On Tue, Jun 14, 2011 at 6:09 PM, Felipe Balbi <balbi@xxxxxx> wrote: > hi, > > On Tue, Jun 14, 2011 at 06:06:16PM +0800, Bob Liu wrote: >> > On Tue, Jun 14, 2011 at 05:52:23PM +0800, Aaron Wu wrote: >> >> On Tue, Jun 14, 2011 at 5:39 PM, Felipe Balbi <balbi@xxxxxx> wrote: >> >> > Hi, >> >> > >> >> > On Tue, Jun 14, 2011 at 05:34:09PM +0800, Aaron Wu wrote: >> >> >> Since the wifi card we are using is of the USB interface. We tried on >> >> >> two diffrent USB controllers, looks like the ISP1362 works OK and >> >> >> problems happens only on MUSB controller. For both USB controllers we >> >> >> are using the same kernel version and USB wifi card. >> >> > >> >> > Yeah, USB host has quite some performance issues to be fixed. To start >> >> > with, it uses only DMA Mode0, which is bound to packet boundaries. BTW, >> >> > which embedded platform are you using ? Which SoC ? >> >> > >> >> >> >> Hi balbi, >> >> Thanks for the quick reply. Our concern is the priority capability, >> >> current problems is, the network data packge should have been >> >> prioritized in the IPV4 stack itself, by IP_TOS option as mentioned, >> >> we are just curious why the priority looks like to have something to >> >> do with the USB controller, I mean by experiment the prioritization >> >> works well on some USB controllers like that on a X86 PC and ISP1362, >> >> but does not work on MUSB. >> >> >> >> We are using PIO mode on Blackfin. Performance is not our concern at >> >> this moment, actually the MUSB has better throughput performance than >> >> the ISP1362, but the prioritization works on 1362. On the 1362 >> >> controller, the second iperf session(high priority) always take up a >> >> large bandwidth while the other two low priority session get only a >> >> little in the test as expected. On MUSB, we see no priority for the >> >> second session, it drops greatly as if it has the same low priority as >> >> the first and third session. >> > >> > Well, I can't answer that question, specially since it's on Blackfin. >> > >> > Maybe Mike can help >> >> I have looked into that problem and have no idea. So i asked Aaron to >> email to maillist for some help. >> >> Is anyone can help verify this problem on other musb platform? Thanks. > > yeah, but if the problem isn't throughput, I don't see how MUSB could be > to blame here. The network priority should have been handled by IP > layer and the controller shouldn't matter, right ? Yes, it's not about performance. We think the same as you do, it should be something handled by the IP layer and the controller does not matter. But expriment shows ISP1362 and X86 PC works, we are curious how this could happen, wired. Appraciate very much if someone can shed some light. -- 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
