linux/net/hsr/hsr_framereg.c

634 lines
17 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0
/* Copyright 2011-2014 Autronica Fire and Security AS
*
* Author(s):
* 2011-2014 Arvid Brodin, arvid.brodin@alten.se
*
* The HSR spec says never to forward the same frame twice on the same
* interface. A frame is identified by its source MAC address and its HSR
* sequence number. This code keeps track of senders and their sequence numbers
* to allow filtering of duplicate frames, and to detect HSR ring errors.
* Same code handles filtering of duplicates for PRP as well.
*/
#include <linux/if_ether.h>
#include <linux/etherdevice.h>
#include <linux/slab.h>
#include <linux/rculist.h>
#include "hsr_main.h"
#include "hsr_framereg.h"
#include "hsr_netlink.h"
/* seq_nr_after(a, b) - return true if a is after (higher in sequence than) b,
* false otherwise.
*/
static bool seq_nr_after(u16 a, u16 b)
{
/* Remove inconsistency where
* seq_nr_after(a, b) == seq_nr_before(a, b)
*/
if ((int)b - a == 32768)
return false;
return (((s16)(b - a)) < 0);
}
#define seq_nr_before(a, b) seq_nr_after((b), (a))
#define seq_nr_before_or_eq(a, b) (!seq_nr_after((a), (b)))
bool hsr_addr_is_self(struct hsr_priv *hsr, unsigned char *addr)
{
struct hsr_self_node *sn;
bool ret = false;
rcu_read_lock();
sn = rcu_dereference(hsr->self_node);
if (!sn) {
WARN_ONCE(1, "HSR: No self node\n");
goto out;
}
if (ether_addr_equal(addr, sn->macaddress_A) ||
ether_addr_equal(addr, sn->macaddress_B))
ret = true;
out:
rcu_read_unlock();
return ret;
}
/* Search for mac entry. Caller must hold rcu read lock.
*/
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
static struct hsr_node *find_node_by_addr_A(struct list_head *node_db,
const unsigned char addr[ETH_ALEN])
{
struct hsr_node *node;
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
list_for_each_entry_rcu(node, node_db, mac_list) {
if (ether_addr_equal(node->macaddress_A, addr))
return node;
}
return NULL;
}
/* Helper for device init; the self_node is used in hsr_rcv() to recognize
* frames from self that's been looped over the HSR ring.
*/
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
int hsr_create_self_node(struct hsr_priv *hsr,
const unsigned char addr_a[ETH_ALEN],
const unsigned char addr_b[ETH_ALEN])
{
struct hsr_self_node *sn, *old;
sn = kmalloc(sizeof(*sn), GFP_KERNEL);
if (!sn)
return -ENOMEM;
ether_addr_copy(sn->macaddress_A, addr_a);
ether_addr_copy(sn->macaddress_B, addr_b);
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_lock_bh(&hsr->list_lock);
old = rcu_replace_pointer(hsr->self_node, sn,
lockdep_is_held(&hsr->list_lock));
spin_unlock_bh(&hsr->list_lock);
if (old)
kfree_rcu(old, rcu_head);
return 0;
}
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
void hsr_del_self_node(struct hsr_priv *hsr)
net: hsr: fix memory leak in hsr_dev_finalize() If hsr_add_port(hsr, hsr_dev, HSR_PT_MASTER) failed to add port, it directly returns res and forgets to free the node that allocated in hsr_create_self_node(), and forgets to delete the node->mac_list linked in hsr->self_node_db. BUG: memory leak unreferenced object 0xffff8881cfa0c780 (size 64): comm "syz-executor.0", pid 2077, jiffies 4294717969 (age 2415.377s) hex dump (first 32 bytes): e0 c7 a0 cf 81 88 ff ff 00 02 00 00 00 00 ad de ................ 