Folks,
Any ideas USB transaction can be screwed ?
Please read problem explanation below.
btw Greg, is there any reason why, in your Bluetooth -> TTY driver, you
submit bulk_in requests with the buf_len exactly
equal to max_packet_len on bulk endpoint instead of a larger buffer that
can hold complete transaction ?
Thanks
Max
>>For DM1/DH1 packet this seems to work ok (a bit slow), but when I
>>increase the packet size lots of kernel messages starts showing up in the
>>recieving end. I've tried on a machine with an Intel controller with
>>uhci.o and usb-uhci.o and on a machine with an Opti controller and
>>ohci.o. They give different error messages, but none of them seems to do
>>the trick.
>>You previously asked me for some dumps, but I havn't gotten arround to
>>installing hcidump, so I hope the following will be enough - otherwise say so.
>>The errors seem to be rather alike, so to me that suggests that the
>>problem is with hci-usb and not the usb drivers...
>hcidump won't help in this case, packets are dropped at usb level.
>Please send me hciconfig -a output.
>
>Dumps indeed look similar but I'm pretty sure that it's not hci_usb
>driver problem.
>Seems like USB transaction got screwed up somehow.
>
>USB folks, any ideas why this could happen ?
>Here is details.
>
>Bluetooth spec defines its own Host Controller Interface (HCI).
>Data packet packets are called ACL packets. Each packet has the following
>header
>typedef struct {
> __u16 handle;
> __u16 dlen;
>} __attribute__ ((packed)) hci_acl_hdr;
>
>Complete ACL packet has to be sent/received in one USB transaction. Only
>the first USB packet in transaction
>has ACL header in it.
>Somehow transaction got completely screwed with this particular device
>(SiliconWave BlueFrog).
>
>Here is a receive dump
>
>hci_usb_rx_complete: hci0 corrupted ACL packet: count 199, dlen 60
>hci_usb: dump, len 199
>hci_usb: 00 10 3C 00 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 00 10 3C 00 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 7F 7F 7F 7F 00 10 3C 00 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F
>hci_usb: 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 7F 00 10 03 00 7F 7F 7F
>
>"00 10 3C 00" is correct ACL header with data len set to 60 bytes. 7F is a
>payload sent by our test tool.
>So as you see somehow bunch of valid ACL packets got merged into one
>transaction.
>Can this be a bug in HCDs or somewhere else in USB stack ?
>
>Max
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel