[PATCH 0/7] bonding: 7 fixes for 2.6.24

[PATCH 0/7] bonding: 7 fixes for 2.6.24

[Jay Vosburgh]

	Following are seven patches to fix locking problems,
silence locking-related warnings and resolve one recent regression
in the current 2.6.24-rc.

	The first three patches are reposts, the rest are new.

	patch 1: fix locking in sysfs primary/active selection

	Call core network functions with expected locks to
eliminate potential deadlock and silence warnings.

	patch 2: fix ASSERT_RTNL that produces spurious warnings

	Relocate ASSERT_RTNL to remove a false warning; after patch,
ASSERT is located in code that holds only RTNL (additional locks were
causing the ASSERT to trip)

	patch 3: fix locking during alb failover and slave removal

	Fix all call paths into alb_fasten_mac_swap to hold only RTNL.
Eliminates potential deadlock and silences warnings.

	patch 4: release slaves when master removed via sysfs

	Insure that all slaves are removed when bond is destroyed via
sysfs.

	patch 5: Fix up parameter parsing

	Recent changes broke parameter parsing; this fixes things.

	patch 6: Fix lock ordering for rtnl and bonding_rwsem

	Resolves some lockdep warnings related to ordering between
rtnl and bonding_rwsem.

	patch 7: Don't hold lock when calling rtnl_unlock

	Since rtnl_unlock can sleep, don't hold any other locks when
calling it.

	Patches are against the current netdev-2.6#upstream branch.

	Please apply for 2.6.24.

	-J

---
	-Jay Vosburgh, IBM Linux Technology Center, fubar@us.ibm.com

[PATCH 1/7] bonding: fix locking in sysfs primary/active selection

[Jay Vosburgh]

	Fix the functions that store the primary and active slave
options via sysfs to hold the correct locks in the correct order.

	The bond_change_active_slave and bond_select_active_slave
functions both require rtnl, bond->lock for read and curr_slave_lock for
write_bh, and no other locks.  This is so that the lower level
mode-specific functions (notably for balance-alb mode) can release locks
down to just rtnl in order to call, e.g., dev_set_mac_address with the
locks it expects (rtnl only).

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
---
 drivers/net/bonding/bond_sysfs.c |   15 ++++++++++-----
 1 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/drivers/net/bonding/bond_sysfs.c b/drivers/net/bonding/bond_sysfs.c
index 11b76b3..28a2d80 100644
--- a/drivers/net/bonding/bond_sysfs.c
+++ b/drivers/net/bonding/bond_sysfs.c
@@ -1075,7 +1075,10 @@ static ssize_t bonding_store_primary(struct device *d,
 	struct slave *slave;
 	struct bonding *bond = to_bond(d);
 
-	write_lock_bh(&bond->lock);
+	rtnl_lock();
+	read_lock(&bond->lock);
+	write_lock_bh(&bond->curr_slave_lock);
+
 	if (!USES_PRIMARY(bond->params.mode)) {
 		printk(KERN_INFO DRV_NAME
 		       ": %s: Unable to set primary slave; %s is in mode %d\n",
@@ -1109,8 +1112,8 @@ static ssize_t bonding_store_primary(struct device *d,
 		}
 	}
 out:
-	write_unlock_bh(&bond->lock);
-
+	write_unlock_bh(&bond->curr_slave_lock);
+	read_unlock(&bond->lock);
 	rtnl_unlock();
 
 	return count;
@@ -1190,7 +1193,8 @@ static ssize_t bonding_store_active_slave(struct device *d,
 	struct bonding *bond = to_bond(d);
 
 	rtnl_lock();
-	write_lock_bh(&bond->lock);
+	read_lock(&bond->lock);
+	write_lock_bh(&bond->curr_slave_lock);
 
 	if (!USES_PRIMARY(bond->params.mode)) {
 		printk(KERN_INFO DRV_NAME
@@ -1247,7 +1251,8 @@ static ssize_t bonding_store_active_slave(struct device *d,
 		}
 	}
 out:
-	write_unlock_bh(&bond->lock);
+	write_unlock_bh(&bond->curr_slave_lock);
+	read_unlock(&bond->lock);
 	rtnl_unlock();
 
 	return count;
-- 
1.5.3.4.206.g58ba4-dirty

[PATCH 2/7] bonding: fix ASSERT_RTNL that produces spurious warnings

[Jay Vosburgh]

	Move an ASSERT_RTNL down to where we should hold only RTNL;
the existing check produces spurious warnings because we hold additional
locks at _bh, tripping a debug warning in spin_lock_mutex().

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
---
 drivers/net/bonding/bond_alb.c |    5 ++---
 1 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c
index 25b8dbf..9b55a12 100644
--- a/drivers/net/bonding/bond_alb.c
+++ b/drivers/net/bonding/bond_alb.c
@@ -1601,9 +1601,6 @@ void bond_alb_handle_active_change(struct bonding *bond, struct slave *new_slave
 	struct slave *swap_slave;
 	int i;
 
-	if (new_slave)
-		ASSERT_RTNL();
-
 	if (bond->curr_active_slave == new_slave) {
 		return;
 	}
@@ -1649,6 +1646,8 @@ void bond_alb_handle_active_change(struct bonding *bond, struct slave *new_slave
 	write_unlock_bh(&bond->curr_slave_lock);
 	read_unlock(&bond->lock);
 
+	ASSERT_RTNL();
+
 	/* curr_active_slave must be set before calling alb_swap_mac_addr */
 	if (swap_slave) {
 		/* swap mac address */
-- 
1.5.3.4.206.g58ba4-dirty

[PATCH 3/7] bonding: fix locking during alb failover and slave removal

[Jay Vosburgh]

	alb_fasten_mac_swap (actually rlb_teach_disabled_mac_on_primary)
requries RTNL and no other locks.  This could cause dev_set_promiscuity
and/or dev_set_mac_address to be called with improper locking.

	Changed callers to hold only RTNL during calls to alb_fasten_mac_swap
or functions calling it.  Updated header comments in affected functions to
reflect proper reality of locking requirements.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
---
 drivers/net/bonding/bond_alb.c  |   18 ++++++++++++------
 drivers/net/bonding/bond_main.c |   14 ++++++++------
 2 files changed, 20 insertions(+), 12 deletions(-)

diff --git a/drivers/net/bonding/bond_alb.c b/drivers/net/bonding/bond_alb.c
index 9b55a12..b57bc94 100644
--- a/drivers/net/bonding/bond_alb.c
+++ b/drivers/net/bonding/bond_alb.c
@@ -979,7 +979,7 @@ static void alb_swap_mac_addr(struct bonding *bond, struct slave *slave1, struct
 /*
  * Send learning packets after MAC address swap.
  *
- * Called with RTNL and bond->lock held for read.
+ * Called with RTNL and no other locks
  */
 static void alb_fasten_mac_swap(struct bonding *bond, struct slave *slave1,
 				struct slave *slave2)
@@ -987,6 +987,8 @@ static void alb_fasten_mac_swap(struct bonding *bond, struct slave *slave1,
 	int slaves_state_differ = (SLAVE_IS_OK(slave1) != SLAVE_IS_OK(slave2));
 	struct slave *disabled_slave = NULL;
 
+	ASSERT_RTNL();
+
 	/* fasten the change in the switch */
 	if (SLAVE_IS_OK(slave1)) {
 		alb_send_learning_packets(slave1, slave1->dev->dev_addr);
@@ -1031,7 +1033,7 @@ static void alb_fasten_mac_swap(struct bonding *bond, struct slave *slave1,
  * a slave that has @slave's permanet address as its current address.
  * We'll make sure that that slave no longer uses @slave's permanent address.
  *
- * Caller must hold bond lock
+ * Caller must hold RTNL and no other locks
  */
 static void alb_change_hw_addr_on_detach(struct bonding *bond, struct slave *slave)
 {
@@ -1542,7 +1544,12 @@ int bond_alb_init_slave(struct bonding *bond, struct slave *slave)
 	return 0;
 }
 
