On Thu, Oct 08, 2009 at 06:54:05PM -0700, Rory Filer wrote: > > -----Original Message----- > > > > Is a struct urb that big? What is the memory footprints you are dealing > > with here that this makes a difference? > > C'mon, work with us here, we're trying to reduce memory consumption especially > for embedded users of the driver! > > It's not size of the URB structure we're concerned about, it's the > transfer buffer part. As it is today in the kernel the driver pre-allocates > N_IN_URB (=8) for all enumerated USB interfaces and then allocates the > Rx_transfer part (=4096 bytes). However the only USB interfaces that really > need this number of URBs are the ones for the data path. Some of our modems > enumerate 7 USB interfaces of which only 3 may be used for data. That leaves > the other 4 interfaces needlessly allocating RAM for high-performance thru- > put they will never need... Ok, this was not explained very well, nor was it obvious from the patch, sorry. > > What about performance issues with smaller number of urbs, this change > > will make some device now go slower, which isn't a nice thing :( > > ... the rest of the interfaces are concerned with things like passing AT commands > and proprietary data, etc. In previous iterations of the driver, they got along > just fine with 4 URBS and that's what this patch is doing; keeping the 8 URBS for > the data paths and reducing the non-data path ones to 4 URBs. > > Hopefully this explanation is a little clearer? Yes it is. Care to redo the patch and rewrite the explaination? > > I see a bugfix snuck in here that you should send separately, as it has > > nothing to do with the memory size changes. > > Well, we tried :) Ugh. Don't do that. Now I need to go and pay special attention to every line of your patches to keep you from doing this again, which takes more time from me, and delays your patch acceptance. Not nice :( Please resend the bugfix, as a separate patch, as it probably should be backported to the -stable tree, right? thanks, greg k-h -- 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
