bonding: Fix race condition between bond_enslave() and bond_3ad_update_lacp_rate()
Subsystems (3)
PATCH 1/3 bonding: Fix race condition between bond_enslave() and bond_3ad_update_lacp_rate()
port->slave can be NULL since it's being initialized in bond_enslave thus dereferencing a NULL pointer in bond_3ad_update_lacp_rate() Also fix a minor bug, which could cause a port not to have AD_STATE_LACP_TIMEOUT since there's no sync between bond_3ad_update_lacp_rate() and bond_3ad_bind_slave(), by changing the read_lock to a write_lock_bh in bond_3ad_update_lacp_rate().
--- drivers/net/bonding/bond_3ad.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/net/bonding/bond_3ad.c +4 -2
@@ -2494,11 +2494,13 @@ void bond_3ad_update_lacp_rate(struct bonding *bond)struct port *port =NULL ;int lacp_fast; -read_lock (&bond->lock); +write_lock_bh (&bond->lock); lacp_fast = bond->params.lacp_fast;bond_for_each_slave (bond, slave, i) { port = &(SLAVE_AD_INFO (slave).port); +if (port->slave ==NULL ) +continue ;__get_state_machine_lock (port);if (lacp_fast) port->actor_oper_port_state |=AD_STATE_LACP_TIMEOUT ;
@@ -2507,5 +2509,5 @@ void bond_3ad_update_lacp_rate(struct bonding *bond)__release_state_machine_lock (port); } -read_unlock (&bond->lock); +write_unlock_bh (&bond->lock); }
Signature · 1 line
1.7.11.7
PATCH 2/3 bonding: Fix initialize after use for 3ad machine state spinlock
The 3ad machine state spinlock can be used before it is inititialized while doing bond_enslave() (and the port is being initialized) since port->slave is set before the lock is prepared, thus causing soft lock-ups and a multitude of other nasty bugs.
--- drivers/net/bonding/bond_3ad.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-)
drivers/net/bonding/bond_3ad.c +4 -5
@@ -389,13 +389,13 @@ static u8 __get_duplex(struct port *port)/** * __initialize_port_locks - initialize a port's STATE machine spinlock -* @port: the port we're looking at +* @port: the slave of the port we're looking at * */ -static inline void __initialize_port_locks (struct port *port) +static inline void __initialize_port_locks (struct slave *port) {// make sure it isn't called twice -spin_lock_init (&(SLAVE_AD_INFO (port->slave).state_machine_lock)); +spin_lock_init (&(SLAVE_AD_INFO (port).state_machine_lock)); }//conversions
@@ -1910,6 +1910,7 @@ int bond_3ad_bind_slave(struct slave *slave)ad_initialize_port (port, bond->params.lacp_fast); +__initialize_port_locks (slave); port->slave = slave; port->actor_port_number =SLAVE_AD_INFO (slave).id;// key is determined according to the link speed, duplex and user key(which is yet not supported)
@@ -1932,8 +1933,6 @@ int bond_3ad_bind_slave(struct slave *slave) port->next_port_in_aggregator =NULL ;__disable_port (port); -__initialize_port_locks (port); -// aggregator initialization aggregator = &(SLAVE_AD_INFO (slave).aggregator);
Signature · 1 line
1.7.11.7
Re: [PATCH net 2/3] bonding: Fix initialize after use for 3ad machine state spinlock
Quoted 4 lines · Nikolay Aleksandrov
The 3ad machine state spinlock can be used before it is inititialized while doing
bond_enslave()(and the port is being initialized) sinceport->slaveis set before the lock is prepared, thus causing soft lock-ups and a multitude of other nasty bugs.
Does this change cause the "uninitialized port" warnings inbond_3ad_state_machine_handler and bond_3ad_rx_indication to intermittently print during the enslavement process? If so (and it looks to me like it will), I think the warnings should be removed, since after this change, port->slave being NULL isn't really an error condition that needs a warning to the log.
Quoted 13 lines
Signed-off-by Nikolay Aleksandrov <[email protected]>---
drivers/net/bonding/bond_3ad.c| 9 ++++-----1 file changed, 4 insertions(+), 5 deletions(-)drivers/net/bonding/bond_3ad.c +0
-0@@ -389,13 +389,13 @@ static u8 __get_duplex(struct port *port)/** * __initialize_port_locks - initialize a port's STATE machine spinlock
drivers/net/bonding/bond_3ad.c static u8 __get_duplex(struct port *port) · line 398
Quoted 10 lines
drivers/net/bonding/bond_3ad.c +3
-3-* @port: the port we're looking at +* @port: the slave of the port we're looking at * */ -static inline void __initialize_port_locks (struct port *port) +static inline void __initialize_port_locks (struct slave *port) {// make sure it isn't called twice -spin_lock_init (&(SLAVE_AD_INFO (port->slave).state_machine_lock)); +spin_lock_init (&(SLAVE_AD_INFO (port).state_machine_lock));
Change the name of the variable here, too, not just the type.
This is confusing.
