On 11/24/21 04:28, Adrian Hunter wrote:
On 19/11/2021 21:57, Bart Van Assche wrote:
Release resources when aborting a command. Make sure that aborted commands
are completed once by clearing the corresponding tag bit from
hba->outstanding_reqs.
Fixes: 7a3e97b0dc4b ("[SCSI] ufshcd: UFS Host controller driver")
Signed-off-by: Bart Van Assche <bvanassche@xxxxxxx>
---
drivers/scsi/ufs/ufshcd.c | 18 ++++++++++++++++--
1 file changed, 16 insertions(+), 2 deletions(-)
diff --git a/drivers/scsi/ufs/ufshcd.c b/drivers/scsi/ufs/ufshcd.c
index 39dcf997a638..7e27d6436889 100644
--- a/drivers/scsi/ufs/ufshcd.c
+++ b/drivers/scsi/ufs/ufshcd.c
@@ -7042,8 +7042,12 @@ static int ufshcd_abort(struct scsi_cmnd *cmd)
ufshcd_hold(hba, false);
reg = ufshcd_readl(hba, REG_UTP_TRANSFER_REQ_DOOR_BELL);
- /* If command is already aborted/completed, return FAILED. */
- if (!(test_bit(tag, &hba->outstanding_reqs))) {
+ /*
+ * If the command is already aborted/completed, return FAILED. This
+ * should never happen since the SCSI core serializes error handling
+ * and scsi_done() calls.
I don't think that is correct. ufshcd_transfer_req_compl() gets called directly
by the interrupt handler.
I will revert this change.
+ /*
+ * Clear the corresponding bit from outstanding_reqs since the command
+ * has been aborted successfully.
+ */
+ spin_lock_irqsave(&hba->outstanding_lock, flags);
+ __clear_bit(tag, &hba->outstanding_reqs);
+ spin_unlock_irqrestore(&hba->outstanding_lock, flags);
+
+ ufshcd_release_scsi_cmd(hba, lrbp);
ufshcd_release_scsi_cmd() must not be called if the bit was already clear
i.e.
spin_lock_irqsave(&hba->outstanding_lock, flags);
rel = __test_and_clear_bit(tag, &hba->outstanding_reqs);
spin_unlock_irqrestore(&hba->outstanding_lock, flags);
if (rel)
ufshcd_release_scsi_cmd(hba, lrbp);
Will look into this.
Thanks,
Bart.
