joeg@dialogic.UUCP (Joe Grecco) (06/27/89)
I am attempting to use AT&T 3.1/386 to do a real time application and after many hours of successful operation, I have run into a problem. The scenario: We have several expansion cards all using the same interrupt vector. The application transfers data from files to the device via reads/writes. Occasionally during heavy activity (ie lots of interrupts) and >50 processes running, the system PANICS. PANIC: Kernel mode trap. Type 0xE Using crash, I find this to be the kernel stack just before the trap. rcopyfau() s5readi() rdwr() read() systrap() sys_call There is also *sometimes* a NOTICE just before the PANIC: NOTICE: getcpages - waiting for 1 contiguous pages My guess is that I may be running out of memory and getting into a situation where the kernel can't cope. (Some guess :-) Not having access to the kernel source, makes this particularly difficult to figure out. Anyone who can figure out WHAT is happening and hopefully give me a way around it, please let me know. PS: I'm currently trying to reproduce this on Interactive/ix 2.0.1 and AT&T release 3.2