freebsd-skq/sys/i4b/capi
Hellmuth Michaelis a986a974e9 It seems that I slipped a nasty bug into the CAPI support. The message
lengths for CONNECT_REQ and CONNECT_IND are incorrect, which causes
dialouts to fail after certain error situations (an invalid -- not
wrong! -- number has been dialed). Since these messages are tagged as
too short, the device reads trailing garbage as the B protocol
parameters; this is OK as long as the garbage consists of zero bytes,
which it usually does, except after the said error.

Another change we have taken into use is to send an explicit Q.850
"normal call clearing" code when a call is ignored using PRI equipment
(specifically AVM T1); the CAPI pseudo-code for ignore, 1, translates
into something at least Ericsson exchanges interpret oddly (message
"this area is not reachable from your number"). NCCLR makes the exchange
give a busy signal, which is the behaviour at least we prefer
(conceivably, the ignore code could be made a sysctl variable).

The attached patch corrects the message length issue. It also includes a
somewhat unpretty solution for the PRI ignore code (if device's number
of channels equals 30, assume PRI and send NCCLR, otherwise send CAPI
ignore). Tested using AVM B1 PCI and T1 PCI.

Submitted by:	Juha-Matti Liukkonen <jml@cubical.fi>
Reviewed by:	hm
MFC after:	1 month
2001-10-21 08:51:54 +00:00
..
iavc
capi_l4if.c
capi_llif.c
capi_msgs.c It seems that I slipped a nasty bug into the CAPI support. The message 2001-10-21 08:51:54 +00:00
capi_msgs.h
capi.h
README

$FreeBSD$

Message-ID: <3AF56886.7D92609A@cubical.fi>
Date: Sun, 06 May 2001 18:06:47 +0300
From: Juha-Matti Liukkonen <jml@cubical.fi>
Organization: Cubical Solutions Ltd

Please find the attached diff and tarball for our support software for
CAPI 2.0 and AVM's active T1 and T1-B (primary rate) and B1 (basic rate)
ISDN adapters for isdn4bsd. The implementation has been made from
scratch by us, based on reverse engineering of the Linux driver provided
by AVM GmbH and available in ftp.avm.de. Cubical Solutions Ltd offers
this implementation for the BSD community free of charge and assuming
absolutely no liabilities whatsoever. Feel free to modify the
implementation any way you see fit, but please retain our one-liner
copyright statement somewhere in the comment headers in the capi and
iavc driver modules.

That said, the attached tarball is i4b-00.96.00-beta with our
modifications integrated, and the diff contains all modifications we
have made to the original (including the new capi files). Our mods add
pseudo-device i4bcapi, which attaches to i4b layer 4, and device iavc0,
which implements a link layer driver for AVM's active B1 and T1 adapters
for i4bcapi. There are also a couple of related improvements to isdnd,
and a number of modifications to isdnd and layer 4 to implement support
for up to 30 channels per adapter (for primary rate use).

We have developed the software explicitly for our telephony application,
to be used with AVM's T1 adapters, and the implementation has been
tested primarily with this functionality in mind. There may be
interesting side effects with eg. the ipr and isppp drivers; we do not
use them and therefore their testing has been cursory at best. The
i4btel driver works well with the T1 (our primary use), and the i4brbch
driver has been verified to work with T1, T1-B and B1 adapters (ftp'd
some files over a dialup PPP connection with each adapter). Only the PCI
versions of the adapters (equipped with the AMCC DMA controller) are
supported, although the basics (PIO mode communication) for the older
ISA model support is in place, so only the bus attachment modules should
be required to support the older hardware.

All of the AVM active adapters use downloadable firmware, which is not
included in the attached package. The firmware files (t1.t4, t1b.t4,
b1.t4) can be found from ftp.avm.de in adapter specific subdirectories,
or from the CDs provided with the adapters (in directory
'cardware/firmware').

Our primary development platform is our own embedded build (we call it
'ebsd') based on FreeBSD 4.2-RELEASE. The implementation has also been
tested on standard FreeBSD 4.2-RELEASE system. The implementation should
not contain any FreeBSD (or even FreeBSD release) specific issues, but
it has not been tested or even compiled on any other platforms;
specifically, only the FreeBSD overinstall.sh script is modified to
install the capi/iavc support in the kernel source tree.

