parker@zaphod.Berkeley.EDU (Ross Parker) (03/06/90)
System: Microvax-II with Emulex QD-32 disk controller and two Fujitsu Eagle disk drives. 13 Mb memory. Running Ultrix 3.0. System supports perhaps 15-20 interactive logins, and perhaps 20 PCs connected via Sun's PC-NFS. The PCs access files using standard NFS on the Microvax. Symptom: One user on a PC can try to bring up a particular file under WordPerfect (version 5.0 or 5.1) on the PC, and, without fail, cause the Microvax to instantly crash. This problem just started happening. No system changes, either hardware or software, have taken place for a number of months. The user does not have a problem with any other files, nor does any other user cause the system to die. The system is also used for NFS operations from other Vaxen, and from some Sun systems, and no problems occur. The crash symptoms are (on the console): Trap Type 9, code = 803771ff, pc = 80034ca0 panic: Protection fault and in the error log: ********************************* ENTRY 29. ********************************* ----- EVENT INFORMATION ----- EVENT CLASS ERROR EVENT OS EVENT TYPE 104. CONTROLLER ERROR SEQUENCE NUMBER 0. OPERATING SYSTEM ULTRIX 32 OCCURRED/LOGGED ON Mon Mar 5 13:44:28 1990 PST OCCURRED ON SYSTEM waters SYSTEM ID x08000000 SYSTYPE REG. x01010000 FIRMWARE REV = 1. PROCESSOR TYPE KA630 ----- UNIT INFORMATION ----- UNIT CLASS ADAPTER/CONTROLLER UNIT TYPE UDA50A CONTROLLER NO. UNIT NO. 0. ERROR SYNDROME CONTROLLER ERROR ********************************* ENTRY 30. ********************************* ----- EVENT INFORMATION ----- EVENT CLASS ERROR EVENT OS EVENT TYPE 200. PANIC SEQUENCE NUMBER 5. OPERATING SYSTEM ULTRIX 32 OCCURRED/LOGGED ON Mon Mar 5 13:42:26 1990 PST OCCURRED ON SYSTEM waters SYSTEM ID x08000000 SYSTYPE REG. x01010000 FIRMWARE REV = 1. PROCESSOR TYPE KA630 PANIC MESSAGE Protection fault ********************************* ENTRY 31. ********************************* ----- EVENT INFORMATION ----- EVENT CLASS ERROR EVENT OS EVENT TYPE 109. EXCEPTION/FAULT SEQUENCE NUMBER 4. OPERATING SYSTEM ULTRIX 32 OCCURRED/LOGGED ON Mon Mar 5 13:42:26 1990 PST OCCURRED ON SYSTEM waters SYSTEM ID x08000000 SYSTYPE REG. x01010000 FIRMWARE REV = 1. PROCESSOR TYPE KA630 ----- UNIT INFORMATION ----- ERROR SYNDROME PROTECTION FAULT ********************************* ENTRY 32. ********************************* ----- EVENT INFORMATION ----- EVENT CLASS OPERATIONAL EVENT OS EVENT TYPE 250. ASCII MSG SEQUENCE NUMBER 7. OPERATING SYSTEM ULTRIX 32 OCCURRED/LOGGED ON Mon Mar 5 13:42:40 1990 PST OCCURRED ON SYSTEM waters SYSTEM ID x08000000 SYSTYPE REG. x01010000 FIRMWARE REV = 1. PROCESSOR TYPE KA630 MESSAGE done ********************************* ENTRY 33. ********************************* ----- EVENT INFORMATION ----- EVENT CLASS OPERATIONAL EVENT OS EVENT TYPE 250. ASCII MSG SEQUENCE NUMBER 6. OPERATING SYSTEM ULTRIX 32 OCCURRED/LOGGED ON Mon Mar 5 13:42:39 1990 PST OCCURRED ON SYSTEM waters SYSTEM ID x08000000 SYSTYPE REG. x01010000 FIRMWARE REV = 1. PROCESSOR TYPE KA630 MESSAGE syncing disks... Now this certainly looks like a probable bad controller, right? Well, we've replaced the controller with a new one, and get an identical problem... down to identical register values in the register dump. We've also run DEC diagnostics and the system passes with no problems, other than (and this I'm mildly worried about) the disk controller... however, I believe the controller is failing because it's a non-DEC controller, and DEC's diags are expecting a KDA50. The controller (the new one) passed the vendor's diags, and the diags included scanning the disks. No problems were found anywhere. In addition, we can read and write any file on both disks locally (not via NFS) and no problems occur, so the problem is possibly related to NFS rather than to disk driver code or whatever. Perusing the resultant crash dumps has not given me much enlightenment, however, I'm not an expert at that, and have misplaced my list of magic incantations to have adb show anything useful. Perhaps someone can enlighten me? Care to bite, George? I'm sure you've done this numerous times. If anyone can shed any light on this, it'd be *much* appreciated. This Monday, the system went down about 7 times before this particular user called us to say that each crash happened exactly when she tried to access this file! Thanks, Ross Parker parker@mpre.mpr.ca (604)293-5495 uunet!ubc-cs!mpre!parker
