bonding: use neighbours instead of own lists
Subsystems (4)
v2 cover bonding: use neighbours instead of own lists
(David, feel free to drop the whole patchset - I know that the window is closed and I'm quite sure that it's not the last version, and even if it is
- I'll easily re-submit it. Sorry for the mess, again :) )
Hi,
v1 -> v2:
Fix a bug when we could remove try to remove a neighbour twice - it was possible if we've had this neighbour already and added another devices which had this neighbour in its lists. Fix it by following the rule - there can be only one neighbour link, so add also _remove_neighbour() functions to the generic _remove()s to handle the neighbour only removing, and fix its refcounting while adding (all in patch 01).
Also, fix a vlan bug when it called unregister netdevice before unlinking - it could possibly lead to a subte race condition - cause it was unregistered but still in global-accessible list.
Rearrange sysfs/vlan patches so that first we fix vlan and only after we add sysfs stuff.
RFC -> v1:
I've added proper, consistent naming for all variables/functions, uninlined some helpers to get better backtraces, just in case (overhead is minimal, no hot paths), rearranged patches for better review, dropped bondings ->prev and bond_for_each_slave_continue() functionality - to be able to RCUify it easier, and renamed slave_* sysfs links to lower_* sysfs links to maintain upper/lower naming. I've also dropped, thanks to bonding cleanup, some heavy and ugly/intrusive patches.
I'm sending it as early as possible, because it's quite a big patchset and some of the approaches I've chosen are not really easy/straightforward. It's, however, quite heavily tested already and works fine.
I'm sending it to gather any feedback possible.
This patchset introduces all the needed infrastructure, on top of current adjacent lists, to be able to remove bond's slave_list/slave->list. The overhead in memory/CPU is minimal, and after the patchset bonding can rely on its slave-related functions, given the proper locking. I've done some netperf benchmarks on a vm, and the delta was about 0.1gbps for 35gbps as a whole, so no speed fluctuations.
It also automatically creates lower/upper and master symlinks in dev's sysfs directory.
The current version works ok, as first tests have shown. I'm testing it further, if anything comes up - I'll update.
The per-patch description has moved from here to per-patch notes.
Thanks in advance.
--- drivers/net/bonding/bond_3ad.c | 54 +-- drivers/net/bonding/bond_alb.c | 81 ++-- drivers/net/bonding/bond_alb.h | 4 +- drivers/net/bonding/bond_main.c | 296 ++++++++------- drivers/net/bonding/bond_procfs.c | 5 +- drivers/net/bonding/bond_sysfs.c | 62 +-- drivers/net/bonding/bonding.h | 74 ++-- drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 3 +- include/linux/netdevice.h | 57 ++- net/8021q/vlan.c | 18 +- net/core/dev.c | 482 +++++++++++++++++++----- 11 files changed, 724 insertions(+), 412 deletions(-)
PATCH v2 1/27 net: add adj_list to save only neighbours
Currently, we distinguish neighbours (first-level linked devices) from non-neighbours by the neighbour bool in the netdev_adjacent. This could be quite time-consuming in case we would like to traverse *only* through neighbours - cause we'd have to traverse through all devices and check for this flag, and in a (quite common) scenario where we have lots of vlans on top of bridge, which is on top of a bond - the bonding would have to go through all those vlans to get its upper neighbour linked devices.
This situation is really unpleasant, cause there are already a lot of cases when a device with slaves needs to go through them in hot path.
To fix this, introduce a new upper/lower device lists structure - adj_list, which contains only the neighbours. It works always in pair with the all_adj_list structure (renamed from upper/lower_dev_list), i.e. both of them contain the same links, only that all_adj_list contains also non-neighbour device links. It's really a small change visible, currently, only for __netdev_adjacent_dev_insert/remove(), and doesn't change the main linked logic at all.
Also, add some comments a fix a name collision in netdev_for_each_upper_dev_rcu() and rework the naming by the following rules:
netdev_(all_)(upper|lower)_*
If "all_" is present, then we work with the whole list of upper/lower devices, otherwise - only with direct neighbours. Uninline functions - to get better stack traces.
---
Notes:
v1 -> v2:
Fix a bug when we've tried to add a non-neighbour dev1 to dev2, while dev2 already had dev1 as a neighbour - we shouldn't care about neighbour's ref_nr at all cause we can only add one - so just add/remove it via different functions. And it became more readable.
