netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Eric Dumazet <dada1@cosmosbay.com>
To: David Miller <davem@davemloft.net>, Dipankar Sarma <dipankar@in.ibm.com>
Cc: netdev@vger.kernel.org
Subject: Re: [IPV4] ROUTE: ip_rt_dump() is unecessary slow
Date: Tue, 08 Jan 2008 07:37:43 +0100	[thread overview]
Message-ID: <47831A37.5000607@cosmosbay.com> (raw)
In-Reply-To: <20080107.221537.178212902.davem@davemloft.net>

[-- Attachment #1: Type: text/plain, Size: 1581 bytes --]

David Miller a écrit :
> From: Eric Dumazet <dada1@cosmosbay.com>
> Date: Tue, 08 Jan 2008 07:11:30 +0100
> 
>> @@ -288,15 +288,15 @@ static struct rtable *rt_cache_get_first(struct seq_file *seq)
>>  
>>  static struct rtable *rt_cache_get_next(struct seq_file *seq, struct rtable *r)
>>  {
>> -	struct rt_cache_iter_state *st = rcu_dereference(seq->private);
>> +	struct rt_cache_iter_state *st = seq->private;
>>  
> 
> Can you explain to me why this rcu_dereference() can be removed?

Very good question, but honestly I really dont see why it was there at the 
first place :

"struct seq_file" is private to this thread, so seq.private is
also private and cannot change while this thread runs rt_cache_get_next().

Reading it once (as guaranted bu rcu_dereference()) or several time if 
compiler really is dumb enough wont change the result...

> 
> The rest of your patch is OK and once I understand the above
> I'll add it to net-2.6, thanks!
> 

Maybe we can first have an Ack from Dipankar Sarma, I can repost the patch
with a more precise ChangeLog and wait his answer ?

[NET] ROUTE: fix rcu_dereference() uses in /proc/net/rt_cache

In rt_cache_get_next(), no need to guard seq->private by a rcu_dereference()
since seq is private to the thread running this function. Reading seq.private
once (as guaranted bu rcu_dereference()) or several time if compiler really is 
dumb enough wont change the result.

But we miss real spots where rcu_dereference() are needed, both in 
rt_cache_get_first() and rt_cache_get_next()

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>


[-- Attachment #2: route_rcu.patch --]
[-- Type: text/plain, Size: 1011 bytes --]

diff --git a/net/ipv4/route.c b/net/ipv4/route.c
index d337706..3b7562f 100644
--- a/net/ipv4/route.c
+++ b/net/ipv4/route.c
@@ -278,7 +278,7 @@ static struct rtable *rt_cache_get_first(struct seq_file *seq)
 
 	for (st->bucket = rt_hash_mask; st->bucket >= 0; --st->bucket) {
 		rcu_read_lock_bh();
-		r = rt_hash_table[st->bucket].chain;
+		r = rcu_dereference(rt_hash_table[st->bucket].chain);
 		if (r)
 			break;
 		rcu_read_unlock_bh();
@@ -288,15 +288,15 @@ static struct rtable *rt_cache_get_first(struct seq_file *seq)
 
 static struct rtable *rt_cache_get_next(struct seq_file *seq, struct rtable *r)
 {
-	struct rt_cache_iter_state *st = rcu_dereference(seq->private);
+	struct rt_cache_iter_state *st = seq->private;
 
-	r = r->u.dst.rt_next;
+	r = rcu_dereference(r->u.dst.rt_next);
 	while (!r) {
 		rcu_read_unlock_bh();
 		if (--st->bucket < 0)
 			break;
 		rcu_read_lock_bh();
-		r = rt_hash_table[st->bucket].chain;
+		r = rcu_dereference(rt_hash_table[st->bucket].chain);
 	}
 	return r;
 }

  reply	other threads:[~2008-01-08  6:38 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-01-07 18:30 [IPV4] ROUTE: ip_rt_dump() is unecessary slow Eric Dumazet
2008-01-08  5:52 ` David Miller
2008-01-08  6:11   ` Eric Dumazet
2008-01-08  6:15     ` David Miller
2008-01-08  6:37       ` Eric Dumazet [this message]
2008-01-09  6:02         ` Herbert Xu
2008-01-09  7:38           ` [NET] ROUTE: fix rcu_dereference() uses in /proc/net/rt_cache Eric Dumazet
2008-01-09  9:46             ` Herbert Xu
2008-01-09 10:37               ` Eric Dumazet
2008-01-09 14:22                 ` Paul E. McKenney
2008-01-09 14:31                   ` David Miller
2008-01-09 14:43                     ` Paul E. McKenney
2008-01-09 17:36                     ` Eric Dumazet
2008-01-10 11:56                 ` David Miller
2008-01-10 14:06                 ` [DECNET] ROUTE: fix rcu_dereference() uses in /proc/net/decnet_cache Eric Dumazet
2008-01-11  6:35                   ` David Miller
2008-01-10 23:10                 ` [NET] ROUTE: fix rcu_dereference() uses in /proc/net/rt_cache Jarek Poplawski
2008-01-10 23:51                   ` Paul E. McKenney
2008-01-11 14:13                     ` Jarek Poplawski
2008-01-11  0:00                   ` Herbert Xu
2008-01-11  8:30                     ` Jarek Poplawski
2008-01-11  9:11                       ` Jarek Poplawski
2008-01-11  9:23                         ` Jarek Poplawski
2008-01-11 10:38                         ` Herbert Xu
2008-01-11 11:30                           ` Jarek Poplawski
2008-01-11 10:37                       ` Herbert Xu
2008-01-11 12:31                         ` Jarek Poplawski

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=47831A37.5000607@cosmosbay.com \
    --to=dada1@cosmosbay.com \
    --cc=davem@davemloft.net \
    --cc=dipankar@in.ibm.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).