hjg@amms4.UUCP (Harry Gross) (07/11/89)
Hello all,
Ok, I have one for the books. I am running vanilla UNIX S5 R2 1.5 on
a Plexus P/60. It has been working just fine. Recently, I switched modems from
a 1200 baud Hayes to a 2400 baud Cardinal (Hayes Smartmodem compatible), and
made the appropriate modification to the L-devices file. The new modem works
just fine. However, from the moment I made the change, the following _strange_
behavior has been observed:
1) If uucico is started by root, it works ok
2) If uucico is started by anyone else (including the mailer and news),
it is unable to open the device and the attempted communication fails
3) uucico is owned by the uucp account, and has permissions of 4111
(i.e. execute for all and setuserid)
4) The device in question does NOT have a getty running on it, is
owned by uucp and has permissions 666 (r/w for all)
5) Now for the real killer - we have 4 machines tied together via uucp
(I know - this isn't the best way to do things, but it's what I have
to work with), and this problem DOES NOT occur when you try to uucp
to one of the other machines, only on the dialout line.
Note that this behavior is also observed if you try to start up uucico
via the uusub command. Note further that the uucp portion of any transaction
appears to work just fine, as all of the appropriate files get created in the
spool directory.
I have been beating my head against this for several days (2 weeks,
actually), and had hoped that I could figure things out for myself. At this
point, however, I have no further ideas as to where to look, so I am putting
out this call for help.
Any suggestions, ideas, etc. would be GREATLY appreciated (by e-mail,
please, and I will post the solution if anyone is interested [and assuming one
is found :-) ] ). Please note: I do NOT have a source license :-(, so any ideas
involving hacking (or reading) the source do me no good :-).
Right now, I have to keep kick-starting things as root to get out-going
mail (and news :-) to work properly, and I am getting _REAL TIRED_ :-) of doing
what the machine should be doing! HELP!?!?!?!
Thanks for any and all help,
--
Harry Gross | reserved for
| something really
Internet: hjg@amms4.UUCP (we're working on registering)| clever - any
UUCP: {jyacc, rna, bklyncis}!amms4!hjg | suggestions?hjg@amms4.UUCP (Harry Gross) (07/28/89)
Recently, I posted two calls for help regarding a problem with uucp. Uucico would start up, initiate a file transfer and then appear to clobber itself. Also, upon installing a 2400 baud modem, suddenly uucico was unable to open the tty port to access the modem, unless it was started as root. I received several helpful replies summerized below: +------------ |The bulk of the replies suggested that there might be a problem with the |permissions and/or ownership of various uucp-related files: | |cmcl2!rochester!biar!jhood (John Hood) |uunet!ppgbms!moe!jeremy (Jeremy Levine) |sun!Central!texbell!moray!siswat!buck (A. Lester Buck) |sun!daver!cheers!greg (Greg Onufer) |rutgers!bikini.cis.ufl.edu!novavax!ki4pv!tanner (Dr. T. Andrews) |rutgers!nexus.yorku.ca!xrtll!rsnider (Richard Snider) |uunet!visdc!jiii (John E Van Deusen III) |rutgers!wolves!ggw (Gregory G. Woodbury) | | Jeremy Levine also suggested using uucheck as someone other than root |or uucp, but unfortunately, I don't have uucheck on my system. | | A. Lester Buck also suggested restoring the 1200 baud modem to see if |things would start working again, but I was able to resolve the problems before |resorting to this test. | | Greg Onufer suggested obtaining Plexus TCP/IP, but we can't run that |unless (and until) we get S5 R2 V1.7 (which I am working on :-) | | Greg Woodbury also described a problem he had on his machine which |was somewhat similar, and offered his shell scripts that solved his problem. |However, based on his discription, they would not have helped me. (Thanks |for the offer, anyway :-) +------------ +------------ |A couple of suggestions were received concerning modem line settings (i.e. DCD) |etc.: | |ncc!adec23!mark (Mark Salyzyn) |uunet!pd1.ccd.harris.com!bill (Bill Davis) | | Bill Davis also suggested that since root starting uucico worked, there |was something going on involving privliges - but what it was remained unclear. +------------ +------------ |A suggestion was received which turned out to be the proper work-around! : | |rutgers!xait.xerox.com!lakart!dg (David Goodenough) suggested: | |>This won't solve the problem, but it will allow your system to work |>without manual intervention. Add lines of the following form to |>your crontab: |> |>25 1,4 * * * root /usr/lib/uucp/uucico -r1 -ssystem1 |>25 2,5 * * * root /usr/lib/uucp/uucico -r1 -ssystem2 |>25 3,6 * * * root /usr/lib/uucp/uucico -r1 -ssystem3 |> |>These are BSD style, but the same thing is possible under SYSV. Basically |>get crontab to do the grunt work, but by selecting root as the executor, |>your uucico's will run with root privileges. This is what I call a |>"Band-Aid" fix: it solves the symptoms enough to get you going, allowing |>a total fix to be done with less sense of pressure. +------------ +------------ |One person offered me a desperation solution, sources to a HDB clone that |would require porting. I didn't take him up on it, but I may in the future, |just to have it :-) | |utzoo!lethe!dybbuk!yonatan (Jonathan Hanna) +------------ +------------ |One very well known netter told me that these bugs had been fixed in |S5 R2 V1.7 and recommended obtaining that release. As stated above, I'm |working on it :-) | |allbery@ncoast.ORG (Brandon S. Allbery) (Thanks!) +------------ +------------ |I struck gold, however with this note: | |rutgers!dsinc.dsi.com!syd (Syd Weinstein) | |>Maybe I can help you, I did a lot of the uucp stuff for Plexus |>as well as a lot of other comm stuff and controller stuff for them. |>I want you to try cu to that line, there used to be a bug with the |>status tables. Give me a call voice some day. | |After a lengthy telephone conversation with Syd (thanks again for your time!), |I discovered a couple of crucial things: | | 1. There is a bug in S5 R2 V1.5 uucico which sometimes fails to |reset status bits on the I/O port, thus causing the line to drop as soon |as a connection is made. uucico tries to regain the line, and fails. The |logfiles LOOK like uucico is clobbering itself, but it really had already |failed. The fix for this is to run cu on the line just prior to calling |out with uucico (or get a patch from Motorola - who has taken over support |of the Plexus hardware). First problem solved. | | 2. There is an un-related flaw in S5 R2 V1.5 uucico pertaining to |baud rates. It seems that the good folks at AT&T created a baud rate table |inside of uucico that understands baud rates of 110, 300, 1200 and 9600 baud. |9600 was provided for hard-wire connections, and the others were provided for |modems. Since AT&T didn't have a 2400 baud modem at the time, they didn't |put 2400 baud support into uucico. When I invoke uucico, the table is |consulted, 2400 baud support is not found, and uucico terminates. When uucico |is invoked by root, however, this table is not consulted, and uucico will work. |Second problem solved. +------------ I want to thank everyone who responded. Tracking this down was no fun, but your assistance made it bearable. And at least I was able to create work arounds (since I don't have source to make _real_ fixes :-) Cheers, -- Harry Gross | reserved for | something really Internet: hjg@amms4.UUCP (we're working on registering)| clever - any UUCP: {jyacc, rna, bklyncis}!amms4!hjg | suggestions?