Hi, apologies for dropping a random question on you, but David Eriksson thought you may be able to give us a pointer about this.
We'd observed a few devices giving -100 errors on connection with the usb-rndis-lite module, and John Carr came up with the attached patch by comparing -lite to -ng. This seems to be very effective, but none of us are really device side developers, and John was never completely convinced this was correct. Quoting him from a while back.... "The reason I consider it dirty is that it needs the rndis_get_in_endpoint function which is called every time something calls rndis_command. Ideally that should run once when the driver loads for a given device, but i felt like i was being too invasive when I started doing that and whimped out. The patch exists through careful comparison of the deprecated user mode and the current kernel mode implementations of usb-rndis, the poking of an INT IN endpoint being the only difference i found." I'd love to get something based on this upstream, it's one of the few remaining bugbears for which we haven't pushed a fix to the kernel. Does this ring any bells with you ? Or does it even look sane ? Many thanks Mark
diff -Nurp usb-rndis-lite.bak/rndis_host.c usb-rndis-lite/rndis_host.c
--- usb-rndis-lite.bak/rndis_host.c 2009-07-07 10:35:20.000000000 +0100
+++ usb-rndis-lite/rndis_host.c 2009-07-07 10:49:20.000000000 +0100
@@ -64,6 +64,25 @@ void rndis_status(struct usbnet *dev, st
}
EXPORT_SYMBOL_GPL(rndis_status);
+/* Function ripped from keyspan_remote.c */
+static struct usb_endpoint_descriptor *rndis_get_in_endpoint(struct usb_host_interface *iface)
+{
+
+ struct usb_endpoint_descriptor *endpoint;
+ int i;
+
+ for (i = 0; i < iface->desc.bNumEndpoints; ++i) {
+ endpoint = &iface->endpoint[i].desc;
+
+ if (usb_endpoint_is_int_in(endpoint)) {
+ /* we found our interrupt in endpoint */
+ return endpoint;
+ }
+ }
+
+ return NULL;
+}
+
/*
* RPC done RNDIS-style. Caller guarantees:
* - message is properly byteswapped
@@ -77,11 +96,14 @@ EXPORT_SYMBOL_GPL(rndis_status);
int rndis_command(struct usbnet *dev, struct rndis_msg_hdr *buf, int buflen)
{
struct cdc_state *info = (void *) &dev->data;
+ struct usb_endpoint_descriptor *endpoint;
int master_ifnum;
int retval;
unsigned count;
__le32 rsp;
u32 xid = 0, msg_len, request_id;
+ char int_buf[128];
+ int maxp, pipe, partial;
/* REVISIT when this gets called from contexts other than probe() or
* disconnect(): either serialize, or dispatch responses on xid
@@ -110,6 +132,29 @@ int rndis_command(struct usbnet *dev, st
// we time out and cancel our "get response" requests...
// so, this is fragile. Probably need to poll for status.
+ /* FIXME This feels rancid */
+ endpoint = rndis_get_in_endpoint(info->control->cur_altsetting);
+ pipe = usb_rcvintpipe(dev->udev, endpoint->bEndpointAddress);
+ maxp = usb_maxpacket(dev->udev, pipe, usb_pipeout(pipe));
+
+ retval = usb_interrupt_msg(dev->udev,
+ pipe,
+ int_buf,
+ (maxp > 8 ? 8 : maxp),
+ &partial,
+ RNDIS_CONTROL_TIMEOUT_MS);
+
+ dev_dbg(&info->control->dev,
+ "pipe: %d, maxp: %d, partial: %d, retval: %d\n",
+ pipe,
+ maxp,
+ partial,
+ retval);
+
+ /* I /think/ usb_interrupt_msg blocks and returns < 0 for error */
+ if (unlikely(retval < 0))
+ return retval;
+
/* ignore status endpoint, just poll the control channel;
* the request probably completed immediately
*/
signature.asc
Description: This is a digitally signed message part
------------------------------------------------------------------------------ This SF.Net email is sponsored by the Verizon Developer Community Take advantage of Verizon's best-in-class app development support A streamlined, 14 day to market process makes app distribution fast and easy Join now and get one step closer to millions of Verizon customers http://p.sf.net/sfu/verizon-dev2dev
_______________________________________________ SynCE-Devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/synce-devel
