From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: linux-kernel@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
stable@vger.kernel.org, Robert Shearman <rshearma@brocade.com>,
Alexander Duyck <alexander.h.duyck@intel.com>,
"David S. Miller" <davem@davemloft.net>
Subject: [PATCH 4.8 34/45] ipv4: Drop suffix update from resize code
Date: Fri, 9 Dec 2016 17:21:03 +0100 [thread overview]
Message-ID: <20161209161756.362121813@linuxfoundation.org> (raw)
In-Reply-To: <20161209161754.912203877@linuxfoundation.org>
4.8-stable review patch. If anyone has any objections, please let me know.
------------------
From: Alexander Duyck <alexander.h.duyck@intel.com>
[ Upstream commit a52ca62c4a6771028da9c1de934cdbcd93d54bb4 ]
It has been reported that update_suffix can be expensive when it is called
on a large node in which most of the suffix lengths are the same. The time
required to add 200K entries had increased from around 3 seconds to almost
49 seconds.
In order to address this we need to move the code for updating the suffix
out of resize and instead just have it handled in the cases where we are
pushing a node that increases the suffix length, or will decrease the
suffix length.
Fixes: 5405afd1a306 ("fib_trie: Add tracking value for suffix length")
Reported-by: Robert Shearman <rshearma@brocade.com>
Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Reviewed-by: Robert Shearman <rshearma@brocade.com>
Tested-by: Robert Shearman <rshearma@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
net/ipv4/fib_trie.c | 42 +++++++++++++++++++++---------------------
1 file changed, 21 insertions(+), 21 deletions(-)
--- a/net/ipv4/fib_trie.c
+++ b/net/ipv4/fib_trie.c
@@ -681,6 +681,13 @@ static unsigned char update_suffix(struc
{
unsigned char slen = tn->pos;
unsigned long stride, i;
+ unsigned char slen_max;
+
+ /* only vector 0 can have a suffix length greater than or equal to
+ * tn->pos + tn->bits, the second highest node will have a suffix
+ * length at most of tn->pos + tn->bits - 1
+ */
+ slen_max = min_t(unsigned char, tn->pos + tn->bits - 1, tn->slen);
/* search though the list of children looking for nodes that might
* have a suffix greater than the one we currently have. This is
@@ -698,12 +705,8 @@ static unsigned char update_suffix(struc
slen = n->slen;
i &= ~(stride - 1);
- /* if slen covers all but the last bit we can stop here
- * there will be nothing longer than that since only node
- * 0 and 1 << (bits - 1) could have that as their suffix
- * length.
- */
- if ((slen + 1) >= (tn->pos + tn->bits))
+ /* stop searching if we have hit the maximum possible value */
+ if (slen >= slen_max)
break;
}
@@ -875,21 +878,7 @@ static struct key_vector *resize(struct
return collapse(t, tn);
/* update parent in case halve failed */
- tp = node_parent(tn);
-
- /* Return if at least one deflate was run */
- if (max_work != MAX_WORK)
- return tp;
-
- /* push the suffix length to the parent node */
- if (tn->slen > tn->pos) {
- unsigned char slen = update_suffix(tn);
-
- if (slen > tp->slen)
- tp->slen = slen;
- }
-
- return tp;
+ return node_parent(tn);
}
static void node_pull_suffix(struct key_vector *tn, unsigned char slen)
@@ -1030,6 +1019,7 @@ static int fib_insert_node(struct trie *
}
/* Case 3: n is NULL, and will just insert a new leaf */
+ node_push_suffix(tp, new->fa_slen);
NODE_INIT_PARENT(l, tp);
put_child_root(tp, key, l);
trie_rebalance(t, tp);
@@ -1472,6 +1462,8 @@ static void fib_remove_alias(struct trie
* out parent suffix lengths as a part of trie_rebalance
*/
if (hlist_empty(&l->leaf)) {
+ if (tp->slen == l->slen)
+ node_pull_suffix(tp, tp->pos);
put_child_root(tp, l->key, NULL);
node_free(l);
trie_rebalance(t, tp);
@@ -1755,6 +1747,10 @@ void fib_table_flush_external(struct fib
if (IS_TRIE(pn))
break;
+ /* update the suffix to address pulled leaves */
+ if (pn->slen > pn->pos)
+ update_suffix(pn);
+
/* resize completed node */
pn = resize(t, pn);
cindex = get_index(pkey, pn);
@@ -1830,6 +1826,10 @@ int fib_table_flush(struct fib_table *tb
if (IS_TRIE(pn))
break;
+ /* update the suffix to address pulled leaves */
+ if (pn->slen > pn->pos)
+ update_suffix(pn);
+
/* resize completed node */
pn = resize(t, pn);
cindex = get_index(pkey, pn);
next prev parent reply other threads:[~2016-12-09 16:24 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <CGME20161209162112epcas5p48e365650e03ecaae64d811516a0f837d@epcas5p4.samsung.com>