-/* Caller must hold bond lock for write */
+/*
+ * Remove slave from tlb and rlb hash tables, and fix up MAC addresses
+ * if necessary.
+ *
+ * Caller must hold RTNL and no other locks
+ */
 void bond_alb_deinit_slave(struct bonding *bond, struct slave *slave)
 {
 	if (bond->slave_cnt > 1) {
@@ -1658,12 +1665,11 @@ void bond_alb_handle_active_change(struct bonding *bond, struct slave *new_slave
 				       bond->alb_info.rlb_enabled);
 	}
 
-	read_lock(&bond->lock);
-
 	if (swap_slave) {
 		alb_fasten_mac_swap(bond, swap_slave, new_slave);
+		read_lock(&bond->lock);
 	} else {
-		/* fasten bond mac on new current slave */
+		read_lock(&bond->lock);
 		alb_send_learning_packets(new_slave, bond->dev->dev_addr);
 	}
 
diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index b0b2603..77d004d 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -1746,7 +1746,9 @@ int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
 		 * has been cleared (if our_slave == old_current),
 		 * but before a new active slave is selected.
 		 */
+		write_unlock_bh(&bond->lock);
 		bond_alb_deinit_slave(bond, slave);
+		write_lock_bh(&bond->lock);
 	}
 
 	if (oldcurrent == slave) {
@@ -1905,6 +1907,12 @@ static int bond_release_all(struct net_device *bond_dev)
 		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);
+
 		if ((bond->params.mode == BOND_MODE_TLB) ||
 		    (bond->params.mode == BOND_MODE_ALB)) {
 			/* must be called only after the slave
@@ -1915,12 +1923,6 @@ static int bond_release_all(struct net_device *bond_dev)
 
 		bond_compute_features(bond);
 
-		/* 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);
-
 		bond_destroy_slave_symlinks(bond_dev, slave_dev);
 		bond_del_vlans_from_slave(bond, slave_dev);
 
-- 
1.5.3.4.206.g58ba4-dirty

[PATCH 4/7] bonding: release slaves when master removed via sysfs

[Jay Vosburgh]

	Add a call to bond_release_all in the bonding netdev event
handler for the master.  This releases the slaves for the case of, e.g.,
"echo -bond0 > /sys/class/net/bonding_masters", which otherwise will spin
forever waiting for references to be released.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
---
 drivers/net/bonding/bond_main.c |    4 +---
 1 files changed, 1 insertions(+), 3 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 77d004d..3ede0a2 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -3401,9 +3401,7 @@ static int bond_master_netdev_event(unsigned long event, struct net_device *bond
 	case NETDEV_CHANGENAME:
 		return bond_event_changename(event_bond);
 	case NETDEV_UNREGISTER:
-		/*
-		 * TODO: remove a bond from the list?
-		 */
+		bond_release_all(event_bond->dev);
 		break;
 	default:
 		break;
-- 
1.5.3.4.206.g58ba4-dirty

[PATCH 5/7] bonding: Fix up parameter parsing

[Jay Vosburgh]

	A recent change to add an additional hash policy modified
bond_parse_parm, but it now does not correctly match parameters passed in
via sysfs.

	Rewrote bond_parse_parm to handle (a) parameter matches that
are substrings of one another and (b) user input with whitespace (e.g.,
sysfs input often has a trailing newline).

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
---
 drivers/net/bonding/bond_main.c  |   23 ++++++++++++++++-------
 drivers/net/bonding/bond_sysfs.c |    8 ++++----
 drivers/net/bonding/bonding.h    |    4 +++-
 3 files changed, 23 insertions(+), 12 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 3ede0a2..379c5d8 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -4540,18 +4540,27 @@ static void bond_free_all(void)
 
