Skip to contentSkip to conversation

drm/amdgpu: dump RAS EEPROM table via debugfs

PATCH In review Xiang Liu 2 participants 2 messages Jun 17

Subsystems (2)
lore
Share & tools
lore

Xiang Liu

author
#0 lore raw

PATCH 1/1 drm/amdgpu: dump RAS EEPROM table via debugfs

When the RAS core manages the EEPROM, the eeprom_control is never initialized (amdgpu_ras_init_badpage_info() returns early), so reading ras/ras_eeprom_table in debugfs printed only a zeroed header and no records, even though bad-page records exist in the RAS core EEPROM.

Source the table header and records from the RAS core EEPROM (ras_core->ras_eeprom) in that case, reusing the existing output layout so the debugfs node keeps the same format.

---
 .../gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c    | 80 +++++++++++++++++++
 1 file changed, 80 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +80 -0
index fca2b49bc13b..cc8e13084063 100644
@@ -1398,6 +1398,82 @@ static ssize_t amdgpu_ras_debugfs_table_read(struct file *f, char __user *buf,
 	return res < 0 ? res : orig_size - size;
 }
 
+static ssize_t
+amdgpu_ras_debugfs_table_read_uniras(struct amdgpu_device *adev,
+				     char __user *buf,
+				     size_t size, loff_t *pos)
+{
+	struct amdgpu_ras_mgr *ras_mgr = amdgpu_ras_mgr_get_context(adev);
+	struct ras_core_context *ras_core = ras_mgr ? ras_mgr->ras_core : NULL;
+	struct eeprom_umc_record *records = NULL;
+	struct ras_eeprom_control *control;
+	size_t bufsz, len = 0;
+	u32 num_recs;
+	char *kbuf;
+	ssize_t res;
+	int i;
+
+	if (!ras_core)
+		return 0;
+
+	control = &ras_core->ras_eeprom;
+	num_recs = ras_eeprom_get_record_count(ras_core);
+
+	bufsz = strlen(tbl_hdr_str) + tbl_hdr_fmt_size +
+		strlen(rec_hdr_str) + (size_t)rec_hdr_fmt_size * num_recs + 1;
+
+	kbuf = kvmalloc(bufsz, GFP_KERNEL);
+	if (!kbuf)
+		return -ENOMEM;
+
+	if (num_recs) {
+		records = kvcalloc(num_recs, sizeof(*records), GFP_KERNEL);
+		if (!records) {
+			res = -ENOMEM;
+			goto out;
+		}
+
+		res = ras_eeprom_read(ras_core, records, num_recs);
+		if (res)
+			goto out;
+	}
+
+	len += scnprintf(kbuf + len, bufsz - len, "%s", tbl_hdr_str);
+	len += scnprintf(kbuf + len, bufsz - len, tbl_hdr_fmt,
+				 control->tbl_hdr.header,
+				 control->tbl_hdr.version,
+				 control->tbl_hdr.first_rec_offset,
+				 control->tbl_hdr.tbl_size,
+				 control->tbl_hdr.checksum);
+	len += scnprintf(kbuf + len, bufsz - len, "%s", rec_hdr_str);
+
+	for (i = 0; i < num_recs; i++) {
+		u32 ai = RAS_RI_TO_AI(control, i);
+		int et = records[i].err_type;
+		const char *ets = (et >= 0 && et < AMDGPU_RAS_EEPROM_ERR_COUNT) ?
+				  record_err_type_str[et] : "na";
+
+		len += scnprintf(kbuf + len, bufsz - len, rec_hdr_fmt,
+				 i,
+				 RAS_INDEX_TO_OFFSET(control, ai),
+				 ets,
+				 records[i].bank,
+				 records[i].ts,
+				 records[i].offset,
+				 records[i].mem_channel,
+				 records[i].mcumc_id,
+				 records[i].retired_row_pfn);
+	}
+
+	res = simple_read_from_buffer(buf, size, pos, kbuf, len);
+
+out:
+	kvfree(records);
+	kvfree(kbuf);
+
+	return res;
+}
+
 static ssize_t
 amdgpu_ras_debugfs_eeprom_table_read(struct file *f, char __user *buf,
 				     size_t size, loff_t *pos)
@@ -1411,6 +1487,10 @@ amdgpu_ras_debugfs_eeprom_table_read(struct file *f, char __user *buf,
 	if (!size)
 		return size;
 
+	if (amdgpu_uniras_enabled(adev))
+		return amdgpu_ras_debugfs_table_read_uniras(adev, buf,
+						    size, pos);
+
 	if (!ras || !control) {
 		res = snprintf(data, sizeof(data), "Not supported\n");
 		if (*pos >= res)
Signature · 1 line
2.34.1
Reviewed-by on v1 1/1
#1 lore raw

AMD General

amdgpu_ras_smu_eeprom_supported check should be added, with that fixed, the patch is:

Quoted 126 lines

-----Original Message-----
From: Liu, Xiang(Dean) <[email protected]>
Sent: Wednesday, June 17, 2026 6:01 PM
To: [email protected]
Cc: Zhang, Hawking <[email protected]>; Zhou1, Tao <[email protected]>; Yang, Stanley <[email protected]>; Chai, Thomas <[email protected]>; Liu, Xiang(Dean) <[email protected]>
Subject: [PATCH] drm/amdgpu: dump RAS EEPROM table via debugfs

When the RAS core manages the EEPROM, the eeprom_control is never initialized (amdgpu_ras_init_badpage_info() returns early), so reading ras/ras_eeprom_table in debugfs printed only a zeroed header and no records, even though bad-page records exist in the RAS core EEPROM.

