On Wed, Jan 03, 2007 at 08:58:07PM -0800, David Brownell wrote: > On Monday 01 January 2007 4:52 am, amateur wrote: > > Hi Guys, > > > > I have a problem with bluez on a arm platform. Here is the detail: > > On the arm embedded platform, I invoke this command > > l2ping -s 667 BDADDR > > to test the bluetooth L2CAP connection, where BDADDR is the bluetooth > > address of my Linux/Bluez laptop. At the same time, I use > > hcidump -x -R > > on my laptop to monitor the bluetooth traffic. The problem is that the > > data I sent from the embedded arm platform got corrupted on > > receptioin. > > You're being unclear here ... Linux on both ends? Is the data > corrupted in memory before the ARM sends it? After? There's > some kind of USB dongle on both ends? The same kind?
Yes, I have linux both on the arm embedded platform and on my thinkpad. The data in memory is not corrupted before nor after USB TX. I have checked the transfer buffer associated with the urb. > > > > After a careful check, I found that the error has the > > following pattern: Some bytes of the ACL data packet, which > > corresponding to the L2CAP Echo Request Command, will have some 4 > > bytes of its content changed from valid data to 4 bytes of zero on > > occasion. And this change-to-zero error may occur several times in a > > ACL Packet sometimes. The distance between consecutive errors is > > always a multiple of 64-bytes which is the maxPacketSize of the USB > > pipe of the bluetooth dongle. The offset of the error position in the > > ACL Packet is always 64*i+4-64*i+7(4 bytes), where 'i' is an integer. > > That is, (64 * i) + 4 - (64 * i) + 7 == +11 bytes offset always?? That's (64 * i) + 4 -> (64 * i) + 7, namely, 4 bytes starting from offset 4. Sorry I didn't make it clear. > > Check out the data before and after TX; you can just checksum it. > That will let you verify that it was OK on TX. Do the same kind I didn't do the check on my laptop which runs linux-2.6.18-7. Because I don't think the problem is at the laptop end. > > > I am really confused. I have checked the skb and urb associated with > > the ACL Packet. And they both looks fine(content correct). So where > > could the problem lies in? Has anyone have had such a problem before? > > Any clue or hints are welcome! > > > > Is this an known bug of OHCI HC driver usb-ohci in kernel 2.4.21? What > > makes me more confused is that the Usb Storage Disk works great on the > > embedded system. > > If you're asking about support for USB in 2.4, or ARM on 2.4, you're > pretty much on your own (unless you have a contract with some vendor > to support you). You're asking for support for _both_ of those, as > well as Bluetooth ... all of which adds up to your being really unlikely > to get any useful answers from the Linux community. (Except via paid > support contracts.) > > The OHCI driver has been seriously reworked since 2.4, in part to get > rid of lots of little problems that slower CPUs (like ARM) were very > good at turning up. The 2.6 OHCI code is pretty solid, and works on > a surprisingly large variety of silicon implementations. I know that 2.4.21 is pretty old. But the only kernels the vendor(Cirrus) support are 2.4.21 and 2.6.8-1. I also tried 2.6.8-1, which also failed, and the problem is exactly the same one. > > > I'm using linux-2.4.21 with patch-linux-2.4.21-mh10 and a cirrus edb9312 > > embedded arm platform which has a OHCI-Compatible USB Host Controller. > > The bluetooth dongle is a CSR one which works normally on my laptop > > PC. > > Try upgrading to Linux 2.6.20-rc, and see if that changes anything. > > Don't ask for support for any Linux-ARM 2.4 kernel and expect anything > more than a pointer to the FAQ that says "upgrade to 2.6 and try again". > It's been a few years since starting new projects on 2.4 kernels was a > good idea. Thank you for your time though. > > - Dave > ------------------------- -- Q: What do monsters eat? A: Things. Q: What do monsters drink? A: Coke. (Because Things go better with Coke.) ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share your opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV _______________________________________________ linux-usb-devel@lists.sourceforge.net To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
