Re: [PATCH 1/1] perf annotate: Fix fallback to unparsed disassembler line

[Arnaldo Carvalho de Melo <acme@kernel.org>]

Em Mon, Mar 09, 2015 at 12:09:58PM +0100, Ingo Molnar escreveu:
> 
> * Arnaldo Carvalho de Melo <acme@kernel.org> wrote:
> 
> > [...]
> > 
> > Further investigation is needed to figure out the nature of the 
> > objdump output change so as to make the parser grok it.
> 
> Btw., maybe someone finds this interesting: we could also (re-)use the 
> in-kernel disassembler (on x86 and any other architecture that might 
> have one), which used by live patching facilities (kprobes et al).
> 
> See arch/x86/lib/insn.c.
> 
> The 'visualization' bit is missing entirely: but it does a lot of the 
> hard work of knowing about the instruction format: it knows about 
> essentially all x86 instructions and is able to determine instruction 
> boundaries, and can decode immediate constants.
> 
> Using this in tools/perf/ would have the added advantage that we could 
> then use the dissasembly in kernel oops output (nice feature!) - plus 
> tooling folks would help us fix and extend the kernel's disassembler! 
> ;-)

I think it should provide a good synergy, yes, IIRC this was discussed
already at some point even.

- Arnaldo