Skip to contentSkip to conversation

ip6_vti: fix memleak on netns dismantle

PATCH Applied Florian Westphal 5 participants 6 messages Apr 15, 2024 – Apr 24, 2024

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ip6_vti: fix memleak on netns dismantle

kmemleak reports net_device resources are no longer released, restore needs_free_netdev toggle. Sample backtrace:

unreferenced object 0xffff88810874f000 (size 4096): [..]

    [<00000000a2b8af8b>] __kmalloc_node+0x209/0x290
    [<0000000040b0a1a9>] alloc_netdev_mqs+0x58/0x470
    [<00000000b4be1e78>] vti6_init_net+0x94/0x230
    [<000000008830c1ea>] ops_init+0x32/0xc0
    [<000000006a26fa8f>] setup_net+0x134/0x2e0

[..]

Cc Breno Leitao <[email protected]>
Fixes a9b2d55a8f1e ("ip6_vti: Do not use custom stat allocator")
---
 net/ipv6/ip6_vti.c | 1 +
 1 file changed, 1 insertion(+)
net/ipv6/ip6_vti.c +1 -0
index 4d68a0777b0c..78344cf3867e 100644
@@ -901,6 +901,7 @@ static void vti6_dev_setup(struct net_device *dev)
 {
 	dev->netdev_ops = &vti6_netdev_ops;
 	dev->header_ops = &ip_tunnel_header_ops;
+	dev->needs_free_netdev = true;
 
 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
 	dev->type = ARPHRD_TUNNEL6;
Signature · 1 line
2.43.2
reviewer Reviewed-by on v1 1/1
#1 lore raw

+ Steffen Klassert, Herbert Xu, David Ahern

Quoted 9 lines · Florian Westphal

kmemleak reports net_device resources are no longer released, restore needs_free_netdev toggle. Sample backtrace:

unreferenced object 0xffff88810874f000 (size 4096): [..]

    [<00000000a2b8af8b>] __kmalloc_node+0x209/0x290
    [<0000000040b0a1a9>] alloc_netdev_mqs+0x58/0x470
    [<00000000b4be1e78>] vti6_init_net+0x94/0x230
    [<000000008830c1ea>] ops_init+0x32/0xc0
    [<000000006a26fa8f>] setup_net+0x134/0x2e0
Quoted 5 lines

[..]

Cc Breno Leitao <[email protected]>
Fixes a9b2d55a8f1e ("ip6_vti: Do not use custom stat allocator")
Quoted 20 lines
---
 net/ipv6/ip6_vti.c | 1 +
 1 file changed, 1 insertion(+)
net/ipv6/ip6_vti.c +1 -0
index 4d68a0777b0c..78344cf3867e 100644
@@ -901,6 +901,7 @@ static void vti6_dev_setup(struct net_device *dev)
 {
 	dev->netdev_ops = &vti6_netdev_ops;
 	dev->header_ops = &ip_tunnel_header_ops;
+	dev->needs_free_netdev = true;
 
 	dev->pcpu_stat_type = NETDEV_PCPU_STAT_TSTATS;
 	dev->type = ARPHRD_TUNNEL6;
Signature · 1 line
2.43.2
#2 lore raw
In reply to Florian Westphal
This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This patch was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:

Quoted 4 lines · On Mon, 15 Apr 2024 14:23:44 +0200 you wrote:

kmemleak reports net_device resources are no longer released, restore needs_free_netdev toggle. Sample backtrace:

unreferenced object 0xffff88810874f000 (size 4096): [..]

Quoted 8 lines
    [<00000000a2b8af8b>] __kmalloc_node+0x209/0x290
    [<0000000040b0a1a9>] alloc_netdev_mqs+0x58/0x470
    [<00000000b4be1e78>] vti6_init_net+0x94/0x230
    [<000000008830c1ea>] ops_init+0x32/0xc0
    [<000000006a26fa8f>] setup_net+0x134/0x2e0

[..]

[...]

