#include "amdgpu_ras.h"
#include "amdgpu_atomfirmware.h"
-struct ras_ih_data {
- /* interrupt bottom half */
- struct work_struct ih_work;
- int inuse;
- /* IP callback */
- ras_ih_cb cb;
- /* full of entries */
- unsigned char *ring;
- unsigned int ring_size;
- unsigned int element_size;
- unsigned int aligned_element_size;
- unsigned int rptr;
- unsigned int wptr;
-};
-
-struct ras_fs_data {
- char sysfs_name[32];
- char debugfs_name[32];
-};
-
-struct ras_err_data {
- unsigned long ue_count;
- unsigned long ce_count;
-};
-
-struct ras_err_handler_data {
- /* point to bad pages array */
- struct {
- unsigned long bp;
- struct amdgpu_bo *bo;
- } *bps;
- /* the count of entries */
- int count;
- /* the space can place new entries */
- int space_left;
- /* last reserved entry's index + 1 */
- int last_reserved;
-};
-
-struct ras_manager {
- struct ras_common_if head;
- /* reference count */
- int use;
- /* ras block link */
- struct list_head node;
- /* the device */
- struct amdgpu_device *adev;
- /* debugfs */
- struct dentry *ent;
- /* sysfs */
- struct device_attribute sysfs_attr;
- int attr_inuse;
-
- /* fs node name */
- struct ras_fs_data fs_data;
-
- /* IH data */
- struct ras_ih_data ih_data;
-
- struct ras_err_data err_data;
-};
-
-struct ras_badpage {
- unsigned int bp;
- unsigned int size;
- unsigned int flags;
-};
-
const char *ras_error_string[] = {
"none",
"parity",
#define AMDGPU_RAS_FLAG_INIT_NEED_RESET 2
#define RAS_DEFAULT_FLAGS (AMDGPU_RAS_FLAG_INIT_BY_VBIOS)
+/* inject address is 52 bits */
+#define RAS_UMC_INJECT_ADDR_LIMIT (0x1ULL << 52)
+
static int amdgpu_ras_reserve_vram(struct amdgpu_device *adev,
uint64_t offset, uint64_t size,
struct amdgpu_bo **bo_ptr);
char err[9] = "ue";
int op = -1;
int block_id;
+ uint32_t sub_block;
u64 address, value;
if (*pos)
return -EINVAL;
data->head.block = block_id;
- data->head.type = memcmp("ue", err, 2) == 0 ?
- AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE :
- AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE;
+ /* only ue and ce errors are supported */
+ if (!memcmp("ue", err, 2))
+ data->head.type = AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE;
+ else if (!memcmp("ce", err, 2))
+ data->head.type = AMDGPU_RAS_ERROR__SINGLE_CORRECTABLE;
+ else
+ return -EINVAL;
+
data->op = op;
if (op == 2) {
- if (sscanf(str, "%*s %*s %*s %llu %llu",
- &address, &value) != 2)
- if (sscanf(str, "%*s %*s %*s 0x%llx 0x%llx",
- &address, &value) != 2)
+ if (sscanf(str, "%*s %*s %*s %u %llu %llu",
+ &sub_block, &address, &value) != 3)
+ if (sscanf(str, "%*s %*s %*s 0x%x 0x%llx 0x%llx",
+ &sub_block, &address, &value) != 3)
return -EINVAL;
+ data->head.sub_block_index = sub_block;
data->inject.address = address;
data->inject.value = value;
}
* write the struct to the control node.