-J
Quoted 23 lines
drivers/net/bonding/bond_3ad.c +1
-2}//conversions @@ -1910,6 +1910,7 @@ int bond_3ad_bind_slave(struct slave *slave)ad_initialize_port (port, bond->params.lacp_fast); +__initialize_port_locks (slave); port->slave = slave; port->actor_port_number =SLAVE_AD_INFO (slave).id;// key is determined according to the link speed, duplex and user key(which is yet not supported) @@ -1932,8 +1933,6 @@ int bond_3ad_bind_slave(struct slave *slave) port->next_port_in_aggregator =NULL ;__disable_port (port); -__initialize_port_locks (port); -// aggregator initialization aggregator = &(SLAVE_AD_INFO (slave).aggregator);Signature · 1 line
1.7.11.7
---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]
Re: [PATCH net 2/3] bonding: Fix initialize after use for 3ad machine state spinlock
Quoted 7 lines · Jay Vosburgh
Quoted 4 lines · Nikolay Aleksandrov
The 3ad machine state spinlock can be used before it is inititialized while doing
bond_enslave()(and the port is being initialized) sinceport->slaveis set before the lock is prepared, thus causing soft lock-ups and a multitude of other nasty bugs.
Quoted 7 lines
Does this change cause the "uninitialized port" warnings in
bond_3ad_state_machine_handlerandbond_3ad_rx_indicationto intermittently print during the enslavement process? If so (and it looks to me like it will), I think the warnings should be removed, since after this change,port->slavebeing NULL isn't really an error condition that needs a warning to the log.
This change couldn't cause that, it only initializes the spin lock before the slave is set, currently after the first patch of this series this is no longer a requirement as far as I can tell the only code that can access the lock before the slave is set was that one, but it still is a bug that can manifest later. I don't think it has anything to do with the warnings, the only change is that the spin lock is initialized prior to setting the slave to the port.
Am I missing something here ?
Quoted 27 lines
Quoted 13 lines
Signed-off-by Nikolay Aleksandrov <[email protected]>---
drivers/net/bonding/bond_3ad.c| 9 ++++-----1 file changed, 4 insertions(+), 5 deletions(-)drivers/net/bonding/bond_3ad.c +0
-0@@ -389,13 +389,13 @@ static u8 __get_duplex(struct port *port)/** * __initialize_port_locks - initialize a port's STATE machine spinlock drivers/net/bonding/bond_3ad.c static u8 __get_duplex(struct port *port) · line 398
Quoted 10 lines
drivers/net/bonding/bond_3ad.c +3
-3-* @port: the port we're looking at +* @port: the slave of the port we're looking at * */ -static inline void __initialize_port_locks (struct port *port) +static inline void __initialize_port_locks (struct slave *port) {// make sure it isn't called twice -spin_lock_init (&(SLAVE_AD_INFO (port->slave).state_machine_lock)); +spin_lock_init (&(SLAVE_AD_INFO (port).state_machine_lock));Change the name of the variable here, too, not just the type.
This is confusing.
Quoted 1 line
-J
Thanks, I saw that after posting, I have prepared this change already.
Quoted 28 lines
Quoted 24 lines
drivers/net/bonding/bond_3ad.c +1
-2}//conversions @@ -1910,6 +1910,7 @@ int bond_3ad_bind_slave(struct slave *slave)ad_initialize_port (port, bond->params.lacp_fast); +__initialize_port_locks (slave); port->slave = slave; port->actor_port_number =SLAVE_AD_INFO (slave).id;// key is determined according to the link speed, duplex and user key(which is yet not supported) @@ -1932,8 +1933,6 @@ int bond_3ad_bind_slave(struct slave *slave) port->next_port_in_aggregator =NULL ;__disable_port (port); -__initialize_port_locks (port); -// aggregator initialization aggregator = &(SLAVE_AD_INFO (slave).aggregator);Signature · 1 line
1.7.11.7---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]
David Miller
maintainerRe: [PATCH net 2/3] bonding: Fix initialize after use for 3ad machine state spinlock
In reply to Jay VosburghFrom: Jay Vosburgh <[email protected]>
Date: Mon, 18 Feb 2013 13:33:10 -0800
Quoted 15 lines
Quoted 4 lines · Nikolay Aleksandrov
The 3ad machine state spinlock can be used before it is inititialized while doing
bond_enslave()(and the port is being initialized) sinceport->slaveis set before the lock is prepared, thus causing soft lock-ups and a multitude of other nasty bugs.Does this change cause the "uninitialized port" warnings in
bond_3ad_state_machine_handlerandbond_3ad_rx_indicationto intermittently print during the enslavement process? If so (and it looks to me like it will), I think the warnings should be removed, since after this change,port->slavebeing NULL isn't really an error condition that needs a warning to the log.Quoted 1 line
Signed-off-by Nikolay Aleksandrov <[email protected]>
...
Quoted 9 lines
drivers/net/bonding/bond_3ad.c static u8 __get_duplex(struct port *port) · line 398
Quoted 6 lines
drivers/net/bonding/bond_3ad.c +1
-1-static inline void __initialize_port_locks (struct port *port) +static inline void __initialize_port_locks (struct slave *port){ // make sure it isn't called twicedrivers/net/bonding/bond_3ad.c +1
-1-spin_lock_init (&(SLAVE_AD_INFO (port->slave).state_machine_lock)); +spin_lock_init (&(SLAVE_AD_INFO (port).state_machine_lock));Change the name of the variable here, too, not just the type.
This is confusing.
I made this adjustment and applied Nikolay's patch.