RFC -> v1:
rename neighbour_dev_list/all_dev_list to adj_list/all_adj_list and rework the naming of functions - to make the use the pattern netdev_(all_)(lower|upper)_* . Uninline functions to get better stack traces - inline doesn't give much speed improvement, but this way the stack traces are meaningful.
drivers/net/bonding/bond_alb.c | 2 +- drivers/net/bonding/bond_main.c | 10 +- include/linux/netdevice.h | 28 +++-- net/core/dev.c | 219 ++++++++++++++++++++++++++++------------ 4 files changed, 179 insertions(+), 80 deletions(-)
drivers/net/bonding/bond_alb.c +1 -1
@@ -1019,7 +1019,7 @@ static void alb_send_learning_packets(struct slave *slave, u8 mac_addr[])/* loop through vlans and send one packet for each */ rcu_read_lock (); -netdev_for_each_upper_dev_rcu (bond->dev, upper, iter) { +netdev_for_each_all_upper_dev_rcu (bond->dev, upper, iter) {if (upper->priv_flags &IFF_802_1Q_VLAN )alb_send_lp_vid (slave, mac_addr,vlan_dev_vlan_id (upper));
drivers/net/bonding/bond_main.c +6 -4
@@ -2267,7 +2267,7 @@ static bool bond_has_this_ip(struct bonding *bond, __be32 ip)return true;rcu_read_lock (); -netdev_for_each_upper_dev_rcu (bond->dev, upper, iter) { +netdev_for_each_all_upper_dev_rcu (bond->dev, upper, iter) {if (ip ==bond_confirm_addr (upper,0 , ip)) { ret = true;break ;
@@ -2342,10 +2342,12 @@ static void bond_arp_send_all(struct bonding *bond, struct slave *slave)* * TODO: QinQ? */ -netdev_for_each_upper_dev_rcu (bond->dev, vlan_upper, vlan_iter) { +netdev_for_each_all_upper_dev_rcu (bond->dev, vlan_upper, + vlan_iter) {if (!is_vlan_dev (vlan_upper))continue ; -netdev_for_each_upper_dev_rcu (vlan_upper, upper, iter) { +netdev_for_each_all_upper_dev_rcu (vlan_upper, upper, + iter) {if (upper == rt->dst.dev) { vlan_id =vlan_dev_vlan_id (vlan_upper);rcu_read_unlock ();
@@ -2358,7 +2360,7 @@ static void bond_arp_send_all(struct bonding *bond, struct slave *slave)* our upper vlans, then just search for any dev that * matches, and in case it's a vlan - save the id */ -netdev_for_each_upper_dev_rcu (bond->dev, upper, iter) { +netdev_for_each_all_upper_dev_rcu (bond->dev, upper, iter) {if (upper == rt->dst.dev) {/* if it's a vlan - get its VID */ if (is_vlan_dev (upper))
include/linux/netdevice.h +19 -9
@@ -1143,8 +1143,18 @@ struct net_device {struct list_head dev_list;struct list_head napi_list;struct list_head unreg_list; -struct list_head upper_dev_list;/* List of upper devices */ -struct list_head lower_dev_list; + +/* directly linked devices, like slaves for bonding */ +struct { +struct list_head upper; +struct list_head lower; + } adj_list; + +/* all linked devices, *including* neighbours */ +struct { +struct list_head upper; +struct list_head lower; + } all_adj_list;/* currently active device features */
@@ -2813,15 +2823,15 @@ extern int bpf_jit_enable;extern bool netdev_has_upper_dev (struct net_device *dev,struct net_device *upper_dev);extern bool netdev_has_any_upper_dev (struct net_device *dev); -extern struct net_device *netdev_upper_get_next_dev_rcu (struct net_device *dev, -struct list_head **iter); +extern struct net_device *netdev_all_upper_get_next_dev_rcu (struct net_device *dev, +struct list_head **iter);/* iterate through upper list, must be called under RCU read lock */ -#define netdev_for_each_upper_dev_rcu (dev, upper, iter) \ - for (iter = &(dev)->upper_dev_list, \ - upper = netdev_upper_get_next_dev_rcu(dev, &(iter)); \ - upper; \ - upper = netdev_upper_get_next_dev_rcu(dev, &(iter))) +#define netdev_for_each_all_upper_dev_rcu (dev, updev, iter) \ + for (iter = &(dev)->all_adj_list.upper, \ + updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \ + updev; \ + updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))extern struct net_device *netdev_master_upper_dev_get (struct net_device *dev);extern struct net_device *netdev_master_upper_dev_get_rcu (struct net_device *dev);
net/core/dev.c +153 -66
436 lines not on this page · raw message
Signature · 1 line
1.8.4
PATCH v2 2/27 net: add RCU variant to search for netdev_adjacent link
In reply to Veaceslav FalicoCurrently we have only the RTNL flavour, however we can traverse it while holding only RCU, so add the RCU search. Add only one function that will be used further, other functions can be added easily afterwards, if anyone would need them.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
rename neighbour_dev_list/all_dev_list to adj_list/all_adj_list and rework the naming of functions - to make the use the pattern netdev_(all_)(lower|upper)_* . Uninline functions to get better stack traces - inline doesn't give much speed improvement, but this way the stack traces are meaningful. Remove the unused (through the whole patchset) functions - they can easily be added afterwards.
