mail archive of the barebox mailing list
 help / color / mirror / Atom feed
* [PATCH v2] usb: storage: retry for up to 10s on lengthy HDD spin up
@ 2021-04-16 17:26 Ahmad Fatoum
  2021-05-07  8:09 ` Sascha Hauer
  0 siblings, 1 reply; 2+ messages in thread
From: Ahmad Fatoum @ 2021-04-16 17:26 UTC (permalink / raw)
  To: barebox; +Cc: Ahmad Fatoum

Some USB disks take notoriously long to spin up. They are seen by a bus
scan, but they report ready only after a few seconds have passed. This
is not a problem if vbus is enabled early on, so devices have had a
chance to spin up. If vbus is first enabled as part of the usb scan,
not enough time might have passed for the USB disk to be usable.

This issue was observed on an i.MX6QP with following topology:

  usb: USB: scanning bus for devices...
  usb: 5 USB Device(s) found
    1 ID 0000:0000
    |  u-boot EHCI Host Controller
    |
    +-2 ID 0424:2517
    |
    +-5 ID 1058:2621
    |    Western Digital Elements 2621
    ...

Unplugging and replugging the USB disk and doing a second usb scan
made the unit ready test succeed. Increasing the retry count
during initialization has negative consequences for other cases,
like when a device is unplugged while being probed (which already
takes way too long).

Instead, just for the case of a detected USB mass storage device that
couldn't get ready initially: retry for 10s at initialization time
before giving up.

Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
---
v1 -> v2:
  - don't increase retry count, instead just check time and don't
    loop longer than 10s at init time.
---
 drivers/usb/storage/usb.c | 20 ++++++++++++--------
 1 file changed, 12 insertions(+), 8 deletions(-)

diff --git a/drivers/usb/storage/usb.c b/drivers/usb/storage/usb.c
index b417640186a2..b0f789b42a9f 100644
--- a/drivers/usb/storage/usb.c
+++ b/drivers/usb/storage/usb.c
@@ -141,20 +141,23 @@ exit:
 	return ret;
 }
 
-static int usb_stor_test_unit_ready(struct us_blk_dev *usb_blkdev)
+static int usb_stor_test_unit_ready(struct us_blk_dev *usb_blkdev, u64 timeout_ns)
 {
+	u64 start;
 	u8 cmd[6];
 	int ret;
 
 	memset(cmd, 0, sizeof(cmd));
 	cmd[0] = SCSI_TST_U_RDY;
 
-	ret = usb_stor_transport(usb_blkdev, cmd, sizeof(cmd), NULL, 0,
-				 10, 100);
-	if (ret < 0)
-		return -ENODEV;
+	start = get_time_ns();
 
-	return 0;
+	do {
+		ret = usb_stor_transport(usb_blkdev, cmd, sizeof(cmd), NULL, 0,
+					 10, 100);
+	} while (ret < 0 && !is_timeout(start, timeout_ns));
+
+	return ret ? -ENODEV : 0;
 }
 
 static int read_capacity_16(struct us_blk_dev *usb_blkdev)
@@ -282,7 +285,7 @@ static int usb_stor_blk_io(struct block_device *disk_dev,
 
 	/* ensure unit ready */
 	dev_dbg(dev, "Testing for unit ready\n");
-	if (usb_stor_test_unit_ready(pblk_dev)) {
+	if (usb_stor_test_unit_ready(pblk_dev, 0)) {
 		dev_dbg(dev, "Device NOT ready\n");
 		return -EIO;
 	}
@@ -365,7 +368,8 @@ static int usb_stor_init_blkdev(struct us_blk_dev *pblk_dev)
 	/* ensure unit ready */
 	dev_dbg(dev, "Testing for unit ready\n");
 
-	result = usb_stor_test_unit_ready(pblk_dev);
+	/* retry a bit longer than usual as some HDDs take longer to spin up */
+	result = usb_stor_test_unit_ready(pblk_dev, 10ULL * NSEC_PER_SEC);
 	if (result) {
 		dev_dbg(dev, "Device NOT ready\n");
 		return result;
-- 
2.29.2


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-05-07  8:10 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-04-16 17:26 [PATCH v2] usb: storage: retry for up to 10s on lengthy HDD spin up Ahmad Fatoum
2021-05-07  8:09 ` Sascha Hauer

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox