bonding: remove RTNL from three sysfs files
Subsystems (3)
cover bonding: remove RTNL from three sysfs files
First patch might fix a potential deadlock.
sysfs handlers should use rtnl_trylock() instead of rtnl_lock().
Following files can be read without acquiring RTNL :
- /sys/class/net/bonding_masters
- /sys/class/net/<name>/bonding/slaves
- /sys/class/net/<name>/bonding/queue_id
Eric Dumazet (3):
bonding: no longer use RTNL in bonding_show_bonds()
bonding: no longer use RTNL in bonding_show_slaves()
bonding: no longer use RTNL in bonding_show_queue_id()
drivers/net/bonding/bond_main.c | 6 +++--- drivers/net/bonding/bond_netlink.c | 3 ++- drivers/net/bonding/bond_options.c | 2 +- drivers/net/bonding/bond_procfs.c | 2 +- drivers/net/bonding/bond_sysfs.c | 25 ++++++++++++------------- drivers/net/bonding/bond_sysfs_slave.c | 2 +- 6 files changed, 20 insertions(+), 20 deletions(-)
Signature · 1 line
2.44.0.478.gd926399ef9-goog
PATCH 1/3 bonding: no longer use RTNL in bonding_show_bonds()
netdev structures are already RCU protected.
Change bond_init() and bond_uninit() to use RCU
enabled list_add_tail_rcu() and list_del_rcu().
Then bonding_show_bonds() can use rcu_read_lock()
while iterating through bn->dev_list.
--- drivers/net/bonding/bond_main.c | 4 ++-- drivers/net/bonding/bond_sysfs.c | 8 ++++---- 2 files changed, 6 insertions(+), 6 deletions(-)
drivers/net/bonding/bond_main.c +2 -2
@@ -5933,7 +5933,7 @@ static void bond_uninit(struct net_device *bond_dev)bond_set_slave_arr (bond,NULL ,NULL ); -list_del (&bond->bond_list); +list_del_rcu (&bond->bond_list);bond_debug_unregister (bond); }
@@ -6347,7 +6347,7 @@ static int bond_init(struct net_device *bond_dev)spin_lock_init (&bond->stats_lock);netdev_lockdep_set_classes (bond_dev); -list_add_tail (&bond->bond_list, &bn->dev_list); +list_add_tail_rcu (&bond->bond_list, &bn->dev_list);bond_prepare_sysfs_group (bond);
drivers/net/bonding/bond_sysfs.c +4 -4
@@ -37,12 +37,12 @@ static ssize_t bonding_show_bonds(const struct class *cls, {const struct bond_net *bn = container_of_const(attr, struct bond_net, class_attr_bonding_masters); -int res =0 ;struct bonding *bond; +int res =0 ; -rtnl_lock (); +rcu_read_lock (); -list_for_each_entry (bond, &bn->dev_list, bond_list) { +list_for_each_entry_rcu (bond, &bn->dev_list, bond_list) {if (res > (PAGE_SIZE -IFNAMSIZ )) {/* not enough space for another interface name */ if ((PAGE_SIZE - res) >10 )
@@ -55,7 +55,7 @@ static ssize_t bonding_show_bonds(const struct class *cls,if (res) buf[res-1 ] ='\n' ;/* eat the leftover space */ -rtnl_unlock (); +rcu_read_unlock ();return res; }
Signature · 1 line
2.44.0.478.gd926399ef9-goog
Re: [PATCH net-next 1/3] bonding: no longer use RTNL in bonding_show_bonds()
Quoted 9 lines · Eric Dumazet
netdev structures are already RCU protected.
