Re: Re: [PATCH 1/5] ftrace: Synchronize variable setting withbreakpoints

[Steven Rostedt]

On Fri, 2012-06-01 at 00:16 +0900, Masami Hiramatsu wrote:


> Perhaps, kprobes-*jump*-optimization may be better to handle it
> because the target probe is simultaneously working while
> optimizing (modifying code). This means that if someone hits
> breakpoint of such kprobe, it must be handled by kprobes, not
> only just tweaking IP address.

Yeah, ftrace is easier on what the breakpoint does. For kprobes, it's a
bit more invasive. Ftrace only deals with a nop, or a "trace-me" call.
As the modifications are being done, we always treat it as a nop until
the modification is complete.

With kprobes, the text_poke() is a bit more difficult, because it can't
treat the location as a nop, as what is changing actually performs some
task for the kernel. It requires that the breakpoint know what to do
with the text (out-of-line execution, and such). ftrace has the benefit
of just "skip this instruction" and return.

-- Steve


--
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/