[PATCH v6 6/7] arm64: ftrace: Add CALLER_ADDRx macros

[takahiro.akashi@linaro.org (AKASHI Takahiro)]

On 03/14/2014 12:54 AM, Will Deacon wrote:
> On Thu, Mar 13, 2014 at 10:13:49AM +0000, AKASHI Takahiro wrote:
>> CALLER_ADDRx returns caller's address at specified level in call stacks.
>> They are used for several tracers like irqsoff and preemptoff.
>> Strange to say, however, they are refered even without FTRACE.
>>
>> Please note that this implementation assumes that we have frame pointers.
>> (which means kernel should be compiled with -fno-omit-frame-pointer.)
>
> How do you ensure that -fno-omit-frame-pointer is passed?

arm64 selects ARCH_WANT_FRAME_POINTERS, then FRAME_POINTER is on (lib/Kconfig.debug)
and so -fno-omit-frame-pointer is appended (${TOP}/Makefile).
(stacktrace.c also assumes FRAME_POINTER.)

Do you think I should remove the comment above?

>> Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
>> ---
>>   arch/arm64/include/asm/ftrace.h    |   13 ++++++++-
>>   arch/arm64/kernel/Makefile         |    3 +-
>>   arch/arm64/kernel/return_address.c |   55 ++++++++++++++++++++++++++++++++++++
>>   3 files changed, 69 insertions(+), 2 deletions(-)
>>   create mode 100644 arch/arm64/kernel/return_address.c
>>
>> diff --git a/arch/arm64/include/asm/ftrace.h b/arch/arm64/include/asm/ftrace.h
>> index ed5c448..c44c4b1 100644
>> --- a/arch/arm64/include/asm/ftrace.h
>> +++ b/arch/arm64/include/asm/ftrace.h
>> @@ -18,6 +18,7 @@
>>
>>   #ifndef __ASSEMBLY__
>>   extern void _mcount(unsigned long);
>> +extern void *return_address(unsigned int);
>>
>>   struct dyn_arch_ftrace {
>>   	/* No extra data needed for arm64 */
>> @@ -33,6 +34,16 @@ static inline unsigned long ftrace_call_adjust(unsigned long addr)
>>   	 */
>>   	return addr;
>>   }
>> -#endif /* __ASSEMBLY__ */
>> +
>> +#define HAVE_ARCH_CALLER_ADDR
>> +
>> +#define CALLER_ADDR0 ((unsigned long)__builtin_return_address(0))
>> +#define CALLER_ADDR1 ((unsigned long)return_address(1))
>> +#define CALLER_ADDR2 ((unsigned long)return_address(2))
>> +#define CALLER_ADDR3 ((unsigned long)return_address(3))
>> +#define CALLER_ADDR4 ((unsigned long)return_address(4))
>> +#define CALLER_ADDR5 ((unsigned long)return_address(5))
>> +#define CALLER_ADDR6 ((unsigned long)return_address(6))
>
> Could we change the core definitions of these macros (in linux/ftrace.h) to
> use return_address, then provide an overridable version of return_address
> that defaults to __builtin_return_address, instead of copy-pasting this
> sequence?

I think I understand what you mean, and will try to post a separate RFC,
but I also want to hold off this change on this patch since such a change
may raise a small controversy from other archs' maintainers.

>> diff --git a/arch/arm64/kernel/return_address.c b/arch/arm64/kernel/return_address.c
>> new file mode 100644
>> index 0000000..89102a6
>> --- /dev/null
>> +++ b/arch/arm64/kernel/return_address.c
>> @@ -0,0 +1,55 @@
>> +/*
>> + * arch/arm64/kernel/return_address.c
>> + *
>> + * Copyright (C) 2013 Linaro Limited
>> + * Author: AKASHI Takahiro <takahiro.akashi@linaro.org>
>> + *
>> + * This program is free software; you can redistribute it and/or modify
>> + * it under the terms of the GNU General Public License version 2 as
>> + * published by the Free Software Foundation.
>> + */
>> +
>> +#include <linux/export.h>
>> +#include <linux/ftrace.h>
>> +
>> +#include <asm/stacktrace.h>
>> +
>> +struct return_address_data {
>> +	unsigned int level;
>> +	void *addr;
>> +};
>> +
>> +static int save_return_addr(struct stackframe *frame, void *d)
>> +{
>> +	struct return_address_data *data = d;
>> +
>> +	if (!data->level) {
>> +		data->addr = (void *)frame->pc;
>> +		return 1;
>> +	} else {
>> +		--data->level;
>> +		return 0;
>> +	}
>> +}
>> +
>> +void *return_address(unsigned int level)
>> +{
>> +	struct return_address_data data;
>> +	struct stackframe frame;
>> +	register unsigned long current_sp asm ("sp");
>> +
>> +	data.level = level + 2;
>> +	data.addr = NULL;
>> +
>> +	frame.fp = (unsigned long)__builtin_frame_address(0);
>> +	frame.sp = current_sp;
>> +	frame.pc = (unsigned long)return_address; /* dummy */
>> +
>> +	walk_stackframe(&frame, save_return_addr, &data);
>> +
>> +	if (!data.level)
>> +		return data.addr;
>> +	else
>> +		return NULL;
>> +}
>> +EXPORT_SYMBOL_GPL(return_address);
>
> This whole file is basically copied from arch/arm/, but it's not too much
> code. Ideally the toolchain would have made use of the frame pointer, but it
> looks like it doesn't bother.

I confirmed that __builtin_return_address([123456]) doesn't work
even with -fno-omit-frame-pointer.
Keep this as it is.

-Takahiro AKASHI

> Will
>