Re: [RFC][PATCH 1/2] x86: Allow breakpoints to emulate call functions

[Masami Hiramatsu]

On Tue, 7 May 2019 23:13:40 +0900
Masami Hiramatsu <mhiramat@xxxxxxxxxx> wrote:

> On Mon, 6 May 2019 17:45:11 -0400
> Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> 
> > If we go with Peter's patch, I can make this code much more sane, and
> > not have to worry about having &regs->sp be at the top of the stack. I
> > could simply, just push everything in the order of pt_regs and call the
> > handler.
> 
> Hi Steve, I need to catch up with the origin of this series, but it seems
> also good to optprobe which is doing similar trick on pt_regs. If we can
> assume that int3 pt_regs can have a gap, optprobe can also make a gap, and
> it can be also used for storing destination address.

Sorry, I misunderstood. I see the issue ( https://lkml.org/lkml/2019/5/1/497 )
and solutions on the thread. If we really need to fix this trace-livepatch
combination issue, it may be good to backport to stable trees.
>From this viewpoint, Linus's suggestion (no pt_reg changes on x86-32) seems
to have a point.

BTW, even though I think Peter's patch (unifying pt_regs behavior) will also
be good for us for more general reason (not only for fixing actual issue).

Thank you,

-- 
Masami Hiramatsu <mhiramat@xxxxxxxxxx>