00 e6 49 cd 81 88 ff ff c0 9b 87 d0 81 88 ff ff ..I............. backtrace: [<00000000e2ff5070>] hsr_dev_finalize+0x736/0x960 [hsr] [<000000003ed2e597>] hsr_newlink+0x2b2/0x3e0 [hsr] [<000000003fa8c6b6>] __rtnl_newlink+0xf1f/0x1600 net/core/rtnetlink.c:3182 [<000000001247a7ad>] rtnl_newlink+0x66/0x90 net/core/rtnetlink.c:3240 [<00000000e7d1b61d>] rtnetlink_rcv_msg+0x54e/0xb90 net/core/rtnetlink.c:5130 [<000000005556bd3a>] netlink_rcv_skb+0x129/0x340 net/netlink/af_netlink.c:2477 [<00000000741d5ee6>] netlink_unicast_kernel net/netlink/af_netlink.c:1310 [inline] [<00000000741d5ee6>] netlink_unicast+0x49a/0x650 net/netlink/af_netlink.c:1336 [<000000009d56f9b7>] netlink_sendmsg+0x88b/0xdf0 net/netlink/af_netlink.c:1917 [<0000000046b35c59>] sock_sendmsg_nosec net/socket.c:621 [inline] [<0000000046b35c59>] sock_sendmsg+0xc3/0x100 net/socket.c:631 [<00000000d208adc9>] __sys_sendto+0x33e/0x560 net/socket.c:1786 [<00000000b582837a>] __do_sys_sendto net/socket.c:1798 [inline] [<00000000b582837a>] __se_sys_sendto net/socket.c:1794 [inline] [<00000000b582837a>] __x64_sys_sendto+0xdd/0x1b0 net/socket.c:1794 [<00000000c866801d>] do_syscall_64+0x147/0x600 arch/x86/entry/common.c:290 [<00000000fea382d9>] entry_SYSCALL_64_after_hwframe+0x49/0xbe [<00000000e01dacb3>] 0xffffffffffffffff Fixes: c5a759117210 ("net/hsr: Use list_head (and rcu) instead of array for slave devices.") Reported-by: Hulk Robot <hulkci@huawei.com> Signed-off-by: Mao Wenan <maowenan@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-06 22:45:01 +08:00
{
struct hsr_self_node *old;
net: hsr: fix memory leak in hsr_dev_finalize() If hsr_add_port(hsr, hsr_dev, HSR_PT_MASTER) failed to add port, it directly returns res and forgets to free the node that allocated in hsr_create_self_node(), and forgets to delete the node->mac_list linked in hsr->self_node_db. BUG: memory leak unreferenced object 0xffff8881cfa0c780 (size 64): comm "syz-executor.0", pid 2077, jiffies 4294717969 (age 2415.377s) hex dump (first 32 bytes): e0 c7 a0 cf 81 88 ff ff 00 02 00 00 00 00 ad de ................ 00 e6 49 cd 81 88 ff ff c0 9b 87 d0 81 88 ff ff ..I............. backtrace: [<00000000e2ff5070>] hsr_dev_finalize+0x736/0x960 [hsr] [<000000003ed2e597>] hsr_newlink+0x2b2/0x3e0 [hsr] [<000000003fa8c6b6>] __rtnl_newlink+0xf1f/0x1600 net/core/rtnetlink.c:3182 [<000000001247a7ad>] rtnl_newlink+0x66/0x90 net/core/rtnetlink.c:3240 [<00000000e7d1b61d>] rtnetlink_rcv_msg+0x54e/0xb90 net/core/rtnetlink.c:5130 [<000000005556bd3a>] netlink_rcv_skb+0x129/0x340 net/netlink/af_netlink.c:2477 [<00000000741d5ee6>] netlink_unicast_kernel net/netlink/af_netlink.c:1310 [inline] [<00000000741d5ee6>] netlink_unicast+0x49a/0x650 net/netlink/af_netlink.c:1336 [<000000009d56f9b7>] netlink_sendmsg+0x88b/0xdf0 net/netlink/af_netlink.c:1917 [<0000000046b35c59>] sock_sendmsg_nosec net/socket.c:621 [inline] [<0000000046b35c59>] sock_sendmsg+0xc3/0x100 net/socket.c:631 [<00000000d208adc9>] __sys_sendto+0x33e/0x560 net/socket.c:1786 [<00000000b582837a>] __do_sys_sendto net/socket.c:1798 [inline] [<00000000b582837a>] __se_sys_sendto net/socket.c:1794 [inline] [<00000000b582837a>] __x64_sys_sendto+0xdd/0x1b0 net/socket.c:1794 [<00000000c866801d>] do_syscall_64+0x147/0x600 arch/x86/entry/common.c:290 [<00000000fea382d9>] entry_SYSCALL_64_after_hwframe+0x49/0xbe [<00000000e01dacb3>] 0xffffffffffffffff Fixes: c5a759117210 ("net/hsr: Use list_head (and rcu) instead of array for slave devices.") Reported-by: Hulk Robot <hulkci@huawei.com> Signed-off-by: Mao Wenan <maowenan@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-06 22:45:01 +08:00
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_lock_bh(&hsr->list_lock);
old = rcu_replace_pointer(hsr->self_node, NULL,
lockdep_is_held(&hsr->list_lock));
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_unlock_bh(&hsr->list_lock);
if (old)
kfree_rcu(old, rcu_head);
