On Tue, 15 Oct 2024 20:39:06 +0200
Heiko Carstens <h...@linux.ibm.com> wrote:

> On Tue, Oct 15, 2024 at 10:29:17AM +0900, Masami Hiramatsu (Google) wrote:
> > From: Masami Hiramatsu (Google) <mhira...@kernel.org>
> > 
> > Use ftrace_regs instead of fgraph_ret_regs for tracing return value
> > on function_graph tracer because of simplifying the callback interface.
> > 
> > The CONFIG_HAVE_FUNCTION_GRAPH_RETVAL is also replaced by
> > CONFIG_HAVE_FUNCTION_GRAPH_FREGS.
> > 
> > Signed-off-by: Masami Hiramatsu (Google) <mhira...@kernel.org>
> 
> ...
> 
> > diff --git a/arch/s390/kernel/mcount.S b/arch/s390/kernel/mcount.S
> > index 7e267ef63a7f..a9ca56ea0858 100644
> > --- a/arch/s390/kernel/mcount.S
> > +++ b/arch/s390/kernel/mcount.S
> > @@ -134,14 +134,15 @@ SYM_CODE_END(ftrace_common)
> >  SYM_FUNC_START(return_to_handler)
> >     stmg    %r2,%r5,32(%r15)
> >     lgr     %r1,%r15
> > -   aghi    %r15,-(STACK_FRAME_OVERHEAD+__FGRAPH_RET_SIZE)
> > +   aghi    %r15,-(STACK_FRAME_OVERHEAD+STACK_FRAME_SIZE_FREGS)
> >     stg     %r1,__SF_BACKCHAIN(%r15)
> > -   la      %r3,STACK_FRAME_OVERHEAD(%r15)
> > -   stg     %r1,__FGRAPH_RET_FP(%r3)
> > -   stg     %r2,__FGRAPH_RET_GPR2(%r3)
> > -   lgr     %r2,%r3
> > +   la      %r4,STACK_FRAME_OVERHEAD(%r15)
> > +   stg     %r2,__PT_R2(%r4)
> > +   stg     %r3,__PT_R3(%r4)
> > +   stg     %r1,__PT_R15(%r4)
> > +   lgr     %r2,%r4
> >     brasl   %r14,ftrace_return_to_handler
> > -   aghi    %r15,STACK_FRAME_OVERHEAD+__FGRAPH_RET_SIZE
> > +   aghi    %r15,STACK_FRAME_SIZE_FREGS
> >     lgr     %r14,%r2
> >     lmg     %r2,%r5,32(%r15)
> >     BR_EX   %r14
> 
> Why didn't you simply merge the addon patch which I provided, and
> which I tested?
> https://lore.kernel.org/all/20240916121656.20933-b-...@linux.ibm.com
> 
> That would make things much simpler... e.g. your new patch is also
> writing r3 to fregs, why? The stackframe allocation is also wrong.
> I didn't try, but I guess the above code would crash instantly.

I thought it is better to focus on replacing fgraph_ret_regs with
ftrace_regs in this patch, but if it is wrong and I should remove
stack_frame, let me fix that.

Thank you,

-- 
Masami Hiramatsu (Google) <mhira...@kernel.org>

Reply via email to