rey@safn2.UUCP (rey) (04/04/91)
I cannot reliably Poll with uucp because it fails about 1/2 the time. The result is usually a success one hour later. An hour later than I need it to be. CODAR told me I have a bad modem. I do not believe that. Modems are Hayes 2400. tty20 and tty21 behave the same. 1200 and 2400 behave the same. Failures are not related to machine load, happening at all hours. There must be something wrong with the hpsio driver. Has anyone had better luck? I have seen several knocks of the hpsio in this group. Can it really be "non-functional"? I have used it for several years on various releases. I have also looked at getty_ps from comp.sources.misc (unix?); but I have not been able to make it work. I do not think this problem has anything to do with getty now; although I once thought that I might solve the problem by having the better modem control of getty_ps. Output from "Uutry -r safn1" follows: mchFind called (safn1) conn(safn1) Device Type ACU wanted mlock tty21 succeeded processdev: calling setdevcfg(uucico, ACU) gdial(hayes12) called expect: ("") got it sendthem () CLOCAL set DELAY <NO CR>AT^M) expect: (OK^M) lost line errno - 0 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^About 50% of the time. set interface UNIX mlock tty20 succeeded processdev: calling setdevcfg(uucico, ACU) gdial(hayes12) called expect: ("") got it sendthem () CLOCAL set DELAY <NO CR>AT^M) expect: (OK^M) AT^M^M^JOK^Mgot it sendthem (<NO CR>AT&D3V1Q0E1S0=1&C1&S1M0&J0B1^M) expect: (OK^M) ^JAT&D3V1Q0E1S0=1&C1&S1M0&J0B1^M^M^JOK^Mgot it sendthem (ECHO CHECK ON <NO CR>A^JATTDDTT,,??????^M^M) expect: (CONNECT) ^M^JCONNECTgot it sendthem () CLOCAL clear <NO CR>) getto ret 6 expect: ("") got it sendthem (^MDELAY ^M^M) expect: (ogin:) Continues to "Successful" Devices file follows: Direct tty21 - 2400 direct ACU tty21,M - 2400 hayes24 \T ACU tty21,M - 1200 hayes12 \T ACU tty20,M - 2400 hayes24 \T ACU tty20,M - 1200 hayes12 \T Dialers file excerpt follows: The \M \m is new in rel 3. \M sets port CLOCAL. # Gold Star 2400 GoldStar =,- "" \M\dAT\r\c OK\r \dAT&D2V1QE1S0=1&C1&S1&JB1\r\c OK\r \EATDT\T\r\c CONNECT \m\c # &C1 and &S1 required for logon. hayes12 =,-, "" \M\dAT\r\c OK\r AT&D3V1Q0E1S0=1&C1&S1M0&J0B1\r\c OK\r \EATDT\T\r\c CONNECT \m\c hayes24 =,-, "" \M\dAT\r\c OK\r AT&D2V1Q0E1S0=1&C1&S1M0&J0B1&W\r\c OK\r \EATDT\T\r\c CONNECT \m\c Thanks in advance for any help. -- Reynolds McClatchey (Southern Aluminum Finishing Co, Atlanta, GA, USA) Architectural Aluminum. Custom Fabrication. Paint, Powder Coating, Anodizing. uunet!safn2!rey MCImail 414-2935
ra@intsys.no (Robert Andersson) (04/04/91)
rey@safn2.UUCP (rey) writes: >I cannot reliably Poll with uucp because it fails about 1/2 the time. >The result is usually a success one hour later. An hour later than I need >it to be. >Dialers file excerpt follows: The \M \m is new in rel 3. \M sets port CLOCAL. >hayes12 =,-, "" \M\dAT\r\c OK\r AT&D3V1Q0E1S0=1&C1&S1M0&J0B1\r\c OK\r \EATDT\T\r\c CONNECT \m\c That explains it. There appears to be a bug in the HPSIO code that makes this CLOCAL fiddling fail. I'm surprised you got it working at all, personally I newer got \M \m to work. I believe NCR inherited this bug from AT&T, since I have also seen it in many other UNIX implementations. I have attached an old Usenet message and a test program that shows the bug. Regards, Robert. From: pim@cti-software.nl (Pim Zandbergen) Date: 21 Aug 90 18:06:53 GMT Newsgroups: comp.unix.i386 Subject: which multiport i/o cards pass this test? Hi netfolk, I am currently looking for a multiport I/O card to replace Specialix, because the Dutch distributor has stopped