Skip to contentSkip to conversation

drm/amdgpu: Update usage for bad page threshold

PATCH In review Hawking Zhang 2 participants 3 messages Jan 22, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/amdgpu: Update usage for bad page threshold

The driver's behavior varies based on
the configuration of amdgpu_bad_page_threshold setting

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c       |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c       | 54 ++++++++++---------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  2 +-
 .../gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c    | 25 +++++----
 4 files changed, 44 insertions(+), 39 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +1 -1
index 99d884e6763a..87ea2e2a062f 100644
@@ -983,7 +983,7 @@ module_param_named(reset_method, amdgpu_reset_method, int, 0644);
  * result in the GPU entering bad status when the number of total
  * faulty pages by ECC exceeds the threshold value.
  */
-MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore threshold (default value), 0 = disable bad page retirement, -2 = driver sets threshold)");
+MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore threshold (default value), 0 = disable bad page retirement, -2 = threshold determined by a formula, 0 < threshold < max records, user-defined threshold)");
 module_param_named(bad_page_threshold, amdgpu_bad_page_threshold, int, 0444);
 
 MODULE_PARM_DESC(num_kcq, "number of kernel compute queue user want to setup (8 if set to greater than 8 or less than 0, only affect gfx 8+)");
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +28 -26

71 lines not on this page · raw message

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -1
index 82db986c36a0..cc4586581dba 100644
@@ -65,7 +65,7 @@ struct amdgpu_iv_entry;
 
 /* Reserve 8 physical dram row for possible retirement.
  * In worst cases, it will lose 8 * 2MB memory in vram domain */
-#define AMDGPU_RAS_RESERVED_VRAM_SIZE	(16ULL << 20)
+#define AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT	(16ULL << 20)
 /* The high three bits indicates socketid */
 #define AMDGPU_RAS_GET_FEATURES(val)  ((val) & ~AMDGPU_RAS_FEATURES_SOCKETID_MASK)
 
drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +14 -11
index 0d824f016916..bd9ed86d8fde 100644
@@ -1428,8 +1428,9 @@ int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
 
 		res = __verify_ras_table_checksum(control);
 		if (res)
-			DRM_ERROR("RAS table incorrect checksum or error:%d\n",
-				  res);
+			dev_err(adev->dev,
+				"RAS table incorrect checksum or error:%d\n",
+				res);
 
 		/* Warn if we are at 90% of the threshold or above
 		 */
@@ -1447,8 +1448,9 @@ int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
 
 		res = __verify_ras_table_checksum(control);
 		if (res) {
-			dev_err(adev->dev, "RAS Table incorrect checksum or error:%d\n",
-				  res);
+			dev_err(adev->dev,
+				"RAS Table incorrect checksum or error:%d\n",
+				res);
 			return -EINVAL;
 		}
 		if (ras->bad_page_cnt_threshold > control->ras_num_bad_pages) {
@@ -1466,17 +1468,18 @@ int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
 			res = amdgpu_ras_eeprom_correct_header_tag(control,
 								   RAS_TABLE_HDR_VAL);
 		} else {
-			dev_err(adev->dev, "RAS records:%d exceed threshold:%d",
+			dev_warn(adev->dev,
+				"RAS records:%d exceed threshold:%d\n",
 				control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
-			if (amdgpu_bad_page_threshold == -1) {
-				dev_warn(adev->dev, "GPU will be initialized due to bad_page_threshold = -1.");
+			if ((amdgpu_bad_page_threshold == -1) ||
+			    (amdgpu_bad_page_threshold == -2)) {
 				res = 0;
+				dev_warn(adev->dev,
+					 "Please consult AMD Service Action Guide (SAG) for appropriate service procedures\n");
 			} else {
 				ras->is_rma = true;
-				dev_err(adev->dev,
-					"RAS records:%d exceed threshold:%d, "
-					"GPU will not be initialized. Replace this GPU or increase the threshold",
-					control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
+				dev_warn(adev->dev,
+					 "User defined threshold is set, runtime service will be halt when threshold is reached\n");
 			}
 		}
 	} else {
Signature · 1 line
2.17.1
#1 lore raw

[AMD Official Use Only - AMD Internal Distribution Only]

Please ignore this one. will send out a new one soon

Regards,
Hawking

-----Original Message-----
From: Zhang, Hawking <[email protected]>
Sent: Wednesday, January 22, 2025 18:30
To: [email protected]; Zhou1, Tao <[email protected]>
Cc: Zhang, Hawking <[email protected]>
Subject: [PATCH] drm/amdgpu: Update usage for bad page threshold

