On Sun, 6 Sep 2015, Peter Chen wrote: > On Wed, Sep 02, 2015 at 10:23:39AM -0400, Alan Stern wrote: > > On Wed, 2 Sep 2015, Peter Chen wrote: > > > > > The bulk queue tests are used to show 'best performance' for bulk > > > transfer, we are often asked this question by users. > > > > > > It's result should be very close to IC simulation, in order > > > to get that, the device side should also need to prepare enough > > > queue. > > > > > > We have got the 'best performance' (IN: ~41MB, OUT: ~39MB) at i.mx platform > > > (USB2, ARM Cortex A9) with below command: > > > > > > Host side: > > > modprobe usbtest > > > ./testusb -a -t 27 -g 64 -s 16384 > > > ./testusb -a -t 28 -g 64 -s 16384 > > > Gadget side: > > > modprobe g_zero loopdefault=1 qlen=64 buflen=16384 > > > > > > Signed-off-by: Peter Chen <peter.chen@xxxxxxxxxxxxx> > > > --- > > > > > > I am not sure if it is good to reuse iso structure, so I take > > > it as RFC. If reuse is accepted, I will rename the iso > > > structures to common one. > > > > > > drivers/usb/misc/usbtest.c | 88 ++++++++++++++++++++++++++++++++++++---------- > > > 1 file changed, 69 insertions(+), 19 deletions(-) > > > > > > diff --git a/drivers/usb/misc/usbtest.c b/drivers/usb/misc/usbtest.c > > > index ad6dd4a..f1faeb9 100644 > > > --- a/drivers/usb/misc/usbtest.c > > > +++ b/drivers/usb/misc/usbtest.c > > > @@ -17,6 +17,7 @@ > > > static int override_alt = -1; > > > module_param_named(alt, override_alt, int, 0644); > > > MODULE_PARM_DESC(alt, ">= 0 to override altsetting selection"); > > > +static void iso_callback(struct urb *urb); > > > > > > /*-------------------------------------------------------------------------*/ > > > > > > @@ -239,7 +240,8 @@ static struct urb *usbtest_alloc_urb( > > > unsigned long bytes, > > > unsigned transfer_flags, > > > unsigned offset, > > > - u8 bInterval) > > > + u8 bInterval, > > > + bool queue_callback) > > > > Should this be named "is_iso" instead? It would be be more clear. > > > > ISO transfer uses different function iso_alloc_urb to allocate urb, > the reason we need one more parameter at usbtest_alloc_urb is we > use different callback for simple polling bulk transfer and > interrupt-triggered bulk transfer. Sorry, I'm a little slow today. I can understand that you want to use different callback routines for the different tests, but I don't understand why you want to use iso_callback for a bulk test. Do you plan to rename the callback as well as the structure? Would it make more sense for the extra argument here to be a pointer to the callback function, instead of a flag? > So, do you agree I reuse the iso structures to implement it? I don't see anything bad about using a common structure for two different purposes, so long as the comments explain what is happening. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
