From:   Jiri Olsa <olsajiri@gmail.com>
Date:   Sun, 19 Mar 2023 16:29:22 +0100
To:     Florent Revest <revest@chromium.org>
Cc:     linux-kernel@vger.kernel.org, linux-trace-kernel@vger.kernel.org,
        rostedt@goodmis.org, mhiramat@kernel.org, mark.rutland@arm.com,
        ast@kernel.org, daniel@iogearbox.net, kpsingh@kernel.org
Subject: Re: [PATCH 5/7] ftrace: Store direct called addresses in their ops
Message-ID: <ZBcqUoUTZSNyIjLx@krava>
References: <20230316173811.1223508-1-revest@chromium.org>
 <20230316173811.1223508-6-revest@chromium.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230316173811.1223508-6-revest@chromium.org>
Precedence: bulk

On Thu, Mar 16, 2023 at 06:38:09PM +0100, Florent Revest wrote:
> All direct calls are now registered using the register_ftrace_direct API
> so each ops can jump to only one direct-called trampoline.
> 
> By storing the direct called trampoline address directly in the ops we
> can save one hashmap lookup in the direct call ops and implement arm64
> direct calls on top of call ops.
> 
> Signed-off-by: Florent Revest <revest@chromium.org>
> ---
>  include/linux/ftrace.h | 3 +++
>  kernel/trace/ftrace.c  | 7 +++++--
>  2 files changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
> index abee60865fc7..6a532dd6789e 100644
> --- a/include/linux/ftrace.h
> +++ b/include/linux/ftrace.h
> @@ -321,6 +321,9 @@ struct ftrace_ops {
>  	unsigned long			trampoline_size;
>  	struct list_head		list;
>  	ftrace_ops_func_t		ops_func;
> +#ifdef CONFIG_DYNAMIC_FTRACE_WITH_DIRECT_CALLS
> +	unsigned long			direct_call;
> +#endif
>  #endif
>  };
>  
> diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
> index 33530198d1ca..66c91fa4b6ab 100644
> --- a/kernel/trace/ftrace.c
> +++ b/kernel/trace/ftrace.c
> @@ -2582,9 +2582,8 @@ ftrace_add_rec_direct(unsigned long ip, unsigned long addr,
>  static void call_direct_funcs(unsigned long ip, unsigned long pip,
>  			      struct ftrace_ops *ops, struct ftrace_regs *fregs)
>  {
> -	unsigned long addr;
> +	unsigned long addr = ops->direct_call;

nice, should it be read with READ_ONCE ?

jirka

>  
> -	addr = ftrace_find_rec_direct(ip);
>  	if (!addr)
>  		return;
>  
> @@ -5380,6 +5379,7 @@ int register_ftrace_direct(struct ftrace_ops *ops, unsigned long addr)
>  	ops->func = call_direct_funcs;
>  	ops->flags = MULTI_FLAGS;
>  	ops->trampoline = FTRACE_REGS_ADDR;
> +	ops->direct_call = addr;
>  
>  	err = register_ftrace_function_nolock(ops);
>  
> @@ -5454,6 +5454,7 @@ __modify_ftrace_direct(struct ftrace_ops *ops, unsigned long addr)
>  	/* Enable the tmp_ops to have the same functions as the direct ops */
>  	ftrace_ops_init(&tmp_ops);
>  	tmp_ops.func_hash = ops->func_hash;
> +	tmp_ops.direct_call = addr;
>  
>  	err = register_ftrace_function_nolock(&tmp_ops);
>  	if (err)
> @@ -5475,6 +5476,8 @@ __modify_ftrace_direct(struct ftrace_ops *ops, unsigned long addr)
>  			entry->direct = addr;
>  		}
>  	}
> +	/* Prevent store tearing if a trampoline concurrently accesses the value */
> +	WRITE_ONCE(ops->direct_call, addr);
>  
>  	mutex_unlock(&ftrace_lock);
>  
> -- 
> 2.40.0.rc2.332.ga46443480c-goog
>