From: Radu Nicolau <radu.nicolau@intel.com>
To: zhangsha.zhang@huawei.com, dev@dpdk.org, declan.doherty@intel.com
Cc: jerry.lilijun@huawei.com, zhoujingbin@huawei.com, caihe@huawei.com
Subject: Re: [dpdk-dev] [PATCH v3] bonding: fix the segfault caused by the race condition between master thread and eal-intr-thread
Date: Mon, 4 Sep 2017 12:54:48 +0100 [thread overview]
Message-ID: <eb497e53-e943-fda2-96e0-57a863fcf8f8@intel.com> (raw)
In-Reply-To: <1501063992-10704-1-git-send-email-zhangsha.zhang@huawei.com>
Hi,
Wouldn't be possible to treat the section of code that segfaults as a
critical one, i.e. use the lock/unlock instead of triggering alarms?
On 7/26/2017 11:13 AM, zhangsha.zhang@huawei.com wrote:
> From: Sha Zhang <zhangsha.zhang@huawei.com>
>
> Function slave_configure calls functions bond_ethdev_lsc_event_callback and
> slave_eth_dev->dev_ops->link_update to fix updating slave link status.
> But there is a low probability that process may be crashed if the master
> thread, which create bonding-device, adds the active_slave_count of the
> bond to nozero while the rx_ring or tx_ring of it haven't been created.
>
> This patch moves the functions bond_ethdev_lsc_event_callback and
> slave_eth_dev->dev_ops->link_update to eal-intr-thread to aviod the
> competition.
>
> Fixes: 210903803f6e ("net/bonding: fix updating slave link status")
>
> Signed-off-by: Sha Zhang <zhangsha.zhang@huawei.com>
> ---
> drivers/net/bonding/rte_eth_bond_pmd.c | 58 +++++++++++++++++++++++++++++-----
> 1 file changed, 50 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/net/bonding/rte_eth_bond_pmd.c b/drivers/net/bonding/rte_eth_bond_pmd.c
> index 383e27c..bc0ee7f 100644
> --- a/drivers/net/bonding/rte_eth_bond_pmd.c
> +++ b/drivers/net/bonding/rte_eth_bond_pmd.c
> @@ -53,6 +53,7 @@
>
> #define REORDER_PERIOD_MS 10
> #define DEFAULT_POLLING_INTERVAL_10_MS (10)
> +#define BOND_LSC_DELAY_TIME_US (10 * 1000)
>
> #define HASH_L4_PORTS(h) ((h)->src_port ^ (h)->dst_port)
>
> @@ -1800,14 +1801,6 @@ struct bwg_slave {
> }
> }
>
> - /* If lsc interrupt is set, check initial slave's link status */
> - if (slave_eth_dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC) {
> - slave_eth_dev->dev_ops->link_update(slave_eth_dev, 0);
> - bond_ethdev_lsc_event_callback(slave_eth_dev->data->port_id,
> - RTE_ETH_EVENT_INTR_LSC, &bonded_eth_dev->data->port_id,
> - NULL);
> - }
> -
> return 0;
> }
>
> @@ -1878,6 +1871,51 @@ struct bwg_slave {
> static void
> bond_ethdev_promiscuous_enable(struct rte_eth_dev *eth_dev);
>
> +static void
> +bond_ethdev_slave_lsc_delay(void *cb_arg)
> +{
> + struct rte_eth_dev *bonded_ethdev, *slave_dev;
> + struct bond_dev_private *internals;
> +
> + /* Default value for polling slave found is true as we don't
> + * want todisable the polling thread if we cannot get the lock.
> + */
> + int i = 0;
> +
> + if (!cb_arg)
> + return;
> +
> + bonded_ethdev = (struct rte_eth_dev *)cb_arg;
> + if (!bonded_ethdev->data->dev_started)
> + return;
> +
> + internals = (struct bond_dev_private *)bonded_ethdev->data->dev_private;
> + if (!rte_spinlock_trylock(&internals->lock)) {
> + rte_eal_alarm_set(BOND_LSC_DELAY_TIME_US * 10,
> + bond_ethdev_slave_lsc_delay,
> + (void *)&rte_eth_devices[internals->port_id]);
> + return;
> + }
> +
> + for (i = 0; i < internals->slave_count; i++) {
> + slave_dev = &(rte_eth_devices[internals->slaves[i].port_id]);
> + if (slave_dev->data->dev_conf.intr_conf.lsc != 0) {
> + if (slave_dev->dev_ops &&
> + slave_dev->dev_ops->link_update)
> + slave_dev->dev_ops->link_update(slave_dev, 0);
> + bond_ethdev_lsc_event_callback(
> + internals->slaves[i].port_id,
> + RTE_ETH_EVENT_INTR_LSC,
> + &bonded_ethdev->data->port_id, NULL);
> + }
> + }
> + rte_spinlock_unlock(&internals->lock);
> + RTE_LOG(INFO, EAL,
> + "bond %s(%u): slave num %d, current active slave num %d\n",
> + bonded_ethdev->data->name, bonded_ethdev->data->port_id,
> + internals->slave_count, internals->active_slave_count);
> +}
> +
> static int
> bond_ethdev_start(struct rte_eth_dev *eth_dev)
> {
> @@ -1953,6 +1991,10 @@ struct bwg_slave {
> if (internals->slaves[i].link_status_poll_enabled)
> internals->link_status_polling_enabled = 1;
> }
> +
> + rte_eal_alarm_set(BOND_LSC_DELAY_TIME_US, bond_ethdev_slave_lsc_delay,
> + (void *)&rte_eth_devices[internals->port_id]);
> +
> /* start polling if needed */
> if (internals->link_status_polling_enabled) {
> rte_eal_alarm_set(
next prev parent reply other threads:[~2017-09-04 11:54 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-07-26 10:13 zhangsha.zhang
2017-09-04 11:54 ` Radu Nicolau [this message]
2017-11-29 17:45 ` 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=eb497e53-e943-fda2-96e0-57a863fcf8f8@intel.com \
--to=radu.nicolau@intel.com \
--cc=caihe@huawei.com \
--cc=declan.doherty@intel.com \
--cc=dev@dpdk.org \
--cc=jerry.lilijun@huawei.com \
--cc=zhangsha.zhang@huawei.com \
--cc=zhoujingbin@huawei.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).