 /*
  * Convert string input module parms.  Accept either the
- * number of the mode or its string name.
+ * number of the mode or its string name.  A bit complicated because
+ * some mode names are substrings of other names, and calls from sysfs
+ * may have whitespace in the name (trailing newlines, for example).
  */
-int bond_parse_parm(char *mode_arg, struct bond_parm_tbl *tbl)
+int bond_parse_parm(const char *buf, struct bond_parm_tbl *tbl)
 {
-	int i;
+	int mode = -1, i, rv;
+	char modestr[BOND_MAX_MODENAME_LEN + 1] = { 0, };
+
+	rv = sscanf(buf, "%d", &mode);
+	if (!rv) {
+		rv = sscanf(buf, "%20s", modestr);
+		if (!rv)
+			return -1;
+	}
 
 	for (i = 0; tbl[i].modename; i++) {
-		if ((isdigit(*mode_arg) &&
-		     tbl[i].mode == simple_strtol(mode_arg, NULL, 0)) ||
-		    (strcmp(mode_arg, tbl[i].modename) == 0)) {
+		if (mode == tbl[i].mode)
+			return tbl[i].mode;
+		if (strcmp(modestr, tbl[i].modename) == 0)
 			return tbl[i].mode;
-		}
 	}
 
 	return -1;
diff --git a/drivers/net/bonding/bond_sysfs.c b/drivers/net/bonding/bond_sysfs.c
index 28a2d80..bff4f2b 100644
--- a/drivers/net/bonding/bond_sysfs.c
+++ b/drivers/net/bonding/bond_sysfs.c
@@ -423,7 +423,7 @@ static ssize_t bonding_store_mode(struct device *d,
 		goto out;
 	}
 
-	new_value = bond_parse_parm((char *)buf, bond_mode_tbl);
+	new_value = bond_parse_parm(buf, bond_mode_tbl);
 	if (new_value < 0)  {
 		printk(KERN_ERR DRV_NAME
 		       ": %s: Ignoring invalid mode value %.*s.\n",
@@ -478,7 +478,7 @@ static ssize_t bonding_store_xmit_hash(struct device *d,
 		goto out;
 	}
 
-	new_value = bond_parse_parm((char *)buf, xmit_hashtype_tbl);
+	new_value = bond_parse_parm(buf, xmit_hashtype_tbl);
 	if (new_value < 0)  {
 		printk(KERN_ERR DRV_NAME
 		       ": %s: Ignoring invalid xmit hash policy value %.*s.\n",
@@ -518,7 +518,7 @@ static ssize_t bonding_store_arp_validate(struct device *d,
 	int new_value;
 	struct bonding *bond = to_bond(d);
 
-	new_value = bond_parse_parm((char *)buf, arp_validate_tbl);
+	new_value = bond_parse_parm(buf, arp_validate_tbl);
 	if (new_value < 0) {
 		printk(KERN_ERR DRV_NAME
 		       ": %s: Ignoring invalid arp_validate value %s\n",
@@ -941,7 +941,7 @@ static ssize_t bonding_store_lacp(struct device *d,
 		goto out;
 	}
 
-	new_value = bond_parse_parm((char *)buf, bond_lacp_tbl);
+	new_value = bond_parse_parm(buf, bond_lacp_tbl);
 
 	if ((new_value == 1) || (new_value == 0)) {
 		bond->params.lacp_fast = new_value;
diff --git a/drivers/net/bonding/bonding.h b/drivers/net/bonding/bonding.h
index e1e4734..6d83be4 100644
--- a/drivers/net/bonding/bonding.h
+++ b/drivers/net/bonding/bonding.h
@@ -141,6 +141,8 @@ struct bond_parm_tbl {
 	int mode;
 };
 
+#define BOND_MAX_MODENAME_LEN 20
+
 struct vlan_entry {
 	struct list_head vlan_list;
 	__be32 vlan_ip;
@@ -314,7 +316,7 @@ void bond_mii_monitor(struct work_struct *);
 void bond_loadbalance_arp_mon(struct work_struct *);
 void bond_activebackup_arp_mon(struct work_struct *);
 void bond_set_mode_ops(struct bonding *bond, int mode);
-int bond_parse_parm(char *mode_arg, struct bond_parm_tbl *tbl);
+int bond_parse_parm(const char *mode_arg, struct bond_parm_tbl *tbl);
 void bond_select_active_slave(struct bonding *bond);
 void bond_change_active_slave(struct bonding *bond, struct slave *new_active);
 void bond_register_arp(struct bonding *);
-- 
1.5.3.4.206.g58ba4-dirty

