public inbox for [email protected]
 help / color / mirror / Atom feed
From: David Vernet <[email protected]>
To: Wan Jiabing <[email protected]>
Cc: Steven Rostedt <[email protected]>,
	Ingo Molnar <[email protected]>,
	Alexei Starovoitov <[email protected]>,
	Daniel Borkmann <[email protected]>,
	Andrii Nakryiko <[email protected]>,
	Martin KaFai Lau <[email protected]>, Song Liu <[email protected]>,
	Yonghong Song <[email protected]>,
	John Fastabend <[email protected]>,
	KP Singh <[email protected]>,
	[email protected], [email protected],
	[email protected]
Subject: Re: [PATCH 1/3] bpf: use 'error_xxx' tags in bpf_kprobe_multi_link_attach
Date: Thu, 12 May 2022 10:05:22 -0700	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

On Thu, May 12, 2022 at 10:17:08PM +0800, Wan Jiabing wrote:
> Use 'error_addrs', 'error_cookies' and 'error_link' tags to make error
> handling more efficient.

Can you add a bit more context to this commit summary? The added goto
labels aren't what make the code more performant, it's the avoidance of
unnecessary free calls on NULL pointers that (supposedly) does.

> 
> Signed-off-by: Wan Jiabing <[email protected]>
> ---
>  kernel/trace/bpf_trace.c | 20 +++++++++++---------
>  1 file changed, 11 insertions(+), 9 deletions(-)
> 
> diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> index 2eaac094caf8..3a8b69ef9a0d 100644
> --- a/kernel/trace/bpf_trace.c
> +++ b/kernel/trace/bpf_trace.c
> @@ -2467,20 +2467,20 @@ int bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
>  	if (uaddrs) {
>  		if (copy_from_user(addrs, uaddrs, size)) {
>  			err = -EFAULT;
> -			goto error;
> +			goto error_addrs;
>  		}
>  	} else {
>  		struct user_syms us;
>  
>  		err = copy_user_syms(&us, usyms, cnt);
>  		if (err)
> -			goto error;
> +			goto error_addrs;
>  
>  		sort(us.syms, cnt, sizeof(*us.syms), symbols_cmp, NULL);
>  		err = ftrace_lookup_symbols(us.syms, cnt, addrs);
>  		free_user_syms(&us);
>  		if (err)
> -			goto error;
> +			goto error_addrs;
>  	}
>  
>  	ucookies = u64_to_user_ptr(attr->link_create.kprobe_multi.cookies);
> @@ -2488,18 +2488,18 @@ int bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
>  		cookies = kvmalloc(size, GFP_KERNEL);
>  		if (!cookies) {
>  			err = -ENOMEM;
> -			goto error;
> +			goto error_addrs;
>  		}
>  		if (copy_from_user(cookies, ucookies, size)) {
>  			err = -EFAULT;
> -			goto error;
> +			goto error_cookies;
>  		}
>  	}
>  
>  	link = kzalloc(sizeof(*link), GFP_KERNEL);
>  	if (!link) {
>  		err = -ENOMEM;
> -		goto error;
> +		goto error_cookies;
>  	}
>  
>  	bpf_link_init(&link->link, BPF_LINK_TYPE_KPROBE_MULTI,
> @@ -2507,7 +2507,7 @@ int bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
>  
>  	err = bpf_link_prime(&link->link, &link_primer);
>  	if (err)
> -		goto error;
> +		goto error_link;
>  
>  	if (flags & BPF_F_KPROBE_MULTI_RETURN)
>  		link->fp.exit_handler = kprobe_multi_link_handler;
> @@ -2539,10 +2539,12 @@ int bpf_kprobe_multi_link_attach(const union bpf_attr *attr, struct bpf_prog *pr
>  
>  	return bpf_link_settle(&link_primer);
>  
> -error:
> +error_link:
>  	kfree(link);
> -	kvfree(addrs);
> +error_cookies:
>  	kvfree(cookies);
> +error_addrs:
> +	kvfree(addrs);
>  	return err;
>  }
>  #else /* !CONFIG_FPROBE */
> -- 
> 2.35.1
> 

Could you clarify what performance gains you observed from doing this? I
wouldn't have expected avoiding a couple of calls and NULL checks to have a
measurable impact on performance, and I'm wondering whether the complexity
from having multiple goto labels is really worth any supposed performance
gains.

Thanks,
David


  parent reply	other threads:[~2022-05-12 17:05 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-05-12 14:17 [PATCH 0/3] bpf: optimize the bpf_kprobe_multi_link_attach function Wan Jiabing
2022-05-12 14:17 ` [PATCH 1/3] bpf: use 'error_xxx' tags in bpf_kprobe_multi_link_attach Wan Jiabing
2022-05-12 16:40   ` Song Liu
2022-05-12 17:05   ` David Vernet [this message]
2022-05-12 14:17 ` [PATCH 2/3] bpf: simplify if-if to if " Wan Jiabing
2022-05-12 16:44   ` Song Liu
2022-05-12 14:17 ` [PATCH 3/3] bpf: use vmemdup_user instead of kvmalloc and copy_from_user Wan Jiabing
2022-05-12 16:45   ` Song Liu

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220512170522.3e47hwj53plhr4qq@dev0025.ash9.facebook.com \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox