rws@cs.brown.edu (Richard W. Sabourin) (09/16/90)
I have recently installed Sendmail 5.64+ w/ IDA on our central mail machine, a Sun 4/360 running 4.1. (I don't have the IDA revision # on hand, but it was just last month ftp'ed from uunet.) It compiled fine, and uses MX records fine. The .cf file is rather a mishmash, but I debugged it and it does what we want. Anyway, this machine has recently begun panicking a couple times a week, always at night (figures), with the following: -------------------------------------------------- panic: Data fault BAD TRAP pid 2680, `sendmail': Data fault kernel read fault at addr=0xf8238ff8, pme=0x0 Bus Error Reg 80<INVALID> pc=0xf80a4074, sp=0xf8178558, psr=0x118018c3, context=0x0 g1-g7: 11001ae6, e30002e7, ffffffff, 6, f824b000, f8121000, f8121000 Begin traceback... [etc.] -------------------------------------------------- Sendmail is always listed as the culprit, and always a Bad Trap. After these panics, mail delivery enters a weird failure mode, in which mail to all local users bounces with "User unknown". Re-freezing the fc file and restarting fixes this. Anybody else seen either failure? Coincidence? Conspiracy? Don't worry, I'm not one of those "Pls fix bug asap urg help thx" posters; just drawing a blank. Thanks in advance, Rick Sabourin Systems Slave
mills@ccu.umanitoba.ca (Gary Mills) (09/16/90)
rws@cs.brown.edu (Richard W. Sabourin) writes: >I have recently installed Sendmail 5.64+ w/ IDA on our central mail machine, >a Sun 4/360 running 4.1. [...] >Anyway, this machine has recently begun panicking a couple times a week, [...] >panic: Data fault [...] >Sendmail is always listed as the culprit, and always a Bad Trap. I believe it's a kernel bug. We had a very similar problem, but with a user directory server witten in perl. I don't know what provokes it. We stopped running the server. Now, two or three times a week, the machine goes to sleep, and has to be re-booted. A ps listing from the dump always shows a particular process getting virtually all the cpu. -- -Gary Mills- -University of Manitoba- -Winnipeg-
mills@ccu.umanitoba.ca (Gary Mills) (09/16/90)
This _is_ a kernel bug. The bug ID is 1029939, and Sun has a patch for it. -- -Gary Mills- -University of Manitoba- -Winnipeg-
rws@cs.brown.edu (Richard W. Sabourin) (09/19/90)
In article <49991@brunix.UUCP> I wrote: |> |>I have recently installed Sendmail 5.64+ w/ IDA on our central mail machine, |>a Sun 4/360 running 4.1. (I don't have the IDA revision # on hand, but it |>was just last month ftp'ed from uunet.) It compiled fine, and uses MX |>records fine. The .cf file is rather a mishmash, but I debugged it and |>it does what we want. |> |>Anyway, this machine has recently begun panicking a couple times a week, |>always at night (figures), with the following: |> |>-------------------------------------------------- |>panic: Data fault |>BAD TRAP |>pid 2680, `sendmail': Data fault |>kernel read fault at addr=0xf8238ff8, pme=0x0 |>Bus Error Reg 80<INVALID> |>[etc] |>-------------------------------------------------- |>[...] Here is a summary of the more useful email responses I've received. (IDA revision is 1.3.4.) From "Mark D. Baushke" <mdb@ESD.3Com.COM>: >Did you remember to build with the -Bstatic switch turned on? I have >only had trouble with 5.64+IDA when using dynamic libraries. Even then >it never caused a panic, it just did not deliver the mail correctly. ... I didn't do that. I'll probably do that next. From rickert@cs.niu.edu: > Never overlook the possibility of a hardware failure. If it happens at >night, 'sendmail' may be the main program running, so the fact that it >always takes the hit might not mean very much. From Paul Pomes - UofIllinois CSO <paul@uxc.cso.uiuc.edu>: >Does the error occur when the sendmail.fc file is removed? Piet Berteema >has found a bug with setdefuser() munging the malloc pointers when the >frozen config file is used. ... I have removed the .fc file, and the mail problems have not recurred in 3 days. So I hope for the best. Lastly, mills@ccu.umanitoba.ca (Gary Mills) posted: >This _is_ a kernel bug. The bug ID is 1029939, and Sun has a patch for it. Thank yous to all who replied. Now I turn my head to Sun's /bin/mail which has also been acting odd, but that's for another group... Rick Sabourin <rws@cs.brown.edu>