_block, AMD_PG_STATE_GATE);
+
+	amdgpu_gfx_enforce_isolation_ring_end_use(ring);
+}
+
 static const struct amd_ip_funcs gfx_v9_0_ip_funcs = {
 	.name = "gfx_v9_0",
 	.early_init = gfx_v9_0_early_init,
@@ -7613,8 +7645,8 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = {
 	.emit_wave_limit = gfx_v9_0_emit_wave_limit,
 	.reset = gfx_v9_0_reset_kcq,
 	.emit_cleaner_shader = gfx_v9_0_ring_emit_cleaner_shader,
-	.begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use,
+	.begin_use = gfx_v9_0_ring_begin_use_compute,
-	.end_use = amdgpu_gfx_enforce_isolation_ring_end_use,
+	.end_use = gfx_v9_0_ring_end_use_compute,
 };
 
 static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = {

drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h

@@ -4121,7 +4121,8 @@ static const uint32_t cwsr_trap_gfx12_hex[] = {
 	0x0000ffff, 0x8bfe7e7e,
 	0x8bea6a6a, 0xb97af804,
 	0xbe804ec2, 0xbf94fffe,
-	0xbe804a6c, 0xbfb10000,
+	0xbe804a6c, 0xbe804ec2,
+	0xbf94fffe, 0xbfb10000,
 	0xbf9f0000, 0xbf9f0000,
 	0xbf9f0000, 0xbf9f0000,
 	0xbf9f0000, 0x00000000,

drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx12.asm

@@ -1049,6 +1049,10 @@ L_SKIP_BARRIER_RESTORE:
 	s_rfe_b64	s_restore_pc_lo						//Return to the main shader program and resume execution
 
 L_END_PGM:
+	// Make sure that no wave of the workgroup can exit the trap handler
+	// before the workgroup barrier state is saved.
+	s_barrier_signal	-2
+	s_barrier_wait	-2
 	s_endpgm_saved
 end
 

drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c

@@ -300,7 +300,7 @@ static int init_user_queue(struct process_queue_manager *pqm,
 	return 0;
 
 free_gang_ctx_bo:
-	amdgpu_amdkfd_free_gtt_mem(dev->adev, (*q)->gang_ctx_bo);
+	amdgpu_amdkfd_free_gtt_mem(dev->adev, &(*q)->gang_ctx_bo);
 cleanup:
 	uninit_queue(*q);
 	*q = NULL;

drivers/gpu/drm/amd/pm/amdgpu_dpm.c

@@ -78,7 +78,7 @@ int amdgpu_dpm_set_powergating_by_smu(struct amdgpu_device *adev,
 	int ret = 0;
 	const struct amd_pm_funcs *pp_funcs = adev->powerplay.pp_funcs;
 	enum ip_power_state pwr_state = gate ? POWER_STATE_OFF : POWER_STATE_ON;
-	bool is_vcn = (block_type == AMD_IP_BLOCK_TYPE_UVD || block_type == AMD_IP_BLOCK_TYPE_VCN);
+	bool is_vcn = block_type == AMD_IP_BLOCK_TYPE_VCN;
 
 	if (atomic_read(&adev->pm.pwr_state[block_type]) == pwr_state &&
 			(!is_vcn || adev->vcn.num_vcn_inst == 1)) {

drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c

@@ -612,7 +612,8 @@ static int smu_sys_set_pp_table(void *handle,
 		return -EIO;
 	}
 
-	if (!smu_table->hardcode_pptable) {
+	if (!smu_table->hardcode_pptable || smu_table->power_play_table_size < size) {
+		kfree(smu_table->hardcode_pptable);
 		smu_table->hardcode_pptable = kzalloc(size, GFP_KERNEL);
 		if (!smu_table->hardcode_pptable)
 			return -ENOMEM;