lkml.org 
[lkml]   [2022]   [Oct]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 1/4] ftrace: pass fregs to arch_ftrace_set_direct_caller()
On Mon, Oct 24, 2022 at 06:04:49PM +0100, Mark Rutland wrote:
> On Mon, Oct 24, 2022 at 10:48:45AM -0400, Steven Rostedt wrote:
> > On Mon, 24 Oct 2022 15:08:43 +0100
> > Mark Rutland <mark.rutland@arm.com> wrote:
> >
> > > --- a/include/linux/ftrace.h
> > > +++ b/include/linux/ftrace.h
> > > @@ -429,6 +429,7 @@ static inline int modify_ftrace_direct_multi_nolock(struct ftrace_ops *ops, unsi
> > > }
> > > #endif /* CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS */
> > >
> > > +#ifdef CONFIG_FUNCTION_TRACER
> >
> > Instead of adding the above preprocessor check, the below chunk should be
> > moved into the CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS block above.
>
> Sure; but note that doing that naively means 'struct ftrace_regs' won't always
> be declared (e.g. if !CONFIG_FUNCTION_TRACER), and will result in warnings, e.g.
>
> | CC arch/x86/kernel/asm-offsets.s
> | In file included from ./include/linux/kvm_host.h:32,
> | from arch/x86/kernel/../kvm/vmx/vmx.h:5,
> | from arch/x86/kernel/asm-offsets.c:22:
> | ./include/linux/ftrace.h:444:57: error: ‘struct ftrace_regs’ declared inside parameter list will not be visible outside of this definition or declaration [-Werror]
> | 444 | static inline void arch_ftrace_set_direct_caller(struct ftrace_regs *fregs,
> | | ^~~~~~~~~~~
> | cc1: all warnings being treated as errors
> | make[1]: *** [scripts/Makefile.build:118: arch/x86/kernel/asm-offsets.s] Error 1
> | make: *** [Makefile:1270: prepare0] Error 2
>
> ... so I'll either need to add some ifdeffery, for CONFIG_FUNCTION_TRACER, or I
> can hoist the declaration of 'struct ftrace_regs' to not depend on
> CONFIG_FUNCTION_TRACER.
>
> I guess the latter is preferable, e.g.
>
> | diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
> | index 2b34fec40a39..f201fcbfffb0 100644
> | --- a/include/linux/ftrace.h
> | +++ b/include/linux/ftrace.h
> | @@ -37,9 +37,10 @@ extern void ftrace_boot_snapshot(void);
> | static inline void ftrace_boot_snapshot(void) { }
> | #endif
> |
> | -#ifdef CONFIG_FUNCTION_TRACER
> | struct ftrace_ops;
> | struct ftrace_regs;
> | +
> | +#ifdef CONFIG_FUNCTION_TRACER
> | /*
> | * If the arch's mcount caller does not support all of ftrace's
> | * features, then it must call an indirect function that
>
> ... so I've done that locally for now.

The kbuild robot complained that this led to x86 getting multiple definitions
of arch_ftrace_set_direct_caller(), so I've also added the below:

| diff --git a/arch/x86/include/asm/ftrace.h b/arch/x86/include/asm/ftrace.h
| index 788e7c1f6463c..6f9b9fee04132 100644
| --- a/arch/x86/include/asm/ftrace.h
| +++ b/arch/x86/include/asm/ftrace.h
| @@ -59,6 +59,7 @@ void ftrace_graph_func(unsigned long ip, unsigned long parent_ip,
| #define FTRACE_GRAPH_TRAMP_ADDR FTRACE_GRAPH_ADDR
| #endif
|
| +#ifdef CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS
| /*
| * When a ftrace registered caller is tracing a function that is
| * also set by a register_ftrace_direct() call, it needs to be
| @@ -71,6 +72,7 @@ static inline void arch_ftrace_set_direct_caller(struct ftrace_regs *fregs, unsi
| /* Emulate a call */
| fregs->regs.orig_ax = addr;
| }
| +#endif
|
| #ifdef CONFIG_DYNAMIC_FTRACE

... AFAICT s390 doesn't need similar treatment.

Thanks,
Mark.

\
 
 \ /
  Last update: 2022-10-28 18:28    [W:0.113 / U:0.092 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site