Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux

bonding: make tx_rebalance_counter an atomic

KCSAN reported a data-race [1] around tx_rebalance_counter
which can be accessed from different contexts, without
the protection of a lock/mutex.

[1]
BUG: KCSAN: data-race in bond_alb_init_slave / bond_alb_monitor

write to 0xffff888157e8ca24 of 4 bytes by task 7075 on cpu 0:
bond_alb_init_slave+0x713/0x860 drivers/net/bonding/bond_alb.c:1613
bond_enslave+0xd94/0x3010 drivers/net/bonding/bond_main.c:1949
do_set_master net/core/rtnetlink.c:2521 [inline]
__rtnl_newlink net/core/rtnetlink.c:3475 [inline]
rtnl_newlink+0x1298/0x13b0 net/core/rtnetlink.c:3506
rtnetlink_rcv_msg+0x745/0x7e0 net/core/rtnetlink.c:5571
netlink_rcv_skb+0x14e/0x250 net/netlink/af_netlink.c:2491
rtnetlink_rcv+0x18/0x20 net/core/rtnetlink.c:5589
netlink_unicast_kernel net/netlink/af_netlink.c:1319 [inline]
netlink_unicast+0x5fc/0x6c0 net/netlink/af_netlink.c:1345
netlink_sendmsg+0x6e1/0x7d0 net/netlink/af_netlink.c:1916
sock_sendmsg_nosec net/socket.c:704 [inline]
sock_sendmsg net/socket.c:724 [inline]
____sys_sendmsg+0x39a/0x510 net/socket.c:2409
___sys_sendmsg net/socket.c:2463 [inline]
__sys_sendmsg+0x195/0x230 net/socket.c:2492
__do_sys_sendmsg net/socket.c:2501 [inline]
__se_sys_sendmsg net/socket.c:2499 [inline]
__x64_sys_sendmsg+0x42/0x50 net/socket.c:2499
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x44/0xd0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x44/0xae

read to 0xffff888157e8ca24 of 4 bytes by task 1082 on cpu 1:
bond_alb_monitor+0x8f/0xc00 drivers/net/bonding/bond_alb.c:1511
process_one_work+0x3fc/0x980 kernel/workqueue.c:2298
worker_thread+0x616/0xa70 kernel/workqueue.c:2445
kthread+0x2c7/0x2e0 kernel/kthread.c:327
ret_from_fork+0x1f/0x30

value changed: 0x00000001 -> 0x00000064

Reported by Kernel Concurrency Sanitizer on:
CPU: 1 PID: 1082 Comm: kworker/u4:3 Not tainted 5.16.0-rc3-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Workqueue: bond1 bond_alb_monitor

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

authored by

Eric Dumazet and committed by
David S. Miller
dac8e00f 03cfda4f

+9 -7
+8 -6
drivers/net/bonding/bond_alb.c
··· 1501 1501 struct slave *slave; 1502 1502 1503 1503 if (!bond_has_slaves(bond)) { 1504 - bond_info->tx_rebalance_counter = 0; 1504 + atomic_set(&bond_info->tx_rebalance_counter, 0); 1505 1505 bond_info->lp_counter = 0; 1506 1506 goto re_arm; 1507 1507 } 1508 1508 1509 1509 rcu_read_lock(); 1510 1510 1511 - bond_info->tx_rebalance_counter++; 1511 + atomic_inc(&bond_info->tx_rebalance_counter); 1512 1512 bond_info->lp_counter++; 1513 1513 1514 1514 /* send learning packets */ ··· 1530 1530 } 1531 1531 1532 1532 /* rebalance tx traffic */ 1533 - if (bond_info->tx_rebalance_counter >= BOND_TLB_REBALANCE_TICKS) { 1533 + if (atomic_read(&bond_info->tx_rebalance_counter) >= BOND_TLB_REBALANCE_TICKS) { 1534 1534 bond_for_each_slave_rcu(bond, slave, iter) { 1535 1535 tlb_clear_slave(bond, slave, 1); 1536 1536 if (slave == rcu_access_pointer(bond->curr_active_slave)) { ··· 1540 1540 bond_info->unbalanced_load = 0; 1541 1541 } 1542 1542 } 1543 - bond_info->tx_rebalance_counter = 0; 1543 + atomic_set(&bond_info->tx_rebalance_counter, 0); 1544 1544 } 1545 1545 1546 1546 if (bond_info->rlb_enabled) { ··· 1610 1610 tlb_init_slave(slave); 1611 1611 1612 1612 /* order a rebalance ASAP */ 1613 - bond->alb_info.tx_rebalance_counter = BOND_TLB_REBALANCE_TICKS; 1613 + atomic_set(&bond->alb_info.tx_rebalance_counter, 1614 + BOND_TLB_REBALANCE_TICKS); 1614 1615 1615 1616 if (bond->alb_info.rlb_enabled) 1616 1617 bond->alb_info.rlb_rebalance = 1; ··· 1648 1647 rlb_clear_slave(bond, slave); 1649 1648 } else if (link == BOND_LINK_UP) { 1650 1649 /* order a rebalance ASAP */ 1651 - bond_info->tx_rebalance_counter = BOND_TLB_REBALANCE_TICKS; 1650 + atomic_set(&bond_info->tx_rebalance_counter, 1651 + BOND_TLB_REBALANCE_TICKS); 1652 1652 if (bond->alb_info.rlb_enabled) { 1653 1653 bond->alb_info.rlb_rebalance = 1; 1654 1654 /* If the updelay module parameter is smaller than the
+1 -1
include/net/bond_alb.h
··· 126 126 struct alb_bond_info { 127 127 struct tlb_client_info *tx_hashtbl; /* Dynamically allocated */ 128 128 u32 unbalanced_load; 129 - int tx_rebalance_counter; 129 + atomic_t tx_rebalance_counter; 130 130 int lp_counter; 131 131 /* -------- rlb parameters -------- */ 132 132 int rlb_enabled;