summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorAndrey Konovalov <andreyknvl@google.com>2020-05-07 20:06:58 +0300
committerFelipe Balbi <balbi@kernel.org>2020-05-14 12:30:18 +0300
commit61d2658db48af0c563d126cccc5c1105137b1ec9 (patch)
tree03c61388a0ef230de689a82d977510b79361a085
parentc61769bd4777a922952aed0d042a2572e5bd9b74 (diff)
downloadlinux-61d2658db48af0c563d126cccc5c1105137b1ec9.tar.xz
usb: raw-gadget: documentation updates
Mention the issue with fixed UDC addresses. Links external examples and test suite. Add more implmenetation details and potential improvements. Signed-off-by: Andrey Konovalov <andreyknvl@google.com> Signed-off-by: Felipe Balbi <balbi@kernel.org>
-rw-r--r--Documentation/usb/raw-gadget.rst30
1 files changed, 28 insertions, 2 deletions
diff --git a/Documentation/usb/raw-gadget.rst b/Documentation/usb/raw-gadget.rst
index 3b3d78e850b2..68d879a8009e 100644
--- a/Documentation/usb/raw-gadget.rst
+++ b/Documentation/usb/raw-gadget.rst
@@ -49,10 +49,36 @@ The typical usage of Raw Gadget looks like:
Raw Gadget and react to those depending on what kind of USB device
needs to be emulated.
+Note, that some UDC drivers have fixed addresses assigned to endpoints, and
+therefore arbitrary endpoint addresses can't be used in the descriptors.
+Nevertheles, Raw Gadget provides a UDC-agnostic way to write USB gadgets.
+Once a USB_RAW_EVENT_CONNECT event is received via USB_RAW_IOCTL_EVENT_FETCH,
+the USB_RAW_IOCTL_EPS_INFO ioctl can be used to find out information about
+endpoints that the UDC driver has. Based on that information, the user must
+chose UDC endpoints that will be used for the gadget being emulated, and
+properly assign addresses in endpoint descriptors.
+
+You can find usage examples (along with a test suite) here:
+
+https://github.com/xairy/raw-gadget
+
+Internal details
+~~~~~~~~~~~~~~~~
+
+Currently every endpoint read/write ioctl submits a USB request and waits until
+its completion. This is the desired mode for coverage-guided fuzzing (as we'd
+like all USB request processing happen during the lifetime of a syscall),
+and must be kept in the implementation. (This might be slow for real world
+applications, thus the O_NONBLOCK improvement suggestion below.)
+
Potential future improvements
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-- Reporting more events (suspend, resume, etc.) through
- USB_RAW_IOCTL_EVENT_FETCH.
+- Report more events (suspend, resume, etc.) through USB_RAW_IOCTL_EVENT_FETCH.
- Support O_NONBLOCK I/O.
+
+- Support USB 3 features (accept SS endpoint companion descriptor when
+ enabling endpoints; allow providing stream_id for bulk transfers).
+
+- Support ISO transfer features (expose frame_number for completed requests).