Skip to contentSkip to conversation

net: hsr: clean up the slave VLAN filters when deleting a port

PATCH In review Jiayuan Chen 2 participants 3 messages Jul 14

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

PATCH 1/1 net: hsr: clean up the slave VLAN filters when deleting a port

Our internal syzkaller reported this warning:

netdevsim netdevsim10 eth0: entered promiscuous mode netdevsim netdevsim10 eth1: entered promiscuous mode netdevsim netdevsim10 eth1 (unregistering): left promiscuous mode

  ------------[ cut here ]------------
  WARNING: drivers/net/netdevsim/netdev.c:1208 at nsim_destroy+0x276/0x6e0, CPU#3: 1/46
  Modules linked in:
  CPU: 3 UID: 0 PID: 46 Comm: kworker/u16:1 Not tainted 7.2.0-rc2+ #286 PREEMPT
  Hardware name: QEMU Ubuntu 24.04 PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
  Workqueue: netns cleanup_net
  RIP: 0010:nsim_destroy (drivers/net/netdevsim/netdev.c:1031 drivers/net/netdevsim/netdev.c:1201)
  RSP: 0018:ffffc90000337898 EFLAGS: 00010293
  RAX: 0000000000000001 RBX: ffff8881036c8b00 RCX: 0000000000000000
  RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
  RBP: ffffc900003378f0 R08: 0000000000000000 R09: 0000000000000000
  R10: ffffffffad4186a8 R11: ffffffffad41c3db R12: ffff8881036c9268
  R13: dffffc0000000000 R14: ffffed10206d9164 R15: 0000000000000001
  FS:  0000000000000000(0000) GS:ffff888160866000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 000055bcf1a2e0c4 CR3: 0000000105376003 CR4: 0000000000770ef0
  PKRU: 55555554
  Call Trace:
  <TASK>
  __nsim_dev_port_del (drivers/net/netdevsim/dev.c:1547)
  nsim_dev_reload_destroy (drivers/net/netdevsim/dev.c:1561 drivers/net/netdevsim/dev.c:1785)
  nsim_dev_reload_down (drivers/net/netdevsim/dev.c:1038)
  devlink_reload (net/devlink/dev.c:462)
  devlink_pernet_pre_exit (net/devlink/core.c:578)
  ops_undo_list (net/core/net_namespace.c:161 net/core/net_namespace.c:235)
  cleanup_net (net/core/net_namespace.c:706)
  process_one_work (kernel/workqueue.c:3322)
  worker_thread (kernel/workqueue.c:3405 kernel/workqueue.c:3486)
  kthread (kernel/kthread.c:436)
  ret_from_fork (arch/x86/kernel/process.c:158)
  ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
  </TASK>
  ---[ end trace 0000000000000000 ]---
  netdevsim netdevsim10 eth0 (unregistering): left promiscuous mode

vlan_vids_add_by_dev() is what passes the VID filter info down to the real device. A virtual device does no filtering itself, so it has to push the VIDs down to its slaves. HSR already does this correctly on the add side:
hsr_ndo_vlan_rx_add_vid() calls vlan_vid_add() on each slave.

We just never clean the slaves up on delete. hsr_del_port() drops the port and leaves the VIDs sitting on the slave, so the filter entries leak. netdevsim keeps a bitmap of the VIDs it was told to filter and checks that it is empty when the netdev goes away, which is what trips the warning above.

So do the cleanup in hsr_del_port(). We follow hsr_ndo_vlan_rx_add_vid()
here: it only pushes the VIDs to HSR_PT_SLAVE_A and HSR_PT_SLAVE_B, so the del side only touches those two as well.

Reproducer:
ip netns add ns0
ip netns exec ns0 sh -c 'echo "10 2" > /sys/bus/netdevsim/new_device' ip netns exec ns0 ip link add hsr0 type hsr slave1 eth0 slave2 eth1 ip netns exec ns0 ip link add link hsr0 name hsr0.1 type vlan id 1 ip netns del ns0

Fixes 1a8a63a5305e ("net: hsr: Add VLAN CTAG filter support")
---
 net/hsr/hsr_slave.c | 7 +++++++
 1 file changed, 7 insertions(+)
net/hsr/hsr_slave.c +7 -0
index d9af9e65f72f..b1baefb65c7a 100644
@@ -237,6 +237,13 @@ void hsr_del_port(struct hsr_port *port)
 	list_del_rcu(&port->port_list);
 
 	if (port != master) {
+		/* Drop the VLAN ids that hsr_ndo_vlan_rx_add_vid() pushed down
+		 * to this slave. Only slave ports get them, see there.
+		 */
+		if (port->type == HSR_PT_SLAVE_A ||
+		    port->type == HSR_PT_SLAVE_B)
+			vlan_vids_del_by_dev(port->dev, master->dev);
+
 		netdev_update_features(master->dev);
 		dev_set_mtu(master->dev, hsr_get_max_mtu(hsr));
 		netdev_rx_handler_unregister(port->dev);
Signature · 1 line
2.43.0
#1 lore raw
Quoted 55 lines · Jiayuan Chen

Our internal syzkaller reported this warning:

netdevsim netdevsim10 eth0: entered promiscuous mode netdevsim netdevsim10 eth1: entered promiscuous mode netdevsim netdevsim10 eth1 (unregistering): left promiscuous mode

  ------------[ cut here ]------------
  WARNING: drivers/net/netdevsim/netdev.c:1208 at nsim_destroy+0x276/0x6e0, CPU#3: 1/46
  Modules linked in:
  CPU: 3 UID: 0 PID: 46 Comm: kworker/u16:1 Not tainted 7.2.0-rc2+ #286 PREEMPT
  Hardware name: QEMU Ubuntu 24.04 PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
  Workqueue: netns cleanup_net
  RIP: 0010:nsim_destroy (drivers/net/netdevsim/netdev.c:1031 drivers/net/netdevsim/netdev.c:1201)
  RSP: 0018:ffffc90000337898 EFLAGS: 00010293
  RAX: 0000000000000001 RBX: ffff8881036c8b00 RCX: 0000000000000000
  RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
  RBP: ffffc900003378f0 R08: 0000000000000000 R09: 0000000000000000
  R10: ffffffffad4186a8 R11: ffffffffad41c3db R12: ffff8881036c9268
  R13: dffffc0000000000 R14: ffffed10206d9164 R15: 0000000000000001
  FS:  0000000000000000(0000) GS:ffff888160866000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 000055bcf1a2e0c4 CR3: 0000000105376003 CR4: 0000000000770ef0
  PKRU: 55555554
  Call Trace:
  <TASK>
  __nsim_dev_port_del (drivers/net/netdevsim/dev.c:1547)
  nsim_dev_reload_destroy (drivers/net/netdevsim/dev.c:1561 drivers/net/netdevsim/dev.c:1785)
  nsim_dev_reload_down (drivers/net/netdevsim/dev.c:1038)
  devlink_reload (net/devlink/dev.c:462)
  devlink_pernet_pre_exit (net/devlink/core.c:578)
  ops_undo_list (net/core/net_namespace.c:161 net/core/net_namespace.c:235)
  cleanup_net (net/core/net_namespace.c:706)
  process_one_work (kernel/workqueue.c:3322)
  worker_thread (kernel/workqueue.c:3405 kernel/workqueue.c:3486)
  kthread (kernel/kthread.c:436)
  ret_from_fork (arch/x86/kernel/process.c:158)
  ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
  </TASK>
  ---[ end trace 0000000000000000 ]---
  netdevsim netdevsim10 eth0 (unregistering): left promiscuous mode

vlan_vids_add_by_dev() is what passes the VID filter info down to the real device. A virtual device does no filtering itself, so it has to push the VIDs down to its slaves. HSR already does this correctly on the add side:
hsr_ndo_vlan_rx_add_vid() calls vlan_vid_add() on each slave.

We just never clean the slaves up on delete. hsr_del_port() drops the port and leaves the VIDs sitting on the slave, so the filter entries leak. netdevsim keeps a bitmap of the VIDs it was told to filter and checks that it is empty when the netdev goes away, which is what trips the warning above.

So do the cleanup in hsr_del_port(). We follow hsr_ndo_vlan_rx_add_vid()
here: it only pushes the VIDs to HSR_PT_SLAVE_A and HSR_PT_SLAVE_B, so the del side only touches those two as well.

Quoted 6 lines

Reproducer:
ip netns add ns0
ip netns exec ns0 sh -c 'echo "10 2" > /sys/bus/netdevsim/new_device' ip netns exec ns0 ip link add hsr0 type hsr slave1 eth0 slave2 eth1 ip netns exec ns0 ip link add link hsr0 name hsr0.1 type vlan id 1 ip netns del ns0

Thank you for your patch! The explanation and the patch look very similar to a fix that Eric posted a week ago [1]. Can you check if Eric's patch also fixes you reproducer? And if it does, please add your Tested-by and/or Reviewed-by to the thread?

