Skip to contentSkip to conversation

staging: qlge: replace msleep with usleep_range

PATCH In review Arun Vijayshankar 2 participants 3 messages Jun 27, 2022 – Jul 1, 2022 1 message not in the archive

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 staging: qlge: replace msleep with usleep_range

qlge_close uses msleep for 1ms, in which case (1 - 20ms) it is preferable to use usleep_range in non-atomic contexts, as documented in Documentation/timers/timers-howto.rst. A range of 1 - 1.05ms is specified here, in case the device takes slightly longer than 1ms to recover from reset.

---
 drivers/staging/qlge/qlge_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/staging/qlge/qlge_main.c +1 -1
index 689a87d58f27..3cc4f1902c80 100644
@@ -3886,7 +3886,7 @@ static int qlge_close(struct net_device *ndev)
 	 * (Rarely happens, but possible.)
 	 */
 	while (!test_bit(QL_ADAPTER_UP, &qdev->flags))
-		msleep(1);
+		usleep_range(1000, 1050);
 
 	/* Make sure refill_work doesn't re-enable napi */
 	for (i = 0; i < qdev->rss_ring_count; i++)
Signature · 1 line
2.34.1
maintainer
#1 lore raw
Quoted 16 lines · Arun Vijayshankar

qlge_close uses msleep for 1ms, in which case (1 - 20ms) it is preferable to use usleep_range in non-atomic contexts, as documented in Documentation/timers/timers-howto.rst. A range of 1 - 1.05ms is specified here, in case the device takes slightly longer than 1ms to recover from reset.

---
 drivers/staging/qlge/qlge_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/staging/qlge/qlge_main.c +0 -0
index 689a87d58f27..3cc4f1902c80 100644

drivers/staging/qlge/qlge_main.c static int qlge_close(struct net_device *ndev) · line 3889

Quoted 5 lines
drivers/staging/qlge/qlge_main.c +1 -1
@@ -3886,7 +3886,7 @@ static int qlge_close(struct net_device *ndev)
 	 * (Rarely happens, but possible.)
 	 */
 	while (!test_bit(QL_ADAPTER_UP, &qdev->flags))
-		msleep(1);
+		usleep_range(1000, 1050);

Have you tested this with a device? Doing these types of changes without access to the hardware isn't a good idea.

Also, a loop that has the chance to never end should be fixed up, don't you think?

thanks,

greg k-h

A message that is not in the archive (sent off-list or not yet received): Yrnr4yYm5mo+PXKU@sug

#3 lore raw
Quoted 41 lines · Greg KH
Quoted 25 lines · Arun Vijayshankar
Quoted 16 lines

qlge_close uses msleep for 1ms, in which case (1 - 20ms) it is preferable to use usleep_range in non-atomic contexts, as documented in Documentation/timers/timers-howto.rst. A range of 1 - 1.05ms is specified here, in case the device takes slightly longer than 1ms to recover from reset.

---
 drivers/staging/qlge/qlge_main.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/staging/qlge/qlge_main.c +0 -0
index 689a87d58f27..3cc4f1902c80 100644

drivers/staging/qlge/qlge_main.c static int qlge_close(struct net_device *ndev) · line 3889

Quoted 5 lines
drivers/staging/qlge/qlge_main.c +1 -1
@@ -3886,7 +3886,7 @@ static int qlge_close(struct net_device *ndev)
 	 * (Rarely happens, but possible.)
 	 */
 	while (!test_bit(QL_ADAPTER_UP, &qdev->flags))
-		msleep(1);
+		usleep_range(1000, 1050);

Have you tested this with a device? Doing these types of changes without access to the hardware isn't a good idea.

Quoted 6 lines

Also, a loop that has the chance to never end should be fixed up, don't you think?

thanks,

greg k-h

I have not tested the change with the hardware. I have checked that the module loads on my PC, but I do not have the actual adapter to test it out. Would you mind sharing any pointers on testing changes on the actual device if I don't have the specific device? I tried to find an emulator or virtual device for the QL adapter, but there doesn't seem to be any. Sorry if this is a very basic question. This is my first patch ever. If there isn't a good way to test the changes with hardware, would you recommend withdrawing it?

Thanks for pointing out the potential runaway loop. I'm a little embarrased on having missed it. I can add a timeout if the device doesn't recover within a given time, but I was wondering if there is a better way of doing it?

Best regards
Arun Vijayshankar