Skip to contentSkip to conversation

RTNL: assertion failed at net/core/dev.c (4494) and RTNL: assertion failed at net/core/rtnetlink.c (940)

DISCUSSION Thomas Glanzmann 4 participants 9 messages Feb 6, 2014 – Feb 8, 2014

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

Hello,
this morning I checked out Linus tip and compiled it after booting my dmesg is full of:

[    8.944991] RTNL: assertion failed at net/core/dev.c (4494)
[    8.950640] CPU: 3 PID: 388 Comm: kworker/u24:4 Not tainted 3.14.0-rc1+ #3
[    8.950642] Hardware name: Supermicro X9SRD-F/X9SRD-F, BIOS 1.0a 10/15/2012
[    8.950654] Workqueue: bond0 bond_3ad_state_machine_handler [bonding]
[    8.950658]  0000000000000000 ffff881020c88000 ffffffff8138e219 ffff881020c88000
[    8.950664]  ffffffff812d3091 ffff881023961040 ffffffff812e3132 0000000000000246
[    8.950670]  0000000000000020 ffff881020ab1be8 0000000020ab1ba8 0000000000000000
[    8.950675] Call Trace:
[    8.950686]  [<ffffffff8138e219>] ? dump_stack+0x41/0x51
[    8.950694]  [<ffffffff812d3091>] ? netdev_master_upper_dev_get+0x2a/0x4d
[    8.950699]  [<ffffffff812e3132>] ? rtnl_fill_ifinfo+0x2c/0xac4
[    8.950707]  [<ffffffff81072211>] ? print_time.part.5+0x50/0x54
[    8.950715]  [<ffffffff812caf94>] ? __kmalloc_reserve.isra.42+0x2a/0x6d
[    8.950721]  [<ffffffff81102040>] ? ksize+0x12/0x1e
[    8.950726]  [<ffffffff812cb2b7>] ? __alloc_skb+0xb5/0x1a9
[    8.950731]  [<ffffffff812e4626>] ? rtmsg_ifinfo+0x6c/0xd6
[    8.950739]  [<ffffffffa035f4f9>] ? __enable_port.isra.17+0x51/0x5a [bonding]
[    8.950747]  [<ffffffffa0360463>] ? ad_agg_selection_logic+0x3d3/0x3ed [bonding]
[    8.950754]  [<ffffffffa0360d40>] ? bond_3ad_state_machine_handler+0x555/0x918 [bonding]
[    8.950761]  [<ffffffff8104db2d>] ? process_one_work+0x191/0x293
[    8.950766]  [<ffffffff8104dfde>] ? worker_thread+0x121/0x1e7
[    8.950770]  [<ffffffff8104debd>] ? rescuer_thread+0x269/0x269
[    8.950777]  [<ffffffff810527b6>] ? kthread+0x99/0xa1
[    8.950782]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59
[    8.950789]  [<ffffffff8139733c>] ? ret_from_fork+0x7c/0xb0
[    8.950794]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59
[    8.950797] RTNL: assertion failed at net/core/rtnetlink.c (940)
[    8.956863] CPU: 3 PID: 388 Comm: kworker/u24:4 Not tainted 3.14.0-rc1+ #3
[    8.956871] Hardware name: Supermicro X9SRD-F/X9SRD-F, BIOS 1.0a 10/15/2012
[    8.956877] Workqueue: bond0 bond_3ad_state_machine_handler [bonding]
[    8.956879]  0000000000000000 ffff881020c88000 ffffffff8138e219 ffff881023961040
[    8.956884]  ffffffff812e315e 0000000000000246 0000000000000020 ffff881020ab1be8
[    8.956890]  0000000020ab1ba8 0000000000000000 ffffffff817ef530 0000000000000008
[    8.956895] Call Trace:
[    8.956899]  [<ffffffff8138e219>] ? dump_stack+0x41/0x51
[    8.956903]  [<ffffffff812e315e>] ? rtnl_fill_ifinfo+0x58/0xac4
[    8.956909]  [<ffffffff81072211>] ? print_time.part.5+0x50/0x54
[    8.956915]  [<ffffffff812caf94>] ? __kmalloc_reserve.isra.42+0x2a/0x6d
[    8.956920]  [<ffffffff81102040>] ? ksize+0x12/0x1e
[    8.956925]  [<ffffffff812cb2b7>] ? __alloc_skb+0xb5/0x1a9
[    8.956929]  [<ffffffff812e4626>] ? rtmsg_ifinfo+0x6c/0xd6
[    8.956936]  [<ffffffffa035f4f9>] ? __enable_port.isra.17+0x51/0x5a [bonding]
[    8.956942]  [<ffffffffa0360463>] ? ad_agg_selection_logic+0x3d3/0x3ed [bonding]
[    8.956949]  [<ffffffffa0360d40>] ? bond_3ad_state_machine_handler+0x555/0x918 [bonding]
[    8.956954]  [<ffffffff8104db2d>] ? process_one_work+0x191/0x293
[    8.956958]  [<ffffffff8104dfde>] ? worker_thread+0x121/0x1e7
[    8.956962]  [<ffffffff8104debd>] ? rescuer_thread+0x269/0x269
[    8.956968]  [<ffffffff810527b6>] ? kthread+0x99/0xa1
[    8.956973]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59
[    8.956978]  [<ffffffff8139733c>] ? ret_from_fork+0x7c/0xb0
[    8.956983]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59

