linux-kernel - Re: [PATCH] bpf: added the account of BPF running time

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 2 Sep 2022 17:38:51 +0200
From:   Daniel Borkmann <daniel@...earbox.net>
To:     Yunhui Cui <cuiyunhui@...edance.com>, corbet@....net,
        ast@...nel.org, andrii@...nel.org, martin.lau@...ux.dev,
        song@...nel.org, yhs@...com, john.fastabend@...il.com,
        kpsingh@...nel.org, sdf@...gle.com, haoluo@...gle.com,
        jolsa@...nel.org, akpm@...ux-foundation.org, hannes@...xchg.org,
        david@...hat.com, mail@...istoph.anton.mitterer.name,
        ccross@...gle.com, vincent.whitchurch@...s.com,
        paul.gortmaker@...driver.com, peterz@...radead.org,
        edumazet@...gle.com, joshdon@...gle.com
Cc:     linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org,
        linux-doc@...r.kernel.org, bpf@...r.kernel.org
Subject: Re: [PATCH] bpf: added the account of BPF running time

On 9/2/22 12:12 PM, Yunhui Cui wrote:
[...]
> index a5f21dc3c432..9cb072f9e32b 100644
> --- a/include/linux/filter.h
> +++ b/include/linux/filter.h
> @@ -565,6 +565,12 @@ struct sk_filter {
>   	struct bpf_prog	*prog;
>   };
>   
> +struct bpf_account {
> +	u64_stats_t nsecs;
> +	struct u64_stats_sync syncp;
> +};
> +DECLARE_PER_CPU(struct bpf_account, bpftime);
> +
>   DECLARE_STATIC_KEY_FALSE(bpf_stats_enabled_key);
>   
>   typedef unsigned int (*bpf_dispatcher_fn)(const void *ctx,
> @@ -577,12 +583,14 @@ static __always_inline u32 __bpf_prog_run(const struct bpf_prog *prog,
>   					  bpf_dispatcher_fn dfunc)
>   {
>   	u32 ret;
> +	struct bpf_account *bact;
> +	unsigned long flags;
> +	u64 start = 0;
>   
>   	cant_migrate();
> +	start = sched_clock();
>   	if (static_branch_unlikely(&bpf_stats_enabled_key)) {
>   		struct bpf_prog_stats *stats;
> -		u64 start = sched_clock();
> -		unsigned long flags;
>   
>   		ret = dfunc(ctx, prog->insnsi, prog->bpf_func);
>   		stats = this_cpu_ptr(prog->stats);
> @@ -593,6 +601,11 @@ static __always_inline u32 __bpf_prog_run(const struct bpf_prog *prog,
>   	} else {
>   		ret = dfunc(ctx, prog->insnsi, prog->bpf_func);
>   	}
> +	bact = this_cpu_ptr(&bpftime);
> +	flags = u64_stats_update_begin_irqsave(&bact->syncp);
> +	u64_stats_add(&bact->nsecs, sched_clock() - start);
> +	u64_stats_update_end_irqrestore(&bact->syncp, flags);
> +
>   	return ret;

The overhead this adds unconditionally is no-go. Have you tried using/improving:

commit 47c09d6a9f6794caface4ad50930460b82d7c670
Author: Song Liu <songliubraving@...com>
Date:   Mon Mar 9 10:32:15 2020 -0700

     bpftool: Introduce "prog profile" command

     With fentry/fexit programs, it is possible to profile BPF program with
     hardware counters. Introduce bpftool "prog profile", which measures key
     metrics of a BPF program.

     bpftool prog profile command creates per-cpu perf events. Then it attaches
     fentry/fexit programs to the target BPF program. The fentry program saves
     perf event value to a map. The fexit program reads the perf event again,
     and calculates the difference, which is the instructions/cycles used by
     the target program.

     Example input and output:

       ./bpftool prog profile id 337 duration 3 cycles instructions llc_misses

             4228 run_cnt
          3403698 cycles                                              (84.08%)
          3525294 instructions   #  1.04 insn per cycle               (84.05%)
               13 llc_misses     #  3.69 LLC misses per million isns  (83.50%)

     This command measures cycles and instructions for BPF program with id
     337 for 3 seconds. The program has triggered 4228 times. The rest of the
     output is similar to perf-stat. [...]

Thanks,
Daniel