[v2] scsi: libsas: Put the disk offline if all recovery actions fail

Commit Message

Xingui Yang Dec. 1, 2023, 2:57 a.m. UTC

Currently, after all recovery actions in sas_eh_handle_sas_errors() fail
for sas disk, we just clear all IO, but the disk is still online. Perhaps
we should free sas_task and continue the subsequent recovery process for IO
that cannot be processed. If it still fails, the disk will be offline in
scsi_eh_ready_devs().

Signed-off-by: Xingui Yang <[e-mail]>
Signed-off-by: Bo Wu <[e-mail]>
---
v1 -> v2:
1. Free the sas_task of the work_q.
---
 drivers/scsi/libsas/sas_scsi_host.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

Patch

diff --git a/drivers/scsi/libsas/sas_scsi_host.c b/drivers/scsi/libsas/sas_scsi_host.c
index 9047cfcd1072..0cd3b92a299f 100644
--- a/drivers/scsi/libsas/sas_scsi_host.c
+++ b/drivers/scsi/libsas/sas_scsi_host.c
@@ -637,8 +637,8 @@  static void sas_eh_handle_sas_errors(struct Scsi_Host *shost, struct list_head *
 			       SAS_ADDR(task->dev->sas_addr),
 			       cmd->device->lun);
 
-			sas_eh_finish_cmd(cmd);
-			goto clear_q;
+			list_move_tail(&cmd->eh_entry, work_q);
+			goto free_q_task;
 		}
 	}
  out:
@@ -646,6 +646,15 @@  static void sas_eh_handle_sas_errors(struct Scsi_Host *shost, struct list_head *
 	list_splice_tail_init(&ha->eh_ata_q, work_q);
 	return;
 
+free_q_task:
+	pr_debug("--- Exit %s -- free_q_task\n", __func__);
+	list_for_each_entry_safe(cmd, n, work_q, eh_entry) {
+		struct sas_task *task = TO_SAS_TASK(cmd);
+
+		sas_end_task(cmd, task);
+	}
+	goto out;
+
  clear_q:
 	pr_debug("--- Exit %s -- clear_q\n", __func__);
 	list_for_each_entry_safe(cmd, n, work_q, eh_entry)