[PATCH 066/109] drm/amd/ras: harden error handling and fix resource/lock issues

From: YiPeng Chai <[email protected]>

Harden error handling and fix resource/lock issues.

Signed-off-by: YiPeng Chai <[email protected]>
Reviewed-by: Hawking Zhang <[email protected]>
Signed-off-by: Alex Deucher <[email protected]>
---
 drivers/gpu/drm/amd/ras/core/eeprom_fw.c | 19 +++++++++++++------
 drivers/gpu/drm/amd/ras/core/ras_mp1.c   |  2 +-
 drivers/gpu/drm/amd/ras/core/ras_umc.c   | 11 ++++++++---
 3 files changed, 22 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/amd/ras/core/eeprom_fw.c b/drivers/gpu/drm/amd/ras/core/eeprom_fw.c
index f7fb2dd7afe72..5f86ff54d9ed7 100644
--- a/drivers/gpu/drm/amd/ras/core/eeprom_fw.c
+++ b/drivers/gpu/drm/amd/ras/core/eeprom_fw.c
@@ -36,10 +36,11 @@ static int fw_eeprom_reset_ras_table(struct ras_core_context *ras_core)
 	mutex_lock(&ctl->record_lock);
 	res = ras_mp1_reset_ras_table(ras_core, &erase_res);
 	if (res || erase_res) {
-		RAS_DEV_WARN(ras_core->dev, "RAS EEPROM reset failed, res:%d result:%d",
-									res, erase_res);
+		RAS_DEV_WARN(ras_core->dev,
+			"RAS EEPROM reset failed, res:%d result:%d\n", res, erase_res);
 		if (!res)
 			res = -EIO;
+		goto out;
 	}
 
 	ctl->record_count = 0;
@@ -51,6 +52,8 @@ static int fw_eeprom_reset_ras_table(struct ras_core_context *ras_core)
 	ctl->rma_status = 0;
 	ctl->bad_channel_bitmap = 0;
 
+out:
+	mutex_unlock(&ctl->record_lock);
 	return res;
 }
 
@@ -73,6 +76,11 @@ static int fw_eeprom_sync_data(struct ras_core_context *ras_core,
 		RAS_DEV_ERR(ras_core->dev, "EEPROM ECC error count mismatch!\n");
 		ret = -EFAULT;
 		goto out;
+	} else if (fw_err_rec_num > MAX_EEPROM_ERR_RECORD_NUM) {
+		RAS_DEV_ERR(ras_core->dev,
+			"Invalid EEPROM error count:0x%x\n", fw_err_rec_num);
+		ret = -EOVERFLOW;
+		goto out;
 	}
 
 	for (idx = ctl->record_count;
@@ -141,9 +149,6 @@ static int fw_eeprom_sw_init(struct ras_core_context *ras_core,
 	if (!ctl)
 		return -ENOMEM;
 
-	mgr->ras_eeprom = ctl;
-	memset(ctl, 0, sizeof(*ctl));
-
 	ctl->eeprom_ip_version = param->eeprom_ip_version;
 	ctl->records = kzalloc(sizeof(*ctl->records) * MAX_EEPROM_ERR_RECORD_NUM, GFP_KERNEL);
 	if (!ctl->records) {
@@ -154,10 +159,12 @@ static int fw_eeprom_sw_init(struct ras_core_context *ras_core,
 	ctl->max_record_count = MAX_EEPROM_ERR_RECORD_NUM;
 	mutex_init(&ctl->record_lock);
 
+	mgr->ras_eeprom = ctl;
+
 	return 0;
 
 out:
-	kfree(ctl->records);
+	kfree(ctl);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/amd/ras/core/ras_mp1.c b/drivers/gpu/drm/amd/ras/core/ras_mp1.c
index 34578806ef3c0..2452448806f65 100644
--- a/drivers/gpu/drm/amd/ras/core/ras_mp1.c
+++ b/drivers/gpu/drm/amd/ras/core/ras_mp1.c
@@ -148,7 +148,7 @@ int ras_mp1_get_record_count(struct ras_core_context *ras_core, u32 *count)
 	int ret;
 
 	if (!count || !mp1->ip_func || !mp1->ip_func->get_record_count)
-		return 0;
+		return -EOPNOTSUPP;
 
 	mutex_lock(&mp1->op_mutex);
 	ret = mp1->ip_func->get_record_count(ras_core, count);
diff --git a/drivers/gpu/drm/amd/ras/core/ras_umc.c b/drivers/gpu/drm/amd/ras/core/ras_umc.c
index 0e6a341f0729a..3558d4633a4e7 100644
--- a/drivers/gpu/drm/amd/ras/core/ras_umc.c
+++ b/drivers/gpu/drm/amd/ras/core/ras_umc.c
@@ -488,6 +488,7 @@ static int ras_umc_log_record(struct ras_core_context *ras_core,
 				struct eeprom_umc_record *record)
 {
 	struct eeprom_umc_record *rec;
+	int ret;
 
 	rec = kzalloc(sizeof(*rec), GFP_KERNEL);
 	if (!rec)
@@ -495,7 +496,11 @@ static int ras_umc_log_record(struct ras_core_context *ras_core,
 
 	memcpy(rec, record, sizeof(*rec));
 
-	return ras_umc_log_ecc(ras_core, rec->cur_nps_retired_row_pfn, rec);
+	ret = ras_umc_log_ecc(ras_core, rec->cur_nps_retired_row_pfn, rec);
+	if (ret)
+		kfree(rec);
+
+	return ret;
 }
 
 /* alloc/realloc bps array */
@@ -537,7 +542,7 @@ static int ras_umc_update_eeprom_rom_data(struct ras_core_context *ras_core,
 
 	/* update bad channel bitmap */
 	if (bps->mem_channel < BITS_PER_TYPE(data->umc_channel_bitmap))
-		data->umc_channel_bitmap |= 1 << bps->mem_channel;
+		data->umc_channel_bitmap |= 0x1ULL << bps->mem_channel;
 
 	return 0;
 }
@@ -585,7 +590,7 @@ static int ras_umc_update_eeprom_ram_data(struct ras_core_context *ras_core,
 
 	/* update bad channel bitmap */
 	if (bps->mem_channel < BITS_PER_TYPE(data->umc_channel_bitmap))
-		data->umc_channel_bitmap |= 1 << bps->mem_channel;
+		data->umc_channel_bitmap |= 0x1ULL << bps->mem_channel;
 
 	return 0;
 }
-- 
2.55.0