net: hsr: fix memory leak in hsr_dev_finalize() If hsr_add_port(hsr, hsr_dev, HSR_PT_MASTER) failed to add port, it directly returns res and forgets to free the node that allocated in hsr_create_self_node(), and forgets to delete the node->mac_list linked in hsr->self_node_db. BUG: memory leak unreferenced object 0xffff8881cfa0c780 (size 64): comm "syz-executor.0", pid 2077, jiffies 4294717969 (age 2415.377s) hex dump (first 32 bytes): e0 c7 a0 cf 81 88 ff ff 00 02 00 00 00 00 ad de ................ 00 e6 49 cd 81 88 ff ff c0 9b 87 d0 81 88 ff ff ..I............. backtrace: [<00000000e2ff5070>] hsr_dev_finalize+0x736/0x960 [hsr] [<000000003ed2e597>] hsr_newlink+0x2b2/0x3e0 [hsr] [<000000003fa8c6b6>] __rtnl_newlink+0xf1f/0x1600 net/core/rtnetlink.c:3182 [<000000001247a7ad>] rtnl_newlink+0x66/0x90 net/core/rtnetlink.c:3240 [<00000000e7d1b61d>] rtnetlink_rcv_msg+0x54e/0xb90 net/core/rtnetlink.c:5130 [<000000005556bd3a>] netlink_rcv_skb+0x129/0x340 net/netlink/af_netlink.c:2477 [<00000000741d5ee6>] netlink_unicast_kernel net/netlink/af_netlink.c:1310 [inline] [<00000000741d5ee6>] netlink_unicast+0x49a/0x650 net/netlink/af_netlink.c:1336 [<000000009d56f9b7>] netlink_sendmsg+0x88b/0xdf0 net/netlink/af_netlink.c:1917 [<0000000046b35c59>] sock_sendmsg_nosec net/socket.c:621 [inline] [<0000000046b35c59>] sock_sendmsg+0xc3/0x100 net/socket.c:631 [<00000000d208adc9>] __sys_sendto+0x33e/0x560 net/socket.c:1786 [<00000000b582837a>] __do_sys_sendto net/socket.c:1798 [inline] [<00000000b582837a>] __se_sys_sendto net/socket.c:1794 [inline] [<00000000b582837a>] __x64_sys_sendto+0xdd/0x1b0 net/socket.c:1794 [<00000000c866801d>] do_syscall_64+0x147/0x600 arch/x86/entry/common.c:290 [<00000000fea382d9>] entry_SYSCALL_64_after_hwframe+0x49/0xbe [<00000000e01dacb3>] 0xffffffffffffffff Fixes: c5a759117210 ("net/hsr: Use list_head (and rcu) instead of array for slave devices.") Reported-by: Hulk Robot <hulkci@huawei.com> Signed-off-by: Mao Wenan <maowenan@huawei.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-03-06 22:45:01 +08:00
}
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
void hsr_del_nodes(struct list_head *node_db)
{
struct hsr_node *node;
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
struct hsr_node *tmp;
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
list_for_each_entry_safe(node, tmp, node_db, mac_list)
kfree(node);
}
void prp_handle_san_frame(bool san, enum hsr_port_type port,
struct hsr_node *node)
{
/* Mark if the SAN node is over LAN_A or LAN_B */
if (port == HSR_PT_SLAVE_A) {
node->san_a = true;
return;
}
if (port == HSR_PT_SLAVE_B)
node->san_b = true;
}
/* Allocate an hsr_node and add it to node_db. 'addr' is the node's address_A;
* seq_out is used to initialize filtering of outgoing duplicate frames
* originating from the newly added node.
*/
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
static struct hsr_node *hsr_add_node(struct hsr_priv *hsr,
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
struct list_head *node_db,
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
unsigned char addr[],
u16 seq_out, bool san,
enum hsr_port_type rx_port)
{
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
struct hsr_node *new_node, *node;
unsigned long now;
int i;
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
new_node = kzalloc(sizeof(*new_node), GFP_ATOMIC);
if (!new_node)
return NULL;
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
ether_addr_copy(new_node->macaddress_A, addr);
spin_lock_init(&new_node->seq_out_lock);
/* We are only interested in time diffs here, so use current jiffies
* as initialization. (0 could trigger an spurious ring error warning).
