Re: [PATCH][RFC] per isr in_progress markers

[Rogier Wolff <R.E.Wolff@BitWizard.nl>]

On Mon, Sep 09, 2002 at 11:53:40AM -0700, Linus Torvalds wrote:
> 
> (Btw, if there is, that would also allow us to notice the "constantly
> screaming PCI interrupt" without help from the low-level isrs)

OH! That'd be nice: instead of a lockup if a PCI device's interrupt
isn't serviced, you get a nice message and a machine that might
still work!

On the other hand, you have a possibility for disaster if the 
threshold isn't set right. 

I have written serial drivers where the card will limit the interrupt
rate to max 100 per second. I then build in a detection: if my IRQ 
handler gets called more than 10 times in a jiffy, we're in trouble.

Turns out that I left this in "in the field" and some people put the
serial card on the same interrupt line as a SCSI controller. The
scsi controller can generate more than 1000 interrupts per second ->
my driver shuts down.... 

Something similar may happen if say you net-spray a sligtly 
under-powered machine with a Gigabit ethernet card: The GBE card may
indeed have a new packet ready by the time the interrupt tries to 
return. Leads to an interesting DOS: just send a bunch of packets
in quick succession and the machine drops off the internet... 

				Roger. 

-- 
** R.E.Wolff@BitWizard.nl ** http://www.BitWizard.nl/ ** +31-15-2137555 **
*-- BitWizard writes Linux device drivers for any device you may have! --*
* The Worlds Ecosystem is a stable system. Stable systems may experience *
* excursions from the stable situation. We are currenly in such an       * 
* excursion: The stable situation does not include humans. ***************