mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Julian Anastasov <ja@ssi.bg>
To: David Laight <David.Laight@ACULAB.COM>
Cc: "netdev@vger.kernel.org" <netdev@vger.kernel.org>,
	"'Naresh Kamboju'" <naresh.kamboju@linaro.org>,
	"'Dan Carpenter'" <dan.carpenter@linaro.org>,
	"'pablo@netfilter.org'" <pablo@netfilter.org>,
	"'open list'" <linux-kernel@vger.kernel.org>,
	"'lkft-triage@lists.linaro.org'" <lkft-triage@lists.linaro.org>,
	"'Linux Regressions'" <regressions@lists.linux.dev>,
	"'Linux ARM'" <linux-arm-kernel@lists.infradead.org>,
	"'netfilter-devel@vger.kernel.org'"
	<netfilter-devel@vger.kernel.org>,
	"'Arnd Bergmann'" <arnd@arndb.de>,
	"'Anders Roxell'" <anders.roxell@linaro.org>,
	"'Johannes Berg'" <johannes.berg@intel.com>,
	"'toke@kernel.org'" <toke@kernel.org>,
	"'Al Viro'" <viro@zeniv.linux.org.uk>,
	"'kernel@jfarr.cc'" <kernel@jfarr.cc>,
	"'kees@kernel.org'" <kees@kernel.org>
Subject: Re: [PATCH net] Fix clamp() of ip_vs_conn_tab on small memory systems.
Date: Fri, 6 Dec 2024 14:19:21 +0200 (EET)	[thread overview]
Message-ID: <5ec10e7c-d050-dab8-1f1b-d0ca2d922eef@ssi.bg> (raw)
In-Reply-To: <33893212b1cc4a418cec09aeeed0a9fc@AcuMS.aculab.com>


	Hello,

On Fri, 6 Dec 2024, David Laight wrote:

> The intention of the code seems to be that the minimum table
> size should be 256 (1 << min).
> However the code uses max = clamp(20, 5, max_avail) which implies

	Actually, it tries to reduce max=20 (max possible) below
max_avail: [8 .. max_avail]. Not sure what 5 is here...

> the author thought max_avail could be less than 5.
> But clamp(val, min, max) is only well defined for max >= min.
> If max < min whether is returns min or max depends on the order of
> the comparisons.

	Looks like max_avail goes below 8 ? What value you see
for such small system?

> Change to clamp(max_avail, 5, 20) which has the expected behaviour.

	It should be clamp(max_avail, 8, 20)

> 
> Replace the clamp_val() on the line below with clamp().
> clamp_val() is just 'an accident waiting to happen' and not needed here.

	OK

> Fixes: 4f325e26277b6
> (Although I actually doubt the code is used on small memory systems.)
> 
> Detected by compile time checks added to clamp(), specifically:
> minmax.h: use BUILD_BUG_ON_MSG() for the lo < hi test in clamp()

	Existing or new check? Does it happen that max_avail
is a constant, so that a compile check triggers?

> 
> Signed-off-by: David Laight <david.laight@aculab.com>

	The code below looks ok to me but can you change the
comments above to more correctly specify the values and if the
problem is that max_avail goes below 8 (min).

> ---
>  net/netfilter/ipvs/ip_vs_conn.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/netfilter/ipvs/ip_vs_conn.c b/net/netfilter/ipvs/ip_vs_conn.c
> index 98d7dbe3d787..c0289f83f96d 100644
> --- a/net/netfilter/ipvs/ip_vs_conn.c
> +++ b/net/netfilter/ipvs/ip_vs_conn.c
> @@ -1495,8 +1495,8 @@ int __init ip_vs_conn_init(void)
>  	max_avail -= 2;		/* ~4 in hash row */
>  	max_avail -= 1;		/* IPVS up to 1/2 of mem */
>  	max_avail -= order_base_2(sizeof(struct ip_vs_conn));

	More likely we can additionally clamp max_avail here:

	max_avail = max(min, max_avail);

	But your solution solves the problem with less lines.

> -	max = clamp(max, min, max_avail);
> -	ip_vs_conn_tab_bits = clamp_val(ip_vs_conn_tab_bits, min, max);
> +	max = clamp(max_avail, min, max);
> +	ip_vs_conn_tab_bits = clamp(ip_vs_conn_tab_bits, min, max);
>  	ip_vs_conn_tab_size = 1 << ip_vs_conn_tab_bits;
>  	ip_vs_conn_tab_mask = ip_vs_conn_tab_size - 1;
>  
> -- 
> 2.17.1

Regards

--
Julian Anastasov <ja@ssi.bg>


  reply	other threads:[~2024-12-06 12:19 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-12-06 10:27 David Laight
2024-12-06 12:19 ` Julian Anastasov [this message]
2024-12-06 13:07   ` David Laight
2024-12-06 16:23     ` Julian Anastasov
2024-12-06 17:53       ` David Laight
2024-12-06 20:58         ` Julian Anastasov

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=5ec10e7c-d050-dab8-1f1b-d0ca2d922eef@ssi.bg \
    --to=ja@ssi.bg \
    --cc=David.Laight@ACULAB.COM \
    --cc=anders.roxell@linaro.org \
    --cc=arnd@arndb.de \
    --cc=dan.carpenter@linaro.org \
    --cc=johannes.berg@intel.com \
    --cc=kees@kernel.org \
    --cc=kernel@jfarr.cc \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=lkft-triage@lists.linaro.org \
    --cc=naresh.kamboju@linaro.org \
    --cc=netdev@vger.kernel.org \
    --cc=netfilter-devel@vger.kernel.org \
    --cc=pablo@netfilter.org \
    --cc=regressions@lists.linux.dev \
    --cc=toke@kernel.org \
    --cc=viro@zeniv.linux.org.uk \
    /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

all inboxes | Powered by JetHome®