*/
now = jiffies;
for (i = 0; i < HSR_PT_PORTS; i++) {
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
new_node->time_in[i] = now;
new_node->time_out[i] = now;
}
for (i = 0; i < HSR_PT_PORTS; i++)
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
new_node->seq_out[i] = seq_out;
if (san && hsr->proto_ops->handle_san_frame)
hsr->proto_ops->handle_san_frame(san, rx_port, new_node);
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_lock_bh(&hsr->list_lock);
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
list_for_each_entry_rcu(node, node_db, mac_list,
lockdep_is_held(&hsr->list_lock)) {
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
if (ether_addr_equal(node->macaddress_A, addr))
goto out;
if (ether_addr_equal(node->macaddress_B, addr))
goto out;
}
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
list_add_tail_rcu(&new_node->mac_list, node_db);
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_unlock_bh(&hsr->list_lock);
return new_node;
out:
spin_unlock_bh(&hsr->list_lock);
kfree(new_node);
return node;
}
void prp_update_san_info(struct hsr_node *node, bool is_sup)
{
if (!is_sup)
return;
node->san_a = false;
node->san_b = false;
}
/* Get the hsr_node from which 'skb' was sent.
*/
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
struct hsr_node *hsr_get_node(struct hsr_port *port, struct list_head *node_db,
struct sk_buff *skb, bool is_sup,
enum hsr_port_type rx_port)
{
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
struct hsr_priv *hsr = port->hsr;
struct hsr_node *node;
struct ethhdr *ethhdr;
struct prp_rct *rct;
bool san = false;
u16 seq_out;
if (!skb_mac_header_was_set(skb))
return NULL;
ethhdr = (struct ethhdr *)skb_mac_header(skb);
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
list_for_each_entry_rcu(node, node_db, mac_list) {
if (ether_addr_equal(node->macaddress_A, ethhdr->h_source)) {
if (hsr->proto_ops->update_san_info)
hsr->proto_ops->update_san_info(node, is_sup);
return node;
}
if (ether_addr_equal(node->macaddress_B, ethhdr->h_source)) {
if (hsr->proto_ops->update_san_info)
hsr->proto_ops->update_san_info(node, is_sup);
return node;
}
}
/* Everyone may create a node entry, connected node to a HSR/PRP
* device.
*/
if (ethhdr->h_proto == htons(ETH_P_PRP) ||
ethhdr->h_proto == htons(ETH_P_HSR)) {
/* Use the existing sequence_nr from the tag as starting point
* for filtering duplicate frames.
*/
seq_out = hsr_get_skb_sequence_nr(skb) - 1;
} else {
rct = skb_get_PRP_rct(skb);
if (rct && prp_check_lsdu_size(skb, rct, is_sup)) {
seq_out = prp_get_skb_sequence_nr(rct);
} else {
if (rx_port != HSR_PT_MASTER)
san = true;
seq_out = HSR_SEQNR_START;
}
}
return hsr_add_node(hsr, node_db, ethhdr->h_source, seq_out,
san, rx_port);
}
/* Use the Supervision frame's info about an eventual macaddress_B for merging
* nodes that has previously had their macaddress_B registered as a separate
* node.
*/
void hsr_handle_sup_frame(struct hsr_frame_info *frame)
{
struct hsr_node *node_curr = frame->node_src;
struct hsr_port *port_rcv = frame->port_rcv;
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
struct hsr_priv *hsr = port_rcv->hsr;
struct hsr_sup_payload *hsr_sp;
struct hsr_sup_tlv *hsr_sup_tlv;
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
struct hsr_node *node_real;
struct sk_buff *skb = NULL;
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
struct list_head *node_db;
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
struct ethhdr *ethhdr;
int i;
unsigned int pull_size = 0;
unsigned int total_pull_size = 0;
/* Here either frame->skb_hsr or frame->skb_prp should be
* valid as supervision frame always will have protocol
* header info.
