Help with line disciplines

[Steve Nuchia]

In article <763 at dms.UUCP> shepperd at dms.UUCP (Dave Shepperd) writes:
>What does the kernel expect the line discipline routines to do EXACTLY?
>I mean what bits in what structures does it want twiddled and, more
>importantly, what bits does it want left alone? I'm concerned the most

This is not documented in any obvious place, and as far as I know
isn't *documented* anywhere.  It can be deduced within the limits
of the correctness of the examples by examining existing drivers
and LD code, if you have access and choose to use it.

>about the l_open, l_close and the l_ioctl functions. I have made
>assumptions about what the others do and apparently guessed right
>because the editor works. This question may be best answered by suggesting
>a book on the subject.

If you find one PLEASE let me know.  I have every book that even
mentions Unix driver writing, and LD interfacing is never covered
in any kind of detail.  I've had to do pretty much the same
reverse-engineering job you did for my 286/386 serial driver.

>Is it always correct for the l_open and l_ioctl routines to assume the
>u structure has meaningful data in it (or that it is even accessible)?

Yes.  The u area is nailed down and correct for anything that
isn't called at interrupt time.  Unless something in streams
breaks this?  I haven't played with stream from the inside yet.

>How does a driver, given the process group number, access the list of
>processes belonging to that group? (ps does it somehow). I believe I will
>need to access this list at interrupt time.

The only thing you can do to the process group is send a signal
to all members; the signal routine has a loop over the process
table that tickles every process that matches.  If you want
to do something else (what? sounds dangerous) you need to
loop similarly.

-- 
Steve Nuchia	      South Coast Computing Services
uunet!nuchat!steve    POB 890952  Houston, Texas  77289
(713) 964 2462	      Consultation & Systems, Support for PD Software.