ssize_t s;
char val[128];
- if (amdgpu_ras_error_query(obj->adev, &info))
+ if (amdgpu_ras_query_error_status(obj->adev, &info))
return -EINVAL;
s = snprintf(val, sizeof(val), "%s: %lu\n%s: %lu\n",
return snprintf(buf, PAGE_SIZE,
"Query currently inaccessible\n");
- if (amdgpu_ras_error_query(obj->adev, &info))
+ if (amdgpu_ras_query_error_status(obj->adev, &info))
return -EINVAL;
return snprintf(buf, PAGE_SIZE, "%s: %lu\n%s: %lu\n",
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
struct ras_manager *obj;
- if (!con)
+ if (!adev->ras_features || !con)
return NULL;
if (head->block >= AMDGPU_RAS_BLOCK_COUNT)
struct ras_manager *obj;
int i;
- if (!con)
+ if (!adev->ras_features || !con)
return NULL;
if (head) {
con->features |= BIT(head->block);
} else {
if (obj && amdgpu_ras_is_feature_enabled(adev, head)) {
- con->features &= ~BIT(head->block);
+ /* skip clean gfx ras context feature for VEGA20 Gaming.
+ * will clean later
+ */
+ if (!(!adev->ras_features && con->features & BIT(AMDGPU_RAS_BLOCK__GFX)))
+ con->features &= ~BIT(head->block);
put_obj(obj);
}
}
if (ret)
return ret;
+ /* gfx block ras dsiable cmd must send to ras-ta */
+ if (head->block == AMDGPU_RAS_BLOCK__GFX)
+ con->features |= BIT(head->block);
+
ret = amdgpu_ras_feature_enable(adev, head, 0);
}
} else
/* feature ctl end */
/* query/inject/cure begin */
-int amdgpu_ras_error_query(struct amdgpu_device *adev,
- struct ras_query_if *info)
+int amdgpu_ras_query_error_status(struct amdgpu_device *adev,
+ struct ras_query_if *info)
{
struct ras_manager *obj = amdgpu_ras_find_obj(adev, &info->head);
struct ras_err_data err_data = {0, 0, 0, NULL};
case AMDGPU_RAS_BLOCK__GFX:
if (adev->gfx.funcs->query_ras_error_count)
adev->gfx.funcs->query_ras_error_count(adev, &err_data);
+
+ if (adev->gfx.funcs->query_ras_error_status)
+ adev->gfx.funcs->query_ras_error_status(adev);
break;
case AMDGPU_RAS_BLOCK__MMHUB:
if (adev->mmhub.funcs->query_ras_error_count)
adev->mmhub.funcs->query_ras_error_count(adev, &err_data);
+
+ if (adev->mmhub.funcs->query_ras_error_status)
+ adev->mmhub.funcs->query_ras_error_status(adev);
break;
case AMDGPU_RAS_BLOCK__PCIE_BIF:
if (adev->nbio.funcs->query_ras_error_count)
return 0;
}
+int amdgpu_ras_reset_error_status(struct amdgpu_device *adev,
+ enum amdgpu_ras_block block)
+{
+ if (!amdgpu_ras_is_supported(adev, block))
+ return -EINVAL;
+
+ switch (block) {
+ case AMDGPU_RAS_BLOCK__GFX:
+ if (adev->gfx.funcs->reset_ras_error_count)
+ adev->gfx.funcs->reset_ras_error_count(adev);
+
+ if (adev->gfx.funcs->reset_ras_error_status)
+ adev->gfx.funcs->reset_ras_error_status(adev);
+ break;
+ case AMDGPU_RAS_BLOCK__MMHUB:
+ if (adev->mmhub.funcs->reset_ras_error_count)
+ adev->mmhub.funcs->reset_ras_error_count(adev);
+ break;
+ case AMDGPU_RAS_BLOCK__SDMA:
+ if (adev->sdma.funcs->reset_ras_error_count)
+ adev->sdma.funcs->reset_ras_error_count(adev);
+ break;
+ default:
+ break;
+ }
+
+ return 0;
+}
+
/* Trigger XGMI/WAFL error */
static int amdgpu_ras_error_inject_xgmi(struct amdgpu_device *adev,
struct ta_ras_trigger_error_input *block_info)
struct ras_manager *obj;
struct ras_err_data data = {0, 0};
- if (!con)
+ if (!adev->ras_features || !con)
return 0;
list_for_each_entry(obj, &con->head, node) {
.head = obj->head,
};
- if (amdgpu_ras_error_query(adev, &info))
+ if (amdgpu_ras_query_error_status(adev, &info))
return 0;
data.ce_count += info.ce_count;
*
*/
/* debugfs begin */
-static void amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev)
+static struct dentry *amdgpu_ras_debugfs_create_ctrl_node(struct amdgpu_device *adev)
{
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+ struct dentry *dir;
struct drm_minor *minor = adev_to_drm(adev)->primary;
- con->dir = debugfs_create_dir(RAS_FS_NAME, minor->debugfs_root);
- debugfs_create_file("ras_ctrl", S_IWUGO | S_IRUGO, con->dir,
- adev, &amdgpu_ras_debugfs_ctrl_ops);
- debugfs_create_file("ras_eeprom_reset", S_IWUGO | S_IRUGO, con->dir,
- adev, &amdgpu_ras_debugfs_eeprom_ops);
+ dir = debugfs_create_dir(RAS_FS_NAME, minor->debugfs_root);
+ debugfs_create_file("ras_ctrl", S_IWUGO | S_IRUGO, dir, adev,
+ &amdgpu_ras_debugfs_ctrl_ops);
+ debugfs_create_file("ras_eeprom_reset", S_IWUGO | S_IRUGO, dir, adev,
+ &amdgpu_ras_debugfs_eeprom_ops);
/*
* After one uncorrectable error happens, usually GPU recovery will
* ERREVENT_ATHUB_INTERRUPT generated. Normal GPU recovery routine
* will never be called.