Change
bond_init()andbond_uninit()to use RCU enabledlist_add_tail_rcu()andlist_del_rcu().Then
bonding_show_bonds()can usercu_read_lock()while iterating throughbn->dev_list.Signed-off-by Eric Dumazet <[email protected]>
Quoted 60 lines
---
drivers/net/bonding/bond_main.c| 4 ++--drivers/net/bonding/bond_sysfs.c| 8 ++++----2 files changed, 6 insertions(+), 6 deletions(-)drivers/net/bonding/bond_main.c +2
-2@@ -5933,7 +5933,7 @@ static void bond_uninit(struct net_device *bond_dev)bond_set_slave_arr (bond,NULL ,NULL ); -list_del (&bond->bond_list); +list_del_rcu (&bond->bond_list);bond_debug_unregister (bond); }@@ -6347,7 +6347,7 @@ static int bond_init(struct net_device *bond_dev)spin_lock_init (&bond->stats_lock);netdev_lockdep_set_classes (bond_dev); -list_add_tail (&bond->bond_list, &bn->dev_list); +list_add_tail_rcu (&bond->bond_list, &bn->dev_list);bond_prepare_sysfs_group (bond);drivers/net/bonding/bond_sysfs.c +4
-4@@ -37,12 +37,12 @@ static ssize_t bonding_show_bonds(const struct class *cls, {const struct bond_net *bn = container_of_const(attr, struct bond_net, class_attr_bonding_masters); -int res =0 ;struct bonding *bond; +int res =0 ; -rtnl_lock (); +rcu_read_lock (); -list_for_each_entry (bond, &bn->dev_list, bond_list) { +list_for_each_entry_rcu (bond, &bn->dev_list, bond_list) {if (res > (PAGE_SIZE -IFNAMSIZ )) {/* not enough space for another interface name */ if ((PAGE_SIZE - res) >10 )@@ -55,7 +55,7 @@ static ssize_t bonding_show_bonds(const struct class *cls,if (res) buf[res-1 ] ='\n' ;/* eat the leftover space */ -rtnl_unlock (); +rcu_read_unlock ();return res; }Signature · 1 line
2.44.0.478.gd926399ef9-goog
PATCH 2/3 bonding: no longer use RTNL in bonding_show_slaves()
In reply to Eric DumazetSlave devices are already RCU protected, simply
switch to bond_for_each_slave_rcu(),
--- drivers/net/bonding/bond_sysfs.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/net/bonding/bond_sysfs.c +3 -4
@@ -170,10 +170,9 @@ static ssize_t bonding_show_slaves(struct device *d,struct slave *slave;int res =0 ; -if (!rtnl_trylock ()) -return restart_syscall (); +rcu_read_lock (); -bond_for_each_slave (bond, slave, iter) { +bond_for_each_slave_rcu (bond, slave, iter) {if (res > (PAGE_SIZE -IFNAMSIZ )) {/* not enough space for another interface name */ if ((PAGE_SIZE - res) >10 )
@@ -184,7 +183,7 @@ static ssize_t bonding_show_slaves(struct device *d, res +=sysfs_emit_at (buf, res,"%s " , slave->dev->name); } -rtnl_unlock (); +rcu_read_unlock ();if (res) buf[res-1 ] ='\n' ;/* eat the leftover space */
Signature · 1 line
2.44.0.478.gd926399ef9-goog
Re: [PATCH net-next 2/3] bonding: no longer use RTNL in bonding_show_slaves()
Quoted 4 lines · Eric Dumazet
Slave devices are already RCU protected, simply
switch tobond_for_each_slave_rcu(),Signed-off-by Eric Dumazet <[email protected]>
Quoted 34 lines
---
drivers/net/bonding/bond_sysfs.c| 7 +++----1 file changed, 3 insertions(+), 4 deletions(-)drivers/net/bonding/bond_sysfs.c +3
-4@@ -170,10 +170,9 @@ static ssize_t bonding_show_slaves(struct device *d,struct slave *slave;int res =0 ; -if (!rtnl_trylock ()) -return restart_syscall (); +rcu_read_lock (); -bond_for_each_slave (bond, slave, iter) { +bond_for_each_slave_rcu (bond, slave, iter) {if (res > (PAGE_SIZE -IFNAMSIZ )) {/* not enough space for another interface name */ if ((PAGE_SIZE - res) >10 )@@ -184,7 +183,7 @@ static ssize_t bonding_show_slaves(struct device *d, res +=sysfs_emit_at (buf, res,"%s " , slave->dev->name); } -rtnl_unlock (); +rcu_read_unlock ();if (res) buf[res-1 ] ='\n' ;/* eat the leftover space */ Signature · 1 line
2.44.0.478.gd926399ef9-goog
PATCH 3/3 bonding: no longer use RTNL in bonding_show_queue_id()
In reply to Eric DumazetAnnotate lockless reads of slave->queue_id.