PATCH 3/3 bonding: fix bond_release_all inconsistencies
In reply to Nikolay AleksandrovThis patch fixes the following inconsistencies in bond_release_all:
- IFF_BONDING flag is not stripped from slaves
- MTU is not restored
- no netdev notifiers are sent
Instead of trying to keep bond_release and bond_release_all in sync I think we can re-use bond_release as the environment for calling it is correct (RTNL is held). I have been running tests for the past week and they came out successful. The only way for bond_release to fail is for the slave to be attached in a different bond or to not be a slave but that cannot happen as RTNL is held and no slave manipulations can be achieved.
--- drivers/net/bonding/bond_main.c | 106 ++-------------------------------------- 1 file changed, 5 insertions(+), 101 deletions(-)
drivers/net/bonding/bond_main.c +5 -101
@@ -2140,113 +2140,17 @@ static int bond_release_and_destroy(struct net_device *bond_dev,/* * This function releases all slaves. */ -static int bond_release_all (struct net_device *bond_dev) +static void bond_release_all (struct net_device *bond_dev) {struct bonding *bond =netdev_priv (bond_dev); -struct slave *slave; -struct net_device *slave_dev; -struct sockaddr addr; - -write_lock_bh (&bond->lock); - -netif_carrier_off (bond_dev);if (bond->slave_cnt ==0 ) - goto out; - - bond->current_arp_slave =NULL ; - bond->primary_slave =NULL ; -bond_change_active_slave (bond,NULL ); - -while ((slave = bond->first_slave) !=NULL ) { -/* Inform AD package of unbinding of slave -* before slave is detached from the list. -*/ -if (bond->params.mode ==BOND_MODE_8023AD ) -bond_3ad_unbind_slave (slave); - - slave_dev = slave->dev; -bond_detach_slave (bond, slave); - -/* now that the slave is detached, unlock and perform -* all the undo steps that should not be called from -* within a lock. -*/ -write_unlock_bh (&bond->lock); - -/* unregister rx_handler early so bond_handle_frame wouldn't -* be called for this slave anymore. -*/ -netdev_rx_handler_unregister (slave_dev); -synchronize_net (); - -if (bond_is_lb (bond)) { -/* must be called only after the slave -* has been detached from the list -*/ -bond_alb_deinit_slave (bond, slave); - } - -bond_destroy_slave_symlinks (bond_dev, slave_dev); -bond_del_vlans_from_slave (bond, slave_dev); - -/* If the mode USES_PRIMARY, then we should only remove its -* promisc and mc settings if it was the curr_active_slave, but that was -* already taken care of above when we detached the slave -*/ -if (!USES_PRIMARY (bond->params.mode)) { -/* unset promiscuity level from slave */ -if (bond_dev->flags &IFF_PROMISC ) -dev_set_promiscuity (slave_dev,-1 ); - -/* unset allmulti level from slave */ -if (bond_dev->flags &IFF_ALLMULTI ) -dev_set_allmulti (slave_dev,-1 ); - -/* flush master's mc_list from slave */ -netif_addr_lock_bh (bond_dev); -bond_mc_list_flush (bond_dev, slave_dev); -netif_addr_unlock_bh (bond_dev); - } - -bond_upper_dev_unlink (bond_dev, slave_dev); - -slave_disable_netpoll (slave); - -/* close slave before restoring its mac address */ -dev_close (slave_dev); - -if (!bond->params.fail_over_mac) { -/* restore original ("permanent") mac address*/ -memcpy (addr.sa_data, slave->perm_hwaddr,ETH_ALEN ); - addr.sa_family = slave_dev->type; -dev_set_mac_address (slave_dev, &addr); - } - -kfree (slave); - -/* re-acquire the lock before getting the next slave */ -write_lock_bh (&bond->lock); - } - -eth_hw_addr_random (bond_dev); - bond->dev_addr_from_first = true; - -if (bond_vlan_used (bond)) { -pr_warning ("%s: Warning: clearing HW address of %s while it still has VLANs.\n" , - bond_dev->name, bond_dev->name); -pr_warning ("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n" , - bond_dev->name); - } - +return ; +while (bond->first_slave !=NULL ) +bond_release (bond_dev, bond->first_slave->dev);pr_info ("%s: released all slaves\n" , bond_dev->name); -out: -write_unlock_bh (&bond->lock); - -bond_compute_features (bond); - -return 0 ; +return ; }/*
Signature · 1 line
1.7.11.7
Re: [PATCH net-next 3/3] bonding: fix bond_release_all inconsistencies
Quoted 11 lines · Nikolay Aleksandrov
This patch fixes the following inconsistencies in
bond_release_all:
-IFF_BONDINGflag is not stripped from slaves
- MTU is not restored
- no netdev notifiers are sent
Instead of trying to keepbond_releaseandbond_release_allin sync I think we can re-usebond_releaseas the environment for calling it is correct (RTNL is held). I have been running tests for the past week and they came out successful. The only way forbond_releaseto fail is for the slave to be attached in a different bond or to not be a slave but that cannot happen as RTNL is held and no slave manipulations can be achieved.
It might be worthwhile to add an "all" argument to bond_release
that skips some things that don't make sense if all slaves are being released. I'm thinking in particular of this block:
if (oldcurrent == slave) {
/*
* Note that we hold RTNL over this sequence, so there
* is no concern that another slave add/remove event
* will interfere.
