On Thu, Sep 10, 2009 at 4:29 PM, Antti Palosaari<crope@xxxxxx> wrote: > Eh, not all needed, but we need some kind of rule of thumb which URB size is > suitable for bandwidth used. 512, 8k, 16k etc. It is not wise at all set it > to only 512 bytes when streaming whole TS example 22Mbit/sec. I have tested > Anysee (Cypress FX2), AF9015, CE6230, RTL2831U and all those allowed to set > URB rather freely. If you want to pick bridges that are important to you and take the time to optimize them better, by all means be my guest. This is the sort of thing that would have to be discussed with the individual maintainers of those bridges, so you can understand what logic was used in making the original decision (ensuring the original logic was not done to work around some bug, etc). > I haven't seen yet device which forces to use just one > size - though it is possible there is. Well, it depends on the chip. Selecting too small a value can result in packets getting dropped (this was a problem on em28xx until I fixed it a few months ago). > And no datasheet even needed, you can > see from debug log or error code if URB is not suitable. Well, this assumes the bridge fails gracefully, returning a failure. Take Patrick's example, where the device returns success but then proceed to not send back any URBs. > Why not set it some good value when possible? And also adding module > parameter which overrides driver default is not hard to add, just look value > user gives as param and round it to nearest suitable one. Frankly, I'm not really confident this provides much value. End-users should not really be playing around with these sorts of settings. If the values are wrong, a patch should be submitted and the maintainer should fix the driver. Cheers, Devin -- Devin J. Heitmueller - Kernel Labs http://www.kernellabs.com -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