[PATCH 6/7] bonding: fix lock ordering for rtnl and bonding_rwsem

[Jay Vosburgh]

Fix the handling of rtnl and the bonding_rwsem to always be acquired
in a consistent order (rtnl, then bonding_rwsem).

The existing code sometimes acquired them in this order, and sometimes
in the opposite order, which opens a window for deadlock between ifenslave
and sysfs.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
---
 drivers/net/bonding/bond_main.c  |   19 ++++++++++++++++
 drivers/net/bonding/bond_sysfs.c |   43 ++++++++++++++-----------------------
 2 files changed, 35 insertions(+), 27 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 379c5d8..2c6da49 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -4874,9 +4874,22 @@ static struct lock_class_key bonding_netdev_xmit_lock_key;
 int bond_create(char *name, struct bond_params *params, struct bonding **newbond)
 {
 	struct net_device *bond_dev;
+	struct bonding *bond, *nxt;
 	int res;
 
 	rtnl_lock();
+	down_write(&bonding_rwsem);
+
+	/* Check to see if the bond already exists. */
+	list_for_each_entry_safe(bond, nxt, &bond_dev_list, bond_list)
+		if (strnicmp(bond->dev->name, name, IFNAMSIZ) == 0) {
+			printk(KERN_ERR DRV_NAME
+			       ": cannot add bond %s; it already exists\n",
+			       name);
+			res = -EPERM;
+			goto out_rtnl;
+		}
+
 	bond_dev = alloc_netdev(sizeof(struct bonding), name ? name : "",
 				ether_setup);
 	if (!bond_dev) {
@@ -4915,10 +4928,12 @@ int bond_create(char *name, struct bond_params *params, struct bonding **newbond
 
 	netif_carrier_off(bond_dev);
 
+	up_write(&bonding_rwsem);
 	rtnl_unlock(); /* allows sysfs registration of net device */
 	res = bond_create_sysfs_entry(bond_dev->priv);
 	if (res < 0) {
 		rtnl_lock();
+		down_write(&bonding_rwsem);
 		goto out_bond;
 	}
 
@@ -4929,6 +4944,7 @@ out_bond:
 out_netdev:
 	free_netdev(bond_dev);
 out_rtnl:
+	up_write(&bonding_rwsem);
 	rtnl_unlock();
 	return res;
 }
@@ -4949,6 +4965,9 @@ static int __init bonding_init(void)
 #ifdef CONFIG_PROC_FS
 	bond_create_proc_dir();
 #endif
+
+	init_rwsem(&bonding_rwsem);
+
 	for (i = 0; i < max_bonds; i++) {
 		res = bond_create(NULL, &bonding_defaults, NULL);
 		if (res)
diff --git a/drivers/net/bonding/bond_sysfs.c b/drivers/net/bonding/bond_sysfs.c
index bff4f2b..90a1f31 100644
--- a/drivers/net/bonding/bond_sysfs.c
+++ b/drivers/net/bonding/bond_sysfs.c
@@ -109,11 +109,10 @@ static ssize_t bonding_store_bonds(struct class *cls, const char *buffer, size_t
 {
 	char command[IFNAMSIZ + 1] = {0, };
 	char *ifname;
-	int res = count;
+	int rv, res = count;
 	struct bonding *bond;
 	struct bonding *nxt;
 
