lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
Open Source and information security mailing list archives
| ||
|
Message-ID: <6d92a9bb-f8ce-3d89-9048-039426480416@blackwall.org> Date: Tue, 28 Mar 2023 11:14:00 +0300 From: Nikolay Aleksandrov <razor@...ckwall.org> To: sujing <sujing@...inos.cn>, davem@...emloft.net Cc: edumazet@...gle.com, kuba@...nel.org, pabeni@...hat.com, andy@...yhouse.net, j.vosburgh@...il.com, netdev@...r.kernel.org, linux-kernel@...r.kernel.org Subject: Re: [PATCH] net: bonding: avoid use-after-free with tx_hashtbl/rx_hashtbl On 28/03/2023 06:40, sujing wrote: > In bonding mode 6 (Balance-alb), > there are some potential race conditions between the 'bond_close' process > and the tx/rx processes that use tx_hashtbl/rx_hashtbl, > which may lead to use-after-free. Potential and may? Have you seen it happen and have a trace? > > For instance, when the bond6 device is in the 'bond_close' process > while some backlogged packets from upper level are transmitted How exactly would that happen? Queues get properly disabled before ndo_stop is called. > to 'bond_start_xmit', there is a spinlock contention between > 'tlb_deinitialize' and 'tlb_choose_channel'. > > If 'tlb_deinitialize' preempts the lock before 'tlb_choose_channel', > a NULL pointer kernel panic will be triggered. > > Here's the timeline: > > bond_close ------------------ bond_start_xmit > bond_alb_deinitialize ------- __bond_start_xmit > tlb_deinitialize ------------ bond_alb_xmit > spin_lock_bh ---------------- bond_xmit_alb_slave_get > tx_hashtbl = NULL ----------- tlb_choose_channel > spin_unlock_bh -------------- //wait for spin_lock_bh > ------------------------------ spin_lock_bh > ------------------------------ __tlb_choose_channel > causing kernel panic ========> tx_hashtbl[hash_index].tx_slave > ------------------------------ spin_unlock_bh I don't see how bond_close() can be called in parallel with bond_start_xmit. Tx queues are disabled and there's a synchronize_rcu() before the device's ndo_stop is called. > > Signed-off-by: sujing <sujing@...inos.cn> > --- > drivers/net/bonding/bond_alb.c | 32 +++++++++------------------ > drivers/net/bonding/bond_main.c | 39 +++++++++++++++++++++++++++------ > include/net/bond_alb.h | 5 ++++- > 3 files changed, 46 insertions(+), 30 deletions(-) > > diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c > index b9dbad3a8af8..f6ff5ea835c4 100644 > --- a/drivers/net/bonding/bond_alb.c > +++ b/drivers/net/bonding/bond_alb.c > @@ -71,7 +71,7 @@ static inline u8 _simple_hash(const u8 *hash_start, int hash_size) > > /*********************** tlb specific functions ***************************/ > > -static inline void tlb_init_table_entry(struct tlb_client_info *entry, int save_load) > +void tlb_init_table_entry(struct tlb_client_info *entry, int save_load) > { > if (save_load) { > entry->load_history = 1 + entry->tx_bytes / > @@ -269,8 +269,8 @@ static void rlb_update_entry_from_arp(struct bonding *bond, struct arp_pkt *arp) > spin_unlock_bh(&bond->mode_lock); > } > > -static int rlb_arp_recv(const struct sk_buff *skb, struct bonding *bond, > - struct slave *slave) > +int rlb_arp_recv(const struct sk_buff *skb, struct bonding *bond, > + struct slave *slave) > { > struct arp_pkt *arp, _arp; > > @@ -756,7 +756,7 @@ static void rlb_init_table_entry_src(struct rlb_client_info *entry) > entry->src_next = RLB_NULL_INDEX; > } > > -static void rlb_init_table_entry(struct rlb_client_info *entry) > +void rlb_init_table_entry(struct rlb_client_info *entry) > { > memset(entry, 0, sizeof(struct rlb_client_info)); > rlb_init_table_entry_dst(entry); > @@ -874,9 +874,6 @@ static int rlb_initialize(struct bonding *bond) > > spin_unlock_bh(&bond->mode_lock); > > - /* register to receive ARPs */ > - bond->recv_probe = rlb_arp_recv; > - > return 0; > } > > @@ -888,7 +885,6 @@ static void rlb_deinitialize(struct bonding *bond) > > kfree(bond_info->rx_hashtbl); > bond_info->rx_hashtbl = NULL; > - bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX; > > spin_unlock_bh(&bond->mode_lock); > } > @@ -1303,7 +1299,7 @@ static bool alb_determine_nd(struct sk_buff *skb, struct bonding *bond) > > /************************ exported alb functions ************************/ > > -int bond_alb_initialize(struct bonding *bond, int rlb_enabled) > +int bond_alb_initialize(struct