Annotate writes of slave->queue_id.
Switch bonding_show_queue_id() to rcu_read_lock()
and bond_for_each_slave_rcu().
--- drivers/net/bonding/bond_main.c | 2 +- drivers/net/bonding/bond_netlink.c | 3 ++- drivers/net/bonding/bond_options.c | 2 +- drivers/net/bonding/bond_procfs.c | 2 +- drivers/net/bonding/bond_sysfs.c | 10 +++++----- drivers/net/bonding/bond_sysfs_slave.c | 2 +- 6 files changed, 11 insertions(+), 10 deletions(-)
drivers/net/bonding/bond_main.c +1 -1
@@ -5245,7 +5245,7 @@ static inline int bond_slave_override(struct bonding *bond,/* Find out if any slaves have the same mapping as this skb. */ bond_for_each_slave_rcu (bond, slave, iter) { -if (slave->queue_id ==skb_get_queue_mapping (skb)) { +if (READ_ONCE (slave->queue_id) ==skb_get_queue_mapping (skb)) {if (bond_slave_is_up (slave) && slave->link ==BOND_LINK_UP ) {bond_dev_queue_xmit (bond, skb, slave->dev);
drivers/net/bonding/bond_netlink.c +2 -1
@@ -51,7 +51,8 @@ static int bond_fill_slave_info(struct sk_buff *skb, slave_dev->addr_len, slave->perm_hwaddr)) goto nla_put_failure; -if (nla_put_u16 (skb,IFLA_BOND_SLAVE_QUEUE_ID , slave->queue_id)) +if (nla_put_u16 (skb,IFLA_BOND_SLAVE_QUEUE_ID , +READ_ONCE (slave->queue_id))) goto nla_put_failure;if (nla_put_s32 (skb,IFLA_BOND_SLAVE_PRIO , slave->prio))
drivers/net/bonding/bond_options.c +1 -1
@@ -1589,7 +1589,7 @@ static int bond_option_queue_id_set(struct bonding *bond, goto err_no_cmd;/* Actually set the qids for the slave */ - update_slave->queue_id = qid; +WRITE_ONCE (update_slave->queue_id, qid); out:return ret;
drivers/net/bonding/bond_procfs.c +1 -1
@@ -209,7 +209,7 @@ static void bond_info_show_slave(struct seq_file *seq,seq_printf (seq,"Permanent HW addr: %*phC\n" , slave->dev->addr_len, slave->perm_hwaddr); -seq_printf (seq,"Slave queue ID: %d\n" , slave->queue_id); +seq_printf (seq,"Slave queue ID: %d\n" ,READ_ONCE (slave->queue_id));if (BOND_MODE (bond) ==BOND_MODE_8023AD ) {const struct port *port = &SLAVE_AD_INFO (slave)->port;
drivers/net/bonding/bond_sysfs.c +5 -5
@@ -625,10 +625,9 @@ static ssize_t bonding_show_queue_id(struct device *d,struct slave *slave;int res =0 ; -if (!rtnl_trylock ()) -return restart_syscall (); +rcu_read_lock (); -bond_for_each_slave (bond, slave, iter) { +bond_for_each_slave_rcu (bond, slave, iter) {if (res > (PAGE_SIZE -IFNAMSIZ -6 )) {/* not enough space for another interface_name:queue_id pair */ if ((PAGE_SIZE - res) >10 )
@@ -637,12 +636,13 @@ static ssize_t bonding_show_queue_id(struct device *d,break ; } res +=sysfs_emit_at (buf, res,"%s:%d " , - slave->dev->name, slave->queue_id); + slave->dev->name, +READ_ONCE (slave->queue_id)); }if (res) buf[res-1 ] ='\n' ;/* eat the leftover space */ -rtnl_unlock (); +rcu_read_unlock ();return res; }
drivers/net/bonding/bond_sysfs_slave.c +1 -1
@@ -53,7 +53,7 @@ static SLAVE_ATTR_RO(perm_hwaddr);static ssize_t queue_id_show (struct slave *slave,char *buf) { -return sysfs_emit (buf,"%d\n" , slave->queue_id); +return sysfs_emit (buf,"%d\n" ,READ_ONCE (slave->queue_id)); }static SLAVE_ATTR_RO (queue_id);
Signature · 1 line
2.44.0.478.gd926399ef9-goog
Re: [PATCH net-next 3/3] bonding: no longer use RTNL in bonding_show_queue_id()
Quoted 6 lines · Eric Dumazet
Annotate lockless reads of
slave->queue_id.Annotate writes of
slave->queue_id.Switch
bonding_show_queue_id()torcu_read_lock()
andbond_for_each_slave_rcu().