*/
write_unlock_bh(&bond->lock);
read_lock(&bond->lock);
write_lock_bh(&bond->curr_slave_lock);
bond_select_active_slave(bond);
write_unlock_bh(&bond->curr_slave_lock);
read_unlock(&bond->lock);
write_lock_bh(&bond->lock);
}
as it's written now, for the release all case, the code may goto the trouble of assigning a new active slave each time one slave is removed (including various log messages, maybe sending IGMPs, etc). If all slaves are being removed, that's pointless. This could be something like:
if (release_all) {
bond->curr_active_slave = NULL;
} else if (oldcurrent == slave) {
[ the current block of stuff ]
}
it's safe here to unconditionally set curr_active_slave to NULLbecause we hold bond->lock for write. The lock dance stuff for the bond_select_active_slave() call is to satisfy its locking requirements.
-J
Quoted 130 lines
Signed-off-by Nikolay Aleksandrov <[email protected]>---
drivers/net/bonding/bond_main.c| 106 ++--------------------------------------1 file changed, 5 insertions(+), 101 deletions(-)drivers/net/bonding/bond_main.c +5
-101@@ -2140,113 +2140,17 @@ static int bond_release_and_destroy(struct net_device *bond_dev,/* * This function releases all slaves. */ -static int bond_release_all (struct net_device *bond_dev) +static void bond_release_all (struct net_device *bond_dev) {struct bonding *bond =netdev_priv (bond_dev); -struct slave *slave; -struct net_device *slave_dev; -struct sockaddr addr; - -write_lock_bh (&bond->lock); - -netif_carrier_off (bond_dev);if (bond->slave_cnt ==0 ) - goto out; - - bond->current_arp_slave =NULL ; - bond->primary_slave =NULL ; -bond_change_active_slave (bond,NULL ); - -while ((slave = bond->first_slave) !=NULL ) { -/* Inform AD package of unbinding of slave -* before slave is detached from the list. -*/ -if (bond->params.mode ==BOND_MODE_8023AD ) -bond_3ad_unbind_slave (slave); - - slave_dev = slave->dev; -bond_detach_slave (bond, slave); - -/* now that the slave is detached, unlock and perform -* all the undo steps that should not be called from -* within a lock. -*/ -write_unlock_bh (&bond->lock); - -/* unregister rx_handler early so bond_handle_frame wouldn't -* be called for this slave anymore. -*/ -netdev_rx_handler_unregister (slave_dev); -synchronize_net (); - -if (bond_is_lb (bond)) { -/* must be called only after the slave -* has been detached from the list -*/ -bond_alb_deinit_slave (bond, slave); - } - -bond_destroy_slave_symlinks (bond_dev, slave_dev); -bond_del_vlans_from_slave (bond, slave_dev); - -/* If the mode USES_PRIMARY, then we should only remove its -* promisc and mc settings if it was the curr_active_slave, but that was -* already taken care of above when we detached the slave -*/ -if (!USES_PRIMARY (bond->params.mode)) { -/* unset promiscuity level from slave */ -if (bond_dev->flags &IFF_PROMISC ) -dev_set_promiscuity (slave_dev,-1 ); - -/* unset allmulti level from slave */ -if (bond_dev->flags &IFF_ALLMULTI ) -dev_set_allmulti (slave_dev,-1 ); - -/* flush master's mc_list from slave */ -netif_addr_lock_bh (bond_dev); -bond_mc_list_flush (bond_dev, slave_dev); -netif_addr_unlock_bh (bond_dev); - } - -bond_upper_dev_unlink (bond_dev, slave_dev); - -slave_disable_netpoll (slave); - -/* close slave before restoring its mac address */ -dev_close (slave_dev); - -if (!bond->params.fail_over_mac) { -/* restore original ("permanent") mac address*/ -memcpy (addr.sa_data, slave->perm_hwaddr,ETH_ALEN ); - addr.sa_family = slave_dev->type; -dev_set_mac_address (slave_dev, &addr); - } - -kfree (slave); - -/* re-acquire the lock before getting the next slave */ -write_lock_bh (&bond->lock); - } - -eth_hw_addr_random (bond_dev); - bond->dev_addr_from_first = true; - -if (bond_vlan_used (bond)) { -pr_warning ("%s: Warning: clearing HW address of %s while it still has VLANs.\n" , - bond_dev->name, bond_dev->name); -pr_warning ("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n" , - bond_dev->name); - } - +return ; +while (bond->first_slave !=NULL ) +bond_release (bond_dev, bond->first_slave->dev);pr_info ("%s: released all slaves\n" , bond_dev->name); -out: -write_unlock_bh (&bond->lock); - -bond_compute_features (bond); - -return 0 ; +return ; }/* Signature · 1 line
1.7.11.7
---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]
Re: [PATCH net-next 3/3] bonding: fix bond_release_all inconsistencies
Quoted 48 lines · Jay Vosburgh
Quoted 11 lines · Nikolay Aleksandrov
This patch fixes the following inconsistencies in
bond_release_all:
-IFF_BONDINGflag is not stripped from slaves
- MTU is not restored
- no netdev notifiers are sent
Instead of trying to keepbond_releaseandbond_release_allin sync I think we can re-usebond_releaseas the environment for calling it is correct (RTNL is held). I have been running tests for the past week and they came out successful. The only way forbond_releaseto fail is for the slave to be attached in a different bond or to not be a slave but that cannot happen as RTNL is held and no slave manipulations can be achieved.It might be worthwhile to add an "all" argument to
bond_release
that skips some things that don't make sense if all slaves are being released. I'm thinking in particular of this block:if (oldcurrent == slave) { /* * Note that we hold RTNL over this sequence, so there * is no concern that another slave add/remove event * will interfere. */ write_unlock_bh(&bond->lock); read_lock(&bond->lock); write_lock_bh(&bond->curr_slave_lock); bond_select_active_slave(bond); write_unlock_bh(&bond->curr_slave_lock); read_unlock(&bond->lock); write_lock_bh(&bond->lock); } as it's written now, for the release all case, the code may goto the trouble of assigning a new active slave each time one slave is removed (including various log messages, maybe sending IGMPs, etc). If all slaves are being removed, that's pointless. This could be something like:
if (release_all) { bond->curr_active_slave = NULL; } else if (oldcurrent == slave) { [ the current block of stuff ] } it's safe here to unconditionally set curr_active_slave to NULL
Quoted 4 lines
because we hold
bond->lockfor write. The lock dance stuff for thebond_select_active_slave()call is to satisfy its locking requirements.-J
I see your point and I agree. I will prepare another version that incorporates it, although I can't add it as an argument since bond_release is used as ndo_del_slave. I'll have to make it a global variable.
