On Wed, Mar 29, 2000 at 09:20:06PM +0200, Manuel Bouyer wrote:
> On Tue, Mar 28, 2000 at 12:16:41PM -0500, Thor Lancelot Simon wrote:
> > Interestingly, I have a system here with a parallel printer attached that
> > has 64MB of buffer memory.  I have seen over *40,000* IRQs/sec from the
> > lpt device on this system, while the system feels completely usable.  It
> > can't just be the number of IRQs.
> 
> My home machine, cyrix 133 with a promise Ultra/33 IDE adapter can handle
> 4000 IDE IRQ/s without troubles. I'll have to run my bench on my system at
> work tomorow.
> 
> > This might explain why some SCSI controllers avoid this problem: when you
> > get an interrupt, it's quite likely you may find out that multiple commands
> > have completed.
> 
> Yes ...
> 
> > 
> > > > With SCSI disks, they don't
> > > > seem to appear in the first place.  I'd suspect some kind of odd barrier
> > > > condition with !B_ASYNC buffers, but since we don't do disconnection or
> > > > multiple command queueing on IDE that doesn't seem likely, either.
> > > 
> > > What SCSI controller do you use ?
> > 
> > A variety of them: ahc, bha, and adw.  I haven't seen the problem we're
> > discussing with any of them.  I run 'ahc' with tagged queueing turned on,
> > BTW.
> 
> Interesting, both systems where I notice the problem are K6/II running on a
> P5A motherboard. One of them is IDE (Ultra/33) the other SCSI (aha2940UW).

I've seen it with Pentium IIs and Celerons on Intel motherboards (mostly
440BX or LX) with various UDMA/33 IDE drives on the built-in IDE controllers.

I have *also* seen it -- and it seems particularly severe -- with UDMA/66
drives (which are used as UDMA/33 of course) on an Asus K7M motherboard
which uses the AMD chipset with the VIA IDE controller and an additional
Promise UDMA/66 controller.  In fact, that system is almost useless if I do
heavy I/O to a FFS on any of its drives... with LFS it's much better.

The very first system I noticed this on was an Intel server motherboard
(440LX, I *think*) with a single Seagate UDMA/33 disk.  As I upgraded this
system from 1.3F (was it 1.3F where IDE DMA first showed up?  Around there,
anyway) or thereabouts towards 1.4, I noticed that with one particular
upgrade this problem appeared.  The only other devices on the machine were
com (unused) lpt (unused), vga, and fxp (not heavily used).