mike@erix.UUCP (Mike Williams) (03/14/84)
Normally when UNIX panics we get a dump which we sometimes look at to
see what happened.
The other day our VAX just died. It continued to run, echoed text from
terminals etc. but just hung if you tried to give it any commands. It
turned out that one of our disk controllers was playing up. We swap and
page on two disks (hp0 and hp1) and hp1 just gave up. This was quickly
repaired and we were up again after an hour.
Is there any way to force a dump in these conditions? Why did the VAX
just play dumb?
Mike Williams
{decvax,philabs}!mcvax!enea!erix!mike
or
mike@erix.UUCPwls@astrovax.UUCP (William L. Sebok) (03/16/84)
> Is there any way to force a dump in these conditions? ... > > Mike Williams I would like to see "take a coredump" as one of the options of the reboot system call. In fact I think I'll hack that in. There have been times when the system was very sick that the only cure was rebooting yet one wanted a core dump to study. Such a time was a year ago when I was searching for why a process running versatec would hang in the pagein() routine. I solved it at the time by putting a "doomsday" ioctl call (suser only, of course) into the dz driver which I happened to be playing with at the time anyway. It was rather fun to type "crash" and actually have it happen. -- Bill Sebok Princeton University, Astrophysics {allegra,akgua,burl,cbosgd,decvax,ihnp4,kpno,princeton,vax135}!astrovax!wls
dave@uwvax.ARPA (03/16/84)
> Is there any way to force a dump in these conditions? ... > > Mike Williams On an 11/780, you can force a dump by typing (in console mode): S C00 The equivalent works on the 750 and 730 flavors also. Dave Cohrs ...!{seismo,allegra,ihnp4}!uwvax!dave dave@wisc-rsch.arpa
ed@unisoft.UUCP (03/18/84)
A dump can be forced on a VAX-11/750 running 4.1 or 4.2 by putting the "power on action" switch in the "restart" position and pushing the reset button. I don't know how to force a dump on a 730 or 780. -- Ed Gould ucbvax!mtxinu!ed
bloom%ucbshadow%berkeley@sri-unix.UUCP (03/19/84)
From: Jim Bloom <bloom%ucbshadow@berkeley> Starting the cpu at c00 to get a core dump usually works. It does not work if you have more than one mba or uba (I don't remember if it is one or the other or both). The address is different in that case. I would recommend looking at the namelist for /vmunix and noting the address of doadump. This is the correct address at which to start for your system. Jim Bloom bloom@berkeley.ARPA ucbvax!bloom
arwhite@watmath.UUCP (Alex White) (03/19/84)
Our 780 all had on the console floppy which came from dec a file called CRASH which is for crashing vms -- it prints several registers, then stuffs negative one into the pc, kernel at ipl 31 into the psl, and continues. This immediately causes a trap, which unless things are screwed up in the various disks should cause your dump. To use, on the console ^P HALT @CRASH
joe@fluke.UUCP (Joe Kelsey) (03/19/84)
There is a command file on every VAX-11/780 standard distribution floppy (and standard distribution cassette for 730s, 750s will have to punt here!) called CRASH.CMD. This routine simply displays the PC, PSL and all stack pointers, replaces the PC with -1, the PSL with 1F0000 (kernel mode, IPL 31) and CONTINUEs. Bang! down comes the system with an appropriate dump file. On a 750, if you have 4.2, you could include CRASH.CMD in your console cassette and kludge something or you could type up a short list of commands which (almost) anyone could follow to do the same thing. Try it! /Joe
dmmartindale@watcgl.UUCP (Dave Martindale) (03/20/84)
If you system just hung, it was probably waiting for the disk controller to interrupt it indicating that the last-issued transfer or seek was complete. The HP disk drivers do no timeouts to detect lost controllers. A drive that goes dead should cause an interrupt, so it's only in the case of a dead controller that you should see this sort of hang.
dmmartindale@watcgl.UUCP (Dave Martindale) (03/20/84)
If you have a 780 and want a dump, try typing "^P" and then "@crash" on the console. It deposits -1 into the PC, which takes the machine down rather quickly.
swine@orac.UUCP (Peter Swain) (04/10/84)
> Is there any way to force a dump in these conditions? ... > > Mike Williams > There have been times when the system was very sick that the only cure was > rebooting yet one wanted a core dump to study. > > Bill Sebok If the system is alive enough to panic, but ignoring the world, give it something to panic about ... on a vax (for example): ^P ! get it's attention H ! halt a 780 (ignored on smaller vaxen) E/G F ! remember PC for later analysis D/G F C0000000 ! the beginning of the illegal gigabyte ... ! -1 is easier to type, but 7[35]0's don't like `-'. C ! continue at silly addr extreme embarrasment produces a nice dump -- buffers are sync'd & whole state preserved (except for the clobbered pc, which was prob the idle loop anyway). Peter Swain