hpg@slu70.UUCP (hpg) (09/22/84)
I recently installed 2.9 on a new 11/73 system. Relevant parts: 73 processor, 1M PEBX ECC memory, DILOG 228 disk controller with Fujitsu Eagle, DILOG 132 tape coupler with Cipher 891, DLV11J SLU all in a BA11SA box. After careful mods to the boot and kernel 2.9 runs like a champ ... except one in every three or four user processes dies just after birth with bus error, segmentation fault, IOT trap, or illegal instruction and the requisite core dump (bus error and seg fault most common). The frequency is load dependent. I've substituted the processor and memory; no fix. Others around me are not having the problem; they have Emulex controllers and Fujitsu 160's, though. The kernel has NEVER crashed in one month of operation. Has anyone seen this problem before? All suggestions welcome. My forehead is becoming badly calloused. Please respond to ...!ihnp4!wuibc!hpg who is: Herschel Goldstein, Washington University, Campus Box 8115, St. Louis, MO 63110. (314) 362-1219
gtaylor@lasspvax.UUCP (10/12/84)
In article <> ksp@lasspvax.UUCP (Keith S. Pickens) writes: > I recently installed 2.9 on a new 11/73 system. Relevant parts: > 73 processor, 1M PEBX ECC memory, DILOG 228 disk controller with Fujitsu Eagle, > DILOG 132 tape coupler with Cipher 891, DLV11J SLU all in a BA11SA box. > After careful mods to the boot and kernel 2.9 runs like a champ ... except > one in every three or four user processes dies just after birth with > bus error, segmentation fault, IOT trap, or illegal instruction and > the requisite core dump (bus error and seg fault most common). > The frequency is load dependent. I've substituted the > processor and memory; no fix. Others around me are > not having the problem; they have Emulex controllers and Fujitsu 160's, > though. The kernel has NEVER crashed in one month of operation. I have run 2.9.1 and 2.9-beta on a 11/73 system. The system contains: 11/73 processor, 1Mb PEBX ECC memory, DILOG 228 disk controller with Fujitsu 160, DILOG 132 tape coupler with Cipher 891, Emulex CS01 (dh11) in a MDB box. I have seen no problems of any type with this system. The system runs with the DILOG 132 in TSV05 mode. The only work required on 2.9.1 was for the 22-bit address space (no UNIBUS map) and for the TSV05 (it is almost a ts11). So the question seems to be: Can you run with Fujitsu Eagles on a Q-bus system? If you are doing so would you please mail me information on your system configuration, and any problems you may have encountered. The Eagle is a very nice drive but can a Q-bus based system handle the high transfer rate? I will summerize the replies I get for the net. -keith cornell!lasspvax!ksp lasspvax.ksp@cornell.arpa (607) 256-5560 Keith S. Pickens LASSP Clark Hall Cornell University Ithaca, New York 14853
brad@bradley.UUCP (10/16/84)
Here is a message that I recieved on that very same subject. The exchange program through DEC is going on until Dec, or so I understand. You really have to get your DEC salesman to dig to help you, most don't know or even heard about it, according to DEC the letters went out in July, so if you haven't heard better given them a call. Brad Smith Text Processing/Bradley Unversity -forward message- There is a bug in some (9%) early J-11 chips, check the date code on your CPU chip if your chip as made before May 1 1984, you may have the problem. The date code is on the CPU end of the J-11 chip and is of the form YYWW where YY is the year of Mfgr. and WW is the week, codes earlier than 8419 may have the fault. The problem is very hard to reproduce as it involves integer add/sub with certain values while a Cache miss or DMA Xfer is in progress. If your CPU fits the date code, get in touch with DEC. I'm in the process of getting a copy of the actual letter posted to a 11/73 owning friend of mine, I hope to get it on monday. Whenever I get it, I'll post a copy to the net as a whole. Keep in touch with me on this as there are at least 3 11/73s on the net now, you 2 and cepu!ucla-an (just getting going with 2.9, and a whole bunch of diskdrives).