Source the table header and records from the RAS core EEPROM (ras_core->ras_eeprom) in that case, reusing the existing output layout so the debugfs node keeps the same format.

---
 .../gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c    | 80 +++++++++++++++++++
 1 file changed, 80 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +53 -0
index fca2b49bc13b..cc8e13084063 100644
@@ -1398,6 +1398,82 @@ static ssize_t amdgpu_ras_debugfs_table_read(struct
 file *f, char __user *buf,
      return res < 0 ? res : orig_size - size;  }
 
+static ssize_t
+amdgpu_ras_debugfs_table_read_uniras(struct amdgpu_device *adev,
+                                  char __user *buf,
+                                  size_t size, loff_t *pos)
+{
+     struct amdgpu_ras_mgr *ras_mgr = amdgpu_ras_mgr_get_context(adev);
+     struct ras_core_context *ras_core = ras_mgr ? ras_mgr->ras_core : NULL;
+     struct eeprom_umc_record *records = NULL;
+     struct ras_eeprom_control *control;
+     size_t bufsz, len = 0;
+     u32 num_recs;
+     char *kbuf;
+     ssize_t res;
+     int i;
+
+     if (!ras_core)
+             return 0;
+
+     control = &ras_core->ras_eeprom;
+     num_recs = ras_eeprom_get_record_count(ras_core);
+
+     bufsz = strlen(tbl_hdr_str) + tbl_hdr_fmt_size +
+             strlen(rec_hdr_str) + (size_t)rec_hdr_fmt_size * num_recs + 1;
+
+     kbuf = kvmalloc(bufsz, GFP_KERNEL);
+     if (!kbuf)
+             return -ENOMEM;
+
+     if (num_recs) {
+             records = kvcalloc(num_recs, sizeof(*records), GFP_KERNEL);
+             if (!records) {
+                     res = -ENOMEM;
+                     goto out;
+             }
+
+             res = ras_eeprom_read(ras_core, records, num_recs);
+             if (res)
+                     goto out;
+     }
+
+     len += scnprintf(kbuf + len, bufsz - len, "%s", tbl_hdr_str);
+     len += scnprintf(kbuf + len, bufsz - len, tbl_hdr_fmt,
+                              control->tbl_hdr.header,
+                              control->tbl_hdr.version,
+                              control->tbl_hdr.first_rec_offset,
+                              control->tbl_hdr.tbl_size,
+                              control->tbl_hdr.checksum);
+     len += scnprintf(kbuf + len, bufsz - len, "%s", rec_hdr_str);
+
+     for (i = 0; i < num_recs; i++) {
+             u32 ai = RAS_RI_TO_AI(control, i);
+             int et = records[i].err_type;
+             const char *ets = (et >= 0 && et <

AMDGPU_RAS_EEPROM_ERR_COUNT) ?

drivers/gpu/drm/amd/amdgpu/amdgpu_ras_eeprom.c +27 -0
+                               record_err_type_str[et] : "na";
+
+             len += scnprintf(kbuf + len, bufsz - len, rec_hdr_fmt,
+                              i,
+                              RAS_INDEX_TO_OFFSET(control, ai),
+                              ets,
+                              records[i].bank,
+                              records[i].ts,
+                              records[i].offset,
+                              records[i].mem_channel,
+                              records[i].mcumc_id,
+                              records[i].retired_row_pfn);
+     }
+
+     res = simple_read_from_buffer(buf, size, pos, kbuf, len);
+
+out:
+     kvfree(records);
+     kvfree(kbuf);
+
+     return res;
+}
+
 static ssize_t
 amdgpu_ras_debugfs_eeprom_table_read(struct file *f, char __user *buf,
                                   size_t size, loff_t *pos)
@@ -1411,6 +1487,10 @@ amdgpu_ras_debugfs_eeprom_table_read(struct file *f,
 char __user *buf,
      if (!size)
              return size;
 
+     if (amdgpu_uniras_enabled(adev))
+             return amdgpu_ras_debugfs_table_read_uniras(adev, buf,
+                                                 size, pos);
+
      if (!ras || !control) {
              res = snprintf(data, sizeof(data), "Not supported\n");
              if (*pos >= res)

--
2.34.1