On Sat, 2011-11-19 at 00:35 +0100, Marius Strobl wrote: > On Fri, Nov 18, 2011 at 09:34:11PM +0000, Ben Hutchings wrote: > > sfxge: Remove interrupt self-test code > > > > It's not currently used; it didn't build on 32-bit and the previous > > build fix is incorrect. If we really implement self-tests we can do > > this again properly. > > Yes, I've also already noticed that this part of r227640 wasn't quite > correct. However Philip suggested to just leave it in for now until > we figure out what on earth the code actually is supposed to do and as > the atomic_cmpset_ptr(9) also works on LP64 and isn't more broken than > the atomic_cmpset_long(9) that was in there before (actually this should > have been atomic_cmpset_64(9) for an uint64_t, which isn't necessarily > available on ILP32 including i386 though). Probably this should have > been converted to be of type cpuset_t and to use the accessors from > <sys/cpuset.h> as nowadays we also support more than 64 CPUs. I'm also > fine with just nuking the interrupt self-test altogether though.
The hardware RX flow hash indirection table has 6-bit entries so it's not possible to use more than 64 RX queues without some kind of flow steering. So for the time being this driver sets a limit of 64 interrupts. Ben. -- Ben Hutchings, Staff Engineer, Solarflare Not speaking for my employer; that's the marketing department's job. They asked us to note that Solarflare product names are trademarked. _______________________________________________ [email protected] mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-net To unsubscribe, send any mail to "[email protected]"