*/
- debugfs_create_bool("auto_reboot", S_IWUGO | S_IRUGO, con->dir,
- &con->reboot);
+ debugfs_create_bool("auto_reboot", S_IWUGO | S_IRUGO, dir, &con->reboot);
/*
* User could set this not to clean up hardware's error count register
* of RAS IPs during ras recovery.
*/
- debugfs_create_bool("disable_ras_err_cnt_harvest", 0644,
- con->dir, &con->disable_ras_err_cnt_harvest);
+ debugfs_create_bool("disable_ras_err_cnt_harvest", 0644, dir,
+ &con->disable_ras_err_cnt_harvest);
+ return dir;
}
static void amdgpu_ras_debugfs_create(struct amdgpu_device *adev,
- struct ras_fs_if *head)
+ struct ras_fs_if *head,
+ struct dentry *dir)
{
- struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
struct ras_manager *obj = amdgpu_ras_find_obj(adev, &head->head);
- if (!obj || obj->ent)
+ if (!obj || !dir)
return;
get_obj(obj);
head->debugfs_name,
sizeof(obj->fs_data.debugfs_name));
- obj->ent = debugfs_create_file(obj->fs_data.debugfs_name,
- S_IWUGO | S_IRUGO, con->dir, obj,
- &amdgpu_ras_debugfs_ops);
+ debugfs_create_file(obj->fs_data.debugfs_name, S_IWUGO | S_IRUGO, dir,
+ obj, &amdgpu_ras_debugfs_ops);
}
void amdgpu_ras_debugfs_create_all(struct amdgpu_device *adev)
{
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+ struct dentry *dir;
struct ras_manager *obj;
struct ras_fs_if fs_info;
if (!IS_ENABLED(CONFIG_DEBUG_FS) || !con)
return;
- amdgpu_ras_debugfs_create_ctrl_node(adev);
+ dir = amdgpu_ras_debugfs_create_ctrl_node(adev);
list_for_each_entry(obj, &con->head, node) {
if (amdgpu_ras_is_supported(adev, obj->head.block) &&
sprintf(fs_info.debugfs_name, "%s_err_inject",
ras_block_str(obj->head.block));
fs_info.head = obj->head;
- amdgpu_ras_debugfs_create(adev, &fs_info);
+ amdgpu_ras_debugfs_create(adev, &fs_info, dir);
}
}
}
-static void amdgpu_ras_debugfs_remove(struct amdgpu_device *adev,
- struct ras_common_if *head)
-{
- struct ras_manager *obj = amdgpu_ras_find_obj(adev, head);
-
- if (!obj || !obj->ent)
- return;
-
- obj->ent = NULL;
- put_obj(obj);
-}
-
-static void amdgpu_ras_debugfs_remove_all(struct amdgpu_device *adev)
-{
- struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
- struct ras_manager *obj, *tmp;
-
- list_for_each_entry_safe(obj, tmp, &con->head, node) {
- amdgpu_ras_debugfs_remove(adev, &obj->head);
- }
-
- con->dir = NULL;
-}
/* debugfs end */
/* ras fs */
static int amdgpu_ras_fs_fini(struct amdgpu_device *adev)
{
- if (IS_ENABLED(CONFIG_DEBUG_FS))
- amdgpu_ras_debugfs_remove_all(adev);
+ struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
+ struct ras_manager *con_obj, *ip_obj, *tmp;
+
+ if (IS_ENABLED(CONFIG_DEBUG_FS)) {
+ list_for_each_entry_safe(con_obj, tmp, &con->head, node) {
+ ip_obj = amdgpu_ras_find_obj(adev, &con_obj->head);
+ if (ip_obj)
+ put_obj(ip_obj);
+ }
+ }
+
amdgpu_ras_sysfs_remove_all(adev);
return 0;
}
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
struct ras_manager *obj;
- if (!con)
+ if (!adev->ras_features || !con)
return;
list_for_each_entry(obj, &con->head, node) {
if (info.head.block == AMDGPU_RAS_BLOCK__PCIE_BIF)
continue;
- amdgpu_ras_error_query(adev, &info);
+ amdgpu_ras_query_error_status(adev, &info);
}
}