This code is not under active development here since the functionality
we use (i4btel, T1) has been working since the beginning of March. We
are also not planning on any further development (no need seen at this
point), but I am naturally interested on whatever bugs and development
ideas pop up on the community and will keep a keen eye on the isdn
mailing list. I personally may be available for consultation, debugging
and possibly development projects, but with notable reservations on my
time (the current IT industry recession seems to be pushing even more
work for us, which tends to keep us pretty busy these days).

Here are some specific technical notes:

* isdnd supports new keyword 'firmware=</path/to/file>' in section
'controller'. This keyword is supported for all controller types, and
causes I4B_CTRL_DOWNLOAD ioctl to be invoked with the specified file as
an argument. In systems equipped with both active and passive adapters,
and the passive cards being detected first, dummy 'controller' entries
are required for the passive cards to get the correct firmwares to
correct adapters. (I hope you did not have other uses for this ioctl in
mind?)

* isdnd supports new keyword 'clone=<entry name>' in section 'entry'.
This causes the entry to be copied from the existing named entry. At
least entry specific 'name' and 'usrdeviceunit' values should be
specified after a 'clone'. (Makes configuring 30 or 60 entries way much
easier.)

* a bug in i4btel driver read routine corrected. The conditions in the
while() clause caused the receive queue to be referenced before checking
if a channel is connected, leading to kernel panic (do a 'dd
if=/dev/i4btel0 of=/dev/null' on an unconnected tel device, panic will
follow). Correction was to reorder the while clause conditions to check
for connectedness first.

* isdnd and i4b layer 4 support up to CHANNELS_MAX (=30) channels per
adapter. The msg_ctrl_info_req_t message reports the number of channels
provided by the adapter, the number is stored in the nbchan field of the
controller state structure. The fixed stateb1 and stateb2 entries in
controller state stuctures are replaced with an array, and all fixed
references there are replaced with loops up to nbchan. Passive stack
layer 1 and 2 are not modified, layer 3 sets this field to fixed value 2
for all adapters (but it could be delegated to the layer 1 driver's
attach request).

* the i4bcapi driver maps i4b channels to logical channels identified
with PLCI/NCCI values in the CAPI protocol using the sc_bchan[] array.
The PLCI/NCCI handling is merged in the channel mapping and is greatly
simplified from the reference state machine model, because in practice
there can be only one PLCI/NCCI per channel active at any given time.

* the i4bcapi driver does not provide any kind of user level interface
(such as the /dev/capi20 interface provided by the linux driver), but
could relatively easily be extended to do so (and if done, interface
compatibility with the linux implementation would probably be a good
goal).

* there are some gritty details in the iavc driver, inherited from the
linux code. Despite us being a legitimate company in the telecom
business, AVM failed to produce any programming reference material for
us (at least in a reasonable time frame), so some guesswork remains due
to classic reverse engineering process (particularly there are a few
magic numbers in the card initialization sequence whose meaning I do not
know).

* pseudo-devices i4bq931, i4bq921 and some passive stack layer 1 driver
(such as iwic) are required to compile, as the required ctrl_desc[]
array is in layer 3, which requires layer 2, which requires layer 1.
Some architectural cleanup would come in handy here, but we did not want
to start making any major changes (and we use iwic in test setups
anyway, so we simply always compile it in).

To summarize: unpack, overinstall, add the following lines (with the
usual passive stack configuration including at least one L1 driver) to
your kernel configuration file:

pseudo-device	"i4bcapi"
device		iavc0

...and the following to your isdnd.rc:

controller
firmware = /usr/lib/isdn/b1.t4

...compile your new kernel, make sure the firmware file is in
/usr/lib/isdn, and your B1 adapter should boot up and Just Work (tm). If
you have multiple adapters, you need a 'controller' section for each to
have them loaded and booted on isdnd startup.

Have fun -- and let me know if there are any complications, or if I can
be of further assistance,

	- Juha
--
Juha-Matti Liukkonen, Cubical Solutions Ltd
Phone: +358(0)405280142
Email: jml@cubical.fi