The driver's behavior varies based on
the configuration of amdgpu_bad_page_threshold setting

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c       |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c       | 54 ++++++++++---------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  2 +-
 .../gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c    | 25 +++++----
 4 files changed, 44 insertions(+), 39 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +4 -1
index 99d884e6763a..87ea2e2a062f 100644
@@ -983,7 +983,7 @@ module_param_named(reset_method, amdgpu_reset_method, int, 0644);
  * result in the GPU entering bad status when the number of total
  * faulty pages by ECC exceeds the threshold value.
  */
-MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore threshold (default value), 0 = disable bad page retirement, -2 = driver sets threshold)");
+MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore
+threshold (default value), 0 = disable bad page retirement, -2 =
+threshold determined by a formula, 0 < threshold < max records,
+user-defined threshold)");

module_param_named(bad_page_threshold, amdgpu_bad_page_threshold, int, 0444);

MODULE_PARM_DESC(num_kcq, "number of kernel compute queue user want to setup (8 if set to greater than 8 or less than 0, only affect gfx 8+)"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 960476e6124b..5676ffe5c43a 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +29 -26

71 lines not on this page · raw message

                break;
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -1
index 82db986c36a0..cc4586581dba 100644
@@ -65,7 +65,7 @@ struct amdgpu_iv_entry;
 
 /* Reserve 8 physical dram row for possible retirement.
  * In worst cases, it will lose 8 * 2MB memory in vram domain */
-#define AMDGPU_RAS_RESERVED_VRAM_SIZE  (16ULL << 20)
+#define AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT  (16ULL << 20)
 /* The high three bits indicates socketid */  #define AMDGPU_RAS_GET_FEATURES(val)  ((val) & ~AMDGPU_RAS_FEATURES_SOCKETID_MASK)
 
drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +6 -4
index 0d824f016916..bd9ed86d8fde 100644
@@ -1428,8 +1428,9 @@ int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
 
                res = __verify_ras_table_checksum(control);
                if (res)
-                       DRM_ERROR("RAS table incorrect checksum or error:%d\n",
-                                 res);
+                       dev_err(adev->dev,
+                               "RAS table incorrect checksum or error:%d\n",
+                               res);
 
                /* Warn if we are at 90% of the threshold or above
                 */
@@ -1447,8 +1448,9 @@ int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
 
                res = __verify_ras_table_checksum(control);
                if (res) {
-                       dev_err(adev->dev, "RAS Table incorrect checksum or error:%d\n",
-                                 res);
+                       dev_err(adev->dev,
+                               "RAS Table incorrect checksum or error:%d\n",
+                               res);
                        return -EINVAL;
                }
                if (ras->bad_page_cnt_threshold > control->ras_num_bad_pages) { @@ -1466,17 +1468,18 @@ int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
                        res = amdgpu_ras_eeprom_correct_header_tag(control,
                                                                   RAS_TABLE_HDR_VAL);
                } else {
-                       dev_err(adev->dev, "RAS records:%d exceed threshold:%d",
+                       dev_warn(adev->dev,
+                               "RAS records:%d exceed threshold:%d\n",
                                control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
-                       if (amdgpu_bad_page_threshold == -1) {
-                               dev_warn(adev->dev, "GPU will be initialized due to bad_page_threshold = -1.");
+                       if ((amdgpu_bad_page_threshold == -1) ||
+                           (amdgpu_bad_page_threshold == -2)) {
                                res = 0;
+                               dev_warn(adev->dev,
+                                        "Please consult AMD Service Action Guide (SAG) for appropriate
+service procedures\n");
                        } else {
                                ras->is_rma = true;
-                               dev_err(adev->dev,
-                                       "RAS records:%d exceed threshold:%d, "
-                                       "GPU will not be initialized. Replace this GPU or increase the threshold",
-                                       control->ras_num_bad_pages, ras->bad_page_cnt_threshold);
+                               dev_warn(adev->dev,
+                                        "User defined threshold is set, runtime service will be halt when
+threshold is reached\n");
                        }
                }
        } else {

--
2.17.1

#2 lore raw
In reply to Hawking Zhang

[AMD Official Use Only - AMD Internal Distribution Only]