net/core/dev.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
net/core/dev.c +27 -0
@@ -4380,6 +4380,33 @@ struct netdev_adjacent {struct rcu_head rcu; }; +static struct netdev_adjacent *__netdev_find_adj_rcu (struct net_device *dev, +struct net_device *adj_dev, +bool upper,bool neighbour) +{ +struct netdev_adjacent *adj; +struct list_head *adj_list; + +if (neighbour) + adj_list = upper ? &dev->adj_list.upper : + &dev->adj_list.lower; +else + adj_list = upper ? &dev->all_adj_list.upper : + &dev->all_adj_list.lower; + +list_for_each_entry_rcu (adj, adj_list, list) { +if (adj->dev == adj_dev) +return adj; + } +return NULL ; +} + +static struct netdev_adjacent *__netdev_lower_find_rcu (struct net_device *dev, +struct net_device *ldev) +{ +return __netdev_find_adj_rcu (dev, ldev, false, true); +} +static struct netdev_adjacent *__netdev_find_adj (struct net_device *dev,struct net_device *adj_dev,bool upper,bool neighbour)
Signature · 1 line
1.8.4
PATCH v2 3/27 net: uninline netdev neighbour functions
In reply to Veaceslav FalicoThey don't give almost any speed/size advantage, however it's really useful to have them in the backtrace.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
New patch.net/core/dev.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-)
net/core/dev.c +11 -11
@@ -4684,25 +4684,25 @@ void __netdev_adjacent_dev_remove(struct net_device *dev,kfree_rcu (adj, rcu); } -static inline void __netdev_upper_dev_remove (struct net_device *dev, -struct net_device *udev) +static void __netdev_upper_dev_remove (struct net_device *dev, +struct net_device *udev) {return __netdev_adjacent_dev_remove (dev, udev, true, false); } -static inline void __netdev_upper_dev_remove_neighbour (struct net_device *dev, +static void __netdev_upper_dev_remove_neighbour (struct net_device *dev,struct net_device *udev) {return __netdev_adjacent_dev_remove (dev, udev, true, true); } -static inline void __netdev_lower_dev_remove (struct net_device *dev, -struct net_device *ldev) +static void __netdev_lower_dev_remove (struct net_device *dev, +struct net_device *ldev) {return __netdev_adjacent_dev_remove (dev, ldev, false, false); } -static inline void __netdev_lower_dev_remove_neighbour (struct net_device *dev, +static void __netdev_lower_dev_remove_neighbour (struct net_device *dev,struct net_device *ldev) {return __netdev_adjacent_dev_remove (dev, ldev, false, true);
@@ -4727,15 +4727,15 @@ int __netdev_adjacent_dev_insert_link(struct net_device *dev,return 0 ; } -static inline int __netdev_adjacent_dev_link (struct net_device *dev, -struct net_device *udev) +static int __netdev_adjacent_dev_link (struct net_device *dev, +struct net_device *udev) {return __netdev_adjacent_dev_insert_link (dev, udev, false, false); } -static inline int __netdev_adjacent_dev_link_neighbour (struct net_device *dev, -struct net_device *udev, -bool master) +static int __netdev_adjacent_dev_link_neighbour (struct net_device *dev, +struct net_device *udev, +bool master) {return __netdev_adjacent_dev_insert_link (dev, udev, master, true); }
Signature · 1 line
1.8.4
PATCH v2 4/27 net: add netdev_adjacent->private and allow to use it
In reply to Veaceslav FalicoCurrently, even though we can access any linked device, we can't attach anything to it, which is vital to properly manage them.
To fix this, add a new void *private to netdev_adjacent and functions setting/getting it (per link), so that we can save, per example, bonding's slave structures there, per slave device.
netdev_master_upper_dev_link_private(dev, upper_dev, private) links dev to upper dev and populates the neighbour link only with private.