Full dmesg is at: http://pbot.rmdir.de/dJZsX0d71RFPaZAAjcTBWQ

I'm running Debian Wheezy and my Interface Konfiguration is:

auto lo
iface lo inet loopback

auto bond0
iface bond0 inet static

       address 10.100.4.62
       netmask 255.255.0.0
       gateway 10.100.0.1
       slaves eth0 eth1
       bond-mode 802.3ad
       bond-miimon 100

auto bond0.101
iface bond0.101 inet static

       address 10.101.99.4
       netmask 255.255.0.0

auto bond0.102
iface bond0.102 inet static

       address 10.102.99.4
       netmask 255.255.0.0

auto bond0.103
iface bond0.103 inet static

       address 10.103.99.4
       netmask 255.255.0.0

auto bond1
iface bond1 inet static

       address 10.100.5.62
       netmask 255.255.0.0
       slaves eth2 eth3
       bond-mode 802.3ad
       bond-miimon 100

auto bond1.101
iface bond1.101 inet static

       address 10.101.99.5
       netmask 255.255.0.0

auto bond1.102
iface bond1.102 inet static

       address 10.102.99.5
       netmask 255.255.0.0

auto bond1.103
iface bond1.103 inet static

       address 10.103.99.5
       netmask 255.255.0.0

My IPv4 interface Konfiguration is:

(node-62) [~/work/linux-2.6] ip -4 a s
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN
inet 127.0.0.1/8 scope host lo
valid_lft forever preferred_lft forever
6: bond0: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
inet 10.100.4.62/16 brd 10.100.255.255 scope global bond0
valid_lft forever preferred_lft forever
7: bond0.101@bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
inet 10.101.99.4/16 brd 10.101.255.255 scope global bond0.101
valid_lft forever preferred_lft forever
8: bond0.102@bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
inet 10.102.99.4/16 brd 10.102.255.255 scope global bond0.102
valid_lft forever preferred_lft forever
9: bond0.103@bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
inet 10.103.99.4/16 brd 10.103.255.255 scope global bond0.103
valid_lft forever preferred_lft forever
10: bond1: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
inet 10.100.5.62/16 brd 10.100.255.255 scope global bond1
valid_lft forever preferred_lft forever
11: bond1.101@bond1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
inet 10.101.99.5/16 brd 10.101.255.255 scope global bond1.101
valid_lft forever preferred_lft forever
12: bond1.102@bond1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
inet 10.102.99.5/16 brd 10.102.255.255 scope global bond1.102
valid_lft forever preferred_lft forever
13: bond1.103@bond1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP
inet 10.103.99.5/16 brd 10.103.255.255 scope global bond1.103
valid_lft forever preferred_lft forever

I also have IPv6 with link-local and global addresses configured.

Kernel Config is at: http://pbot.rmdir.de/VTAnhVv8ECP7a7SPaxMsFA

I'm available for testing fixes or providing ssh access to the system if you want to do further tests. I have the same config running on 3.13.0 on the same machine without any problems whatsoever. If I should bisect it, let me know.

Cheers,
Thomas

#1 lore raw
Quoted 22 lines · Thomas Glanzmann

Hello,
this morning I checked out Linus tip and compiled it after booting my dmesg is full of:

[    8.944991] RTNL: assertion failed at net/core/dev.c (4494)
[    8.950640] CPU: 3 PID: 388 Comm: kworker/u24:4 Not tainted 3.14.0-rc1+ #3
[    8.950642] Hardware name: Supermicro X9SRD-F/X9SRD-F, BIOS 1.0a 10/15/2012
[    8.950654] Workqueue: bond0 bond_3ad_state_machine_handler [bonding]
[    8.950658]  0000000000000000 ffff881020c88000 ffffffff8138e219 ffff881020c88000
[    8.950664]  ffffffff812d3091 ffff881023961040 ffffffff812e3132 0000000000000246
[    8.950670]  0000000000000020 ffff881020ab1be8 0000000020ab1ba8 0000000000000000
[    8.950675] Call Trace:
[    8.950686]  [<ffffffff8138e219>] ? dump_stack+0x41/0x51
[    8.950694]  [<ffffffff812d3091>] ? netdev_master_upper_dev_get+0x2a/0x4d
[    8.950699]  [<ffffffff812e3132>] ? rtnl_fill_ifinfo+0x2c/0xac4
[    8.950707]  [<ffffffff81072211>] ? print_time.part.5+0x50/0x54
[    8.950715]  [<ffffffff812caf94>] ? __kmalloc_reserve.isra.42+0x2a/0x6d
[    8.950721]  [<ffffffff81102040>] ? ksize+0x12/0x1e
[    8.950726]  [<ffffffff812cb2b7>] ? __alloc_skb+0xb5/0x1a9
[    8.950731]  [<ffffffff812e4626>] ? rtmsg_ifinfo+0x6c/0xd6
[    8.950739]  [<ffffffffa035f4f9>] ? __enable_port.isra.17+0x51/0x5a [bonding]
[    8.950747]  [<ffffffffa0360463>] ? ad_agg_selection_logic+0x3d3/0x3ed [bonding]
Quoted 8 lines
[    8.950754]  [<ffffffffa0360d40>] ? bond_3ad_state_machine_handler+0x555/0x918 [bonding]
[    8.950761]  [<ffffffff8104db2d>] ? process_one_work+0x191/0x293
[    8.950766]  [<ffffffff8104dfde>] ? worker_thread+0x121/0x1e7
[    8.950770]  [<ffffffff8104debd>] ? rescuer_thread+0x269/0x269
[    8.950777]  [<ffffffff810527b6>] ? kthread+0x99/0xa1
[    8.950782]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59
[    8.950789]  [<ffffffff8139733c>] ? ret_from_fork+0x7c/0xb0
[    8.950794]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59

Hmm, rtmsg_ifinfo() should be called with rtnl lock, but __enable_port() is called
with rcu_read_lock() which means we can't block inside it, therefore we probably should take rtnl lock outside:

drivers/net/bonding/bond_3ad.c +0 -0
index cce1f1b..3c09ffa 100644
@@ -2065,6 +2065,7 @@ void bond_3ad_state_machine_handler(struct

work_struct *work)

        struct slave *slave;
        struct port *port;
+       rtnl_lock();
        read_lock(&bond->lock);
        rcu_read_lock();

@@ -2123,6 +2124,7 @@ void bond_3ad_state_machine_handler(struct work_struct *work)
re_arm:

        rcu_read_unlock();
        read_unlock(&bond->lock);
+       rtnl_unlock();
        queue_delayed_work(bond->wq, &bond->ad_work, ad_delta_in_ticks);

}

#2 lore raw
Quoted 54 lines · Cong Wang
Quoted 22 lines · Thomas Glanzmann

Hello,
this morning I checked out Linus tip and compiled it after booting my dmesg is full of:

[    8.944991] RTNL: assertion failed at net/core/dev.c (4494)
[    8.950640] CPU: 3 PID: 388 Comm: kworker/u24:4 Not tainted 3.14.0-rc1+ #3
[    8.950642] Hardware name: Supermicro X9SRD-F/X9SRD-F, BIOS 1.0a 10/15/2012
[    8.950654] Workqueue: bond0 bond_3ad_state_machine_handler [bonding]
[    8.950658]  0000000000000000 ffff881020c88000 ffffffff8138e219 ffff881020c88000
[    8.950664]  ffffffff812d3091 ffff881023961040 ffffffff812e3132 0000000000000246
[    8.950670]  0000000000000020 ffff881020ab1be8 0000000020ab1ba8 0000000000000000
[    8.950675] Call Trace:
[    8.950686]  [<ffffffff8138e219>] ? dump_stack+0x41/0x51
[    8.950694]  [<ffffffff812d3091>] ? netdev_master_upper_dev_get+0x2a/0x4d
[    8.950699]  [<ffffffff812e3132>] ? rtnl_fill_ifinfo+0x2c/0xac4
[    8.950707]  [<ffffffff81072211>] ? print_time.part.5+0x50/0x54
[    8.950715]  [<ffffffff812caf94>] ? __kmalloc_reserve.isra.42+0x2a/0x6d
[    8.950721]  [<ffffffff81102040>] ? ksize+0x12/0x1e
[    8.950726]  [<ffffffff812cb2b7>] ? __alloc_skb+0xb5/0x1a9
[    8.950731]  [<ffffffff812e4626>] ? rtmsg_ifinfo+0x6c/0xd6
[    8.950739]  [<ffffffffa035f4f9>] ? __enable_port.isra.17+0x51/0x5a [bonding]
[    8.950747]  [<ffffffffa0360463>] ? ad_agg_selection_logic+0x3d3/0x3ed [bonding]
Quoted 8 lines
[    8.950754]  [<ffffffffa0360d40>] ? bond_3ad_state_machine_handler+0x555/0x918 [bonding]
[    8.950761]  [<ffffffff8104db2d>] ? process_one_work+0x191/0x293
[    8.950766]  [<ffffffff8104dfde>] ? worker_thread+0x121/0x1e7
[    8.950770]  [<ffffffff8104debd>] ? rescuer_thread+0x269/0x269
[    8.950777]  [<ffffffff810527b6>] ? kthread+0x99/0xa1
[    8.950782]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59
[    8.950789]  [<ffffffff8139733c>] ? ret_from_fork+0x7c/0xb0
[    8.950794]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59

Hmm, rtmsg_ifinfo() should be called with rtnl lock, but __enable_port() is called
with rcu_read_lock() which means we can't block inside it, therefore we probably should take rtnl lock outside:

drivers/net/bonding/bond_3ad.c +1 -0
index cce1f1b..3c09ffa 100644
@@ -2065,6 +2065,7 @@ void bond_3ad_state_machine_handler(struct
 work_struct *work)
       struct slave *slave;
       struct port *port;
 
+       rtnl_lock();
       read_lock(&bond->lock);
       rcu_read_lock();
@@ -2123,6 +2124,7 @@ void bond_3ad_state_machine_handler(struct
 work_struct *work)

re_arm:

Quoted 5 lines
       rcu_read_unlock();
       read_unlock(&bond->lock);
+       rtnl_unlock();
       queue_delayed_work(bond->wq, &bond->ad_work, ad_delta_in_ticks);
}

That would eliminate the warning, but is suboptimal. Acquiring
RTNL is not necessary on the vast majority of state machine runs (because no state changes take place, i.e., no ports are disabled or enabled). The above change would add 10 round trips per second to RTNL, which seems excessive.

Also, we cannot unconditionally acquire RTNL in this function,
as it would race with the call to cancel_delayed_work_sync from bond_close (via bond_work_cancel_all).

	-J

---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]

#3 lore raw
Quoted 68 lines · Jay Vosburgh
Quoted 54 lines · Cong Wang
Quoted 22 lines · Thomas Glanzmann

Hello,
this morning I checked out Linus tip and compiled it after booting my dmesg is full of:

[    8.944991] RTNL: assertion failed at net/core/dev.c (4494)
[    8.950640] CPU: 3 PID: 388 Comm: kworker/u24:4 Not tainted 3.14.0-rc1+ #3
[    8.950642] Hardware name: Supermicro X9SRD-F/X9SRD-F, BIOS 1.0a 10/15/2012
[    8.950654] Workqueue: bond0 bond_3ad_state_machine_handler [bonding]
[    8.950658]  0000000000000000 ffff881020c88000 ffffffff8138e219 ffff881020c88000
[    8.950664]  ffffffff812d3091 ffff881023961040 ffffffff812e3132 0000000000000246
[    8.950670]  0000000000000020 ffff881020ab1be8 0000000020ab1ba8 0000000000000000
[    8.950675] Call Trace:
[    8.950686]  [<ffffffff8138e219>] ? dump_stack+0x41/0x51
[    8.950694]  [<ffffffff812d3091>] ? netdev_master_upper_dev_get+0x2a/0x4d
[    8.950699]  [<ffffffff812e3132>] ? rtnl_fill_ifinfo+0x2c/0xac4
[    8.950707]  [<ffffffff81072211>] ? print_time.part.5+0x50/0x54
[    8.950715]  [<ffffffff812caf94>] ? __kmalloc_reserve.isra.42+0x2a/0x6d
[    8.950721]  [<ffffffff81102040>] ? ksize+0x12/0x1e
[    8.950726]  [<ffffffff812cb2b7>] ? __alloc_skb+0xb5/0x1a9
[    8.950731]  [<ffffffff812e4626>] ? rtmsg_ifinfo+0x6c/0xd6
[    8.950739]  [<ffffffffa035f4f9>] ? __enable_port.isra.17+0x51/0x5a [bonding]
[    8.950747]  [<ffffffffa0360463>] ? ad_agg_selection_logic+0x3d3/0x3ed [bonding]
Quoted 8 lines
[    8.950754]  [<ffffffffa0360d40>] ? bond_3ad_state_machine_handler+0x555/0x918 [bonding]
[    8.950761]  [<ffffffff8104db2d>] ? process_one_work+0x191/0x293
[    8.950766]  [<ffffffff8104dfde>] ? worker_thread+0x121/0x1e7
[    8.950770]  [<ffffffff8104debd>] ? rescuer_thread+0x269/0x269
[    8.950777]  [<ffffffff810527b6>] ? kthread+0x99/0xa1
[    8.950782]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59
[    8.950789]  [<ffffffff8139733c>] ? ret_from_fork+0x7c/0xb0
[    8.950794]  [<ffffffff8105271d>] ? __kthread_parkme+0x59/0x59

Hmm, rtmsg_ifinfo() should be called with rtnl lock, but __enable_port() is called
with rcu_read_lock() which means we can't block inside it, therefore we probably should take rtnl lock outside:

drivers/net/bonding/bond_3ad.c +1 -0
index cce1f1b..3c09ffa 100644
@@ -2065,6 +2065,7 @@ void bond_3ad_state_machine_handler(struct
 work_struct *work)
       struct slave *slave;
       struct port *port;
 
+       rtnl_lock();
       read_lock(&bond->lock);
       rcu_read_lock();
@@ -2123,6 +2124,7 @@ void bond_3ad_state_machine_handler(struct
 work_struct *work)

re_arm:

Quoted 5 lines
       rcu_read_unlock();
       read_unlock(&bond->lock);
+       rtnl_unlock();
       queue_delayed_work(bond->wq, &bond->ad_work, ad_delta_in_ticks);
}

That would eliminate the warning, but is suboptimal. Acquiring
RTNL is not necessary on the vast majority of state machine runs (because no state changes take place, i.e., no ports are disabled or enabled). The above change would add 10 round trips per second to RTNL, which seems excessive.

Quoted 3 lines

Also, we cannot unconditionally acquire RTNL in this function,
as it would race with the call to cancel_delayed_work_sync from bond_close (via bond_work_cancel_all).

Thought of one more problem: we can't hold a regular lock while
calling rtmsg_ifinfo, as it may sleep in alloc_skb. The rtmsg_ifinfo call has to be RTNL and nothing else.

	-J

---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]

#4 lore raw
Quoted 11 lines · Jay Vosburgh
Quoted 9 lines · Jay Vosburgh

Cong Wang <[email protected]> wrote:

That would eliminate the warning, but is suboptimal. Acquiring
RTNL is not necessary on the vast majority of state machine runs (because no state changes take place, i.e., no ports are disabled or enabled). The above change would add 10 round trips per second to RTNL, which seems excessive.

Quoted 3 lines
Quoted 3 lines

Also, we cannot unconditionally acquire RTNL in this function,
as it would race with the call to cancel_delayed_work_sync from bond_close (via bond_work_cancel_all).

OK.

Quoted 5 lines

Thought of one more problem: we can't hold a regular lock while
calling rtmsg_ifinfo, as it may sleep in alloc_skb. The rtmsg_ifinfo call has to be RTNL and nothing else.

s/GFP_KERNEL/GFP_ATOMIC/

#5 lore raw
Quoted 19 lines · Cong Wang
Quoted 11 lines · Jay Vosburgh
Quoted 9 lines · Jay Vosburgh

Cong Wang <[email protected]> wrote:

That would eliminate the warning, but is suboptimal. Acquiring
RTNL is not necessary on the vast majority of state machine runs (because no state changes take place, i.e., no ports are disabled or enabled). The above change would add 10 round trips per second to RTNL, which seems excessive.

Quoted 3 lines
Quoted 3 lines

Also, we cannot unconditionally acquire RTNL in this function,
as it would race with the call to cancel_delayed_work_sync from bond_close (via bond_work_cancel_all).

OK.

Quoted 1 line
Quoted 6 lines
Quoted 4 lines

Thought of one more problem: we can't hold a regular lock while
calling rtmsg_ifinfo, as it may sleep in alloc_skb. The rtmsg_ifinfo call has to be RTNL and nothing else.

s/GFP_KERNEL/GFP_ATOMIC/

Yah, that would help with extra locks, but not totally solve
things. I'm looking around, and seeing a number of other places that will end up at one of these rtmsg_ifinfo calls with incorrect locking:

bond_ab_arp_probe calls via bond_set_slave_active_flags and
bond_set_slave_inactive_flags without RTNL.

bond_change_active_slave calls via bond_set_slave_inactive_flags
and bond_set_slave_active_flags with other locks held, and maybe without RTNL; I'm not sure if bond_option_active_slave_set holds RTNL when it calls bond_select_active_slave.

bond_open calls via bond_set_slave_active_flags and
bond_set_slave_inactive_flags with RTNL, but also with other locks held.

bond_loadbalance_arp_mon calls bond_set_active_slave and
bond_set_backup_slave without RTNL.

This is in addition to the cases in the 802.3ad code from
__enable_port and __disable_port calls.

	-J

---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]

#6 lore raw
Quoted 41 lines · Jay Vosburgh
Quoted 19 lines · Cong Wang
Quoted 11 lines · Jay Vosburgh
Quoted 9 lines · Jay Vosburgh

Cong Wang <[email protected]> wrote:

That would eliminate the warning, but is suboptimal. Acquiring
RTNL is not necessary on the vast majority of state machine runs (because no state changes take place, i.e., no ports are disabled or enabled). The above change would add 10 round trips per second to RTNL, which seems excessive.

Quoted 3 lines
Quoted 3 lines

Also, we cannot unconditionally acquire RTNL in this function,
as it would race with the call to cancel_delayed_work_sync from bond_close (via bond_work_cancel_all).

OK.

Quoted 1 line
Quoted 6 lines
Quoted 4 lines

Thought of one more problem: we can't hold a regular lock while
calling rtmsg_ifinfo, as it may sleep in alloc_skb. The rtmsg_ifinfo call has to be RTNL and nothing else.

s/GFP_KERNEL/GFP_ATOMIC/

Yah, that would help with extra locks, but not totally solve
things. I'm looking around, and seeing a number of other places that will end up at one of these rtmsg_ifinfo calls with incorrect locking:

bond_ab_arp_probe calls via bond_set_slave_active_flags and
bond_set_slave_inactive_flags without RTNL.

bond_change_active_slave calls via bond_set_slave_inactive_flags
and bond_set_slave_active_flags with other locks held, and maybe without RTNL; I'm not sure if bond_option_active_slave_set holds RTNL when it calls bond_select_active_slave.

Quoted 8 lines

bond_open calls via bond_set_slave_active_flags and
bond_set_slave_inactive_flags with RTNL, but also with other locks held.

bond_loadbalance_arp_mon calls bond_set_active_slave and
bond_set_backup_slave without RTNL.

This is in addition to the cases in the 802.3ad code from
__enable_port and __disable_port calls.

Just an update in case anybody else is looking into this, and
some questions for Scott.

Acquiring RTNL for the __enable_port and __disable_port cases is
difficult, as those calls generally already hold the state machine lock, and cannot unconditionally call rtnl_lock because either they already hold RTNL (for calls via bond_3ad_unbind_slave) or due to the potential for deadlock with bond_3ad_adapter_speed_changed, bond_3ad_adapter_duplex_changed, bond_3ad_link_change, or bond_3ad_update_lacp_rate. All four of those are called with RTNL held, and acquire the state machine lock second. The calling contexts for __enable_port and __disable_port already hold the state machine lock, and may or may not need RTNL.

Scott: you added these calls, so can you explain what they're
for? I'm asking for two reasons:

First, if they do not occur synchronously is it going to be a
problem? E.g., for the 802.3ad case, if the rtmsg_ifinfo is called either at the end of the state machine run, or for non-state machine events, at the next run of the state machine (which is every 100 ms), would that be a problem? Setting a flag in the slave somewhere that an rtmsg_ifinfo is needed should be doable for the 802.3ad case.

Second, what do the messages mean? That the slave is now
"active and usable"? I'm asking because I suspect the bond_ab_arp_probe usage wherein it adjusts the flags and curr_active_slave should not actually call rtmsg_ifinfo, as the slave there is not really "up." What's going on there is that the ARP monitor cycles through each slave one by one, and tests to see if that slave works. If it does work, then it is set as the active elsewhere in the monitor code. This function adjusts the flags so that the ARP monitor will treat the "testing" slave as "active" for purposes of determining whether or not it is up. I suspect this adjustment to the flags should not actually generate an rtmsg_ifinfo.

I think the remaining cases can be dealt with, but clarification
on the above two questions would be very helpful.

	-J

---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]

#7 lore raw
Quoted 55 lines · Jay Vosburgh
Quoted 41 lines · Jay Vosburgh
Quoted 19 lines · Cong Wang
Quoted 11 lines · Jay Vosburgh
Quoted 9 lines · Jay Vosburgh

Cong Wang <[email protected]> wrote:

That would eliminate the warning, but is suboptimal. Acquiring
RTNL is not necessary on the vast majority of state machine runs (because no state changes take place, i.e., no ports are disabled or enabled). The above change would add 10 round trips per second to RTNL, which seems excessive.

Quoted 3 lines
Quoted 3 lines

Also, we cannot unconditionally acquire RTNL in this function,
as it would race with the call to cancel_delayed_work_sync from bond_close (via bond_work_cancel_all).

OK.

Quoted 1 line
Quoted 6 lines
Quoted 4 lines

Thought of one more problem: we can't hold a regular lock while
calling rtmsg_ifinfo, as it may sleep in alloc_skb. The rtmsg_ifinfo call has to be RTNL and nothing else.

s/GFP_KERNEL/GFP_ATOMIC/

Yah, that would help with extra locks, but not totally solve
things. I'm looking around, and seeing a number of other places that will end up at one of these rtmsg_ifinfo calls with incorrect locking:

bond_ab_arp_probe calls via bond_set_slave_active_flags and
bond_set_slave_inactive_flags without RTNL.

bond_change_active_slave calls via bond_set_slave_inactive_flags
and bond_set_slave_active_flags with other locks held, and maybe without RTNL; I'm not sure if bond_option_active_slave_set holds RTNL when it calls bond_select_active_slave.

Quoted 8 lines

bond_open calls via bond_set_slave_active_flags and
bond_set_slave_inactive_flags with RTNL, but also with other locks held.

bond_loadbalance_arp_mon calls bond_set_active_slave and
bond_set_backup_slave without RTNL.

This is in addition to the cases in the 802.3ad code from
__enable_port and __disable_port calls.

Just an update in case anybody else is looking into this, and
some questions for Scott.

Quoted 10 lines

Acquiring RTNL for the __enable_port and __disable_port cases is
difficult, as those calls generally already hold the state machine lock, and cannot unconditionally call rtnl_lock because either they already hold RTNL (for calls via bond_3ad_unbind_slave) or due to the potential for deadlock with bond_3ad_adapter_speed_changed, bond_3ad_adapter_duplex_changed, bond_3ad_link_change, or bond_3ad_update_lacp_rate. All four of those are called with RTNL held, and acquire the state machine lock second. The calling contexts for __enable_port and __disable_port already hold the state machine lock, and may or may not need RTNL.

Agree, it is hard to add RTNL here, deadlock is easily happened.