-	down_write(&(bonding_rwsem));
 	sscanf(buffer, "%16s", command); /* IFNAMSIZ*/
 	ifname = command + 1;
 	if ((strlen(command) <= 1) ||
@@ -121,39 +120,28 @@ static ssize_t bonding_store_bonds(struct class *cls, const char *buffer, size_t
 		goto err_no_cmd;
 
 	if (command[0] == '+') {
-
-		/* Check to see if the bond already exists. */
-		list_for_each_entry_safe(bond, nxt, &bond_dev_list, bond_list)
-			if (strnicmp(bond->dev->name, ifname, IFNAMSIZ) == 0) {
-				printk(KERN_ERR DRV_NAME
-					": cannot add bond %s; it already exists\n",
-					ifname);
-				res = -EPERM;
-				goto out;
-			}
-
 		printk(KERN_INFO DRV_NAME
 			": %s is being created...\n", ifname);
-		if (bond_create(ifname, &bonding_defaults, &bond)) {
-			printk(KERN_INFO DRV_NAME
-			": %s interface already exists. Bond creation failed.\n",
-			ifname);
-			res = -EPERM;
+		rv = bond_create(ifname, &bonding_defaults, &bond);
+		if (rv) {
+			printk(KERN_INFO DRV_NAME ": Bond creation failed.\n");
+			res = rv;
 		}
 		goto out;
 	}
 
 	if (command[0] == '-') {
+		rtnl_lock();
+		down_write(&bonding_rwsem);
+
 		list_for_each_entry_safe(bond, nxt, &bond_dev_list, bond_list)
 			if (strnicmp(bond->dev->name, ifname, IFNAMSIZ) == 0) {
-				rtnl_lock();
 				/* check the ref count on the bond's kobject.
 				 * If it's > expected, then there's a file open,
 				 * and we have to fail.
 				 */
 				if (atomic_read(&bond->dev->dev.kobj.kref.refcount)
 							> expected_refcount){
-					rtnl_unlock();
 					printk(KERN_INFO DRV_NAME
 						": Unable remove bond %s due to open references.\n",
 						ifname);
@@ -164,6 +152,7 @@ static ssize_t bonding_store_bonds(struct class *cls, const char *buffer, size_t
 					": %s is being deleted...\n",
 					bond->dev->name);
 				bond_destroy(bond);
+				up_write(&bonding_rwsem);
 				rtnl_unlock();
 				goto out;
 			}
@@ -171,6 +160,8 @@ static ssize_t bonding_store_bonds(struct class *cls, const char *buffer, size_t
 		printk(KERN_ERR DRV_NAME
 			": unable to delete non-existent bond %s\n", ifname);
 		res = -ENODEV;
+		up_write(&bonding_rwsem);
+		rtnl_unlock();
 		goto out;
 	}
 
@@ -183,7 +174,6 @@ err_no_cmd:
 	 * get called forever, which is bad.
 	 */
 out:
-	up_write(&(bonding_rwsem));
 	return res;
 }
 /* class attribute for bond_masters file.  This ends up in /sys/class/net */
@@ -271,6 +261,9 @@ static ssize_t bonding_store_slaves(struct device *d,
 
 	/* Note:  We can't hold bond->lock here, as bond_create grabs it. */
 
+	rtnl_lock();
+	down_write(&(bonding_rwsem));
+
 	sscanf(buffer, "%16s", command); /* IFNAMSIZ*/
 	ifname = command + 1;
 	if ((strlen(command) <= 1) ||
@@ -336,12 +329,10 @@ static ssize_t bonding_store_slaves(struct device *d,
 				dev->mtu = bond->dev->mtu;
 			}
 		}
-		rtnl_lock();
 		res = bond_enslave(bond->dev, dev);
 		bond_for_each_slave(bond, slave, i)
 			if (strnicmp(slave->dev->name, ifname, IFNAMSIZ) == 0)
 				slave->original_mtu = original_mtu;
-		rtnl_unlock();
 		if (res) {
 			ret = res;
 		}
@@ -359,12 +350,10 @@ static ssize_t bonding_store_slaves(struct device *d,
 		if (dev) {
 			printk(KERN_INFO DRV_NAME ": %s: Removing slave %s\n",
 				bond->dev->name, dev->name);
-			rtnl_lock();
 			if (bond->setup_by_slave)
 				res = bond_release_and_destroy(bond->dev, dev);
 			else
 				res = bond_release(bond->dev, dev);
-			rtnl_unlock();
 			if (res) {
 				ret = res;
 				goto out;
@@ -389,6 +378,8 @@ err_no_cmd:
 	ret = -EPERM;
 
 out:
+	up_write(&(bonding_rwsem));
+	rtnl_unlock();
 	return ret;
 }
 
@@ -1423,8 +1414,6 @@ int bond_create_sysfs(void)
 	int ret = 0;
 	struct bonding *firstbond;
 