Nik
Re: [PATCH net-next 3/3] bonding: fix bond_release_all inconsistencies
Quoted 52 lines · Nikolay Aleksandrov
Quoted 51 lines · Jay Vosburgh
Quoted 11 lines · Nikolay Aleksandrov
This patch fixes the following inconsistencies in
bond_release_all:
-IFF_BONDINGflag is not stripped from slaves
- MTU is not restored
- no netdev notifiers are sent
Instead of trying to keepbond_releaseandbond_release_allin sync I think we can re-usebond_releaseas the environment for calling it is correct (RTNL is held). I have been running tests for the past week and they came out successful. The only way forbond_releaseto fail is for the slave to be attached in a different bond or to not be a slave but that cannot happen as RTNL is held and no slave manipulations can be achieved.It might be worthwhile to add an "all" argument to
bond_release
that skips some things that don't make sense if all slaves are being released. I'm thinking in particular of this block:if (oldcurrent == slave) { /* * Note that we hold RTNL over this sequence, so there * is no concern that another slave add/remove event * will interfere. */ write_unlock_bh(&bond->lock); read_lock(&bond->lock); write_lock_bh(&bond->curr_slave_lock); bond_select_active_slave(bond); write_unlock_bh(&bond->curr_slave_lock); read_unlock(&bond->lock); write_lock_bh(&bond->lock); } as it's written now, for the release all case, the code may goto the trouble of assigning a new active slave each time one slave is removed (including various log messages, maybe sending IGMPs, etc). If all slaves are being removed, that's pointless. This could be something like:
if (release_all) { bond->curr_active_slave = NULL; } else if (oldcurrent == slave) { [ the current block of stuff ] } it's safe here to unconditionally set curr_active_slave to NULLbecause we hold
bond->lockfor write. The lock dance stuff for thebond_select_active_slave()call is to satisfy its locking requirements.
Quoted 5 lines
Quoted 1 line
-JI see your point and I agree. I will prepare another version that incorporates it, although I can't add it as an argument since
bond_releaseis used asndo_del_slave. I'll have to make it a global variable.
No, just rename the current bond_release to __bond_release_one,
add the extra argument, and create a new bond_release .ndo_del_slave that calls __bond_release_one with "all=0". Then, bond_release_all calls __bond_release_one with all=1.
Also, there's only one caller of bond_release_all, and since the
new & improved bond_release_all is trivial, it could be open coded into bond_uninit, eliminating bond_release_all as a function.
-J
---
-Jay Vosburgh, IBM Linux Technology Center, [email protected]
Quoted 6 lines · Nikolay Aleksandrov
port->slavecan be NULL since it's being initialized inbond_enslavethus dereferencing a NULL pointer inbond_3ad_update_lacp_rate()Also fix a minor bug, which could cause a port not to haveAD_STATE_LACP_TIMEOUTsince there's no sync betweenbond_3ad_update_lacp_rate()andbond_3ad_bind_slave(), by changing theread_lockto awrite_lock_bhinbond_3ad_update_lacp_rate().
Quoted 34 lines
Signed-off-by Nikolay Aleksandrov <[email protected]>---
drivers/net/bonding/bond_3ad.c| 6 ++++--1 file changed, 4 insertions(+), 2 deletions(-)drivers/net/bonding/bond_3ad.c +4
-2@@ -2494,11 +2494,13 @@ void bond_3ad_update_lacp_rate(struct bonding *bond)struct port *port =NULL ;int lacp_fast; -read_lock (&bond->lock); +write_lock_bh (&bond->lock); lacp_fast = bond->params.lacp_fast;bond_for_each_slave (bond, slave, i) { port = &(SLAVE_AD_INFO (slave).port); +if (port->slave ==NULL ) +continue ;__get_state_machine_lock (port);if (lacp_fast) port->actor_oper_port_state |=AD_STATE_LACP_TIMEOUT ;@@ -2507,5 +2509,5 @@ void bond_3ad_update_lacp_rate(struct bonding *bond)__release_state_machine_lock (port); } -read_unlock (&bond->lock); +write_unlock_bh (&bond->lock); }Signature · 1 line
1.7.11.7
From: Jay Vosburgh <[email protected]>
Date: Mon, 18 Feb 2013 13:09:13 -0800
Quoted 12 lines
Quoted 6 lines · Nikolay Aleksandrov
port->slavecan be NULL since it's being initialized inbond_enslavethus dereferencing a NULL pointer inbond_3ad_update_lacp_rate()Also fix a minor bug, which could cause a port not to haveAD_STATE_LACP_TIMEOUTsince there's no sync betweenbond_3ad_update_lacp_rate()andbond_3ad_bind_slave(), by changing theread_lockto awrite_lock_bhinbond_3ad_update_lacp_rate().Signed-off-by Jay Vosburgh <[email protected]>Quoted 1 line
Signed-off-by Nikolay Aleksandrov <[email protected]>
Applied
PATCH v2 3/3 bonding: fix bond_release_all inconsistencies
In reply to Nikolay AleksandrovThis patch fixes the following inconsistencies in bond_release_all:
- IFF_BONDING flag is not stripped from slaves
- MTU is not restored
- no netdev notifiers are sent
Instead of trying to keep bond_release and bond_release_all in sync I think we can re-use bond_release as the environment for calling it is correct (RTNL is held). I have been running tests for the past week and they came out successful. The only way for bond_release to fail is for the slave to be attached in a different bond or to not be a slave but that cannot happen as RTNL is held and no slave manipulations can be achieved.
