> It seems that the XactErr bit gets set occasionally but does > not always lead to hard failure: But then it seems that eventually it occasionally does: [17182959.448000] ehci_hcd 0000:03:0c.2: XactErr, token=0000cd08 [17186270.500000] ehci_hcd 0000:03:0c.2: XactErr, token=00004d08 [17192204.628000] ehci_hcd 0000:03:0c.2: XactErr, token=00004d08 [17207488.564000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17207534.812000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17210149.544000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17213184.564000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17234794.928000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17269865.128000] ehci_hcd 0000:03:0c.2: XactErr, token=00004d08 [17289071.784000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17290546.404000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17293014.016000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17293428.860000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17293474.040000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17298119.660000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17299546.240000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17300465.852000] ehci_hcd 0000:03:0c.2: XactErr, token=80004d08 [17300788.188000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17300925.768000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17300972.004000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17301063.372000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17301155.772000] ehci_hcd 0000:03:0c.2: XactErr, token=8000cd08 [17301248.172000] ehci_hcd 0000:03:0c.2: XactErr, token=8038c948 [17301248.172000] ehci_hcd 0000:03:0c.2: XactErr, token=a8002148 [17301248.172000] ehci_hcd 0000:03:0c.2: devpath 1 ep3in 3strikes,t=a8002148 The last transaction error has also the Halted bit set for the status field of the token. The EHCI spec says: "This can be caused by babble, the error counter counting down to zero, or reception of the STALL handshake from the device during a transaction.". The "Babble Detected" bit is not set whereas the CERR is zero (but that we knew already from the 3strikes debug print). The previous transaction error has CERR==2, but there's no transaction error with CERR==1.. hmm.. Also the other transaction errors have all CERR==3 and transfer length of zero. I put the usbmon log to http://www.hut.fi/~apm/usb-disconnect3.txt.gz If my reasoning about the timestamps is correct the last transaction error is at timestamp 30344747. There is a transfer length 10240 at that timestamp which would match the token value a8002148. I suppose I'll go reading the USB2 spec now.. :-) -- http://www.iki.fi/~ananaza/ _______________________________________________ linux-dvb mailing list linux-dvb@xxxxxxxxxxx http://www.linuxtv.org/cgi-bin/mailman/listinfo/linux-dvb