Here is the summary with links:
- [net-next] ip6_vti: fix memleak on netns dismantle

    https://git.kernel.org/netdev/net-next/c/86600ea11dc1

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
#3 lore raw
In reply to Florian Westphal

Hello Florian,

Quoted 23 lines · Florian Westphal

kmemleak reports net_device resources are no longer released, restore needs_free_netdev toggle. Sample backtrace:

unreferenced object 0xffff88810874f000 (size 4096): [..]

    [<00000000a2b8af8b>] __kmalloc_node+0x209/0x290
    [<0000000040b0a1a9>] alloc_netdev_mqs+0x58/0x470
    [<00000000b4be1e78>] vti6_init_net+0x94/0x230
    [<000000008830c1ea>] ops_init+0x32/0xc0
    [<000000006a26fa8f>] setup_net+0x134/0x2e0

[..]

Cc Breno Leitao <[email protected]>
Fixes a9b2d55a8f1e ("ip6_vti: Do not use custom stat allocator")
---
 net/ipv6/ip6_vti.c | 1 +
 1 file changed, 1 insertion(+)
net/ipv6/ip6_vti.c +0 -0
index 4d68a0777b0c..78344cf3867e 100644

net/ipv6/ip6_vti.c static void vti6_dev_setup(struct net_device *dev) · line 904

Quoted 4 lines
net/ipv6/ip6_vti.c +1 -0
@@ -901,6 +901,7 @@ static void vti6_dev_setup(struct net_device *dev)
 {
 	dev->netdev_ops = &vti6_netdev_ops;
 	dev->header_ops = &ip_tunnel_header_ops;
+	dev->needs_free_netdev = true;

Thanks for the fix!

Could you help me to understand how needs_free_netdev will trigger the free()here?

I _though_ that any device that is being unregistered would have the stats freed.

This is the flow I am reading:

1) When the device is unregistered, then it is marked as todo:

	unregister_netdevice_many_notify() {
		list_for_each_entry(dev, head, unreg_list) {
			net_set_todo(dev);
		}
	}

2) Then, "run_todo" will run later, and it does:

	netdev_run_todo() {
		list_for_each_entry_safe(dev, tmp, &list, todo_list) {
		if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
			netdev_WARN(dev, "run_todo but not unregistering\n");
			list_del(&dev->todo_list);
			continue;
		}

		while (!list_empty(&list)) {
			netdev_do_free_pcpu_stats(dev);
		}

	}

Thank you!

#4 lore raw
Quoted 51 lines · Breno Leitao

Hello Florian,

Quoted 23 lines · Florian Westphal

kmemleak reports net_device resources are no longer released, restore needs_free_netdev toggle. Sample backtrace:

unreferenced object 0xffff88810874f000 (size 4096): [..]

    [<00000000a2b8af8b>] __kmalloc_node+0x209/0x290
    [<0000000040b0a1a9>] alloc_netdev_mqs+0x58/0x470
    [<00000000b4be1e78>] vti6_init_net+0x94/0x230
    [<000000008830c1ea>] ops_init+0x32/0xc0
    [<000000006a26fa8f>] setup_net+0x134/0x2e0

[..]

Cc Breno Leitao <[email protected]>
Fixes a9b2d55a8f1e ("ip6_vti: Do not use custom stat allocator")
---
 net/ipv6/ip6_vti.c | 1 +
 1 file changed, 1 insertion(+)
net/ipv6/ip6_vti.c +0 -0
index 4d68a0777b0c..78344cf3867e 100644

net/ipv6/ip6_vti.c static void vti6_dev_setup(struct net_device *dev) · line 904

Quoted 4 lines
net/ipv6/ip6_vti.c +1 -0
@@ -901,6 +901,7 @@ static void vti6_dev_setup(struct net_device *dev)
 {
      dev->netdev_ops = &vti6_netdev_ops;
      dev->header_ops = &ip_tunnel_header_ops;
+     dev->needs_free_netdev = true;

Thanks for the fix!

Could you help me to understand how needs_free_netdev will trigger the free()here?

I _though_ that any device that is being unregistered would have the stats freed.

