]> www.infradead.org Git - nvme.git/commitdiff
drm/amdgpu: flush all cached ras bad pages to eeprom
authorYiPeng Chai <YiPeng.Chai@amd.com>
Tue, 2 Jul 2024 10:16:52 +0000 (18:16 +0800)
committerAlex Deucher <alexander.deucher@amd.com>
Wed, 10 Jul 2024 14:13:35 +0000 (10:13 -0400)
Before uninstalling gpu driver, flush all cached ras
bad pages to eeprom.

v2:
  Put the same code into a function and reuse the function.

Signed-off-by: YiPeng Chai <YiPeng.Chai@amd.com>
Reviewed-by: Tao Zhou <tao.zhou1@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/amdgpu/amdgpu_ras.c

index 1ba18655176dcdc155639e70c0cd918072a5554e..64bee125f17a0d10ff7b3e5c0bda61303bea7254 100644 (file)
@@ -124,6 +124,8 @@ const char *get_ras_block_str(struct ras_common_if *ras_block)
 
 #define AMDGPU_RAS_RETIRE_PAGE_INTERVAL 100  //ms
 
+#define MAX_FLUSH_RETIRE_DWORK_TIMES  100
+
 enum amdgpu_ras_retire_page_reservation {
        AMDGPU_RAS_RETIRE_PAGE_RESERVED,
        AMDGPU_RAS_RETIRE_PAGE_PENDING,
@@ -2907,6 +2909,23 @@ static void amdgpu_ras_ecc_log_fini(struct ras_ecc_log_info *ecc_log)
        ecc_log->prev_de_queried_count = 0;
 }
 
+static bool amdgpu_ras_schedule_retirement_dwork(struct amdgpu_ras *con,
+                               uint32_t delayed_ms)
+{
+       int ret;
+
+       mutex_lock(&con->umc_ecc_log.lock);
+       ret = radix_tree_tagged(&con->umc_ecc_log.de_page_tree,
+                       UMC_ECC_NEW_DETECTED_TAG);
+       mutex_unlock(&con->umc_ecc_log.lock);
+
+       if (ret)
+               schedule_delayed_work(&con->page_retirement_dwork,
+                       msecs_to_jiffies(delayed_ms));
+
+       return ret ? true : false;
+}
+
 static void amdgpu_ras_do_page_retirement(struct work_struct *work)
 {
        struct amdgpu_ras *con = container_of(work, struct amdgpu_ras,
@@ -2928,12 +2947,8 @@ static void amdgpu_ras_do_page_retirement(struct work_struct *work)
        if (err_cnt && con->is_rma)
                amdgpu_ras_reset_gpu(adev);
 
-       mutex_lock(&con->umc_ecc_log.lock);
-       if (radix_tree_tagged(&con->umc_ecc_log.de_page_tree,
-                               UMC_ECC_NEW_DETECTED_TAG))
-               schedule_delayed_work(&con->page_retirement_dwork,
-                       msecs_to_jiffies(AMDGPU_RAS_RETIRE_PAGE_INTERVAL));
-       mutex_unlock(&con->umc_ecc_log.lock);
+       amdgpu_ras_schedule_retirement_dwork(con,
+                       AMDGPU_RAS_RETIRE_PAGE_INTERVAL);
 }
 
 static int amdgpu_ras_poison_creation_handler(struct amdgpu_device *adev,
@@ -3237,11 +3252,19 @@ static int amdgpu_ras_recovery_fini(struct amdgpu_device *adev)
 {
        struct amdgpu_ras *con = amdgpu_ras_get_context(adev);
        struct ras_err_handler_data *data = con->eh_data;
+       int max_flush_timeout = MAX_FLUSH_RETIRE_DWORK_TIMES;
+       bool ret;
 
        /* recovery_init failed to init it, fini is useless */
        if (!data)
                return 0;
 
+       /* Save all cached bad pages to eeprom */
+       do {
+               flush_delayed_work(&con->page_retirement_dwork);
+               ret = amdgpu_ras_schedule_retirement_dwork(con, 0);
+       } while (ret && max_flush_timeout--);
+
        if (con->page_retirement_thread)
                kthread_stop(con->page_retirement_thread);