*/
if (frame->skb_hsr)
skb = frame->skb_hsr;
else if (frame->skb_prp)
skb = frame->skb_prp;
else if (frame->skb_std)
skb = frame->skb_std;
if (!skb)
return;
/* Leave the ethernet header. */
pull_size = sizeof(struct ethhdr);
skb_pull(skb, pull_size);
total_pull_size += pull_size;
ethhdr = (struct ethhdr *)skb_mac_header(skb);
/* And leave the HSR tag. */
if (ethhdr->h_proto == htons(ETH_P_HSR)) {
pull_size = sizeof(struct ethhdr);
skb_pull(skb, pull_size);
total_pull_size += pull_size;
}
/* And leave the HSR sup tag. */
pull_size = sizeof(struct hsr_tag);
skb_pull(skb, pull_size);
total_pull_size += pull_size;
/* get HSR sup payload */
hsr_sp = (struct hsr_sup_payload *)skb->data;
/* Merge node_curr (registered on macaddress_B) into node_real */
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
node_db = &port_rcv->hsr->node_db;
node_real = find_node_by_addr_A(node_db, hsr_sp->macaddress_A);
if (!node_real)
/* No frame received from AddrA of this node yet */
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
node_real = hsr_add_node(hsr, node_db, hsr_sp->macaddress_A,
HSR_SEQNR_START - 1, true,
port_rcv->type);
if (!node_real)
goto done; /* No mem */
if (node_real == node_curr)
/* Node has already been merged */
goto done;
/* Leave the first HSR sup payload. */
pull_size = sizeof(struct hsr_sup_payload);
skb_pull(skb, pull_size);
total_pull_size += pull_size;
/* Get second supervision tlv */
hsr_sup_tlv = (struct hsr_sup_tlv *)skb->data;
/* And check if it is a redbox mac TLV */
if (hsr_sup_tlv->HSR_TLV_type == PRP_TLV_REDBOX_MAC) {
/* We could stop here after pushing hsr_sup_payload,
* or proceed and allow macaddress_B and for redboxes.
*/
/* Sanity check length */
if (hsr_sup_tlv->HSR_TLV_length != 6)
goto done;
/* Leave the second HSR sup tlv. */
pull_size = sizeof(struct hsr_sup_tlv);
skb_pull(skb, pull_size);
total_pull_size += pull_size;
/* Get redbox mac address. */
hsr_sp = (struct hsr_sup_payload *)skb->data;
/* Check if redbox mac and node mac are equal. */
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
if (!ether_addr_equal(node_real->macaddress_A, hsr_sp->macaddress_A)) {
/* This is a redbox supervision frame for a VDAN! */
goto done;
}
}
ether_addr_copy(node_real->macaddress_B, ethhdr->h_source);
spin_lock_bh(&node_real->seq_out_lock);
for (i = 0; i < HSR_PT_PORTS; i++) {
if (!node_curr->time_in_stale[i] &&
time_after(node_curr->time_in[i], node_real->time_in[i])) {
node_real->time_in[i] = node_curr->time_in[i];
node_real->time_in_stale[i] =
node_curr->time_in_stale[i];
}
if (seq_nr_after(node_curr->seq_out[i], node_real->seq_out[i]))
node_real->seq_out[i] = node_curr->seq_out[i];
}
spin_unlock_bh(&node_real->seq_out_lock);
node_real->addr_B_port = port_rcv->type;
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_lock_bh(&hsr->list_lock);
if (!node_curr->removed) {
list_del_rcu(&node_curr->mac_list);
node_curr->removed = true;
kfree_rcu(node_curr, rcu_head);
}
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_unlock_bh(&hsr->list_lock);
done:
/* Push back here */
skb_push(skb, total_pull_size);
}
/* 'skb' is a frame meant for this host, that is to be passed to upper layers.
*
* If the frame was sent by a node's B interface, replace the source
* address with that node's "official" address (macaddress_A) so that upper
* layers recognize where it came from.
*/
void hsr_addr_subst_source(struct hsr_node *node, struct sk_buff *skb)
{
if (!skb_mac_header_was_set(skb)) {
WARN_ONCE(1, "%s: Mac header not set\n", __func__);
return;
}
memcpy(&eth_hdr(skb)->h_source, node->macaddress_A, ETH_ALEN);
}
/* 'skb' is a frame meant for another host.
* 'port' is the outgoing interface
*
* Substitute the target (dest) MAC address if necessary, so the it matches the
* recipient interface MAC address, regardless of whether that is the
* recipient's A or B interface.
* This is needed to keep the packets flowing through switches that learn on
* which "side" the different interfaces are.
*/
void hsr_addr_subst_dest(struct hsr_node *node_src, struct sk_buff *skb,
struct hsr_port *port)
{
struct hsr_node *node_dst;
if (!skb_mac_header_was_set(skb)) {
WARN_ONCE(1, "%s: Mac header not set\n", __func__);
return;
}
if (!is_unicast_ether_addr(eth_hdr(skb)->h_dest))
return;
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
node_dst = find_node_by_addr_A(&port->hsr->node_db,
eth_hdr(skb)->h_dest);
if (!node_dst) {
if (net_ratelimit() && port->hsr->prot_version != PRP_V1)
netdev_err(skb->dev, "%s: Unknown node\n", __func__);
return;
}
if (port->type != node_dst->addr_B_port)
return;
if (is_valid_ether_addr(node_dst->macaddress_B))
ether_addr_copy(eth_hdr(skb)->h_dest, node_dst->macaddress_B);
}
void hsr_register_frame_in(struct hsr_node *node, struct hsr_port *port,
u16 sequence_nr)
{
/* Don't register incoming frames without a valid sequence number. This
* ensures entries of restarted nodes gets pruned so that they can
* re-register and resume communications.