This is combining two logical changes into one patch, isn't it?
The annotation change isn't part of what's stated in the Subject.
-J
Quoted 113 lines
Signed-off-by Eric Dumazet <[email protected]>---
drivers/net/bonding/bond_main.c | 2 +- drivers/net/bonding/bond_netlink.c | 3 ++- drivers/net/bonding/bond_options.c | 2 +- drivers/net/bonding/bond_procfs.c | 2 +- drivers/net/bonding/bond_sysfs.c | 10 +++++----- drivers/net/bonding/bond_sysfs_slave.c | 2 +- 6 files changed, 11 insertions(+), 10 deletions(-)drivers/net/bonding/bond_main.c +1
-1@@ -5245,7 +5245,7 @@ static inline int bond_slave_override(struct bonding *bond,/* Find out if any slaves have the same mapping as this skb. */ bond_for_each_slave_rcu (bond, slave, iter) { -if (slave->queue_id ==skb_get_queue_mapping (skb)) { +if (READ_ONCE (slave->queue_id) ==skb_get_queue_mapping (skb)) {if (bond_slave_is_up (slave) && slave->link ==BOND_LINK_UP ) {bond_dev_queue_xmit (bond, skb, slave->dev);drivers/net/bonding/bond_netlink.c +2
-1@@ -51,7 +51,8 @@ static int bond_fill_slave_info(struct sk_buff *skb, slave_dev->addr_len, slave->perm_hwaddr)) goto nla_put_failure; -if (nla_put_u16 (skb,IFLA_BOND_SLAVE_QUEUE_ID , slave->queue_id)) +if (nla_put_u16 (skb,IFLA_BOND_SLAVE_QUEUE_ID , +READ_ONCE (slave->queue_id))) goto nla_put_failure;if (nla_put_s32 (skb,IFLA_BOND_SLAVE_PRIO , slave->prio))drivers/net/bonding/bond_options.c +1
-1@@ -1589,7 +1589,7 @@ static int bond_option_queue_id_set(struct bonding *bond, goto err_no_cmd;/* Actually set the qids for the slave */ - update_slave->queue_id = qid; +WRITE_ONCE (update_slave->queue_id, qid);out:
return ret;drivers/net/bonding/bond_procfs.c +1
-1@@ -209,7 +209,7 @@ static void bond_info_show_slave(struct seq_file *seq,seq_printf (seq,"Permanent HW addr: %*phC\n" , slave->dev->addr_len, slave->perm_hwaddr); -seq_printf (seq,"Slave queue ID: %d\n" , slave->queue_id); +seq_printf (seq,"Slave queue ID: %d\n" ,READ_ONCE (slave->queue_id));if (BOND_MODE (bond) ==BOND_MODE_8023AD ) {const struct port *port = &SLAVE_AD_INFO (slave)->port;drivers/net/bonding/bond_sysfs.c +5
-5@@ -625,10 +625,9 @@ static ssize_t bonding_show_queue_id(struct device *d,struct slave *slave;int res =0 ; -if (!rtnl_trylock ()) -return restart_syscall (); +rcu_read_lock (); -bond_for_each_slave (bond, slave, iter) { +bond_for_each_slave_rcu (bond, slave, iter) {if (res > (PAGE_SIZE -IFNAMSIZ -6 )) {/* not enough space for another interface_name:queue_id pair */ if ((PAGE_SIZE - res) >10 )@@ -637,12 +636,13 @@ static ssize_t bonding_show_queue_id(struct device *d,break ; } res +=sysfs_emit_at (buf, res,"%s:%d " , - slave->dev->name, slave->queue_id); + slave->dev->name, +READ_ONCE (slave->queue_id)); }if (res) buf[res-1 ] ='\n' ;/* eat the leftover space */ -rtnl_unlock (); +rcu_read_unlock ();return res; }drivers/net/bonding/bond_sysfs_slave.c +1