V2: As suggested bond_release is renamed to __bond_release_one with a new parameter "all" introduced so to avoid calling unnecessary code while destroying a bond, and a wrapper for it called bond_release is created because of ndo_del_link. bond_release_all() is removed.
--- drivers/net/bonding/bond_main.c | 135 ++++++---------------------------------- 1 file changed, 18 insertions(+), 117 deletions(-)
drivers/net/bonding/bond_main.c +18 -117
@@ -1937,7 +1937,8 @@ err_undo_flags:/* * Try to release the slave device <slave> from the bond device <master> * It is legal to access curr_active_slave without a lock because all the function -* is write-locked. +* is write-locked. If "all" is true it means that the function is being called +* while destroying a bond interface and all slaves are being released. * * The rules for slave state should be: * for Active/Backup:
@@ -1945,7 +1946,9 @@ err_undo_flags:* for Bonded connections: * The first up interface should be left on and all others downed. */ -int bond_release (struct net_device *bond_dev,struct net_device *slave_dev) +static int __bond_release_one (struct net_device *bond_dev, +struct net_device *slave_dev, +bool all) {struct bonding *bond =netdev_priv (bond_dev);struct slave *slave, *oldcurrent;
@@ -1982,7 +1985,7 @@ int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)synchronize_net ();write_lock_bh (&bond->lock); -if (!bond->params.fail_over_mac) { +if (!all && !bond->params.fail_over_mac) {if (ether_addr_equal (bond_dev->dev_addr, slave->perm_hwaddr) && bond->slave_cnt >1 ) pr_warning("%s: Warning: the permanent HWaddr of %s - %pM - is still in use by %s. Set the HWaddr of %s to a different address to avoid conflicts.\n",
@@ -2028,7 +2031,9 @@ int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)write_lock_bh (&bond->lock); } -if (oldcurrent == slave) { +if (all) { + bond->curr_active_slave =NULL ; + }else if (oldcurrent == slave) {/* * Note that we hold RTNL over this sequence, so there * is no concern that another slave add/remove event
@@ -2117,6 +2122,12 @@ int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)return 0 ;/* deletion OK */ } +/* A wrapper used because of ndo_del_link */ +int bond_release (struct net_device *bond_dev,struct net_device *slave_dev) +{ +return __bond_release_one (bond_dev, slave_dev, false); +} +/* * First release a slave and then destroy the bond if no more slaves are left. * Must be under rtnl_lock when this function is called.
@@ -2138,118 +2149,6 @@ static int bond_release_and_destroy(struct net_device *bond_dev, }/* -* This function releases all slaves. -*/ -static int bond_release_all (struct net_device *bond_dev) -{ -struct bonding *bond =netdev_priv (bond_dev); -struct slave *slave; -struct net_device *slave_dev; -struct sockaddr addr; - -write_lock_bh (&bond->lock); - -netif_carrier_off (bond_dev); - -if (bond->slave_cnt ==0 ) - goto out; - - bond->current_arp_slave =NULL ; - bond->primary_slave =NULL ; -bond_change_active_slave (bond,NULL ); - -while ((slave = bond->first_slave) !=NULL ) { -/* Inform AD package of unbinding of slave -* before slave is detached from the list. -*/ -if (bond->params.mode ==BOND_MODE_8023AD ) -bond_3ad_unbind_slave (slave); - - slave_dev = slave->dev; -bond_detach_slave (bond, slave); - -/* now that the slave is detached, unlock and perform -* all the undo steps that should not be called from -* within a lock. -*/ -write_unlock_bh (&bond->lock); - -/* unregister rx_handler early so bond_handle_frame wouldn't -* be called for this slave anymore. -*/ -netdev_rx_handler_unregister (slave_dev); -synchronize_net (); - -if (bond_is_lb (bond)) { -/* must be called only after the slave -* has been detached from the list -*/ -bond_alb_deinit_slave (bond, slave); - } - -bond_destroy_slave_symlinks (bond_dev, slave_dev); -bond_del_vlans_from_slave (bond, slave_dev); - -/* If the mode USES_PRIMARY, then we should only remove its -* promisc and mc settings if it was the curr_active_slave, but that was -* already taken care of above when we detached the slave -*/ -if (!USES_PRIMARY (bond->params.mode)) { -/* unset promiscuity level from slave */ -if (bond_dev->flags &IFF_PROMISC ) -dev_set_promiscuity (slave_dev,-1 ); - -/* unset allmulti level from slave */ -if (bond_dev->flags &IFF_ALLMULTI ) -dev_set_allmulti (slave_dev,-1 ); - -/* flush master's mc_list from slave */ -netif_addr_lock_bh (bond_dev); -bond_mc_list_flush (bond_dev, slave_dev); -netif_addr_unlock_bh (bond_dev); - } - -bond_upper_dev_unlink (bond_dev, slave_dev); - -slave_disable_netpoll (slave); - -/* close slave before restoring its mac address */ -dev_close (slave_dev); - -if (!bond->params.fail_over_mac) { -/* restore original ("permanent") mac address*/ -memcpy (addr.sa_data, slave->perm_hwaddr,ETH_ALEN ); - addr.sa_family = slave_dev->type; -dev_set_mac_address (slave_dev, &addr); - } - -kfree (slave); - -/* re-acquire the lock before getting the next slave */ -write_lock_bh (&bond->lock); - } - -eth_hw_addr_random (bond_dev); - bond->dev_addr_from_first = true; - -if (bond_vlan_used (bond)) { -pr_warning ("%s: Warning: clearing HW address of %s while it still has VLANs.