johnl@esegue.segue.boston.ma.us (John R. Levine) (02/17/90)
I have been having minor but persistent problems with my PC Unix system
that appear to be related to missing interrupts, and I'm not sure how to
go about properly diagnosing or fixing it.
Hardware:
Intel 302 (25MHz cached motherboard, 8MB ram)
WD1007 disk controller w/ priam 300mb disk, two floppies
Telebit internal modem, with 8250 replaced by 16550A
Archive internal 150MB tape drive with 402 controller
WD8003 Ethernet
WD VGA controller
Software:
Interactive 386/ix 2.0.2 with X5 async driver update
The first problem is that about half of the time when I'm backing up to
tape, the tape drive will hang. A little poking at the driver shows that it
seems to be hung waiting for the interrupt routine to give back buffers.
Nothing short of reboot will reset it. Other users with similar setups
report no such problem. I've replaced both the drive and the controller
(for free one year minus two days after I bought it when something else went
wrong, Archive was great about it) with no change in symptoms.
The other is that the input throughput on my modem is awful, 400-600 cps for
incoming news. Output is over 1000, like it should be. Other people tell
me that incoming uucp with a 16550A-ized Telebit is very fast.
I have carefully checked for interrupt and DMA collisions. Here's the
interrupt list:
1 Keyboard
2 (cascaded, see below)
3 Telebit com2
4 internal com1 for mouse
5 tape drive
6 floppy
7 printer, not in use
8 internal clock
9 ethernet
10
11
12
13 reserved for 387
14 hard disk
15
The tape is DMA 1, the floppy is DMA 2. I don't use the floppy when I'm
running the tape. Switching to DMA 3 didn't make any difference.
Both of these problems would be explained if it were sometimes losing
interrupts. It's not clear to me how I diagnose this problem, much less fix
it. There's no problem with the disk, and I believe that if I were losing
disk interrupts all sorts of awful things would happen. I suppose a
motherboard transplant is a possibility, but that's expensive and would put
me out of commission for a week or two while motherboards flew across the
country.
Any suggestions?
--
John R. Levine, Segue Software, POB 349, Cambridge MA 02238, +1 617 864 9650
johnl@esegue.segue.boston.ma.us, {ima|lotus|spdcc}!esegue!johnl
"Now, we are all jelly doughnuts."keithe@tekgvs.LABS.TEK.COM (Keith Ericson) (02/21/90)
In article <1990Feb16.170751.4059@esegue.segue.boston.ma.us> johnl@esegue.segue.boston.ma.us (John R. Levine) writes: >I have been having minor but persistent problems with my PC Unix system >that appear to be related to missing interrupts, and I'm not sure how to >go about properly diagnosing or fixing it. NOTE: The following may well NOT be the cause of the problem. I'm only suggesting it to give you "something else" to try... Some of the AMD (I thing it was) second-source versions of the Intel interrupt controller (the 8259? I can't remember) ended up with a too-high valued internal pull-up resistor when the chip went through die-shrink. But the symptom in that case seemed to be too many (i.e., extraneous) interrupts and the machine would appear to be running very slowly, especially if it were runnin UNIX. The cure was to tie some appx. 1k-ohm resistors from the interrupt-controller's IRQ-out pins to +5v. My guess that the bad parts would've been weeded out by the time Intel was building 302 motherboards: we had the problem with some 301's. kEITHe