ata: libata-eh: Clarify ata_eh_qc_retry() behavior at call site

While the function documentation for ata_eh_qc_retry() is clear,
from simply reading the single function that calls ata_eh_qc_retry(),
it is not clear that ata_eh_qc_retry() might not retry the command.

Add a comment in the single function that calls ata_eh_qc_retry() to
clarify the behavior.

[Damien] Added curly braces to "if () else" with multi-line comment.

Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>

1 files changed, 15 insertions, 2 deletions

diff --git a/drivers/ata/libata-eh.c b/drivers/ata/libata-eh.c
index a6c901811802..c7336a0a884d 100644
--- a/drivers/ata/libata-eh.c
+++ b/drivers/ata/libata-eh.c
@@ -3813,16 +3813,29 @@ void ata_eh_finish(struct ata_port *ap)
 			 * generate sense data in this function,
 			 * considering both err_mask and tf.
 			 */
-			if (qc->flags & ATA_QCFLAG_RETRY)
+			if (qc->flags & ATA_QCFLAG_RETRY) {
+				/*
+				 * Since qc->err_mask is set, ata_eh_qc_retry()
+				 * will not increment scmd->allowed, so upper
+				 * layer will only retry the command if it has
+				 * not already been retried too many times.
+				 */
 				ata_eh_qc_retry(qc);
-			else
+			} else {
 				ata_eh_qc_complete(qc);
+			}
 		} else {
 			if (qc->flags & ATA_QCFLAG_SENSE_VALID) {
 				ata_eh_qc_complete(qc);
 			} else {
 				/* feed zero TF to sense generation */
 				memset(&qc->result_tf, 0, sizeof(qc->result_tf));
+				/*
+				 * Since qc->err_mask is not set,
+				 * ata_eh_qc_retry() will increment
+				 * scmd->allowed, so upper layer is guaranteed
+				 * to retry the command.
+				 */
 				ata_eh_qc_retry(qc);
 			}
 		}