[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <9bfa8866-a90f-41bf-8b22-bf704c01a2e5@linux.dev>
Date: Fri, 1 Aug 2025 07:49:53 -0700
From: Yonghong Song <yonghong.song@...ux.dev>
To: Steven Rostedt <rostedt@...dmis.org>, LKML
<linux-kernel@...r.kernel.org>,
Linux Trace Kernel <linux-trace-kernel@...r.kernel.org>
Cc: Masami Hiramatsu <mhiramat@...nel.org>,
Mathieu Desnoyers <mathieu.desnoyers@...icios.com>, bpf@...r.kernel.org,
Douglas Raillard <douglas.raillard@....com>
Subject: Re: [PATCH] tracing: Have unsigned int function args displayed as
hexadecimal
On 7/31/25 4:31 PM, Steven Rostedt wrote:
> From: Steven Rostedt <rostedt@...dmis.org>
>
> Most function arguments that are passed in as unsigned int or unsigned
> long are better displayed as hexadecimal than normal integer. For example,
> the functions:
>
> static void __create_object(unsigned long ptr, size_t size,
> int min_count, gfp_t gfp, unsigned int objflags);
>
> static bool stack_access_ok(struct unwind_state *state, unsigned long _addr,
> size_t len);
>
> void __local_bh_disable_ip(unsigned long ip, unsigned int cnt);
>
> Show up in the trace as:
>
> __create_object(ptr=-131387050520576, size=4096, min_count=1, gfp=3264, objflags=0) <-kmem_cache_alloc_noprof
> stack_access_ok(state=0xffffc9000233fc98, _addr=-60473102566256, len=8) <-unwind_next_frame
> __local_bh_disable_ip(ip=-2127311112, cnt=256) <-handle_softirqs
>
> Instead, by displaying unsigned as hexadecimal, they look more like this:
>
> __create_object(ptr=0xffff8881028d2080, size=0x280, min_count=1, gfp=0x82820, objflags=0x0) <-kmem_cache_alloc_node_noprof
> stack_access_ok(state=0xffffc90000003938, _addr=0xffffc90000003930, len=0x8) <-unwind_next_frame
> __local_bh_disable_ip(ip=0xffffffff8133cef8, cnt=0x100) <-handle_softirqs
>
> Which is much easier to understand as most unsigned longs are usually just
> pointers. Even the "unsigned int cnt" in __local_bh_disable_ip() looks
> better as hexadecimal as a lot of flags are passed as unsigned.
>
> Signed-off-by: Steven Rostedt (Google) <rostedt@...dmis.org>
LGTM. But it seems some format issue. See below.
Acked-by: Yonghong Song <yonghong.song@...ux.dev>
> ---
> kernel/trace/trace_output.c | 17 ++++++++++++++++-
> 1 file changed, 16 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/trace/trace_output.c b/kernel/trace/trace_output.c
> index 0b3db02030a7..b18393d66a7f 100644
> --- a/kernel/trace/trace_output.c
> +++ b/kernel/trace/trace_output.c
> @@ -690,6 +690,12 @@ int trace_print_lat_context(struct trace_iterator *iter)
> }
>
> #ifdef CONFIG_FUNCTION_TRACE_ARGS
> +
> +static u32 btf_type_int(const struct btf_type *t)
> +{
> + return *(u32 *)(t + 1);
> +}
> +
> void print_function_args(struct trace_seq *s, unsigned long *args,
> unsigned long func)
> {
> @@ -701,6 +707,8 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
> struct btf *btf;
> s32 tid, nr = 0;
> int a, p, x;
> + int int_data;
> + u16 encode;
>
> trace_seq_printf(s, "(");
>
> @@ -744,7 +752,14 @@ void print_function_args(struct trace_seq *s, unsigned long *args,
> trace_seq_printf(s, "0x%lx", arg);
> break;
> case BTF_KIND_INT:
> - trace_seq_printf(s, "%ld", arg);
> + /* Get the INT encodoing */
> + int_data = btf_type_int(t);
> + encode = BTF_INT_ENCODING(int_data);
See different identation between above 'int_data' and 'encode'. The same as below.
> + /* Print unsigned ints as hex */
> + if (encode & BTF_INT_SIGNED)
> + trace_seq_printf(s, "%ld", arg);
> + else
> + trace_seq_printf(s, "0x%lx", arg);
> break;
> case BTF_KIND_ENUM:
> trace_seq_printf(s, "%ld", arg);
Powered by blists - more mailing lists