Re: [PATCH] filter: Optimize instruction revalidation code.

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mardi 16 novembre 2010 à 22:08 +0900, Tetsuo Handa a écrit :
> In the wake of commit 57fe93b3 "filter: make sure filters dont read
> uninitialized memory", I came up with this patch. I don't know how to test.
> Please review and do tests.
> 
> This patch optimizes sk_chk_filter(). Using gcc 4.1.2 on x86_32, the size of
> net/core/filter.o changed from 7416 bytes to 5260 bytes
> ("make allnoconfig" + CONFIG_NET=y).
> 
> By the way, if "struct sock_filter *filter" argument of sk_run_filter() does
> not change after it was revalidated at sk_chk_filter(), can't we detect and
> reject "BPF_S_LD_MEM/BPF_S_LDX_MEM before BPF_S_ST/BPF_S_STX" at
> sk_chk_filter()?

Sure, this was the idea, but seems complex because of branches.

Patches are welcomed ;)

> ----------------------------------------
>  From 2bf36130bd4912590b409b47a6a9a82e2884e035 Mon Sep 17 00:00:00 2001
> From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
> Date: Tue, 16 Nov 2010 21:39:50 +0900
> Subject: [PATCH] filter: Optimize instruction revalidation code.
> 
> Since repeating small value to small value conversion using switch() clause's
> case statement is wasteful, this patch instroduces u16 to u16 mapping table
> and removes most of case statements. As a result, the size of net/core/filter.o
> is reduced by about 30% on x86.
> 
> Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
> ---

Patch seems fine to me, with the 'const' codes[] Michael Tokarev already
spotted.

Thanks