grr@cbmvax.commodore.com (George Robbins) (03/07/90)
In article <2080@kiwi.mpr.ca> parker@zaphod.Berkeley.EDU (Ross Parker) writes: > > Symptom: One user on a PC can try to bring up a particular > file under WordPerfect (version 5.0 or 5.1) on the > PC, and, without fail, cause the Microvax to > instantly crash. This problem just started happening. > No system changes, either hardware or software, have > taken place for a number of months. The user does > not have a problem with any other files, nor does > any other user cause the system to die. The system > is also used for NFS operations from other Vaxen, and > from some Sun systems, and no problems occur. The crash > symptoms are (on the console): Well, from your description there isn't enough info to make a good guess, but I'd suggest two avenues to pursue... 1) look a the crash dump information on the console and a nm -n /vmunix and see what kind of code the pc is in, nfs, ufs or something else. 2) fsck the disk and make sure there are no organizational problems associated with the file. 3) try to read the file/files in question with tar or other utility to make sure that the problem is (perhaps) NFS and not basic filesystem stuff. 4) try to correlate the errors with the UERF output. Is there exactly one of the "controller errors" per crash? Do all the crashes have the same PC value? 5) try the uerf -o full option to see if you get more information on the "controller error" that might give a clue... 6) talk to the support center about known 3.0 NFS problems (I think there are quite few) and either apply the 3.0 NFS / networking patch tape or upgrade to 3.1 and do any appropriate patches there... -- George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr but no way officially representing: domain: grr@cbmvax.commodore.com Commodore, Engineering Department phone: 215-431-9349 (only by moonlite)
parker@zaphod.Berkeley.EDU (Ross Parker) (03/08/90)
In article <10020@cbmvax.commodore.com>, grr@cbmvax.commodore.com (George Robbins) writes: > In article <2080@kiwi.mpr.ca> parker@zaphod.Berkeley.EDU (Ross Parker) writes: > > ... stuff about one of my systems crashing ... > > George suggests a number of things to try to track down the problem... Actually, George, I had tried all that you had suggested, and also finally remembered some of my long-lost adb knowledge to get a stack trace from one of the core dumps. The problem is definitely in the nfs code. My original panicked message was written in the wee hours of the morning with my brain on autopilot, so it wasn't very clear! > > 6) talk to the support center about known 3.0 NFS problems (I think there > are quite few) and either apply the 3.0 NFS / networking patch tape > or upgrade to 3.1 and do any appropriate patches there... > This I did the next day, and (much to my surprise) they seemed to think that the problem was a known one... I've been playing phone tag since then, but I believe an upgrade to 3.1 plus some patches will cure my woes (exactly what George suggests). > George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr > but no way officially representing: domain: grr@cbmvax.commodore.com > Commodore, Engineering Department phone: 215-431-9349 (only by moonlite) Thanks for the relpy, Ross Parker parker@mpre.mpr.ca (604)293-5495 uunet!ubc-cs!mpre!parker
cmaarpc@cc.ic.ac.uk (Peter Churchyard) (03/08/90)
We have a microvax with Fujitsu drives and QD32 controller. When I first tried to install 3.0, the kernel would panic whenever I wrote to the Eagle. (chpt mount fsck etc) I checked with emulex and found that we were running OLD firmware in the QD32. (revision C) we have now moved to revision H and I then had to RE-FORMAT the Eagle. This installs some diag stuff. Put up 3.0 etc and now works great! So Check those revs. ------------------------------------------------------------------------- If you are reading this and you are not a DECUS member, WHY NOT?. Peter Churchyard. DECUS Europe UNiX SIG Chair. Real life: Group Leader Communications Section, Imperial College Comp Centre.