Merge tag 'drm-next-2022-06-03-1' of git://anongit.freedesktop.org/drm/drm

Pull more drm updates from Dave Airlie:
 "This is mostly regular fixes, msm and amdgpu. There is a tegra patch
  that is bit of prep work for a 5.20 feature to avoid some inter-tree
  syncs, and a couple of late addition amdgpu uAPI changes but best to
  get those in early, and the userspace pieces are ready.

  msm:
   - Limiting WB modes to max sspp linewidth
   - Fixing the supported rotations to add 180 back for IGT
   - Fix to handle pm_runtime_get_sync() errors to avoid unclocked
     access in the bind() path for dpu driver
   - Fix the irq_free() without request issue which was a big-time
     hitter in the CI-runs.

  amdgpu:
   - Update fdinfo to the common drm format
   - uapi:
       - Add VM_NOALLOC GPUVM attribute to prevent buffers for going
         into the MALL
       - Add AMDGPU_GEM_CREATE_DISCARDABLE flag to create buffers that
         can be discarded on eviction
       - Mesa code which uses these:
           https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16466
   - Link training fixes
   - DPIA fixes
   - Misc code cleanups
   - Aux fixes
   - Hotplug fixes
   - More FP clean up
   - Misc GFX9/10 fixes
   - Fix a possible memory leak in SMU shutdown
   - SMU 13 updates
   - RAS fixes
   - TMZ fixes
   - GC 11 updates
   - SMU 11 metrics fixes
   - Fix coverage blend mode for overlay plane
   - Note DDR vs LPDDR memory
   - Fuzz fix for CS IOCTL
   - Add new PCI DID

  amdkfd:
   - Clean up hive setup
   - Misc fixes

  tegra:
   - add some prelim 5.20 work to avoid inter-tree mess"

* tag 'drm-next-2022-06-03-1' of git://anongit.freedesktop.org/drm/drm: (57 commits)
  drm/msm/dpu: Move min BW request and full BW disable back to mdss
  drm/msm/dpu: Fix pointer dereferenced before checking
  drm/msm/dpu: Remove unused code
  drm/msm/disp/dpu1: remove superfluous init
  drm/msm/dp: Always clear mask bits to disable interrupts at dp_ctrl_reset_irq_ctrl()
  gpu: host1x: Add context bus
  drm/amdgpu: add drm-client-id to fdinfo v2
  drm/amdgpu: Convert to common fdinfo format v5
  drm/amdgpu: bump minor version number
  drm/amdgpu: add AMDGPU_VM_NOALLOC v2
  drm/amdgpu: add AMDGPU_GEM_CREATE_DISCARDABLE
  drm/amdgpu: add beige goby PCI ID
  drm/amd/pm: Return auto perf level, if unsupported
  drm/amdkfd: fix typo in comment
  drm/amdgpu/gfx: fix typos in comments
  drm/amdgpu/cs: make commands with 0 chunks illegal behaviour.
  drm/amdgpu: differentiate between LP and non-LP DDR memory
  drm/amdgpu: Resolve pcie_bif RAS recovery bug
  drm/amdgpu: clean up asd on the ta_firmware_header_v2_0
  drm/amdgpu/discovery: validate VCN and SDMA instances
  ...

1 files changed, 30 insertions, 12 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 035891ec59d5..2de9309a4193 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -726,7 +726,9 @@ int amdgpu_ras_feature_enable(struct amdgpu_device *adev,
 	/* Do not enable if it is not allowed. */
 	WARN_ON(enable && !amdgpu_ras_is_feature_allowed(adev, head));
 
-	if (!amdgpu_ras_intr_triggered()) {
+	/* Only enable ras feature operation handle on host side */
+	if (!amdgpu_sriov_vf(adev) &&
+		!amdgpu_ras_intr_triggered()) {
 		ret = psp_ras_enable_features(&adev->psp, info, enable);
 		if (ret) {
 			dev_err(adev->dev, "ras %s %s failed poison:%d ret:%d\n",
@@ -1523,7 +1525,9 @@ static int amdgpu_ras_fs_fini(struct amdgpu_device *adev)
  */
 void amdgpu_ras_interrupt_fatal_error_handler(struct amdgpu_device *adev)
 {
-	if (!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF))
+	/* Fatal error events are handled on host side */
+	if (amdgpu_sriov_vf(adev) ||
+		!amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__PCIE_BIF))
 		return;
 
 	if (adev->nbio.ras &&
@@ -2270,10 +2274,14 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev)
 {
 	adev->ras_hw_enabled = adev->ras_enabled = 0;
 
-	if (amdgpu_sriov_vf(adev) || !adev->is_atom_fw ||
+	if (!adev->is_atom_fw ||
 	    !amdgpu_ras_asic_supported(adev))
 		return;
 
+	if (!(amdgpu_sriov_vf(adev) &&
+		(adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 2))))
+		return;
+
 	if (!adev->gmc.xgmi.connected_to_cpu) {
 		if (amdgpu_atomfirmware_mem_ecc_supported(adev)) {
 			dev_info(adev->dev, "MEM ECC is active.\n");
@@ -2285,15 +2293,21 @@ static void amdgpu_ras_check_supported(struct amdgpu_device *adev)
 
 		if (amdgpu_atomfirmware_sram_ecc_supported(adev)) {
 			dev_info(adev->dev, "SRAM ECC is active.\n");
-			adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC |
-						    1 << AMDGPU_RAS_BLOCK__DF);
-
-			if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0))
-				adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN |
-						1 << AMDGPU_RAS_BLOCK__JPEG);
-			else
-				adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN |
-						1 << AMDGPU_RAS_BLOCK__JPEG);
+			if (!amdgpu_sriov_vf(adev)) {
+				adev->ras_hw_enabled |= ~(1 << AMDGPU_RAS_BLOCK__UMC |
+							    1 << AMDGPU_RAS_BLOCK__DF);
+
+				if (adev->ip_versions[VCN_HWIP][0] == IP_VERSION(2, 6, 0))
+					adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__VCN |
+							1 << AMDGPU_RAS_BLOCK__JPEG);
+				else
+					adev->ras_hw_enabled &= ~(1 << AMDGPU_RAS_BLOCK__VCN |
+							1 << AMDGPU_RAS_BLOCK__JPEG);
+			} else {
+				adev->ras_hw_enabled |= (1 << AMDGPU_RAS_BLOCK__PCIE_BIF |
+								1 << AMDGPU_RAS_BLOCK__SDMA |
+								1 << AMDGPU_RAS_BLOCK__GFX);
+			}
 		} else {
 			dev_info(adev->dev, "SRAM ECC is not presented.\n");
 		}
@@ -2637,6 +2651,10 @@ int amdgpu_ras_late_init(struct amdgpu_device *adev)
 	struct amdgpu_ras_block_object *obj;
 	int r;
 
+	/* Guest side doesn't need init ras feature */
+	if (amdgpu_sriov_vf(adev))
+		return 0;
+
 	list_for_each_entry_safe(node, tmp, &adev->ras_list, node) {
 		if (!node->ras_obj) {
 			dev_warn(adev->dev, "Warning: abnormal ras list node.\n");