added kernel detach to avoid error: LIBUSB_ERROR_BUSY #64
Conversation
|
*fixed kernel claim |
firehose.c
Outdated
| @@ -130,13 +130,13 @@ static int firehose_read(struct qdl_device *qdl, int timeout_ms, | |||
| timeradd(&now, &delta, &timeout); | |||
|
|
|||
| for (;;) { | |||
| n = qdl_read(qdl, buf, sizeof(buf), 100); | |||
| n = qdl_read(qdl, buf, sizeof(buf), 700); | |||
There was a problem hiding this comment.
But when we hit this timeout, won't timercmp() below result in the loop to be continued?
There was a problem hiding this comment.
Please see pull request #67.
usb.c
Outdated
| @@ -193,8 +193,10 @@ int qdl_read(struct qdl_device *qdl, void *buf, size_t len, unsigned int timeout | |||
| int ret; | |||
|
|
|||
| ret = libusb_bulk_transfer(qdl->usb_handle, qdl->in_ep, buf, len, &actual, timeout); | |||
| if (ret < 0) | |||
| if (ret < 0){ | |||
| warnx("bulk read failed: %s", libusb_strerror(ret)); | |||
There was a problem hiding this comment.
Will this not result in two timeout prints when a timeout happens?
And for the initial firehose_read() in firehose_run(), we do expect to get a timeout once we've consumed the LOG entries. I would hope we can remove the existing "firehose operation timed out" print.
|
We did some testing and have confirmed that the LIBUSB_ERROR_BUSY is something we want, so I was hoping to merge that. |
|
ok, created separate commit only for LIBUSB_ERROR_BUSY : #68 I got mate, who can reflash over Windows with QFIL without problems, but at linux got constant timeout in random places, usually at big image part 2 gb: FIREHOSE READ: |
No description provided.