Luck, Tony píše v Po 17. 11. 2008 v 12:58 -0800: > > Ignorance being bliss? > > If the module is only hitting unaligned addresses that the processor > can handle ... then ignorance might well be bliss. The cost of > an un-reported misaligned access is anly a few cycles. The cost > of the trap and trip to arch/ia64/kernel/unaligned.c is IIRC > somewhere in the 800-900 cycle range ... more for the printk. > > If we had less of a sledgehammer to report the problem, then I'd > find it a lot easier to say that this should be enabled in > production systems. > > > If this module was developed in an environment where the unaligned > > accesses could not be easily seen how does hiding the messages in > > production get them addressed? Modules being developed in such > > environments would seem to suggest that the messages _should_ be seen in > > production? > > Definitely a downside to my approach ... if the end users > don't see that they have been sold a crummy driver they > won't know that they should complain. Well, then let's make it a boot option (or a sysctl). The default must be psr.ac = 1, but if the admin sees "unaligned access" messages and knows that there is no way of getting a (timely) fix for that particular production system, s/he should be able to turn them off. IMO this approach solves both issues, because: 1. Everebody is aware of the performance hit 2. The cost of it can be minimized Just my two cents, Petr Tesarik -- To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
|