-	init_rwsem(&bonding_rwsem);
-
 	/* get the netdev class pointer */
 	firstbond = container_of(bond_dev_list.next, struct bonding, bond_list);
 	if (!firstbond)
-- 
1.5.3.4.206.g58ba4-dirty

[PATCH 7/7] bonding: Don't hold lock when calling rtnl_unlock

[Jay Vosburgh]

Change bond_mii_monitor to not hold any locks when calling rtnl_unlock,
as rtnl_unlock can sleep (when acquring another mutex in netdev_run_todo).

Bug reported by Makito SHIOKAWA <mshiokawa@miraclelinux.com>, who
included a different patch.

Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
---
 drivers/net/bonding/bond_main.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 2c6da49..49a1982 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -2386,7 +2386,9 @@ void bond_mii_monitor(struct work_struct *work)
 		rtnl_lock();
 		read_lock(&bond->lock);
 		__bond_mii_monitor(bond, 1);
-		rtnl_unlock();
+		read_unlock(&bond->lock);
+		rtnl_unlock();	/* might sleep, hold no other locks */
+		read_lock(&bond->lock);
 	}
 
 	delay = ((bond->params.miimon * HZ) / 1000) ? : 1;
-- 
1.5.3.4.206.g58ba4-dirty

Re: [PATCH 6/7] bonding: fix lock ordering for rtnl and bonding_rwsem

[Andrew Morton]

On Thu, 17 Jan 2008 16:25:02 -0800
Jay Vosburgh <fubar@us.ibm.com> wrote:

> Fix the handling of rtnl and the bonding_rwsem to always be acquired
> in a consistent order (rtnl, then bonding_rwsem).
> 
> The existing code sometimes acquired them in this order, and sometimes
> in the opposite order, which opens a window for deadlock between ifenslave
> and sysfs.
> 
> ...
>
>  int bond_create(char *name, struct bond_params *params, struct bonding **newbond)
>  {
>  	struct net_device *bond_dev;
> +	struct bonding *bond, *nxt;
>  	int res;
>  
>  	rtnl_lock();
> +	down_write(&bonding_rwsem);
> +
> +	/* Check to see if the bond already exists. */
> +	list_for_each_entry_safe(bond, nxt, &bond_dev_list, bond_list)

this could (should) use list_for_each_entry().

> +		if (strnicmp(bond->dev->name, name, IFNAMSIZ) == 0) {
> +			printk(KERN_ERR DRV_NAME
> +			       ": cannot add bond %s; it already exists\n",
> +			       name);
> +			res = -EPERM;
> +			goto out_rtnl;
> +		}
> +
>  	bond_dev = alloc_netdev(sizeof(struct bonding), name ? name : "",
>  				ether_setup);
>  	if (!bond_dev) {
> @@ -4915,10 +4928,12 @@ int bond_create(char *name, struct bond_params *params, struct bonding **newbond
>  
>  	netif_carrier_off(bond_dev);
>  
> +	up_write(&bonding_rwsem);
>  	rtnl_unlock(); /* allows sysfs registration of net device */
>  	res = bond_create_sysfs_entry(bond_dev->priv);
>  	if (res < 0) {
>  		rtnl_lock();
> +		down_write(&bonding_rwsem);
>  		goto out_bond;
>  	}
>  
> @@ -4929,6 +4944,7 @@ out_bond:
>  out_netdev:
>  	free_netdev(bond_dev);
>  out_rtnl:
> +	up_write(&bonding_rwsem);
>  	rtnl_unlock();
>  	return res;
>  }
> @@ -4949,6 +4965,9 @@ static int __init bonding_init(void)
>  #ifdef CONFIG_PROC_FS
>  	bond_create_proc_dir();
>  #endif
> +
> +	init_rwsem(&bonding_rwsem);

It would be better to initialise this at compile time with DECLARE_RWSEM().


