Messages in this thread | | | Date | Fri, 29 Oct 2010 13:48:31 -0700 | From | David Daney <> | Subject | Re: [PATCH 4/8] jump label: Make arch_jump_label_text_poke_early() optional |
| |
On 10/29/2010 01:37 PM, David Miller wrote: > From: Steven Rostedt<rostedt@goodmis.org> > Date: Thu, 28 Oct 2010 09:55:52 -0400 > >> From: David Daney<ddaney@caviumnetworks.com> >> >> For the forthcoming MIPS jump label support, >> arch_jump_label_text_poke_early() is unneeded as the MIPS NOP >> instruction is already optimal. >> >> Supply a default implementation that does nothing. Flag x86 and SPARC >> as having arch_jump_label_text_poke_early(). >> >> Cc: Jason Baron<jbaron@redhat.com> >> Cc: David Miller<davem@davemloft.net> >> Signed-off-by: David Daney<ddaney@caviumnetworks.com> >> LKML-Reference:<1286218615-24011-2-git-send-email-ddaney@caviumnetworks.com> >> Signed-off-by: Steven Rostedt<rostedt@goodmis.org> > > On the SPARC side, it's not that the nop isn't optimal, on sparc it's > always 0x01000000.
That's what I thought. You could remove your arch_jump_label_text_poke_early() function to no determent.
> The issue is that when written an I-cache flush is > necessary using a 'flushi' instruction. > > Does MIPS not need a flush when poking instructions?
Yes, it does.
> I find this hard to believe, although it's been some time since I > last touched that architecture :-)
I just looked at my patch again. I am indeed flushing the I cache after patching the code, so I don't really know what you are talking about.
David Daney
| |