From: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
To: Eric Dumazet <eric.dumazet@gmail.com>
Cc: David Miller <davem@davemloft.net>, netdev <netdev@vger.kernel.org>
Subject: Re: [PATCH net-next-2.6] net: build_ehash_secret() and rt_bind_peer() cleanups
Date: Thu, 19 Aug 2010 12:28:05 -0400 [thread overview]
Message-ID: <20100819162805.GC24357@Krystal> (raw)
In-Reply-To: <1282232815.2549.61.camel@edumazet-laptop>
* Eric Dumazet (eric.dumazet@gmail.com) wrote:
> Now cmpxchg() is available on all arches, we can use it in
> build_ehash_secret() and rt_bind_peer() instead of using spinlocks.
>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> CC: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
> ---
> net/ipv4/af_inet.c | 8 +++-----
> net/ipv4/route.c | 11 +++--------
> 2 files changed, 6 insertions(+), 13 deletions(-)
>
> diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
> index 6a1100c..f581f77 100644
> --- a/net/ipv4/af_inet.c
> +++ b/net/ipv4/af_inet.c
> @@ -227,18 +227,16 @@ EXPORT_SYMBOL(inet_ehash_secret);
>
> /*
> * inet_ehash_secret must be set exactly once
> - * Instead of using a dedicated spinlock, we (ab)use inetsw_lock
> */
> void build_ehash_secret(void)
> {
> u32 rnd;
> +
> do {
> get_random_bytes(&rnd, sizeof(rnd));
> } while (rnd == 0);
> - spin_lock_bh(&inetsw_lock);
> - if (!inet_ehash_secret)
> - inet_ehash_secret = rnd;
> - spin_unlock_bh(&inetsw_lock);
> +
> + cmpxchg(&inet_ehash_secret, 0, rnd);
I'd be more comfortable if you add a comment saying why you can get away
with not testing the cmpxchg failure case. (it's because the failure is
caused by a concurrent CPU setting the ehash secret to a random value,
which actually does the job for us, so no need to retry).
> }
> EXPORT_SYMBOL(build_ehash_secret);
>
> diff --git a/net/ipv4/route.c b/net/ipv4/route.c
> index 3f56b6e..ae3dba7 100644
> --- a/net/ipv4/route.c
> +++ b/net/ipv4/route.c
> @@ -1268,18 +1268,13 @@ skip_hashing:
>
> void rt_bind_peer(struct rtable *rt, int create)
> {
> - static DEFINE_SPINLOCK(rt_peer_lock);
> struct inet_peer *peer;
>
> peer = inet_getpeer(rt->rt_dst, create);
Hrm, peer can be NULL on OOM condition.
>
> - spin_lock_bh(&rt_peer_lock);
> - if (rt->peer == NULL) {
> - rt->peer = peer;
> - peer = NULL;
> - }
> - spin_unlock_bh(&rt_peer_lock);
> - if (peer)
> + peer = cmpxchg(&rt->peer, NULL, peer);
> +
> + if (unlikely(peer))
> inet_putpeer(peer);
And you don't want to put the peer returned by cmpxchg, you want to put
the peer you just allocated.
I'd go for:
if (!peer)
return;
if (unlikely(cmpxchg(&rt->peer, NULL, peer) != NULL))
inet_putpeer(peer);
Thanks,
Mathieu
> }
>
>
>
--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
prev parent reply other threads:[~2010-08-19 16:28 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-08-19 15:46 [PATCH net-next-2.6] net: build_ehash_secret() and rt_bind_peer() cleanups Eric Dumazet
2010-08-19 15:58 ` Changli Gao
2010-08-19 16:10 ` [PATCH net-next-2.6 v2] " Eric Dumazet
2010-08-20 7:51 ` David Miller
2010-08-19 16:10 ` [PATCH net-next-2.6] " Changli Gao
2010-08-19 16:13 ` Mathieu Desnoyers
2010-08-19 16:30 ` Mathieu Desnoyers
2010-08-19 16:28 ` Mathieu Desnoyers [this message]
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=20100819162805.GC24357@Krystal \
--to=mathieu.desnoyers@polymtl.ca \
--cc=davem@davemloft.net \
--cc=eric.dumazet@gmail.com \
--cc=netdev@vger.kernel.org \
/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;
as well as URLs for NNTP newsgroup(s).