narten@purdue.arpa (Thomas Narten) (10/02/86)
We have uVAX II's running 4.3 BSD. Whenever the system crashes, a crash dump is generated which apparently is successful (judging from the message on the console). When /etc/savecore runs to generate the vmcore file, it aborts after reading about 1.8 of the 2meg crash image. The message it gives is simply "read I/O error". Can anyone give pointers as to what the problem might be? Thanks, Thomas Narten narten@purdue.edu ----------
gershon@ccicpg.UUCP ( Gershon Shamay) (10/15/86)
We had similar problems with savecore on a different machine. It turned out to be the machine-dependency of 'savecore' itself. Somewhere in it there was a simple-minded macro which tried to map Unix virtual addresses as seen in memory (they begin with 0x80..... on a VAX 780/etc) to displacements in the swap file (where the core immage is kept). Knowing that there should be a 1-1 mapping, the macro mapped 0x80000000 to 0, 0x80000400 to 400 and so on. It did so by just dropping the MSB of the virtual address it needed. Now on our machine that was wrong, the kernel starts at 0xC0000000, not 0x80000000 so the macro was generating huge displacements in the core file, eventually failing with a read I/O error (the disk driver refused to read there). I don't know about kernel virtual addresses on your machine, but that's a pointer - try to look for this macro in savecore. Gershon Shamay CCI Computer Products Group