2016-12-09 16:20 ` [PATCH 4.8 00/45] 4.8.14-stable review Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 01/45] gro_cells: mark napi struct as not busy poll candidates Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 02/45] virtio-net: add a missing synchronize_net() Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 03/45] net: dsa: b53: Fix VLAN usage and how we treat CPU port Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 04/45] net: check dead netns for peernet2id_alloc() Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 05/45] ip6_tunnel: disable caching when the traffic class is inherited Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 06/45] net: sky2: Fix shutdown crash Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 07/45] af_unix: conditionally use freezable blocking calls in read Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 08/45] rtnetlink: fix FDB size computation Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 09/45] l2tp: fix racy SOCK_ZAPPED flag check in l2tp_ip{,6}_bind() Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 10/45] rtnl: fix the loop index update error in rtnl_dump_ifinfo() Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 11/45] ipv6: bump genid when the IFA_F_TENTATIVE flag is clear Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 12/45] udplite: call proper backlog handlers Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 13/45] net: dsa: bcm_sf2: Ensure we re-negotiate EEE during after link change Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 14/45] net, sched: respect rcu grace period on cls destruction Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 15/45] net: dsa: fix unbalanced dsa_switch_tree reference counting Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 16/45] net/sched: pedit: make sure that offset is valid Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 17/45] netlink: Call cb->done from a worker thread Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 18/45] netlink: Do not schedule work from sk_destruct Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 19/45] net: macb: fix the RX queue reset in macb_rx() Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 20/45] net/dccp: fix use-after-free in dccp_invalid_packet Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 21/45] GSO: Reload iph after pskb_may_pull Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 22/45] packet: fix race condition in packet_set_ring Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 23/45] ip6_offload: check segs for NULL in ipv6_gso_segment Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 25/45] net: bcmgenet: Utilize correct struct device for all DMA operations Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 26/45] sh_eth: remove unchecked interrupts for RZ/A1 Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 28/45] geneve: avoid use-after-free of skb->data Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 29/45] net: avoid signed overflows for SO_{SND|RCV}BUFFORCE Greg Kroah-Hartman
2016-12-09 16:20 ` [PATCH 4.8 30/45] net: ping: check minimum size on ICMP header length Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 31/45] ipv4: Restore fib_trie_flush_external function and fix call ordering Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 32/45] ipv4: Fix memory leak in exception case for splitting tries Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 33/45] ipv4: Drop leaf from suffix pull/push functions Greg Kroah-Hartman
2016-12-09 16:21 ` Greg Kroah-Hartman [this message]
2016-12-09 16:21 ` [PATCH 4.8 35/45] sparc64: Fix find_node warning if numa node cannot be found Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 36/45] sparc64: fix compile warning section mismatch in find_node() Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 37/45] sparc32: Fix inverted invalid_frame_pointer checks on sigreturns Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 38/45] Dont feed anything but regular iovecs to blk_rq_map_user_iov Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 39/45] constify iov_iter_count() and iter_is_iovec() Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 40/45] ipv6: Set skb->protocol properly for local output Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 41/45] ipv4: " Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 42/45] Revert: "ip6_tunnel: Update skb->protocol to ETH_P_IPV6 in ip6_tnl_xmit()" Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 44/45] esp4: Fix integrity verification when ESN are used Greg Kroah-Hartman
2016-12-09 16:21 ` [PATCH 4.8 45/45] esp6: " Greg Kroah-Hartman
2016-12-09 18:24 ` [PATCH 4.8 00/45] 4.8.14-stable review Shuah Khan
2016-12-10 12:17 ` Greg Kroah-Hartman
2016-12-09 22:36 ` Guenter Roeck
2016-12-10 12:17 ` Greg Kroah-Hartman
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=20161209161756.362121813@linuxfoundation.org \
--to=gregkh@linuxfoundation.org \
--cc=alexander.h.duyck@intel.com \
--cc=davem@davemloft.net \
--cc=linux-kernel@vger.kernel.org \
--cc=rshearma@brocade.com \
--cc=stable@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).