While usb_kill_urb is in progress, calls to usb_submit_urb will fail
with -EPERM (documented in Documentation/usb/URB.txt). The UVC driver
does not correctly handle this case -- there is no synchronization
between uvc_v4l2_open / uvc_status_start and uvc_v4l2_release /
uvc_status_stop.
This patch adds a retry / timeout when uvc_status_open / usb_submit_urb
returns -EPERM. This usually means that usb_kill_urb is in progress, and
we just need to wait a while.
BUG=chromium:226216
TEST=Manual. Repeat suspend / resume on Stout many times, verify that
open() on /dev/video0 never fails.
Change-Id: I8a69dab345a18c89e175bd916aa943edd080b6cc
Signed-off-by: Shawn Nematbakhsh <shawnn@chromium.org>
Reviewed-on: https://gerrit.chromium.org/gerrit/48982
Reviewed-by: Vincent Palatin <vpalatin@chromium.org>
{
struct uvc_streaming *stream;
struct uvc_fh *handle;
+ unsigned long timeout;
int ret = 0;
uvc_trace(UVC_TRACE_CALLS, "uvc_v4l2_open\n");
}
if (atomic_inc_return(&stream->dev->users) == 1) {
- ret = uvc_status_start(stream->dev);
+ timeout = jiffies + msecs_to_jiffies(UVC_STATUS_START_TIMEOUT);
+ /* -EPERM means stop in progress, wait for completion */
+ do {
+ ret = uvc_status_start(stream->dev);
+ if (ret == -EPERM)
+ usleep_range(5000, 6000);
+ } while (ret == -EPERM && time_before(jiffies, timeout));
+
if (ret < 0) {
usb_autopm_put_interface(stream->dev->intf);
atomic_dec(&stream->dev->users);
#define UVC_CTRL_CONTROL_TIMEOUT 300
#define UVC_CTRL_STREAMING_TIMEOUT 5000
+#define UVC_STATUS_START_TIMEOUT 100
/* Maximum allowed number of control mappings per device */
#define UVC_MAX_CONTROL_MAPPINGS 1024