home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!cs.utexas.edu!sun-barr!olivea!sgigate!odin!news.csd.sgi.com!rock.csd.sgi.com!mitch
- From: mitch@rock.csd.sgi.com (Tom Mitchell)
- Newsgroups: comp.sys.hp
- Subject: Re: HPUX Device Driver
- Message-ID: <1992Sep5.045139.8066@news.csd.sgi.com>
- Date: 5 Sep 92 04:51:39 GMT
- References: <1992Sep01.202444.3973@Schultz.ON.CA>
- Sender: news@news.csd.sgi.com (Net News)
- Distribution: na
- Organization: Silicon Graphics, Inc. Mountain View, CA
- Lines: 36
-
- In <1992Sep01.202444.3973@Schultz.ON.CA> peter@Schultz.ON.CA (Peter Roeser) writes:
-
- >I am writing a kernal level device driver that interfaces to some
- >hardware that does not support interrupts for the HP9000 series 700
- >EISA bus. It is necessary that a watchdog timer on the hardware be
- >kicked at least once every second
-
- >We currently have the timer reset every one-half second, so that
- >seems like a long time for the driver to be scheduled to run.
-
- >4. Any other ideas?
-
- Two -- wild ones.
-
- The one-half second reset is most likely syncronous with the
- one second timer. Can it be that these tend to interact in
- a way that you do not expect. Sort of like counting fence
- posts one foot apart on a 20 foot fence.
-
- Try kicking the timer three or four times in a second. I am
- not sure how the timer your are setting is implemented.
- Your two times per second also sounds like a sampling
- problem. Should interupts be disabled at the wrong time a
- periodic system clock would pass you by. Half a second
- later and the 1 second timer is passed at the same time your
- 1/2 second counter comes by again.
-
- If the clocks had no relation to each other one might guess
- that there are almost always two ticks within a single
- second. If they are not there is only one -- if it is
- possible to miss it -- you will.
-
-
- --
-
- Tom Mitchell -- mitch@sgi.com or mitch@relay.csd.sgi.com
-