diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index bde0496d2f153961c758d7cdf8e6d758e9ca892c..177d70b70471755d35e6d8fb2e590ae6fe4a3aa8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -4995,6 +4995,20 @@ pci_ers_result_t amdgpu_pci_slot_reset(struct pci_dev *pdev) bool need_full_reset = true; u32 memsize; struct list_head device_list; + struct amdgpu_hive_info *hive; + int hive_ras_recovery = 0; + struct amdgpu_ras *ras; + + /* PCI error slot reset should be skipped During RAS recovery */ + hive = amdgpu_get_xgmi_hive(adev); + if (hive) { + hive_ras_recovery = atomic_read(&hive->ras_recovery); + amdgpu_put_xgmi_hive(hive); + } + ras = amdgpu_ras_get_context(adev); + if ((amdgpu_ip_version(adev, GC_HWIP, 0) == IP_VERSION(9, 4, 3)) && + ras && (atomic_read(&ras->in_recovery) || hive_ras_recovery)) + return PCI_ERS_RESULT_RECOVERED; DRM_INFO("PCI error: slot reset callback!!\n");