[RFC PATCH -next v2 3/4] arm64/ftrace: support dynamically allocated trampolines

[Masami Hiramatsu]

Hi Steve,

On Mon, 9 May 2022 14:22:03 -0400
Steven Rostedt <rostedt at goodmis.org> wrote:

> On Thu, 5 May 2022 12:15:38 +0900
> Masami Hiramatsu <mhiramat at kernel.org> wrote:
> 
> > OK. But my interest is that the ftrace on arm64 can provide a limited
> > access to registers via pt_regs or not. I don't mind the contained values
> > so much because in the most case, 'users' will (most likely) access to the
> > ARGs via BPF or tracefs (and we can just warn users if they try to access
> > the registers which is not saved.) But if the arm64 ftrace only provides
> > a special data structure, arch-independent code must have 2 different access
> > code. That is inefficient. That is my concern.
> > IOW, I'm interested in interface abstraction.
> 
> Note, ftrace now has a ftrace_regs structure that is passed to the
> callbacks for the function tracer.
> 
> It then has an arch dependent helper function ftrace_get_regs(fregs), that
> returns a pt_regs from the fregs only if the fregs has a full pt_regs to
> return. If not, it returns NULL.
>
> This was suggested by both Peter Zijlstra and Thomas Gleixner when I
> introduced FTRACE_WITH_ARGS, where all functions can now get the arguments
> from fregs, but not the full pt_regs.

Hmm, I thought the ftrace_get_regs() is the all-or-nothing interface, or
is there any way to get the arguments from fregs?

> If a ftrace_ops has the REGS flag set
> (using ftrace_regs_caller), the ftrace_get_regs(fregs) will return the
> pt_regs, or it will return NULL if ftrace_regs_caller was not used.
> 
> This way the same parameter can provide full pt_regs or a subset, and have
> an generic interface to tell the difference.

If it can provide a partial (subset of) pt_regs, that could be good for me
too, since at least kprobe-events on ftrace can check the traced register
is in the subset or not and reject it if it doesn't saved.

Thank you,

-- 
Masami Hiramatsu <mhiramat at kernel.org>