*/
if (!(port->dev->features & NETIF_F_HW_HSR_TAG_RM) &&
seq_nr_before(sequence_nr, node->seq_out[port->type]))
return;
node->time_in[port->type] = jiffies;
node->time_in_stale[port->type] = false;
}
/* 'skb' is a HSR Ethernet frame (with a HSR tag inserted), with a valid
* ethhdr->h_source address and skb->mac_header set.
*
* Return:
* 1 if frame can be shown to have been sent recently on this interface,
* 0 otherwise, or
* negative error code on error
*/
int hsr_register_frame_out(struct hsr_port *port, struct hsr_node *node,
u16 sequence_nr)
{
spin_lock_bh(&node->seq_out_lock);
if (seq_nr_before_or_eq(sequence_nr, node->seq_out[port->type]) &&
time_is_after_jiffies(node->time_out[port->type] +
msecs_to_jiffies(HSR_ENTRY_FORGET_TIME))) {
spin_unlock_bh(&node->seq_out_lock);
return 1;
}
node->time_out[port->type] = jiffies;
node->seq_out[port->type] = sequence_nr;
spin_unlock_bh(&node->seq_out_lock);
return 0;
}
static struct hsr_port *get_late_port(struct hsr_priv *hsr,
struct hsr_node *node)
{
if (node->time_in_stale[HSR_PT_SLAVE_A])
return hsr_port_get_hsr(hsr, HSR_PT_SLAVE_A);
if (node->time_in_stale[HSR_PT_SLAVE_B])
return hsr_port_get_hsr(hsr, HSR_PT_SLAVE_B);
if (time_after(node->time_in[HSR_PT_SLAVE_B],
node->time_in[HSR_PT_SLAVE_A] +
msecs_to_jiffies(MAX_SLAVE_DIFF)))
return hsr_port_get_hsr(hsr, HSR_PT_SLAVE_A);
if (time_after(node->time_in[HSR_PT_SLAVE_A],
node->time_in[HSR_PT_SLAVE_B] +
msecs_to_jiffies(MAX_SLAVE_DIFF)))
return hsr_port_get_hsr(hsr, HSR_PT_SLAVE_B);
return NULL;
}
/* Remove stale sequence_nr records. Called by timer every
* HSR_LIFE_CHECK_INTERVAL (two seconds or so).
*/
void hsr_prune_nodes(struct timer_list *t)
{
struct hsr_priv *hsr = from_timer(hsr, t, prune_timer);
struct hsr_node *node;
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
struct hsr_node *tmp;
struct hsr_port *port;
unsigned long timestamp;
unsigned long time_a, time_b;
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_lock_bh(&hsr->list_lock);
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
list_for_each_entry_safe(node, tmp, &hsr->node_db, mac_list) {
/* Don't prune own node. Neither time_in[HSR_PT_SLAVE_A]
* nor time_in[HSR_PT_SLAVE_B], will ever be updated for
* the master port. Thus the master node will be repeatedly
* pruned leading to packet loss.
*/
if (hsr_addr_is_self(hsr, node->macaddress_A))
continue;
/* Shorthand */
time_a = node->time_in[HSR_PT_SLAVE_A];
time_b = node->time_in[HSR_PT_SLAVE_B];
/* Check for timestamps old enough to risk wrap-around */
if (time_after(jiffies, time_a + MAX_JIFFY_OFFSET / 2))
node->time_in_stale[HSR_PT_SLAVE_A] = true;
if (time_after(jiffies, time_b + MAX_JIFFY_OFFSET / 2))
node->time_in_stale[HSR_PT_SLAVE_B] = true;
/* Get age of newest frame from node.