\n" , - bond_dev->name, bond_dev->name); -pr_warning ("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n" , - bond_dev->name); - } - -pr_info ("%s: released all slaves\n" , bond_dev->name); - -out: -write_unlock_bh (&bond->lock); - -bond_compute_features (bond); - -return 0 ; -} - -/* * This function changes the active slave to slave <slave_dev>. * It returns -EINVAL in the following cases. * - <slave_dev> is not found in the list.
@@ -4440,7 +4339,9 @@ static void bond_uninit(struct net_device *bond_dev)bond_netpoll_cleanup (bond_dev);/* Release the bonded slaves */ -bond_release_all (bond_dev); +while (bond->first_slave !=NULL ) +__bond_release_one (bond_dev, bond->first_slave->dev, true); +pr_info ("%s: released all slaves\n" , bond_dev->name);list_del (&bond->bond_list);
Signature · 1 line
1.7.11.7
Re: [PATCH net-next v2 3/3] bonding: fix bond_release_all inconsistencies
Quoted 12 lines · Nikolay Aleksandrov
This patch fixes the following inconsistencies in
bond_release_all:
-IFF_BONDINGflag is not stripped from slaves
- MTU is not restored
- no netdev notifiers are sent
Instead of trying to keepbond_releaseandbond_release_allin sync I think we can re-usebond_releaseas the environment for calling it is correct (RTNL is held). I have been running tests for the past week and they came out successful. The only way forbond_releaseto fail is for the slave to be attached in a different bond or to not be a slave but that cannot happen as RTNL is held and no slave manipulations can be achieved.
Quoted 6 lines
V2: As suggested
bond_releaseis renamed to__bond_release_onewith a new parameter "all" introduced so to avoid calling unnecessary code while destroying a bond, and a wrapper for it calledbond_releaseis created because ofndo_del_link.bond_release_all()is removed.Signed-off-by Nikolay Aleksandrov <[email protected]>
Quoted 200 lines
---
drivers/net/bonding/bond_main.c| 135 ++++++----------------------------------1 file changed, 18 insertions(+), 117 deletions(-)drivers/net/bonding/bond_main.c +18
-117@@ -1937,7 +1937,8 @@ err_undo_flags:/* * Try to release the slave device <slave> from the bond device <master> * It is legal to access curr_active_slave without a lock because all the function -* is write-locked. +* is write-locked. If "all" is true it means that the function is being called +* while destroying a bond interface and all slaves are being released. * * The rules for slave state should be: * for Active/Backup: @@ -1945,7 +1946,9 @@ err_undo_flags:* for Bonded connections: * The first up interface should be left on and all others downed. */ -int bond_release (struct net_device *bond_dev,struct net_device *slave_dev) +static int __bond_release_one (struct net_device *bond_dev, +struct net_device *slave_dev, +bool all) {struct bonding *bond =netdev_priv (bond_dev);struct slave *slave, *oldcurrent;@@ -1982,7 +1985,7 @@ int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)synchronize_net ();write_lock_bh (&bond->lock); -if (!bond->params.fail_over_mac) { +if (!all && !bond->params.fail_over_mac) {if (ether_addr_equal (bond_dev->dev_addr, slave->perm_hwaddr) && bond->slave_cnt >1 ) pr_warning("%s: Warning: the permanent HWaddr of %s - %pM - is still in use by %s. Set the HWaddr of %s to a different address to avoid conflicts.\n",@@ -2028,7 +2031,9 @@ int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)write_lock_bh (&bond->lock); } -if (oldcurrent == slave) { +if (all) { + bond->curr_active_slave =NULL ; + }else if (oldcurrent == slave) {/* * Note that we hold RTNL over this sequence, so there * is no concern that another slave add/remove event @@ -2117,6 +2122,12 @@ int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)return 0 ;/* deletion OK */ } +/* A wrapper used because of ndo_del_link */ +int bond_release (struct net_device *bond_dev,struct net_device *slave_dev) +{ +return __bond_release_one (bond_dev, slave_dev, false); +} +/* * First release a slave and then destroy the bond if no more slaves are left. * Must be under rtnl_lock when this function is called. @@ -2138,118 +2149,6 @@ static int bond_release_and_destroy(struct net_device *bond_dev, }/* -* This function releases all slaves. -*/ -static int bond_release_all (struct net_device *bond_dev) -{ -struct bonding *bond =netdev_priv (bond_dev); -struct slave *slave; -struct net_device *slave_dev; -struct sockaddr addr; - -write_lock_bh (&bond->lock); - -netif_carrier_off (bond_dev); - -if (bond->slave_cnt ==0 ) - goto out; - - bond->current_arp_slave =NULL ; - bond->primary_slave =NULL ; -bond_change_active_slave (bond,NULL ); - -while ((slave = bond->first_slave) !