Quoted 187 lines

-----Original Message-----
From: Zhang, Hawking <[email protected]>
Sent: Wednesday, January 22, 2025 6:30 PM
To: [email protected]; Zhou1, Tao <[email protected]>
Cc: Zhang, Hawking <[email protected]>
Subject: [PATCH] drm/amdgpu: Update usage for bad page threshold

The driver's behavior varies based on
the configuration of amdgpu_bad_page_threshold setting

---
 drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c       |  2 +-
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c       | 54 ++++++++++---------
 drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h       |  2 +-
 .../gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c    | 25 +++++----
 4 files changed, 44 insertions(+), 39 deletions(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +0 -1
index 99d884e6763a..87ea2e2a062f 100644
@@ -983,7 +983,7 @@ module_param_named(reset_method,
 amdgpu_reset_method, int, 0644);
  * result in the GPU entering bad status when the number of total
  * faulty pages by ECC exceeds the threshold value.
  */
-MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore

threshold (default value), 0 = disable bad page retirement, -2 = driver sets threshold)");

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +4 -0
+MODULE_PARM_DESC(bad_page_threshold, "Bad page threshold(-1 = ignore
+threshold (default value), 0 = disable bad page retirement, -2 =
+threshold determined by a formula, 0 < threshold < max records,
+user-defined threshold)");
 module_param_named(bad_page_threshold, amdgpu_bad_page_threshold, int,

0444);

MODULE_PARM_DESC(num_kcq, "number of kernel compute queue user want to
setup (8 if set to greater than 8 or less than 0, only affect gfx 8+)"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c
index 960476e6124b..5676ffe5c43a 100644

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +14 -1
@@ -3071,35 +3071,35 @@ static void amdgpu_ras_validate_threshold(struct
 amdgpu_device *adev,
      struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
 
      /*
-      * Justification of value bad_page_cnt_threshold in ras structure
+      * amdgpu_bad_page_threshold is used to config
+      * the threshold for the number of bad pages.
+      * -1:  Threshold is set to default value
+      *      Driver will issue a warning message when threshold is reached
+      *      and continue runtime services.
+      * 0:   Disable bad page retirement
+      *      Driver will not retire bad pages
+      *      which is intended for debugging purpose.
+      * -2:  Threshold is determined by a formula
+      *      that assumes 1 bad page per 100M of local memory.
+      *      Driver will continue runtime services when threhold is reached.
+      * 0 < threshold < max number of bad page records in EEPROM,
+      *      A user-defined threshold is set
+      *      Driver will halt runtime services when this custom threshold is

reached.

       *
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +7 -17
-      * Generally, 0 <= amdgpu_bad_page_threshold <= max record length
-      * in eeprom or amdgpu_bad_page_threshold == -2, introduce two
-      * scenarios accordingly.
-      *
-      * Bad page retirement enablement:
-      *    - If amdgpu_bad_page_threshold = -2,
-      *      bad_page_cnt_threshold = typical value by formula.
-      *
-      *    - When the value from user is 0 < amdgpu_bad_page_threshold <
-      *      max record length in eeprom, use it directly.
-      *
-      * Bad page retirement disablement:
-      *    - If amdgpu_bad_page_threshold = 0, bad page retirement
-      *      functionality is disabled, and bad_page_cnt_threshold will
-      *      take no effect.
       */
+        if (amdgpu_bad_page_threshold == -2) {
+                u64 val = adev->gmc.mc_vram_size;
 
-     if (amdgpu_bad_page_threshold < 0) {
-             u64 val = adev->gmc.mc_vram_size;
+                do_div(val, RAS_BAD_PAGE_COVER);
+                con->bad_page_cnt_threshold = min(lower_32_bits(val),
+                                                  max_count);
+        } else if (amdgpu_bad_page_threshold == -1) {
+                con->bad_page_cnt_threshold = ((con->reserved_pages_in_bytes) >>

21) << 4;

drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c +7 -8
+        } else {
+                con->bad_page_cnt_threshold = min_t(int, max_count,
+                                                    amdgpu_bad_page_threshold);
+        }
 
-             do_div(val, RAS_BAD_PAGE_COVER);
-             con->bad_page_cnt_threshold = min(lower_32_bits(val),
-                                               max_count);
-     } else {
-             con->bad_page_cnt_threshold = min_t(int, max_count,
-                                                 amdgpu_bad_page_threshold);
-     }
 }
 
