[PATCH 1/6] linux: Fix cancel_transfer return value when cancelling a multi-urb transfer

[Hans de Goede]

If we fail to cancel the last urb of a multi-urb transfer because it
has already completed (errno == EINVAL on DISCARD_URB), then the entire
transfer has already completed, so returning NOT_FOUND is consistent with what
the documentation for libusb_cancel_transfer says.

But if we've successfully cancelled the last urb, and then another urb
fails with errno == EINVAL, this means that we've still cancelled the
transfer, as it has only *partially* completed.

Signed-off-by: Hans de Goede <hdegoede at redhat.com>
---
 libusb/os/linux_usbfs.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/libusb/os/linux_usbfs.c b/libusb/os/linux_usbfs.c
index ad3c37a..703d844 100644
--- a/libusb/os/linux_usbfs.c
+++ b/libusb/os/linux_usbfs.c
@@ -1483,7 +1483,8 @@ static int discard_urbs(struct usbi_transfer *itransfer, int first, int last_plu
 
 		if (EINVAL == errno) {
 			usbi_dbg("URB not found --> assuming ready to be reaped");
-			ret = LIBUSB_ERROR_NOT_FOUND;
+			if (i == (last_plus_one - 1))
+				ret = LIBUSB_ERROR_NOT_FOUND;
 		} else if (ENODEV == errno) {
 			usbi_dbg("Device not found for URB --> assuming ready to be reaped");
 			ret = LIBUSB_ERROR_NO_DEVICE;
-- 
1.7.9.3