Re: [PATCH v2 2/2] ftrace: Consolidate ftrace_regs accessor functions for archs using pt_regs

From: Steven Rostedt
Date: Wed Oct 09 2024 - 09:43:29 EST


On Wed, 9 Oct 2024 13:47:23 +0900
Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx> wrote:

> > --- /dev/null
> > +++ b/include/linux/ftrace_regs.h
> > @@ -0,0 +1,36 @@
> > +/* SPDX-License-Identifier: GPL-2.0 */
> > +#ifndef _LINUX_FTRACE_TYPES_H
> > +#define _LINUX_FTRACE_TYPES_H
> ^^^^^^^^^^^^^^^^ Is this _LINUX_FTRACE_REGS_H?

Ah, I originally called it ftrace_types.h, but later decided that name
didn't really fit. I changed all references to it but this one.

Thanks for catching this.

>
>
> > +
> > +/*
> > + * For archs that just copy pt_regs in ftrace regs, it can use this default.
> > + * If an architecture does not use pt_regs, it must define all the below
> > + * accessor functions.
> > + */
> > +#ifndef HAVE_ARCH_FTRACE_REGS
> > +struct __arch_ftrace_regs {
> > + struct pt_regs regs;
> > +};
> > +
> > +#define arch_ftrace_regs(fregs) ((struct __arch_ftrace_regs *)(fregs))
> > +
> > +struct ftrace_regs;
> > +
> > +#define ftrace_regs_get_instruction_pointer(fregs) \
> > + instruction_pointer(arch_ftrace_get_regs(fregs))
> > +#define ftrace_regs_get_argument(fregs, n) \
> > + regs_get_kernel_argument(arch_ftrace_get_regs(fregs), n)
> > +#define ftrace_regs_get_stack_pointer(fregs) \
> > + kernel_stack_pointer(arch_ftrace_get_regs(fregs))
> > +#define ftrace_regs_return_value(fregs) \
> > + regs_return_value(arch_ftrace_get_regs(fregs))
> > +#define ftrace_regs_set_return_value(fregs, ret) \
> > + regs_set_return_value(arch_ftrace_get_regs(fregs), ret)
> > +#define ftrace_override_function_with_return(fregs) \
> > + override_function_with_return(arch_ftrace_get_regs(fregs))
> > +#define ftrace_regs_query_register_offset(name) \
> > + regs_query_register_offset(name)
> > +
> > +#endif /* HAVE_ARCH_FTRACE_REGS */
> > +
> > +#endif /* _LINUX_FTRACE_TYPES_H */
>
> Ditto.
>
> Others looks good to me.
>
> Acked-by: Masami Hiramatsu (Google) <mhiramat@xxxxxxxxxx>

Thanks,

I'll send a v2 with this update.

-- Steve