=NULL ) { -/* Inform AD package of unbinding of slave -* before slave is detached from the list. -*/ -if (bond->params.mode ==BOND_MODE_8023AD ) -bond_3ad_unbind_slave (slave); - - slave_dev = slave->dev; -bond_detach_slave (bond, slave); - -/* now that the slave is detached, unlock and perform -* all the undo steps that should not be called from -* within a lock. -*/ -write_unlock_bh (&bond->lock); - -/* unregister rx_handler early so bond_handle_frame wouldn't -* be called for this slave anymore. -*/ -netdev_rx_handler_unregister (slave_dev); -synchronize_net (); - -if (bond_is_lb (bond)) { -/* must be called only after the slave -* has been detached from the list -*/ -bond_alb_deinit_slave (bond, slave); - } - -bond_destroy_slave_symlinks (bond_dev, slave_dev); -bond_del_vlans_from_slave (bond, slave_dev); - -/* If the mode USES_PRIMARY, then we should only remove its -* promisc and mc settings if it was the curr_active_slave, but that was -* already taken care of above when we detached the slave -*/ -if (!USES_PRIMARY (bond->params.mode)) { -/* unset promiscuity level from slave */ -if (bond_dev->flags &IFF_PROMISC ) -dev_set_promiscuity (slave_dev,-1 ); - -/* unset allmulti level from slave */ -if (bond_dev->flags &IFF_ALLMULTI ) -dev_set_allmulti (slave_dev,-1 ); - -/* flush master's mc_list from slave */ -netif_addr_lock_bh (bond_dev); -bond_mc_list_flush (bond_dev, slave_dev); -netif_addr_unlock_bh (bond_dev); - } - -bond_upper_dev_unlink (bond_dev, slave_dev); - -slave_disable_netpoll (slave); - -/* close slave before restoring its mac address */ -dev_close (slave_dev); - -if (!bond->params.fail_over_mac) { -/* restore original ("permanent") mac address*/ -memcpy (addr.sa_data, slave->perm_hwaddr,ETH_ALEN ); - addr.sa_family = slave_dev->type; -dev_set_mac_address (slave_dev, &addr); - } - -kfree (slave); - -/* re-acquire the lock before getting the next slave */ -write_lock_bh (&bond->lock); - } - -eth_hw_addr_random (bond_dev); - bond->dev_addr_from_first = true; - -if (bond_vlan_used (bond)) { -pr_warning ("%s: Warning: clearing HW address of %s while it still has VLANs.\n" , - bond_dev->name, bond_dev->name); -pr_warning ("%s: When re-adding slaves, make sure the bond's HW address matches its VLANs'.\n" , - bond_dev->name); - } - -pr_info ("%s: released all slaves\n" , bond_dev->name); - -out: -write_unlock_bh (&bond->lock); - -bond_compute_features (bond); - -return 0 ; -} - -/* * This function changes the active slave to slave <slave_dev>. * It returns -EINVAL in the following cases. * - <slave_dev> is not found in the list. @@ -4440,7 +4339,9 @@ static void bond_uninit(struct net_device *bond_dev)bond_netpoll_cleanup (bond_dev);/* Release the bonded slaves */ -bond_release_all (bond_dev); +while (bond->first_slave !=NULL ) +__bond_release_one (bond_dev, bond->first_slave->dev, true); +pr_info ("%s: released all slaves\n" , bond_dev->name);list_del (&bond->bond_list);Signature · 6 lines
1.7.11.7 -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH net-next v2 3/3] bonding: fix bond_release_all inconsistencies
From: Jay Vosburgh <[email protected]>
Date: Mon, 18 Feb 2013 19:12:01 -0800
Quoted 14 lines
Quoted 12 lines · Nikolay Aleksandrov
This patch fixes the following inconsistencies in
bond_release_all:
-IFF_BONDINGflag is not stripped from slaves
- MTU is not restored
- no netdev notifiers are sent
Instead of trying to keepbond_releaseandbond_release_allin sync I think we can re-usebond_releaseas the environment for calling it is correct (RTNL is held). I have been running tests for the past week and they came out successful. The only way forbond_releaseto fail is for the slave to be attached in a different bond or to not be a slave but that cannot happen as RTNL is held and no slave manipulations can be achieved.
Quoted 8 lines
Quoted 6 lines
V2: As suggested
bond_releaseis renamed to__bond_release_onewith a new parameter "all" introduced so to avoid calling unnecessary code while destroying a bond, and a wrapper for it calledbond_releaseis created because ofndo_del_link.bond_release_all()is removed.Signed-off-by Nikolay Aleksandrov <[email protected]>Signed-off-by Jay Vosburgh <[email protected]>
Applied.