Quoted 23 lines

Scott: you added these calls, so can you explain what they're
for? I'm asking for two reasons:

First, if they do not occur synchronously is it going to be a
problem? E.g., for the 802.3ad case, if the rtmsg_ifinfo is called either at the end of the state machine run, or for non-state machine events, at the next run of the state machine (which is every 100 ms), would that be a problem? Setting a flag in the slave somewhere that an rtmsg_ifinfo is needed should be doable for the 802.3ad case.

Second, what do the messages mean? That the slave is now
"active and usable"? I'm asking because I suspect the bond_ab_arp_probe usage wherein it adjusts the flags and curr_active_slave should not actually call rtmsg_ifinfo, as the slave there is not really "up." What's going on there is that the ARP monitor cycles through each slave one by one, and tests to see if that slave works. If it does work, then it is set as the active elsewhere in the monitor code. This function adjusts the flags so that the ARP monitor will treat the "testing" slave as "active" for purposes of determining whether or not it is up. I suspect this adjustment to the flags should not actually generate an rtmsg_ifinfo.

Quoted 5 lines

I think the remaining cases can be dealt with, but clarification
on the above two questions would be very helpful.

	-J

commit 6fde8f037e604e05df1529 fix the problem for bond_loadbalance_arp_mon(),
and commit 66dd1c077a3f3c130d1 fix the problem for bond_activebackup_arp_mon(),
but we still miss the 3ad monitor, I think if the slave should send the message
by netlink, it is better to refer to fdb_notify() for bridge,I doubts that why we need to send so many message, just slave info is enough, then RTNL is not needed here.

Ding

Quoted 10 lines

---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]

--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

.

#8 lore raw
In reply to Thomas Glanzmann
Quoted 15 lines · Thomas Glanzmann

Hello,
this morning I checked out Linus tip and compiled it after booting my dmesg is full of:

[    8.944991] RTNL: assertion failed at net/core/dev.c (4494)
[    8.950640] CPU: 3 PID: 388 Comm: kworker/u24:4 Not tainted 3.14.0-rc1+ #3
[    8.950642] Hardware name: Supermicro X9SRD-F/X9SRD-F, BIOS 1.0a 10/15/2012
[    8.950654] Workqueue: bond0 bond_3ad_state_machine_handler [bonding]
[    8.950658]  0000000000000000 ffff881020c88000 ffffffff8138e219 ffff881020c88000
[    8.950664]  ffffffff812d3091 ffff881023961040 ffffffff812e3132 0000000000000246
[    8.950670]  0000000000000020 ffff881020ab1be8 0000000020ab1ba8 0000000000000000
[    8.950675] Call Trace:
[    8.950686]  [<ffffffff8138e219>] ? dump_stack+0x41/0x51
[    8.950694]  [<ffffffff812d3091>] ? netdev_master_upper_dev_get+0x2a/0x4d
[    8.950699]  [<ffffffff812e3132>] ? rtnl_fill_ifinfo+0x2c/0xac4
Quoted 8 lines
[    8.950707]  [<ffffffff81072211>] ? print_time.part.5+0x50/0x54
[    8.950715]  [<ffffffff812caf94>] ? __kmalloc_reserve.isra.42+0x2a/0x6d
[    8.950721]  [<ffffffff81102040>] ? ksize+0x12/0x1e
[    8.950726]  [<ffffffff812cb2b7>] ? __alloc_skb+0xb5/0x1a9
[    8.950731]  [<ffffffff812e4626>] ? rtmsg_ifinfo+0x6c/0xd6
[    8.950739]  [<ffffffffa035f4f9>] ? __enable_port.isra.17+0x51/0x5a [bonding]
[    8.950747]  [<ffffffffa0360463>] ? ad_agg_selection_logic+0x3d3/0x3ed [bonding]
[    8.950754]  [<ffffffffa0360d40>] ? bond_3ad_state_machine_handler+0x555/0x918 [bonding]

This was apparently introduced by commit:

commit 1d3ee88ae0d605629bf369ab0b868dae8ca62a48

Author: [email protected] <[email protected]>
Date:   Thu Jan 16 22:57:56 2014 -0800

bonding: add netlink attributes to slave link dev

which adds an "rtmsg_ifinfo" call to bond_set_active_slave and
bond_set_backup_slave, both of which are invoked in several places in the 802.3ad code without RTNL held (via __enable_port and __disable_port).

	Still looking into a fix.

	-J

---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]