*
* bash:
- * echo op block [error [address value]] > .../ras/ras_ctrl
+ * echo op block [error [sub_blcok address value]] > .../ras/ras_ctrl
* op: disable, enable, inject
* disable: only block is needed
* enable: block and error are needed
* error: ue, ce
* ue: multi_uncorrectable
* ce: single_correctable
+ * sub_block: sub block index, pass 0 if there is no sub block
*
* here are some examples for bash commands,
- * echo inject umc ue 0x0 0x0 > /sys/kernel/debug/dri/0/ras/ras_ctrl
- * echo inject umc ce 0 0 > /sys/kernel/debug/dri/0/ras/ras_ctrl
+ * echo inject umc ue 0x0 0x0 0x0 > /sys/kernel/debug/dri/0/ras/ras_ctrl
+ * echo inject umc ce 0 0 0 > /sys/kernel/debug/dri/0/ras/ras_ctrl
* echo disable umc > /sys/kernel/debug/dri/0/ras/ras_ctrl
*
* How to check the result?
{
struct amdgpu_device *adev = (struct amdgpu_device *)file_inode(f)->i_private;
struct ras_debug_if data;
- struct amdgpu_bo *bo;
int ret = 0;
ret = amdgpu_ras_debugfs_ctrl_parse_data(f, buf, size, pos, &data);
ret = amdgpu_ras_feature_enable(adev, &data.head, 1);
break;
case 2:
- ret = amdgpu_ras_reserve_vram(adev,
- data.inject.address, PAGE_SIZE, &bo);
- if (ret) {
- /* address was offset, now it is absolute.*/
- data.inject.address += adev->gmc.vram_start;
- if (data.inject.address > adev->gmc.vram_end)
- break;
- } else
- data.inject.address = amdgpu_bo_gpu_offset(bo);
+ if ((data.inject.address >= adev->gmc.mc_vram_size) ||
+ (data.inject.address >= RAS_UMC_INJECT_ADDR_LIMIT)) {
+ ret = -EINVAL;
+ break;
+ }
+
+ /* data.inject.address is offset instead of absolute gpu address */
ret = amdgpu_ras_error_inject(adev, &data.inject);
- amdgpu_ras_release_vram(adev, &bo);
break;
default:
ret = -EINVAL;
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};
if (!obj)
return -EINVAL;
- /* TODO might read the register to read the count */
+
+ switch (info->head.block) {
+ case AMDGPU_RAS_BLOCK__UMC:
+ if (adev->umc.funcs->query_ras_error_count)
+ adev->umc.funcs->query_ras_error_count(adev, &err_data);
+ /* umc query_ras_error_address is also responsible for clearing
+ * error status
+ */
+ if (adev->umc.funcs->query_ras_error_address)
+ adev->umc.funcs->query_ras_error_address(adev, &err_data);
+ break;
+ case AMDGPU_RAS_BLOCK__GFX:
+ if (adev->gfx.funcs->query_ras_error_count)
+ adev->gfx.funcs->query_ras_error_count(adev, &err_data);
+ break;
+ case AMDGPU_RAS_BLOCK__MMHUB:
+ if (adev->mmhub_funcs->query_ras_error_count)
+ adev->mmhub_funcs->query_ras_error_count(adev, &err_data);
+ break;
+ default:
+ break;
+ }
+
+ obj->err_data.ue_count += err_data.ue_count;
+ obj->err_data.ce_count += err_data.ce_count;
info->ue_count = obj->err_data.ue_count;
info->ce_count = obj->err_data.ce_count;
+ if (err_data.ce_count)
+ dev_info(adev->dev, "%ld correctable errors detected in %s block\n",
+ obj->err_data.ce_count, ras_block_str(info->head.block));
+ if (err_data.ue_count)
+ dev_info(adev->dev, "%ld uncorrectable errors detected in %s block\n",
+ obj->err_data.ue_count, ras_block_str(info->head.block));
+
return 0;
}
if (!obj)
return -EINVAL;
- if (block_info.block_id != TA_RAS_BLOCK__UMC) {
+ switch (info->head.block) {
+ case AMDGPU_RAS_BLOCK__GFX:
+ if (adev->gfx.funcs->ras_error_inject)
