Re: [PATCH] tracing: Have saved_cmdlines arrays all in one allocation

[Tim Chen]

On Mon, 2024-02-12 at 18:09 -0500, Steven Rostedt wrote:
> From: "Steven Rostedt (Google)" <rostedt@xxxxxxxxxxx>
> 
> The saved_cmdlines have three arrays for mapping PIDs to COMMs:
> 
>  - map_pid_to_cmdline[]
>  - map_cmdline_to_pid[]
>  - saved_cmdlines
> 
> The map_pid_to_cmdline[] is PID_MAX_DEFAULT in size and holds the index
> into the other arrays. The map_cmdline_to_pid[] is a mapping back to the
> full pid as it can be larger than PID_MAX_DEFAULT. And the
> saved_cmdlines[] just holds the COMMs associated to the pids.
> 
> Currently the map_pid_to_cmdline[] and saved_cmdlines[] are allocated
> together (in reality the saved_cmdlines is just in the memory of the
> rounding of the allocation of the structure as it is always allocated in
> powers of two). The map_cmdline_to_pid[] array is allocated separately.
> 
> Since the rounding to a power of two is rather large (it allows for 8000
> elements in saved_cmdlines), also include the map_cmdline_to_pid[] array.
> (This drops it to 6000 by default, which is still plenty for most use
> cases). This saves even more memory as the map_cmdline_to_pid[] array
> doesn't need to be allocated.


This patch does make better use of the extra space and make the
previous change better.

Reviewed-by: Tim Chen <tim.c.chen@xxxxxxxxxxxxxxx>
 
> 
> Link: https://lore.kernel.org/linux-trace-kernel/20240212174011.068211d9@xxxxxxxxxxxxxxxxxx/
> 
> Signed-off-by: Steven Rostedt (Google) <rostedt@xxxxxxxxxxx>
> ---
>  kernel/trace/trace_sched_switch.c | 13 ++++---------
>  1 file changed, 4 insertions(+), 9 deletions(-)
> 
> diff --git a/kernel/trace/trace_sched_switch.c b/kernel/trace/trace_sched_switch.c
> index e4fbcc3bede5..210c74dcd016 100644
> --- a/kernel/trace/trace_sched_switch.c
> +++ b/kernel/trace/trace_sched_switch.c
> @@ -201,7 +201,7 @@ static struct saved_cmdlines_buffer *allocate_cmdlines_buffer(unsigned int val)
>  	int order;
>  
>  	/* Figure out how much is needed to hold the given number of cmdlines */
> -	orig_size = sizeof(*s) + val * TASK_COMM_LEN;
> +	orig_size = sizeof(*s) + val * (TASK_COMM_LEN + sizeof(int));

Strictly speaking, *map_cmdline_to_pid is unsigned int so it is more consistent
to use sizeof(unsigned) in line above.  But I'm nitpicking and I'm fine to
leave it as is.

>  	order = get_order(orig_size);
>  	size = 1 << (order + PAGE_SHIFT);
>  	page = alloc_pages(GFP_KERNEL, order);
> @@ -212,16 +212,11 @@ static struct saved_cmdlines_buffer *allocate_cmdlines_buffer(unsigned int val)
>  	memset(s, 0, sizeof(*s));
>  
>  	/* Round up to actual allocation */
> -	val = (size - sizeof(*s)) / TASK_COMM_LEN;
> +	val = (size - sizeof(*s)) / (TASK_COMM_LEN + sizeof(int));
>  	s->cmdline_num = val;
>  
> -	s->map_cmdline_to_pid = kmalloc_array(val,
> -					      sizeof(*s->map_cmdline_to_pid),
> -					      GFP_KERNEL);
> -	if (!s->map_cmdline_to_pid) {
> -		free_saved_cmdlines_buffer(s);
> -		return NULL;
> -	}
> +	/* Place map_cmdline_to_pid array right after saved_cmdlines */
> +	s->map_cmdline_to_pid = (unsigned *)&s->saved_cmdlines[val * TASK_COMM_LEN];
>  
>  	s->cmdline_idx = 0;
>  	memset(&s->map_pid_to_cmdline, NO_CMDLINE_MAP,