/usr/lib/uucp/uucico ignores LCK.. files

Mark E. Mallett mem at zinn.MV.COM
Wed Oct 26 06:31:22 AEST 1988


In article <171 at libove.UUCP> root at libove.UUCP (Jay M. Libove) writes:
> 
> While I am using it manually, if a cron job comes up that calls 
> /usr/lib/uucp/uucico, the LCK.. file in /usr/spool/uucp/ will be ignored
> and I will get the modem commands being fired out on the line while I
> am connected over it! As yet, this has beena nuisance, but I suppose it
> stands to be damaging eventually.

Perhaps I am speaking out of turn here, because this is the xenix group
and I am speaking of the same problem in the Microport uucico, but this
looks like the same problem.  When I installed HDB uucp on my Microport
Sys V/AT system, I was careful (in the face of having NO documentation,
of course) to investigate the format of the new LCK files and change my
local comm programs accordingly.  Under the HDB convention, each lock
file contains the decimal-ASCII number of their owner process - a
10-digit field followed by a newline.  I found that even when my comm
programs were constructing LCK files with this convention, they were
being ignored.

A probable reason did occur to me.  uucico is likely checking to see if
the lock's owner process exists.  It is doing this by sending a zero (0)
signal (i.e., kill) to that process; a successful return from kill()
means that the process does indeed exist.

The problem is that an error return from kill() does not necessarily
mean that the process does not exist.  It can often mean that the
process performing the kill() does not have permission to send the
signal to the process owning the lock.  As proof of this, I tried
running my comm programs under ownership of uucp, and the LCK file that
they created were indeed honored by uucico, which was also running as uucp.

This is not a very good situation.  If this is indeed the answer, I
suggest that SCO (and Microport) fix uucico to be more intelligent in
interpreting the result from the kill() of the LCK owner process.

-mm-
-- 
Mark E. Mallett  Zinn Computer Co/ PO Box 4188/ Manchester NH/ 03103 
Bus. Phone: 603 645 5069    Home: 603 424 8129
uucp: mem at zinn.MV.COM  (  ...{decvax|elrond|harvard}!zinn!mem   )
BIX: mmallett



More information about the Comp.unix.microport mailing list