2.6.19 bluetooth PPP (rfcomm) regression

2.6.19 bluetooth PPP (rfcomm) regression

[Daniel Drake]

Hi,

Testing from Bill Kenworthy indicates that commit 
0a85b964e141a4b8db6eaf500ceace12f8f52f93 introduces a ppp-over-bluetooth 
regression.

https://bugs.gentoo.org/show_bug.cgi?id=159277

Dec 28 21:56:54 rattus hcid[22749]: pin_code_request (sba=00:0A:3A:59:39:38,
dba=00:07:E0:06:AC:7A)
Dec 28 21:57:02 rattus hcid[22749]: link_key_notify (sba=00:0A:3A:59:39:38,
dba=00:07:E0:06:AC:7A)
Dec 28 21:57:37 rattus rfcomm_tty_ioctl: TIOCGSERIAL is not supported
Dec 28 21:57:37 rattus dund[23081]: New connection from 00:07:E0:06:AC:7A
Dec 28 21:57:37 rattus pppd[23094]: pppd 2.4.4 started by root, uid 0
Dec 28 21:57:37 rattus pppd[23094]: Couldn't get channel number: 
Input/output
error
Dec 28 21:57:37 rattus pppd[23094]: Exit.

> diffing net/bluetooth/rfcomm/tty.c between the two kernels shows
> 
> 2.6.19-r2: tty_register_device(rfcomm_tty_driver, dev->id,
> rfcomm_get_device(dev));
> 
> and
> 
> 2.16.18-r2: tty_register_device(rfcomm_tty_driver, dev->id, NULL);
> 
> Changing this line to match the earlier kernel (using NULL) removes the error
> and ppp connects as it should.



I see this has already been reported by Johannes Hoerhan on the bluez 
list. http://thread.gmane.org/gmane.linux.bluez.devel/10148/focus=10232

I don't really see how this could be a udev issue, since it's obviously 
so intricately linked to that line of code.

How can we help diagnose this further - any debugging flags we should 
turn on?

Thanks,
Daniel