On Tue, 10 Jan 2017 15:42:07 -0600 Josh Poimboeuf <jpoim...@redhat.com> wrote:
> On Tue, Jan 10, 2017 at 09:59:23AM +0100, Peter Zijlstra wrote: > > On Mon, Jan 09, 2017 at 11:36:48AM -0600, Josh Poimboeuf wrote: > > > BTW, I think we'll have more problems with generated code if/when we > > > move to an x86 DWARF unwinder, because it won't have any idea how to > > > unwind past generated code. Long term I wonder if it would make sense > > > to create some kind of framework for creating or registering generated > > > code, so we can solve these types of problems in a single place. > > > > Yes, this seems like a good idea. Maybe we could pull the rbtree thing > > from modules and make that a more generic interface for code > > registration. > > Yeah, a generic fast lookup for module+generated code could be useful. > It would also be nice to associate names with generated functions and > integrate that with kallsyms. For DWARF we would also need a way to > associate CFI metadata with address ranges. >From the kprobe point of view, kallsyms and DWARF seems a bit overkill. At least for kprobe insn slots, I think one symbol per page instead of each entry will be better and CFI also should be generated for each call. Thank you, -- Masami Hiramatsu <mhira...@kernel.org>
