[PATCH 2/3] cxlflash: Avoid double mutex unlock

Uma Krishnan ukrishn at linux.vnet.ibm.com
Sat Aug 26 08:18:03 AEST 2017


From: "Matthew R. Ochs" <mrochs at linux.vnet.ibm.com>

The AFU recovery routine uses an interruptible mutex to control the
flow of in-flight recoveries. Upon receiving an interruptible signal
the code branches to a common exit path which wrongly assumes the
mutex is held. Add a local variable to track when the mutex should be
unlocked.

Signed-off-by: Matthew R. Ochs <mrochs at linux.vnet.ibm.com>
Signed-off-by: Uma Krishnan <ukrishn at linux.vnet.ibm.com>
---
 drivers/scsi/cxlflash/superpipe.c | 13 ++++++++++---
 1 file changed, 10 insertions(+), 3 deletions(-)

diff --git a/drivers/scsi/cxlflash/superpipe.c b/drivers/scsi/cxlflash/superpipe.c
index ad0f996..e9ee1d9 100644
--- a/drivers/scsi/cxlflash/superpipe.c
+++ b/drivers/scsi/cxlflash/superpipe.c
@@ -1650,6 +1650,7 @@ static int cxlflash_afu_recover(struct scsi_device *sdev,
 	u64 ctxid = DECODE_CTXID(recover->context_id),
 	    rctxid = recover->context_id;
 	long reg;
+	bool locked = true;
 	int lretry = 20; /* up to 2 seconds */
 	int new_adap_fd = -1;
 	int rc = 0;
@@ -1658,8 +1659,11 @@ static int cxlflash_afu_recover(struct scsi_device *sdev,
 	up_read(&cfg->ioctl_rwsem);
 	rc = mutex_lock_interruptible(mutex);
 	down_read(&cfg->ioctl_rwsem);
-	if (rc)
+	if (rc) {
+		locked = false;
 		goto out;
+	}
+
 	rc = check_state(cfg);
 	if (rc) {
 		dev_err(dev, "%s: Failed state rc=%d\n", __func__, rc);
@@ -1693,8 +1697,10 @@ static int cxlflash_afu_recover(struct scsi_device *sdev,
 				mutex_unlock(mutex);
 				msleep(100);
 				rc = mutex_lock_interruptible(mutex);
-				if (rc)
+				if (rc) {
+					locked = false;
 					goto out;
+				}
 				goto retry_recover;
 			}
 
@@ -1738,7 +1744,8 @@ static int cxlflash_afu_recover(struct scsi_device *sdev,
 out:
 	if (likely(ctxi))
 		put_context(ctxi);
-	mutex_unlock(mutex);
+	if (locked)
+		mutex_unlock(mutex);
 	atomic_dec_if_positive(&cfg->recovery_threads);
 	return rc;
 }
-- 
2.1.0



More information about the Linuxppc-dev mailing list