This is the flow I am reading:

1) When the device is unregistered, then it is marked as todo:

        unregister_netdevice_many_notify() {
                list_for_each_entry(dev, head, unreg_list) {
                        net_set_todo(dev);
                }
        }

2) Then, "run_todo" will run later, and it does:

Quoted 10 lines
        netdev_run_todo() {
                list_for_each_entry_safe(dev, tmp, &list, todo_list) {
                if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
                        netdev_WARN(dev, "run_todo but not unregistering\n");
                        list_del(&dev->todo_list);
                        continue;
                }

                while (!list_empty(&list)) {
                        netdev_do_free_pcpu_stats(dev);

The relevant part is missing:

                         if (dev->needs_free_netdev)
                                  free_netdev(dev);
Quoted 5 lines
                }

        }

Thank you!

I do not think Florian patch has anything to do with free_pcpu_stats()

Please take a look at

git show cf124db566e -- net/ipv6/ip6_vti.c

#5 lore raw

Hello Eric,

Quoted 68 lines · Eric Dumazet
Quoted 51 lines · Breno Leitao

Hello Florian,

Quoted 23 lines · Florian Westphal

kmemleak reports net_device resources are no longer released, restore needs_free_netdev toggle. Sample backtrace:

unreferenced object 0xffff88810874f000 (size 4096): [..]

    [<00000000a2b8af8b>] __kmalloc_node+0x209/0x290
    [<0000000040b0a1a9>] alloc_netdev_mqs+0x58/0x470
    [<00000000b4be1e78>] vti6_init_net+0x94/0x230
    [<000000008830c1ea>] ops_init+0x32/0xc0
    [<000000006a26fa8f>] setup_net+0x134/0x2e0

[..]

Cc Breno Leitao <[email protected]>
Fixes a9b2d55a8f1e ("ip6_vti: Do not use custom stat allocator")
---
 net/ipv6/ip6_vti.c | 1 +
 1 file changed, 1 insertion(+)
net/ipv6/ip6_vti.c +0 -0
index 4d68a0777b0c..78344cf3867e 100644

net/ipv6/ip6_vti.c static void vti6_dev_setup(struct net_device *dev) · line 904

Quoted 4 lines
net/ipv6/ip6_vti.c +1 -0
@@ -901,6 +901,7 @@ static void vti6_dev_setup(struct net_device *dev)
 {
      dev->netdev_ops = &vti6_netdev_ops;
      dev->header_ops = &ip_tunnel_header_ops;
+     dev->needs_free_netdev = true;

Thanks for the fix!

Could you help me to understand how needs_free_netdev will trigger the free()here?

I _though_ that any device that is being unregistered would have the stats freed.

This is the flow I am reading:

1) When the device is unregistered, then it is marked as todo:

        unregister_netdevice_many_notify() {
                list_for_each_entry(dev, head, unreg_list) {
                        net_set_todo(dev);
                }
        }

2) Then, "run_todo" will run later, and it does:

Quoted 10 lines
        netdev_run_todo() {
                list_for_each_entry_safe(dev, tmp, &list, todo_list) {
                if (unlikely(dev->reg_state != NETREG_UNREGISTERING)) {
                        netdev_WARN(dev, "run_todo but not unregistering\n");
                        list_del(&dev->todo_list);
                        continue;
                }

                while (!list_empty(&list)) {
                        netdev_do_free_pcpu_stats(dev);

The relevant part is missing:

                         if (dev->needs_free_netdev)
                                  free_netdev(dev);
Quoted 7 lines
Quoted 5 lines
                }

        }

Thank you!

I do not think Florian patch has anything to do with free_pcpu_stats()

That makes total sense now.

I though the memory that was being leaked was related to pcpu stats, since this is was I was supposedly changing, but, it is not.

The regression is not related to the fact that a9b2d55a8f1e ("ip6_vti: Do not use custom stat allocator") removed the following line wrongly:

	-dev->needs_free_netdev = true;

Everything is clear now.
Thanks!