 #ifdef HAVE_KFIFO_PUT_NON_POINTER
@@ -3852,8 +3852,10 @@ static void amdgpu_ras_init_reserved_vram_size(struct
 amdgpu_device *adev)
      case IP_VERSION(13, 0, 2):
      case IP_VERSION(13, 0, 6):
      case IP_VERSION(13, 0, 12):
+             con->reserved_pages_in_bytes =
 AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT;
+             break;
      case IP_VERSION(13, 0, 14):
-             con->reserved_pages_in_bytes =
 AMDGPU_RAS_RESERVED_VRAM_SIZE;
+             con->reserved_pages_in_bytes =

(AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT

+<< 1);
              break;
      default:
              break;
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.h +1 -1
index 82db986c36a0..cc4586581dba 100644
@@ -65,7 +65,7 @@ struct amdgpu_iv_entry;
 
 /* Reserve 8 physical dram row for possible retirement.
  * In worst cases, it will lose 8 * 2MB memory in vram domain */
-#define AMDGPU_RAS_RESERVED_VRAM_SIZE        (16ULL << 20)
+#define AMDGPU_RAS_RESERVED_VRAM_SIZE_DEFAULT        (16ULL << 20)
 /* The high three bits indicates socketid */  #define

AMDGPU_RAS_GET_FEATURES(val) ((val) &
~AMDGPU_RAS_FEATURES_SOCKETID_MASK)

drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +3 -2
index 0d824f016916..bd9ed86d8fde 100644
@@ -1428,8 +1428,9 @@ int amdgpu_ras_eeprom_check(struct
 amdgpu_ras_eeprom_control *control)
 
              res = __verify_ras_table_checksum(control);
              if (res)
-                     DRM_ERROR("RAS table incorrect checksum or error:%d\n",
-                               res);
+                     dev_err(adev->dev,
+                             "RAS table incorrect checksum or error:%d\n",
+                             res);
 
              /* Warn if we are at 90% of the threshold or above
               */
drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +3 -2
@@ -1447,8 +1448,9 @@ int amdgpu_ras_eeprom_check(struct
 amdgpu_ras_eeprom_control *control)
 
              res = __verify_ras_table_checksum(control);
              if (res) {
-                     dev_err(adev->dev, "RAS Table incorrect checksum or
 error:%d\n",
-                               res);
+                     dev_err(adev->dev,
+                             "RAS Table incorrect checksum or error:%d\n",
+                             res);
                      return -EINVAL;
              }
              if (ras->bad_page_cnt_threshold > control->ras_num_bad_pages)

{ @@ -1466,17 +1468,18 @@ int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control)
res = amdgpu_ras_eeprom_correct_header_tag(control,

RAS_TABLE_HDR_VAL);
              } else {
drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +2 -1
-                     dev_err(adev->dev, "RAS records:%d exceed threshold:%d",
+                     dev_warn(adev->dev,
+                             "RAS records:%d exceed threshold:%d\n",
                              control->ras_num_bad_pages, ras-
Quoted 1 line

bad_page_cnt_threshold);

drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +0 -2
-                     if (amdgpu_bad_page_threshold == -1) {
-                             dev_warn(adev->dev, "GPU will be initialized due to

bad_page_threshold = -1.");

drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c int amdgpu_ras_eeprom_check(struct amdgpu_ras_eeprom_control *control) · line 1475

Quoted 2 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +2 -0
+                     if ((amdgpu_bad_page_threshold == -1) ||
+                         (amdgpu_bad_page_threshold == -2)) {

[Tao] as discussed, besides modprobe, ras->is_rma shouldn't be set in runtime if threshold is -1.

Quoted 22 lines
                              res = 0;
drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +2 -0
+                             dev_warn(adev->dev,
+                                      "Please consult AMD Service Action Guide

(SAG) for appropriate

drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +1 -3
+service procedures\n");
                      } else {
                              ras->is_rma = true;
-                             dev_err(adev->dev,
-                                     "RAS records:%d exceed threshold:%d, "
-                                     "GPU will not be initialized. Replace this GPU

or increase the threshold",
- control->ras_num_bad_pages, ras-

Quoted 1 line

bad_page_cnt_threshold);

drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +2 -0
+                             dev_warn(adev->dev,
+                                      "User defined threshold is set, runtime service

will be halt when

+threshold is reached\n");
                      }
              }
      } else {

--
2.17.1