* At least one time_in is OK here; nodes get pruned long
* before both time_ins can get stale
*/
timestamp = time_a;
if (node->time_in_stale[HSR_PT_SLAVE_A] ||
(!node->time_in_stale[HSR_PT_SLAVE_B] &&
time_after(time_b, time_a)))
timestamp = time_b;
/* Warn of ring error only as long as we get frames at all */
if (time_is_after_jiffies(timestamp +
msecs_to_jiffies(1.5 * MAX_SLAVE_DIFF))) {
rcu_read_lock();
port = get_late_port(hsr, node);
if (port)
hsr_nl_ringerror(hsr, node->macaddress_A, port);
rcu_read_unlock();
}
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
/* Prune old entries */
if (time_is_before_jiffies(timestamp +
msecs_to_jiffies(HSR_NODE_FORGET_TIME))) {
hsr_nl_nodedown(hsr, node->macaddress_A);
if (!node->removed) {
list_del_rcu(&node->mac_list);
node->removed = true;
/* Note that we need to free this entry later: */
kfree_rcu(node, rcu_head);
}
}
}
hsr: fix a race condition in node list insertion and deletion hsr nodes are protected by RCU and there is no write side lock. But node insertions and deletions could be being operated concurrently. So write side locking is needed. Test commands: ip netns add nst ip link add veth0 type veth peer name veth1 ip link add veth2 type veth peer name veth3 ip link set veth1 netns nst ip link set veth3 netns nst ip link set veth0 up ip link set veth2 up ip link add hsr0 type hsr slave1 veth0 slave2 veth2 ip a a 192.168.100.1/24 dev hsr0 ip link set hsr0 up ip netns exec nst ip link set veth1 up ip netns exec nst ip link set veth3 up ip netns exec nst ip link add hsr1 type hsr slave1 veth1 slave2 veth3 ip netns exec nst ip a a 192.168.100.2/24 dev hsr1 ip netns exec nst ip link set hsr1 up for i in {0..9} do for j in {0..9} do for k in {0..9} do for l in {0..9} do arping 192.168.100.2 -I hsr0 -s 00:01:3$i:4$j:5$k:6$l -c1 & done done done done Splat looks like: [ 236.066091][ T3286] list_add corruption. next->prev should be prev (ffff8880a5940300), but was ffff8880a5940d0. [ 236.069617][ T3286] ------------[ cut here ]------------ [ 236.070545][ T3286] kernel BUG at lib/list_debug.c:25! [ 236.071391][ T3286] invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN PTI [ 236.072343][ T3286] CPU: 0 PID: 3286 Comm: arping Tainted: G W 5.5.0-rc1+ #209 [ 236.073463][ T3286] Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 [ 236.074695][ T3286] RIP: 0010:__list_add_valid+0x74/0xd0 [ 236.075499][ T3286] Code: 48 39 da 75 27 48 39 f5 74 36 48 39 dd 74 31 48 83 c4 08 b8 01 00 00 00 5b 5d c3 48 b [ 236.078277][ T3286] RSP: 0018:ffff8880aaa97648 EFLAGS: 00010286 [ 236.086991][ T3286] RAX: 0000000000000075 RBX: ffff8880d4624c20 RCX: 0000000000000000 [ 236.088000][ T3286] RDX: 0000000000000075 RSI: 0000000000000008 RDI: ffffed1015552ebf [ 236.098897][ T3286] RBP: ffff88809b53d200 R08: ffffed101b3c04f9 R09: ffffed101b3c04f9 [ 236.099960][ T3286] R10: 00000000308769a1 R11: ffffed101b3c04f8 R12: ffff8880d4624c28 [ 236.100974][ T3286] R13: ffff8880d4624c20 R14: 0000000040310100 R15: ffff8880ce17ee02 [ 236.138967][ T3286] FS: 00007f23479fa680(0000) GS:ffff8880d9c00000(0000) knlGS:0000000000000000 [ 236.144852][ T3286] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 236.145720][ T3286] CR2: 00007f4a14bab210 CR3: 00000000a61c6001 CR4: 00000000000606f0 [ 236.146776][ T3286] Call Trace: [ 236.147222][ T3286] hsr_add_node+0x314/0x490 [hsr] [ 236.153633][ T3286] hsr_forward_skb+0x2b6/0x1bc0 [hsr] [ 236.154362][ T3286] ? rcu_read_lock_sched_held+0x90/0xc0 [ 236.155091][ T3286] ? rcu_read_lock_bh_held+0xa0/0xa0 [ 236.156607][ T3286] hsr_dev_xmit+0x70/0xd0 [hsr] [ 236.157254][ T3286] dev_hard_start_xmit+0x160/0x740 [ 236.157941][ T3286] __dev_queue_xmit+0x1961/0x2e10 [ 236.158565][ T3286] ? netdev_core_pick_tx+0x2e0/0x2e0 [ ... ] Reported-by: syzbot+3924327f9ad5f4d2b343@syzkaller.appspotmail.com Fixes: f421436a591d ("net/hsr: Add support for the High-availability Seamless Redundancy protocol (HSRv0)") Signed-off-by: Taehee Yoo <ap420073@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-12-22 19:26:54 +08:00
spin_unlock_bh(&hsr->list_lock);
/* Restart timer */
mod_timer(&hsr->prune_timer,
jiffies + msecs_to_jiffies(PRUNE_PERIOD));
}
void *hsr_get_next_node(struct hsr_priv *hsr, void *_pos,