-1@@ -53,7 +53,7 @@ static SLAVE_ATTR_RO(perm_hwaddr);static ssize_t queue_id_show (struct slave *slave,char *buf) { -return sysfs_emit (buf,"%d\n" , slave->queue_id); +return sysfs_emit (buf,"%d\n" ,READ_ONCE (slave->queue_id)); }static SLAVE_ATTR_RO (queue_id);Signature · 1 line
2.44.0.478.gd926399ef9-goog
---
-Jay Vosburgh, [email protected]
Re: [PATCH net-next 3/3] bonding: no longer use RTNL in bonding_show_queue_id()
Quoted 5 lines · Jay Vosburgh
Quoted 2 lines · Eric Dumazet
Annotate lockless reads of
slave->queue_id.
Quoted 7 lines
Quoted 4 lines
Annotate writes of
slave->queue_id.Switch
bonding_show_queue_id()torcu_read_lock()
andbond_for_each_slave_rcu().This is combining two logical changes into one patch, isn't it?
The annotation change isn't part of what's stated in the Subject.
The annotations are really part of this change, otherwise KCSAN might find races.
Jay Vosburgh
Acked-byRe: [PATCH net-next 3/3] bonding: no longer use RTNL in bonding_show_queue_id()
Quoted 8 lines · Eric Dumazet
Quoted 7 lines · Jay Vosburgh
Quoted 4 lines · Eric Dumazet
Annotate lockless reads of
slave->queue_id.Annotate writes of
slave->queue_id.
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
Switch
bonding_show_queue_id()torcu_read_lock()
andbond_for_each_slave_rcu().This is combining two logical changes into one patch, isn't it?
The annotation change isn't part of what's stated in the Subject.The annotations are really part of this change, otherwise KCSAN might find races.
Works for me.
-J
---
-Jay Vosburgh, [email protected]
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Hello:
This series was applied to netdev/net-next.git (main) by Jakub Kicinski <[email protected]>:
Quoted 10 lines · On Mon, 8 Apr 2024 19:04:34 +0000 you wrote:
First patch might fix a potential deadlock.
sysfs handlers should usertnl_trylock()instead ofrtnl_lock().Following files can be read without acquiring RTNL :
-
/sys/class/net/bonding_masters
- /sys/class/net/<name>/bonding/slaves
- /sys/class/net/<name>/bonding/queue_id[...]
Here is the summary with links:
- [net-next,1/3] bonding: no longer use RTNL in bonding_show_bonds()
https://git.kernel.org/netdev/net-next/c/6c5d17143fa4 - [net-next,2/3] bonding: no longer use RTNL in bonding_show_slaves() https://git.kernel.org/netdev/net-next/c/d67fed98caa1 - [net-next,3/3] bonding: no longer use RTNL in bonding_show_queue_id() https://git.kernel.org/netdev/net-next/c/662e451d9a62
You are awesome, thank you!
Signature · 2 lines
Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html