From: Hyong Youb Kim <hyonkim@cisco.com>
To: Ferruh Yigit <ferruh.yigit@intel.com>,
Declan Doherty <declan.doherty@intel.com>,
Chas Williams <chas3@att.com>
Cc: dev@dpdk.org, Hyong Youb Kim <hyonkim@cisco.com>, stable@dpdk.org
Subject: [dpdk-dev] [PATCH 2/2] net/bonding: avoid the next active slave going out of bound
Date: Thu, 10 Jan 2019 02:22:35 -0800 [thread overview]
Message-ID: <20190110102235.1238-3-hyonkim@cisco.com> (raw)
In-Reply-To: <20190110102235.1238-1-hyonkim@cisco.com>
For bonding modes like broadcast that use bond_ethdev_rx_burst(), it
is fairly easy to produce a crash simply by bringing a slave port's
link down. When slave links go down, the driver on one thread reduces
active_slave_count via the LSC callback and deactivate_slave(). At the
same time, bond_ethdev_rx_burst() running on a forwarding thread may
increment active_slave (next active slave) beyond
active_slave_count. Here is a typical sequence of events.
At time 0:
active_slave_count = 3
active_slave = 2
At time 1:
A slave link goes down.
Thread 0 (main) reduces active_slave_count to 2.
At time 2:
Thread 1 (forwarding) executes bond_ethdev_rx_burst().
- Reads active_slave_count = 2.
- Increments active_slave at the end to 3.
>From this point on, everytime bond_ethdev_rx_burst() runs,
active_slave increments by one, eventually going well out of bound of
the active_slaves array and causing a crash.
Make the rx burst function to first check that active_slave is within
bound. If not, reset it to 0 to avoid out-of-range array access.
Fixes: e1110e977648 ("net/bonding: fix Rx slave fairness")
Cc: stable@dpdk.org
Signed-off-by: Hyong Youb Kim <hyonkim@cisco.com>
---
drivers/net/bonding/rte_eth_bond_pmd.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)
diff --git a/drivers/net/bonding/rte_eth_bond_pmd.c b/drivers/net/bonding/rte_eth_bond_pmd.c
index daf2440cd..bc2405e54 100644
--- a/drivers/net/bonding/rte_eth_bond_pmd.c
+++ b/drivers/net/bonding/rte_eth_bond_pmd.c
@@ -68,6 +68,15 @@ bond_ethdev_rx_burst(void *queue, struct rte_mbuf **bufs, uint16_t nb_pkts)
internals = bd_rx_q->dev_private;
slave_count = internals->active_slave_count;
active_slave = internals->active_slave;
+ /*
+ * Reset the active slave index, in case active_slave goes out
+ * of bound. It can hapen when slave links go down, and
+ * another thread (LSC callback) shrinks the slave count.
+ */
+ if (active_slave >= slave_count) {
+ internals->active_slave = 0;
+ active_slave = 0;
+ }
for (i = 0; i < slave_count && nb_pkts; i++) {
uint16_t num_rx_slave;
@@ -273,6 +282,11 @@ bond_ethdev_rx_burst_8023ad_fast_queue(void *queue, struct rte_mbuf **bufs,
active_slave = internals->active_slave;
memcpy(slaves, internals->active_slaves,
sizeof(internals->active_slaves[0]) * slave_count);
+ /* active_slave may go out of bound. See bond_ethdev_rx_burst() */
+ if (active_slave >= slave_count) {
+ internals->active_slave = 0;
+ active_slave = 0;
+ }
for (i = 0; i < slave_count && nb_pkts; i++) {
uint16_t num_rx_slave;
--
2.16.2
next prev parent reply other threads:[~2019-01-10 10:23 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-01-10 10:22 [dpdk-dev] [PATCH 0/2] net/bonding: a couple minor fixes Hyong Youb Kim
2019-01-10 10:22 ` [dpdk-dev] [PATCH 1/2] net/bonding: do not set promisc on non-existent primary port Hyong Youb Kim
2019-02-09 13:16 ` Chas Williams
2019-02-21 14:38 ` Ferruh Yigit
2019-01-10 10:22 ` Hyong Youb Kim [this message]
2019-02-09 13:17 ` [dpdk-dev] [PATCH 2/2] net/bonding: avoid the next active slave going out of bound Chas Williams
[not found] ` <7AE31235A30B41498D1C31348DC858BD5B5329AB@IRSMSX103.ger.corp.intel.com>
2019-02-18 15:25 ` Ferruh Yigit
2019-02-20 16:28 ` [dpdk-dev] [dpdk-stable] " Ferruh Yigit
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=20190110102235.1238-3-hyonkim@cisco.com \
--to=hyonkim@cisco.com \
--cc=chas3@att.com \
--cc=declan.doherty@intel.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=stable@dpdk.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).