unsigned char addr[ETH_ALEN])
{
struct hsr_node *node;
if (!_pos) {
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
node = list_first_or_null_rcu(&hsr->node_db,
struct hsr_node, mac_list);
if (node)
ether_addr_copy(addr, node->macaddress_A);
return node;
}
node = _pos;
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
list_for_each_entry_continue_rcu(node, &hsr->node_db, mac_list) {
ether_addr_copy(addr, node->macaddress_A);
return node;
}
return NULL;
}
int hsr_get_node_data(struct hsr_priv *hsr,
const unsigned char *addr,
unsigned char addr_b[ETH_ALEN],
unsigned int *addr_b_ifindex,
int *if1_age,
u16 *if1_seq,
int *if2_age,
u16 *if2_seq)
{
struct hsr_node *node;
struct hsr_port *port;
unsigned long tdiff;
Revert "net: hsr: use hlist_head instead of list_head for mac addresses" The hlist optimisation (which not only uses hlist_head instead of list_head but also splits hsr_priv::node_db into an array of 256 slots) does not consider the "node merge": Upon starting the hsr network (with three nodes) a packet that is sent from node1 to node3 will also be sent from node1 to node2 and then forwarded to node3. As a result node3 will receive 2 packets because it is not able to filter out the duplicate. Each packet received will create a new struct hsr_node with macaddress_A only set the MAC address it received from (the two MAC addesses from node1). At some point (early in the process) two supervision frames will be received from node1. They will be processed by hsr_handle_sup_frame() and one frame will leave early ("Node has already been merged") and does nothing. The other frame will be merged as portB and have its MAC address written to macaddress_B and the hsr_node (that was created for it as macaddress_A) will be removed. From now on HSR is able to identify a duplicate because both packets sent from one node will result in the same struct hsr_node because hsr_get_node() will find the MAC address either on macaddress_A or macaddress_B. Things get tricky with the optimisation: If sender's MAC address is saved as macaddress_A then the lookup will work as usual. If the MAC address has been merged into macaddress_B of another hsr_node then the lookup won't work because it is likely that the data structure is in another bucket. This results in creating a new struct hsr_node and not recognising a possible duplicate. A way around it would be to add another hsr_node::mac_list_B and attach it to the other bucket to ensure that this hsr_node will be looked up either via macaddress_A _or_ macaddress_B. I however prefer to revert it because it sounds like an academic problem rather than real life workload plus it adds complexity. I'm not an HSR expert with what is usual size of a network but I would guess 40 to 60 nodes. With 10.000 nodes and assuming 60us for pass-through (from node to node) then it would take almost 600ms for a packet to almost wrap around which sounds a lot. Revert the hash MAC addresses optimisation. Fixes: 4acc45db71158 ("net: hsr: use hlist_head instead of list_head for mac addresses") Cc: Juhee Kang <claudiajkang@gmail.com> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de> Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2022-11-30 00:48:08 +08:00
node = find_node_by_addr_A(&hsr->node_db, addr);
if (!node)
return -ENOENT;
ether_addr_copy(addr_b, node->macaddress_B);
tdiff = jiffies - node->time_in[HSR_PT_SLAVE_A];
if (node->time_in_stale[HSR_PT_SLAVE_A])
*if1_age = INT_MAX;
#if HZ <= MSEC_PER_SEC
else if (tdiff > msecs_to_jiffies(INT_MAX))
*if1_age = INT_MAX;
#endif
else
*if1_age = jiffies_to_msecs(tdiff);
tdiff = jiffies - node->time_in[HSR_PT_SLAVE_B];
if (node->time_in_stale[HSR_PT_SLAVE_B])
*if2_age = INT_MAX;
#if HZ <= MSEC_PER_SEC
else if (tdiff > msecs_to_jiffies(INT_MAX))
*if2_age = INT_MAX;
#endif
else
*if2_age = jiffies_to_msecs(tdiff);
/* Present sequence numbers as if they were incoming on interface */
*if1_seq = node->seq_out[HSR_PT_SLAVE_B];
*if2_seq = node->seq_out[HSR_PT_SLAVE_A];
if (node->addr_B_port != HSR_PT_NONE) {
port = hsr_port_get_hsr(hsr, node->addr_B_port);
*addr_b_ifindex = port->dev->ifindex;
} else {
*addr_b_ifindex = -1;
}
return 0;
}