Thanks,
Felix

[1]: https://lore.kernel.org/netdev/[email protected]/

#2 lore raw
Quoted 67 lines · Felix Maurer
Quoted 55 lines · Jiayuan Chen

Our internal syzkaller reported this warning:

   netdevsim netdevsim10 eth0: entered promiscuous mode
   netdevsim netdevsim10 eth1: entered promiscuous mode
   netdevsim netdevsim10 eth1 (unregistering): left promiscuous mode
   ------------[ cut here ]------------
   WARNING: drivers/net/netdevsim/netdev.c:1208 at nsim_destroy+0x276/0x6e0, CPU#3: 1/46
   Modules linked in:
   CPU: 3 UID: 0 PID: 46 Comm: kworker/u16:1 Not tainted 7.2.0-rc2+ #286 PREEMPT
   Hardware name: QEMU Ubuntu 24.04 PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
   Workqueue: netns cleanup_net
   RIP: 0010:nsim_destroy (drivers/net/netdevsim/netdev.c:1031 drivers/net/netdevsim/netdev.c:1201)
   RSP: 0018:ffffc90000337898 EFLAGS: 00010293
   RAX: 0000000000000001 RBX: ffff8881036c8b00 RCX: 0000000000000000
   RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
   RBP: ffffc900003378f0 R08: 0000000000000000 R09: 0000000000000000
   R10: ffffffffad4186a8 R11: ffffffffad41c3db R12: ffff8881036c9268
   R13: dffffc0000000000 R14: ffffed10206d9164 R15: 0000000000000001
   FS:  0000000000000000(0000) GS:ffff888160866000(0000) knlGS:0000000000000000
   CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
   CR2: 000055bcf1a2e0c4 CR3: 0000000105376003 CR4: 0000000000770ef0
   PKRU: 55555554
   Call Trace:
   <TASK>
   __nsim_dev_port_del (drivers/net/netdevsim/dev.c:1547)
   nsim_dev_reload_destroy (drivers/net/netdevsim/dev.c:1561 drivers/net/netdevsim/dev.c:1785)
   nsim_dev_reload_down (drivers/net/netdevsim/dev.c:1038)
   devlink_reload (net/devlink/dev.c:462)
   devlink_pernet_pre_exit (net/devlink/core.c:578)
   ops_undo_list (net/core/net_namespace.c:161 net/core/net_namespace.c:235)
   cleanup_net (net/core/net_namespace.c:706)
   process_one_work (kernel/workqueue.c:3322)
   worker_thread (kernel/workqueue.c:3405 kernel/workqueue.c:3486)
   kthread (kernel/kthread.c:436)
   ret_from_fork (arch/x86/kernel/process.c:158)
   ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
   </TASK>
   ---[ end trace 0000000000000000 ]---
   netdevsim netdevsim10 eth0 (unregistering): left promiscuous mode

vlan_vids_add_by_dev() is what passes the VID filter info down to the real device. A virtual device does no filtering itself, so it has to push the VIDs down to its slaves. HSR already does this correctly on the add side:
hsr_ndo_vlan_rx_add_vid() calls vlan_vid_add() on each slave.

We just never clean the slaves up on delete. hsr_del_port() drops the port and leaves the VIDs sitting on the slave, so the filter entries leak. netdevsim keeps a bitmap of the VIDs it was told to filter and checks that it is empty when the netdev goes away, which is what trips the warning above.

So do the cleanup in hsr_del_port(). We follow hsr_ndo_vlan_rx_add_vid()
here: it only pushes the VIDs to HSR_PT_SLAVE_A and HSR_PT_SLAVE_B, so the del side only touches those two as well.

Quoted 6 lines

Reproducer:
ip netns add ns0
ip netns exec ns0 sh -c 'echo "10 2" > /sys/bus/netdevsim/new_device' ip netns exec ns0 ip link add hsr0 type hsr slave1 eth0 slave2 eth1 ip netns exec ns0 ip link add link hsr0 name hsr0.1 type vlan id 1 ip netns del ns0

Thank you for your patch! The explanation and the patch look very similar to a fix that Eric posted a week ago [1]. Can you check if Eric's patch also fixes you reproducer? And if it does, please add your Tested-by and/or Reviewed-by to the thread?

Quoted 5 lines

Thanks,
Felix

[1]: https://lore.kernel.org/netdev/[email protected]/

Thanks, It seems my patch is duplicated.