From: Eric Dumazet <dada1@cosmosbay.com>
To: "David S. Miller" <davem@davemloft.net>
Cc: netdev@oss.sgi.com
Subject: [BUG] overflow in net/ipv4/route.c rt_check_expire()
Date: Wed, 16 Mar 2005 11:47:34 +0100 [thread overview]
Message-ID: <42380EC6.60100@cosmosbay.com> (raw)
In-Reply-To: <20050315103253.590c8bfc.davem@davemloft.net>
After having hard times tuning /proc/sys/net/ipv4/route/* values, with various crashes of production machines,
I did some investigations...
The rt_check_expire() has a serious problem on machines with large route caches, and a standard HZ value of 1000.
With default values, ie ip_rt_gc_interval = 60*HZ = 60000 ;
the loop count :
for (t = ip_rt_gc_interval << rt_hash_log; t >= 0;
overflows (t is a 31 bit value) as soon rt_hash_log is >= 16 (65536 slots in route cache hash table)
Another problem is the fact that this function has close to 0 effect, because even if ip_rt_gc_interval is changed to 1 HZ,
only 1/300 of the table is scanned every second. And the loop breaks as soon a jiffie is consumed.
We should adapt the loop count based on the actual number of entries in the route cache,
and eventually give more 'jiffies' in some pressure cases.
I am experimenting some changes that I will share when ready.
Thank you
Eric Dumazet
next prev parent reply other threads:[~2005-03-16 10:47 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-03-15 16:13 [PATCH] no more rwlock_t inside tcp_ehash_bucket Eric Dumazet
2005-03-15 18:32 ` David S. Miller
2005-03-16 10:47 ` Eric Dumazet [this message]
2005-03-16 18:05 ` [PATCH] reduce sizeof(struct inet_peer) from 128 to 64 bytes on 64bits architectures Eric Dumazet
2005-03-16 22:10 ` David S. Miller
2005-03-16 22:09 ` [BUG] overflow in net/ipv4/route.c rt_check_expire() David S. Miller
2005-03-17 19:52 ` Eric Dumazet
2005-04-01 6:13 ` David S. Miller
2005-04-01 14:39 ` Eric Dumazet
2005-04-01 15:53 ` Robert Olsson
2005-04-01 16:34 ` Eric Dumazet
2005-04-01 17:26 ` Robert Olsson
2005-04-01 20:28 ` David S. Miller
2005-04-01 21:05 ` Eric Dumazet
2005-04-01 21:08 ` David S. Miller
2005-04-01 21:43 ` Eric Dumazet
2005-04-01 22:34 ` David S. Miller
2005-04-01 23:21 ` Eric Dumazet
2005-04-01 23:54 ` David S. Miller
2005-04-02 8:21 ` Herbert Xu
2005-04-02 9:21 ` Eric Dumazet
2005-04-02 11:23 ` Get rid of rt_check_expire and rt_garbage_collect Herbert Xu
2005-04-02 13:58 ` Eric Dumazet
2005-04-02 14:03 ` Robert Olsson
2005-04-02 21:05 ` jamal
2005-04-03 7:38 ` Herbert Xu
2005-04-03 7:41 ` Herbert Xu
2005-04-02 13:48 ` [BUG] overflow in net/ipv4/route.c rt_check_expire() Robert Olsson
2005-04-02 14:10 ` Eric Dumazet
2005-04-02 14:46 ` Robert Olsson
2005-04-02 20:47 ` jamal
2005-04-02 19:32 ` Herbert Xu
2005-04-02 19:55 ` David S. Miller
2005-04-03 7:43 ` Herbert Xu
2005-04-03 19:57 ` Robert Olsson
2005-04-03 21:45 ` Herbert Xu
2005-04-04 10:27 ` Robert Olsson
2005-04-04 10:38 ` Herbert Xu
2005-04-04 12:29 ` Robert Olsson
2005-04-03 19:36 ` Robert Olsson
2005-04-03 21:43 ` Herbert Xu
2005-04-04 10:38 ` Robert Olsson
2005-04-04 10:48 ` Herbert Xu
2005-04-04 13:17 ` Robert Olsson
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=42380EC6.60100@cosmosbay.com \
--to=dada1@cosmosbay.com \
--cc=davem@davemloft.net \
--cc=netdev@oss.sgi.com \
/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).