accelerometer driver

[Joerg Reisenweber]

Am Di  26. August 2008 schrieb Andy Green:
> Somebody in the thread at some point said:
> | It seems like the accelerometers have the following:
> |
> | sensitive to losing an edge trigger
> 
> We just suspect it, nobody showed it yet... it would probably be a bug
> in the driver handling or s3c stuff rather than accelerometer "feature".
> 
> | occur at a periodic rate
> | access conflicts with other drivers so should be done outside of irq
> 
> I have a nice patch coming that just does it bitbang the whole way using
> IRQs off as locking throughout.
> 
> | Does it make sense to just turn off the irq and have the driver use a
> | delayed work queue to poll at a given rate?
> 
> Interrupts are at least somewhat synchronized to the source, I think it
> will suffer from dropped samples otherwise.  But, at some point we have
> to take more radical measures if we don't solve it...

Like general switch from edge-triggered IRQ to level-triggered.
Edge-triggered IRQs are *NOT* suitable for physical sharing of IRQ-line, ok we 
don't do this. Edge is more sensible to spikes and glitches, ok we don't see 
(really?). But I expect IRQs to be locked/disabled during execution of a 
IRQ-handler (usually that's even implicit in CPU's handling of IRQs), so only 
level-IRQ will be noticed after we exit handler.

I consider usage of edge-trigger a major flaw in Openmoko-design at large. 
Edge is used only where you can't reset the interrupt-source - like 
pushbuttons etc. I.E. async. For "sync" interrupts level is method of choice.

/jOERG
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part.
Url : http://lists.openmoko.org/pipermail/openmoko-kernel/attachments/20080829/77989abb/attachment-0001.pgp