Re: [Bluez-devel] [BUG] hci_usb + uhci vs. usb-uhci

On Fri, Sep 12, 2003 at 06:32:51PM -0400, Johannes Erdfelt wrote:

On Fri, Sep 12, 2003, Jean Tourrilhes [off-list ref] wrote:

quoted

	I just spent some time tracking a problem with hci_usb (BlueZ
USB driver) and uhci. This happens on 2 different PCs with different
hardware and 2.4.X kernels, I describe one here.

UHCI Hardware :
	00:07.2 USB Controller: Intel Corp. 82371AB PIIX4 USB (rev 01)
BlueTooth hardware :
	HCI Ver: 1.1 (0x1) HCI Rev: 0x1bb LMP Ver: 1.1 (0x1) LMP Subver: 0x1bb
	Manufacturer: Cambridge Silicon Radio (10)
Kernel : 2.4.22 UP

	What happens :
	Start BNEP session (i.e. TCP/IP over BT). Start intensive TCP
app (web download, netperf, X-Window). After a few seconds, on the
transmitter the hardware just stops :
--------------------------------------------------------
< ACL data: handle 0x0029 flags 0x01 dlen 192
< ACL data: handle 0x0029 flags 0x01 dlen 192
< ACL data: handle 0x0029 flags 0x01 dlen 192
< ACL data: handle 0x0029 flags 0x01 dlen 192
< ACL data: handle 0x0029 flags 0x01 dlen 192
--------------------------------------------------------
	The "Number of Completed Packets" event never comes back. I
have to physically unplug the USB dongle to recover. I could not see
any error messages in the log, except after a few seconds some BNEP
and ACL timeouts (which are expected).

	I initially though that it was a BlueTooth problem. However,
I've never seen it happening with my OHCI hardware on SMP (which is
what I normally use). And, if I replace uhci.o with usb-uhci.o on the
same hardware, the problem goes away completely.
	So, there is definitely something about uhci.o...

Did you enable zero packet when you compiled the BT driver?

JE

	Nope. The compilation of hci_usb was exactly the same for uhci
and for usb-uhci/ohci.

	Jean