But neither of those things need to be done for 2.6.24.

Re: [PATCH 1/7] bonding: fix locking in sysfs primary/active selection

[Jeff Garzik]

Jay Vosburgh wrote:
> 	Fix the functions that store the primary and active slave
> options via sysfs to hold the correct locks in the correct order.
> 
> 	The bond_change_active_slave and bond_select_active_slave
> functions both require rtnl, bond->lock for read and curr_slave_lock for
> write_bh, and no other locks.  This is so that the lower level
> mode-specific functions (notably for balance-alb mode) can release locks
> down to just rtnl in order to call, e.g., dev_set_mac_address with the
> locks it expects (rtnl only).
> 
> Signed-off-by: Jay Vosburgh <fubar@us.ibm.com>
> Signed-off-by: Andy Gospodarek <andy@greyhouse.net>
> ---
>  drivers/net/bonding/bond_sysfs.c |   15 ++++++++++-----
>  1 files changed, 10 insertions(+), 5 deletions(-)

applied 1-7

Re: [PATCH 0/7] bonding: 7 fixes for 2.6.24

[David Miller]

From: Jay Vosburgh <fubar@us.ibm.com>
Date: Thu, 17 Jan 2008 16:24:56 -0800

> 	Following are seven patches to fix locking problems,
> silence locking-related warnings and resolve one recent regression
> in the current 2.6.24-rc.

Jeff, are you going to merge this stuff in?

I'll drop that rtnl_lock() one-liner if so...

Re: [PATCH 0/7] bonding: 7 fixes for 2.6.24

[Andrew Morton]

On Sun, 20 Jan 2008 17:15:38 -0800 (PST) David Miller <davem@davemloft.net> wrote:

> From: Jay Vosburgh <fubar@us.ibm.com>
> Date: Thu, 17 Jan 2008 16:24:56 -0800
> 
> > 	Following are seven patches to fix locking problems,
> > silence locking-related warnings and resolve one recent regression
> > in the current 2.6.24-rc.
> 
> Jeff, are you going to merge this stuff in?
> 
> I'll drop that rtnl_lock() one-liner if so...

It's all in mainline.  The one-liner should be unneeded.

Re: [PATCH 0/7] bonding: 7 fixes for 2.6.24

[Jeff Garzik]

David Miller wrote:
> From: Jay Vosburgh <fubar@us.ibm.com>
> Date: Thu, 17 Jan 2008 16:24:56 -0800
> 
>> 	Following are seven patches to fix locking problems,
>> silence locking-related warnings and resolve one recent regression
>> in the current 2.6.24-rc.
> 
> Jeff, are you going to merge this stuff in?
> 
> I'll drop that rtnl_lock() one-liner if so...

Linus has this series in linux-2.6.git, even...

	Jeff

Re: [PATCH 0/7] bonding: 7 fixes for 2.6.24

[David Miller]

From: Andrew Morton <akpm@linux-foundation.org>
Date: Sun, 20 Jan 2008 17:36:44 -0800

> On Sun, 20 Jan 2008 17:15:38 -0800 (PST) David Miller <davem@davemloft.net> wrote:
> 
> > From: Jay Vosburgh <fubar@us.ibm.com>
> > Date: Thu, 17 Jan 2008 16:24:56 -0800
> > 
> > > 	Following are seven patches to fix locking problems,
> > > silence locking-related warnings and resolve one recent regression
> > > in the current 2.6.24-rc.
> > 
> > Jeff, are you going to merge this stuff in?
> > 
> > I'll drop that rtnl_lock() one-liner if so...
> 
> It's all in mainline.  The one-liner should be unneeded.

Excellent.