drm/amdgpu: avoid ras error injection for retired page

check whether a page is bad page before umc error injection, bad page
should not be accessed again

Signed-off-by: Tao Zhou <tao.zhou1@amd.com>
Reviewed-by: Guchun Chen <guchun.chen@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 84d8c3342a81a0c7c6abd88042208f22b4e6919d..49bdee9d9fada8beea03fe9593133bafbd8cebff 100644 (file)
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
@@ -71,6 +71,9 @@ const char *ras_block_string[] = {
 
 atomic_t amdgpu_ras_in_intr = ATOMIC_INIT(0);
 
+static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
+                               uint64_t addr);
+
 static ssize_t amdgpu_ras_debugfs_read(struct file *f, char __user *buf,
                                        size_t size, loff_t *pos)
 {
@@ -291,6 +294,14 @@ static ssize_t amdgpu_ras_debugfs_ctrl_write(struct file *f, const char __user *
                        break;
                }
 
+               /* umc ce/ue error injection for a bad page is not allowed */
+               if ((data.head.block == AMDGPU_RAS_BLOCK__UMC) &&
+                   amdgpu_ras_check_bad_page(adev, data.inject.address)) {
+                       DRM_WARN("RAS WARN: 0x%llx has been marked as bad before error injection!\n",
+                                       data.inject.address);
+                       break;
+               }
+
                /* data.inject.address is offset instead of absolute gpu address */
                ret = amdgpu_ras_error_inject(adev, &data.inject);
                break;
@@ -1431,6 +1442,39 @@ out:
        return ret;
 }
 
+/*
+ * check if an address belongs to bad page
+ *
+ * Note: this check is only for umc block
+ */
+static bool amdgpu_ras_check_bad_page(struct amdgpu_device *adev,
+                               uint64_t addr)
+{
+       struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+       struct ras_err_handler_data *data;
+       int i;
+       bool ret = false;
+
+       if (!con || !con->eh_data)
+               return ret;
+
+       mutex_lock(&con->recovery_lock);
+       data = con->eh_data;
+       if (!data)
+               goto out;
+
+       addr >>= AMDGPU_GPU_PAGE_SHIFT;
+       for (i = 0; i < data->count; i++)
+               if (addr == data->bps[i].retired_page) {
+                       ret = true;
+                       goto out;
+               }
+
+out:
+       mutex_unlock(&con->recovery_lock);
+       return ret;
+}
+
 /* called in gpu recovery/init */
 int amdgpu_ras_reserve_bad_pages(struct amdgpu_device *adev)
 {