Re: [PATCH] Mutilated form of Andi Kleen's AMD prefetch errata patch

[Jamie Lokier]

Dave Jones wrote:
> Which gets us back to the question of why this is needed at all ?
> You said earlier "In case you hadn't fully grokked it, my code doesn't
> disable the workaround!"  So why do you need this ?

To change the prefetch workaround from a critical requirement to an
optimisation knob.

X86_USE_3DNOW is very similar: if it's enabled, the kernel has some
extra code to make certain CPUs run faster, but they also run fine
without it.  X86_OOSTORE is another.

What I'd really like your opinion on is the appropriate userspace
behaviour.  If we don't care about fixing up userspace, then
__ex_table is a much tidier workaround for the prefetch bug.  If we do
care about fixing up userspace, then do we need a policy decision that
says it's not acceptable to run on AMD without userspace fixups from
2.6.0 onwards - it must fixup userspace or refuse to run?

-- Jamie
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/