bonding *bond) > { > int res; > > @@ -1311,15 +1307,10 @@ int bond_alb_initialize(struct bonding *bond, int rlb_enabled) > if (res) > return res; > > - if (rlb_enabled) { > - res = rlb_initialize(bond); > - if (res) { > - tlb_deinitialize(bond); > - return res; > - } > - bond->alb_info.rlb_enabled = 1; > - } else { > - bond->alb_info.rlb_enabled = 0; > + res = rlb_initialize(bond); > + if (res) { > + tlb_deinitialize(bond); > + return res; > } > > return 0; > @@ -1327,12 +1318,9 @@ int bond_alb_initialize(struct bonding *bond, int rlb_enabled) > > void bond_alb_deinitialize(struct bonding *bond) > { > - struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); > - > tlb_deinitialize(bond); > > - if (bond_info->rlb_enabled) > - rlb_deinitialize(bond); > + rlb_deinitialize(bond); > } > > static netdev_tx_t bond_do_alb_xmit(struct sk_buff *skb, struct bonding *bond, > diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c > index 236e5219c811..8fcb5d3ac0a2 100644 > --- a/drivers/net/bonding/bond_main.c > +++ b/drivers/net/bonding/bond_main.c > @@ -4217,6 +4217,7 @@ static int bond_open(struct net_device *bond_dev) > struct bonding *bond = netdev_priv(bond_dev); > struct list_head *iter; > struct slave *slave; > + int i; > > if (BOND_MODE(bond) == BOND_MODE_ROUNDROBIN && !bond->rr_tx_counter) { > bond->rr_tx_counter = alloc_percpu(u32); > @@ -4239,11 +4240,29 @@ static int bond_open(struct net_device *bond_dev) > } > > if (bond_is_lb(bond)) { > - /* bond_alb_initialize must be called before the timer > - * is started. > - */ > - if (bond_alb_initialize(bond, (BOND_MODE(bond) == BOND_MODE_ALB))) > - return -ENOMEM; > + struct alb_bond_info *bond_info = &(BOND_ALB_INFO(bond)); > + > + spin_lock_bh(&bond->mode_lock); > + > + for (i = 0; i < TLB_HASH_TABLE_SIZE; i++) > + tlb_init_table_entry(&bond_info->tx_hashtbl[i], 0); > + > + spin_unlock_bh(&bond->mode_lock); > + > + if (BOND_MODE(bond) == BOND_MODE_ALB) { > + bond->alb_info.rlb_enabled = 1; > + spin_lock_bh(&bond->mode_lock); > + > + bond_info->rx_hashtbl_used_head = RLB_NULL_INDEX; > + for (i = 0; i < RLB_HASH_TABLE_SIZE; i++) > + rlb_init_table_entry(bond_info->rx_hashtbl + i); > + > + spin_unlock_bh(&bond->mode_lock); > + bond->recv_probe = rlb_arp_recv; > + } else { > + bond->alb_info.rlb_enabled = 0; > + } > + > if (bond->params.tlb_dynamic_lb || BOND_MODE(bond) == BOND_MODE_ALB) > queue_delayed_work(bond->wq, &bond->alb_work, 0); > } > @@ -4279,8 +4298,6 @@ static int bond_close(struct net_device *bond_dev) > > bond_work_cancel_all(bond); > bond->send_peer_notif = 0; > - if (bond_is_lb(bond)) > - bond_alb_deinitialize(bond); > bond->recv_probe = NULL; > > if (bond_uses_primary(bond)) { > @@ -5854,6 +5871,8 @@ static void bond_uninit(struct net_device *bond_dev) > struct list_head *iter; > struct slave *slave; > > + bond_alb_deinitialize(bond); > + > bond_netpoll_cleanup(bond_dev); > > /* Release the bonded slaves */ > @@ -6295,6 +6314,12 @@ static int bond_init(struct net_device *bond_dev) > bond_dev->addr_assign_type == NET_ADDR_PERM) > eth_hw_addr_random(bond_dev); > > + /* bond_alb_initialize must be called before the timer > + * is started. > + */ > + if (bond_alb_initialize(bond)) > + return -ENOMEM; > + > return 0; > } > > diff --git a/include/net/bond_alb.h b/include/net/bond_alb.h > index 9dc082b2d543..9fd16e20ef82 100644 > --- a/include/net/bond_alb.h > +++ b/include/net/bond_alb.h > @@ -150,7 +150,7 @@ struct alb_bond_info { > */ > }; > > -int bond_alb_initialize(struct bonding *bond, int rlb_enabled); > +int bond_alb_initialize(struct bonding *bond); > void bond_alb_deinitialize(struct bonding *bond); > int bond_alb_init_slave(struct bonding *bond, struct slave *slave); > void bond_alb_deinit_slave(struct bonding *bond, struct slave *slave); > @@ -165,5 +165,8 @@ struct slave *bond_xmit_tlb_slave_get(struct bonding *bond, > void bond_alb_monitor(struct work_struct *); > int bond_alb_set_mac_address(struct net_device *bond_dev, void *addr); > void bond_alb_clear_vlan(struct bonding *bond, unsigned short vlan_id); > +int rlb_arp_recv(const struct sk_buff *skb, struct bonding *bond, struct slave *slave); > +void tlb_init_table_entry(struct tlb_client_info *entry, int save_load); > +void rlb_init_table_entry(struct rlb_client_info *entry); > #endif /* _NET_BOND_ALB_H */ >
Powered by blists - more mailing lists