+ ret = adev->gfx.funcs->ras_error_inject(adev, info);
+ else
+ ret = -EINVAL;
+ break;
+ case AMDGPU_RAS_BLOCK__UMC:
+ case AMDGPU_RAS_BLOCK__MMHUB:
+ ret = psp_ras_trigger_error(&adev->psp, &block_info);
+ break;
+ default:
DRM_INFO("%s error injection is not supported yet\n",
ras_block_str(info->head.block));
- return -EINVAL;
+ ret = -EINVAL;
}
- ret = psp_ras_trigger_error(&adev->psp, &block_info);
if (ret)
DRM_ERROR("RAS ERROR: inject %s error failed ret %d\n",
ras_block_str(info->head.block),
}
/* get the total error counts on all IPs */
-int amdgpu_ras_query_error_count(struct amdgpu_device *adev,
+unsigned long amdgpu_ras_query_error_count(struct amdgpu_device *adev,
bool is_ce)
{
struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
struct ras_err_data data = {0, 0};
if (!con)
- return -EINVAL;
+ return 0;
list_for_each_entry(obj, &con->head, node) {
struct ras_query_if info = {
};
if (amdgpu_ras_error_query(adev, &info))
- return -EINVAL;
+ return 0;
data.ce_count += info.ce_count;
data.ue_count += info.ue_count;
{
struct amdgpu_ras *con =
container_of(attr, struct amdgpu_ras, features_attr);
- struct drm_device *ddev = dev_get_drvdata(dev);
- struct amdgpu_device *adev = ddev->dev_private;
- struct ras_common_if head;
- int ras_block_count = AMDGPU_RAS_BLOCK_COUNT;
- int i;
- ssize_t s;
- struct ras_manager *obj;
-
- s = scnprintf(buf, PAGE_SIZE, "feature mask: 0x%x\n", con->features);
- for (i = 0; i < ras_block_count; i++) {
- head.block = i;
-
- if (amdgpu_ras_is_feature_enabled(adev, &head)) {
- obj = amdgpu_ras_find_obj(adev, &head);
- s += scnprintf(&buf[s], PAGE_SIZE - s,
- "%s: %s\n",
- ras_block_str(i),
- ras_err_str(obj->head.type));
- } else
- s += scnprintf(&buf[s], PAGE_SIZE - s,
- "%s: disabled\n",
- ras_block_str(i));
- }
-
- return s;
+ return scnprintf(buf, PAGE_SIZE, "feature mask: 0x%x\n", con->features);
}
static int amdgpu_ras_sysfs_create_feature_node(struct amdgpu_device *adev)
struct ras_ih_data *data = &obj->ih_data;
struct amdgpu_iv_entry entry;
int ret;
+ struct ras_err_data err_data = {0, 0, 0, NULL};
while (data->rptr != data->wptr) {
rmb();
* from the callback to udpate the error type/count, etc
*/
if (data->cb) {
- ret = data->cb(obj->adev, &entry);
+ ret = data->cb(obj->adev, &err_data, &entry);
/* ue will trigger an interrupt, and in that case
* we need do a reset to recovery the whole system.
* But leave IP do that recovery, here we just dispatch
* the error.
*/
- if (ret == AMDGPU_RAS_UE) {
- obj->err_data.ue_count++;
+ if (ret == AMDGPU_RAS_SUCCESS) {
+ /* these counts could be left as 0 if
+ * some blocks do not count error number
+ */
+ obj->err_data.ue_count += err_data.ue_count;
+ obj->err_data.ce_count += err_data.ce_count;
}
- /* Might need get ce count by register, but not all IP
- * saves ce count, some IP just use one bit or two bits
- * to indicate ce happened.
- */
}
}
}
if (amdgpu_ras_fs_init(adev))
goto fs_out;
+ /* ras init for each ras block */
+ if (adev->umc.funcs->ras_init)
+ adev->umc.funcs->ras_init(adev);
+
DRM_INFO("RAS INFO: ras initialized successfully, "
"hardware ability[%x] ras_mask[%x]\n",
con->hw_supported, con->supported);
projects / linux.git / blobdiff