On Thu, Jun 19, 2003 at 04:51:35PM -0300, Werner Almesberger wrote:
> 4) Losses:
>
> Actually, I'm not so sure what really ought to happen with
> losses. If we have serialization requirements elsewhere,
> proceeding after unrecovered losses would probably mean that
> they're being violated. So if they can be violated then,
> maybe there is some leeway in other circumstances too ?
>
> On the other hand, if any loss means that major surgery is
> needed, the interface should probably have a "in loss" state,
> in which it just shuts down until someone cleans up the mess.
> Also a partial shutdown may be interesting (e.g. implemented
> by the dispatcher), where events with no interdependencies
> with other events would still be processed.
One thing came to my mind:
If you have a sysfs-scanning method for startup, couldn't you
just make the sequence-number-checking daemon reset its state
and redo the sysfs scan on loss of events? (Or even make it
just exec itself and use the exact same code as at startup.)
That way the system recovers from event loss (or a reordering
that gets the earlier event too late and is believed to be a
loss) in a way that needs to work anyway, and isn't a magic
special case.
--
:(){ :|:&};:
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Mon Jun 30 2003 - 22:00:23 EST