importing them. I have evaluated some cards but lots of their drivers have trouble with correctly handling modem control. This prevents them from working with the intelligent modem feature in HDB UUCP: "ttyxx,M" in Devices means open with O_NDELAY set and "\M" and "\m" in Dialers means set and clear the CLOCAL flag. The problem is that with these drivers, the O_NDELAY flag cannot be properly cleared when carrier detect is not present. I have noticed this problem with the stock ISC asy driver, with Specialix drivers older than release 3.03, with the Equinox Megaport 1.8.2 driver (a pity, I like this card) and the Chase Research AT8+ 4.02a driver. In fact, the only working drivers I have seen are the Specialix 3.03 and the PORTS and EPORTS drivers for the AT&T 3B2. Rather than evaluate every card on the market myself I would appreciate very much if you would run my little test program and mail me the result. I *think* this program emulates HDB UUCP behaviour. I only tested it on ISC 386/ix 2.0.2 and on the AT&T 3B2. If any UNIX guru would like to comment, please feel free. The crucial moment in this program is the read(). It should block, waiting for input, but returns immediately with zero bytes read on lots of cards. You should run this program on a port that has no cable whatsoever attached to it, nor should any other process have this port opened (ie. no getty's running). The port should implement modem control, ie. "> /dev/ttyxx" should hang. Thanks in advance for you help. #include <fcntl.h> #include <errno.h> #include <signal.h> #include <termio.h> #include <stdio.h> main(argc, argv) int argc; char *argv[]; { int fcntl_flags; int fd; extern int errno; extern unsigned alarm(); char buf[1]; char *usage = "usage: %s <tty device with modem control>\n"; extern char *sys_errlist[]; struct termio termio_flags; struct termio old_termio_flags; void on_alarm(); extern void exit(); if (argc != 2) { (void) fprintf(stderr, usage, argv[0]); exit(1); } (void) printf("opening %s with O_NDELAY set\n", argv[1]); if ((fd = open(argv[1], O_RDWR | O_NDELAY)) < 0) { (void) fprintf(stderr, "%s: can't open %s: %s\n", argv[0], argv[1], sys_errlist[errno]); exit(2); } (void) printf("clearing O_NDELAY flag\n"); if ((fcntl_flags = fcntl(fd, F_GETFL, 0)) == -1) { (void) fprintf(stderr, "%s: can't get fcntl flags: %s\n", argv[0], sys_errlist[errno]); exit(3); } fcntl_flags &= ~(O_NDELAY); if (fcntl(fd, F_SETFL, fcntl_flags) == -1) { (void) fprintf(stderr, "%s: can't set fcntl flags: %s\n", argv[0], sys_errlist[errno]); exit(4); } (void) printf("setting CLOCAL\n"); if (ioctl(fd, TCGETA, &termio_flags) < 0) { (void) fprintf(stderr, "%s: can't get termio flags: %s\n", argv[0], sys_errlist[errno]); exit(5); } old_termio_flags = termio_flags; /* save original modes */ termio_flags.c_cflag |= CLOCAL; if (ioctl(fd, TCSETAW, &termio_flags) < 0) { (void) fprintf(stderr, "%s: can't set termio flags: %s\n", argv[0], sys_errlist[errno]); exit(6); } (void) printf("reading from %s\n", argv[1]); (void) signal(SIGALRM, on_alarm); (void) alarm((unsigned) 10); if ((read(fd, buf, 1) < 1) && (errno == EINTR)) (void) printf("test succesful\n"); else (void) printf("test failed\n"); (void) printf("clearing CLOCAL\n"); if (ioctl(fd, TCSETAW, &old_termio_flags) < 0) { (void) fprintf(stderr, "%s: can't restore termio flags: %s\n", argv[0], sys_errlist[errno]); exit(7); } (void) printf("exiting\n"); return 0; } void on_alarm() { (void) printf("caught SIGALRM\n"); } -- Robert Andersson Voice +47 2 371055 International Systems A/S ra@intsys.no Fax +47 2 356448 P.O. Box 3356 ...!{uunet,mcsun,nuug}!intsys.no!ra 0405 Oslo 4, NORWAY