From: zhuyj <zyjzyj2000@gmail•com>
To: Jay Vosburgh <jay.vosburgh@canonical•com>
Cc: mkubecek@suse•cz, vfalico@gmail•com, gospo@cumulusnetworks•com,
netdev@vger•kernel.org, boris.shteinbock@windriver•com,
emil.s.tantilov@intel•com
Subject: Re: [PATCH 1/1] bonding: Use notifiers for slave link state detection
Date: Tue, 26 Jan 2016 11:19:45 +0800 [thread overview]
Message-ID: <56A6E5D1.7070807@gmail.com> (raw)
In-Reply-To: <7578.1453769012@famine>
On 01/26/2016 08:43 AM, Jay Vosburgh wrote:
> <zyjzyj2000@gmail•com> wrote:
>
>> From: Zhu Yanjun <zyjzyj2000@gmail•com>
>>
>> Bonding will utilize notifier callbacks to detect slave
>> link state changes. It is intended to be used with miimon
>> set to zero, and does not support the updelay or downdelay
>> options to bonding.
>>
>> Because of link flap from the slave interface, if the notifier
>> is NETDEV_UP while the actual link state is down, it is not
>> necessary to continue.
>>
>> Signed-off-by: Jay Vosburgh <jay.vosburgh@canonical•com>
> I haven't signed off on this patch.
>
> I've just started some testing, but as before immediately get an
> RCU warning; it looks to be coming from bond_miimon_inspect_slave();
>
> [ 316.473050] bond1: Enslaving eth1 as a backup interface with an up link
> [ 316.473059]
> [ 316.473806] ===============================
> [ 316.475630] [ INFO: suspicious RCU usage. ]
> [ 316.477519] 4.4.0+ #38 Not tainted
> [ 316.479094] -------------------------------
> [ 316.480765] drivers/net/bonding/bond_main.c:2024 suspicious rcu_dereference_check() usage!
>
> This is presumably because the "case NETDEV_DOWN" call to
> bond_miimon_inspect_slave does not hold RCU. It does hold RTNL, though,
> which should be safe for this usage (RTNL mutexes changes to the active
> slave). The appended patch on top of the original makes the warning go
> away.
>
> I'm still testing the patch and have no comment about its
> functionality as yet.
>
> diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
> index 9f67948..e3faee9 100644
> --- a/drivers/net/bonding/bond_main.c
> +++ b/drivers/net/bonding/bond_main.c
> @@ -2014,14 +2014,14 @@ static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *in
>
> /*-------------------------------- Monitoring -------------------------------*/
>
> -/* called with rcu_read_lock() */
> +/* called with rcu_read_lock() or RTNL */
> static int bond_miimon_inspect_slave(struct bonding *bond, struct slave *slave,
> unsigned long event)
> {
> int link_state;
> bool ignore_updelay;
>
> - ignore_updelay = !rcu_dereference(bond->curr_active_slave);
> + ignore_updelay = !rcu_dereference_rtnl(bond->curr_active_slave);
Thanks a lot.
Because kernel v4.4 needs this kind of patch, I backport this patch from
net-next to kernel v4.4.
If it is not appropriate, I will revert this patch.
Best Regards!
Zhu Yanjun
>
> slave->new_link = BOND_LINK_NOCHANGE;
>
>
> -J
>
> ---
> -Jay Vosburgh, jay.vosburgh@canonical•com
next prev parent reply other threads:[~2016-01-26 3:19 UTC|newest]
Thread overview: 52+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-12-17 8:03 [PATCH 1/1] bonding: restrict up state in 802.3ad mode zyjzyj2000
2015-12-17 21:57 ` Jay Vosburgh
2015-12-18 4:36 ` zyjzyj2000
2015-12-18 4:36 ` [PATCH 1/1] bonding: delay up state without speed and duplex " zyjzyj2000
2015-12-18 4:54 ` Jay Vosburgh
2015-12-18 13:37 ` Sergei Shtylyov
2015-12-28 8:43 ` [PATCH 1/1] bonding: restrict up state " Michal Kubecek
2015-12-28 9:19 ` zhuyj
2016-01-06 1:26 ` Tantilov, Emil S
2016-01-06 3:05 ` zhuyj
2016-01-07 2:43 ` Tantilov, Emil S
2016-01-07 3:33 ` zhuyj
2016-01-07 5:02 ` Tantilov, Emil S
2016-01-07 6:15 ` zyjzyj2000
2016-01-07 6:22 ` zhuyj
2016-01-07 6:33 ` Jay Vosburgh
2016-01-07 15:27 ` Tantilov, Emil S
2016-01-08 1:28 ` [RFC PATCH net-next] bonding: Use notifiers for slave link state detection Jay Vosburgh
2016-01-08 4:36 ` zhuyj
2016-01-08 6:12 ` Jay Vosburgh
2016-01-08 7:41 ` (unknown), zyjzyj2000
2016-01-08 7:41 ` [PATCH 1/1] bonding: utilize notifier callbacks to detect slave link state changes zyjzyj2000
2016-01-08 10:18 ` zhuyj
2016-01-09 1:35 ` [RFC PATCH net-next] bonding: Use notifiers for slave link state detection Tantilov, Emil S
2016-01-09 2:19 ` Jay Vosburgh
2016-01-11 9:03 ` zhuyj
2016-01-13 2:54 ` zhuyj
2016-01-13 17:03 ` Tantilov, Emil S
2016-01-20 5:13 ` [PATCH 1/1] " zyjzyj2000
2016-01-20 5:13 ` zyjzyj2000
2016-01-21 10:16 ` zyjzyj2000
2016-01-21 10:16 ` zyjzyj2000
2016-01-25 16:37 ` Tantilov, Emil S
2016-01-26 0:43 ` Jay Vosburgh
2016-01-26 3:19 ` zhuyj [this message]
2016-01-26 6:00 ` Jay Vosburgh
2016-01-26 6:26 ` zhuyj
2016-01-26 6:45 ` zhuyj
2016-01-27 20:00 ` Tantilov, Emil S
2016-01-28 8:44 ` zyjzyj2000
2016-01-29 7:05 ` zhuyj
2016-01-25 16:33 ` Tantilov, Emil S
2016-01-25 18:00 ` David Miller
2016-01-25 18:37 ` Tantilov, Emil S
2016-01-08 2:29 ` [PATCH 1/1] bonding: restrict up state in 802.3ad mode zhuyj
2016-01-07 6:53 ` Michal Kubecek
2016-01-07 7:37 ` zhuyj
2016-01-07 7:59 ` Michal Kubecek
2016-01-07 8:35 ` zhuyj
2016-01-07 7:47 ` zhuyj
2016-01-07 18:28 ` Tantilov, Emil S
2016-01-08 6:09 ` zhuyj
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=56A6E5D1.7070807@gmail.com \
--to=zyjzyj2000@gmail$(echo .)com \
--cc=boris.shteinbock@windriver$(echo .)com \
--cc=emil.s.tantilov@intel$(echo .)com \
--cc=gospo@cumulusnetworks$(echo .)com \
--cc=jay.vosburgh@canonical$(echo .)com \
--cc=mkubecek@suse$(echo .)cz \
--cc=netdev@vger$(echo .)kernel.org \
--cc=vfalico@gmail$(echo .)com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
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