Provided by: linux-doc-2.6.15_2.6.15-23.39_all
usb_submit_urb - issue an asynchronous transfer request for an endpoint
int usb_submit_urb (struct urb * urb, gfp_t mem_flags);
urb pointer to the urb describing the request
the type of memory to allocate, see kmalloc for a list of valid
options for this.
This submits a transfer request, and transfers control of the URB
describing that request to the USB subsystem. Request completion will
be indicated later, asynchronously, by calling the completion handler.
The three types of completion are success, error, and unlink (a
software-induced fault, also called ‘‘request cancellation’’).
URBs may be submitted in interrupt context.
The caller must have correctly initialized the URB before submitting
it. Functions such as usb_fill_bulk_urb and usb_fill_control_urb are
available to ensure that most fields are correctly initialized, for the
particular kind of transfer, although they will not initialize any
Successful submissions return 0; otherwise this routine returns a
negative error number. If the submission is successful, the complete
callback from the URB will be called exactly once, when the USB core
and Host Controller Driver (HCD) are finished with the URB. When the
completion function is called, control of the URB is returned to the
device driver which issued the request. The completion handler may then
immediately free or reuse that URB.
With few exceptions, USB device drivers should never access URB fields
provided by usbcore or the HCD until its complete is called. The
exceptions relate to periodic transfer scheduling. For both interrupt
and isochronous urbs, as part of successful URB submission
urb->interval is modified to reflect the actual transfer period used
(normally some power of two units). And for isochronous urbs,
urb->start_frame is modified to reflect when the URB’s transfers were
scheduled to start. Not all isochronous transfer scheduling policies
will work, but most host controller drivers should easily handle ISO
queues going from now until 10-200 msec into the future.
For control endpoints, the synchronous usb_control_msg call is often
used (in non-interrupt context) instead of this call. That is often
used through convenience wrappers, for the requests that are
standardized in the USB 2.0 specification. For bulk endpoints, a
synchronous usb_bulk_msg call is available.
URBs may be submitted to endpoints before previous ones complete, to
minimize the impact of interrupt latencies and system overhead on data
throughput. With that queuing policy, an endpoint’s queue would never
be empty. This is required for continuous isochronous data streams, and
may also be required for some kinds of interrupt transfers. Such
queuing also maximizes bandwidth utilization by letting USB controllers
start work on later requests before driver software has finished the
completion processing for earlier (successful) requests.
As of Linux 2.6, all USB endpoint transfer queues support depths
greater than one. This was previously a HCD-specific behavior, except
for ISO transfers. Non-isochronous endpoint queues are inactive during
cleanup after faults (transfer errors or cancellation).
RESERVED BANDWIDTH TRANSFERS
Periodic transfers (interrupt or isochronous) are performed repeatedly,
using the interval specified in the urb. Submitting the first urb to
the endpoint reserves the bandwidth necessary to make those transfers.
If the USB subsystem can’t allocate sufficient bandwidth to perform the
periodic request, submitting such a periodic request should fail.
Device drivers must explicitly request that repetition, by ensuring
that some URB is always on the endpoint’s queue (except possibly for
short periods during completion callacks). When there is no longer an
urb queued, the endpoint’s bandwidth reservation is canceled. This
means drivers can use their completion handlers to ensure they keep
bandwidth they need, by reinitializing and resubmitting the
just-completed urb until the driver longer needs that periodic
The general rules for how to decide which mem_flags to use are the same
as for kmalloc. There are four different possible values; GFP_KERNEL,
GFP_NOFS, GFP_NOIO and GFP_ATOMIC.
GFP_NOFS is not ever used, as it has not been implemented yet.
GFP_ATOMIC is used when (a) you are inside a completion handler, an
interrupt, bottom half, tasklet or timer, or (b) you are holding a
spinlock or rwlock (does not apply to semaphores), or (c)
current->state != TASK_RUNNING, this is the case only after you’ve
GFP_NOIO is used in the block io path and error handling of storage
All other situations use GFP_KERNEL.
Some more specific rules for mem_flags can be inferred, such as (1)
start_xmit, timeout, and receive methods of network drivers must use
GFP_ATOMIC (they are called with a spinlock held); (2) queuecommand
methods of scsi drivers must use GFP_ATOMIC (also called with a
spinlock held); (3) If you use a kernel thread with a network driver
you must use GFP_NOIO, unless (b) or (c) apply; (4) after you have done
a down you can use GFP_KERNEL, unless (b) or (c) apply or your are in a
storage driver’s block io path; (5) USB probe and disconnect can use
GFP_KERNEL unless (b) or (c) apply; and (6) changing firmware on a
running storage or net device uses GFP_NOIO, unless b) or c) apply