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
struct ras_manager *obj;
- if (!con)
+ if (!adev->ras_features || !con)
return;
list_for_each_entry(obj, &con->head, node) {
bool exc_err_limit = false;
int ret;
- if (con)
+ if (adev->ras_features && con)
data = &con->eh_data;
else
return 0;
max_eeprom_records_len = amdgpu_ras_eeprom_get_record_max_length();
amdgpu_ras_validate_threshold(adev, max_eeprom_records_len);
+ /* Todo: During test the SMU might fail to read the eeprom through I2C
+ * when the GPU is pending on XGMI reset during probe time
+ * (Mostly after second bus reset), skip it now
+ */
+ if (adev->gmc.xgmi.pending_reset)
+ return 0;
ret = amdgpu_ras_eeprom_init(&con->eeprom_control, &exc_err_limit);
/*
* This calling fails when exc_err_limit is true or
return;
if (amdgpu_atomfirmware_mem_ecc_supported(adev)) {
- dev_info(adev->dev, "HBM ECC is active.\n");
+ dev_info(adev->dev, "MEM ECC is active.\n");
*hw_supported |= (1 << AMDGPU_RAS_BLOCK__UMC |
1 << AMDGPU_RAS_BLOCK__DF);
} else
- dev_info(adev->dev, "HBM ECC is not presented.\n");
+ dev_info(adev->dev, "MEM ECC is not presented.\n");
if (amdgpu_atomfirmware_sram_ecc_supported(adev)) {
dev_info(adev->dev, "SRAM ECC is active.\n");
amdgpu_ras_check_supported(adev, &con->hw_supported,
&con->supported);
if (!con->hw_supported || (adev->asic_type == CHIP_VEGA10)) {
+ /* set gfx block ras context feature for VEGA20 Gaming
+ * send ras disable cmd to ras ta during ras late init.
+ */
+ if (!adev->ras_features && adev->asic_type == CHIP_VEGA20) {
+ con->features |= BIT(AMDGPU_RAS_BLOCK__GFX);
+
+ return 0;
+ }
+
r = 0;
goto release_con;
}
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
struct ras_manager *obj, *tmp;
- if (!con)
+ if (!adev->ras_features || !con) {
+ /* clean ras context for VEGA20 Gaming after send ras disable cmd */
+ amdgpu_release_ras_context(adev);
+
return;
+ }
if (con->flags & AMDGPU_RAS_FLAG_INIT_BY_VBIOS) {
/* Set up all other IPs which are not implemented. There is a
{
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
- if (!con)
+ if (!adev->ras_features || !con)
return;
amdgpu_ras_disable_all_features(adev, 0);
{
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
- if (!con)
+ if (!adev->ras_features || !con)
return 0;
/* Need disable ras on all IPs here before ip [hw/sw]fini */
{
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
- if (!con)
+ if (!adev->ras_features || !con)
return 0;
amdgpu_ras_fs_fini(adev);
return false;
}
-bool amdgpu_ras_check_err_threshold(struct amdgpu_device *adev)
+void amdgpu_release_ras_context(struct amdgpu_device *adev)
{
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
- bool exc_err_limit = false;
- if (con && (amdgpu_bad_page_threshold != 0))
- amdgpu_ras_eeprom_check_err_threshold(&con->eeprom_control,
- &exc_err_limit);
+ if (!con)
+ return;
- /*
- * We are only interested in variable exc_err_limit,
- * as it says if GPU is in bad state or not.
- */
- return exc_err_limit;
+ if (!adev->ras_features && con->features & BIT(AMDGPU_RAS_BLOCK__GFX)) {
+ con->features &= ~BIT(AMDGPU_RAS_BLOCK__GFX);
+ amdgpu_ras_set_context(adev, NULL);
+ kfree(con);
+ }
}
projects / linux.git / blobdiff