netdev_lower_dev_get_private{,_rcu}() returns the private, if found.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
rename neighbour_dev_list/all_dev_list to adj_list/all_adj_list and rework
the naming of functions - to make the use the pattern
netdev_(all_)(lower|upper)_* .include/linux/netdevice.h | 7 ++++ net/core/dev.c | 83 ++++++++++++++++++++++++++++++++++++++--------- 2 files changed, 75 insertions(+), 15 deletions(-)
include/linux/netdevice.h +7 -0
@@ -2839,8 +2839,15 @@ extern int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);extern int netdev_master_upper_dev_link (struct net_device *dev,struct net_device *upper_dev); +extern int netdev_master_upper_dev_link_private (struct net_device *dev, +struct net_device *upper_dev, +void *private);extern void netdev_upper_dev_unlink (struct net_device *dev,struct net_device *upper_dev); +extern void *netdev_lower_dev_get_private_rcu (struct net_device *dev, +struct net_device *lower_dev); +extern void *netdev_lower_dev_get_private (struct net_device *dev, +struct net_device *lower_dev);extern int skb_checksum_help (struct sk_buff *skb);extern struct sk_buff *__skb_gso_segment (struct sk_buff *skb,netdev_features_t features,bool tx_path);
net/core/dev.c +68 -15
170 lines not on this page · raw message
Signature · 1 line
1.8.4
PATCH v2 5/27 bonding: populate neighbour's private on enslave
In reply to Veaceslav FalicoUse the new provided function when attaching the lower slave to populate its ->private with struct slave *new_slave. Also, move it to the end to be able to 'find' it only after it was completely initialized, and deinitialize in the first place on release.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Move the patch right after the introduction of ->private, to immediately
use it - easier for review and more logical.drivers/net/bonding/bond_main.c | 30 ++++++++++++++++-------------- 1 file changed, 16 insertions(+), 14 deletions(-)
drivers/net/bonding/bond_main.c +16 -14
@@ -1233,11 +1233,12 @@ static rx_handler_result_t bond_handle_frame(struct sk_buff **pskb) }static int bond_master_upper_dev_link (struct net_device *bond_dev, -struct net_device *slave_dev) +struct net_device *slave_dev, +struct slave *slave) {int err; - err =netdev_master_upper_dev_link (slave_dev, bond_dev); + err =netdev_master_upper_dev_link_private (slave_dev, bond_dev, slave);if (err)return err; slave_dev->flags |=IFF_SLAVE ;
@@ -1413,17 +1414,11 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) } } - res =bond_master_upper_dev_link (bond_dev, slave_dev); -if (res) { -pr_debug ("Error %d calling bond_master_upper_dev_link\n" , res); - goto err_restore_mac; - } -/* open the slave since the application closed it */ res =dev_open (slave_dev);if (res) {pr_debug ("Opening slave %s failed\n" , slave_dev->name); - goto err_unset_master; + goto err_restore_mac; } new_slave->bond = bond;
@@ -1637,6 +1632,13 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) goto err_dest_symlinks; } + res =bond_master_upper_dev_link (bond_dev, slave_dev, new_slave); +if (res) { +pr_debug ("Error %d calling bond_master_upper_dev_link\n" , res); + goto err_unregister; + } + + pr_info("%s: enslaving %s as a%s interface with a%s link.\n", bond_dev->name, slave_dev->name, bond_is_active_slave(new_slave) ? "n active" : " backup",
@@ -1646,6 +1648,9 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)return 0 ;/* Undo stages on error */ +err_unregister: +netdev_rx_handler_unregister (slave_dev); + err_dest_symlinks:bond_destroy_slave_symlinks (bond_dev, slave_dev);
@@ -1675,9 +1680,6 @@ err_close: slave_dev->priv_flags &= ~IFF_BONDING ;dev_close (slave_dev); -err_unset_master: -bond_upper_dev_unlink (bond_dev, slave_dev); - err_restore_mac:if (!bond->params.fail_over_mac) {/* XXX TODO - fom follow mode needs to change master's
@@ -1748,6 +1750,8 @@ static int __bond_release_one(struct net_device *bond_dev, }write_unlock_bh (&bond->lock); + +bond_upper_dev_unlink (bond_dev, slave_dev);/* unregister rx_handler early so bond_handle_frame wouldn't be called * for this slave anymore. */
@@ -1866,8 +1870,6 @@ static int __bond_release_one(struct net_device *bond_dev,bond_hw_addr_flush (bond_dev, slave_dev); } -bond_upper_dev_unlink (bond_dev, slave_dev); -slave_disable_netpoll (slave);/* close slave before restoring its mac address */
Signature · 1 line
1.8.4
PATCH v2 6/27 bonding: modify bond_get_slave_by_dev() to use neighbours
In reply to Veaceslav FalicoIt should be used under rtnl/bonding lock, so use the non-RCU version.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Move the patch right after we've populated ->private with bonding.drivers/net/bonding/bonding.h | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-)
drivers/net/bonding/bonding.h +2 -7
@@ -275,13 +275,8 @@ struct bonding {static inline struct slave *bond_get_slave_by_dev (struct bonding *bond,struct net_device *slave_dev) { -struct slave *slave =NULL ; - -bond_for_each_slave (bond, slave) -if (slave->dev == slave_dev) -return slave; - -return NULL ; +return (struct slave *)netdev_lower_dev_get_private (bond->dev, + slave_dev); }static inline struct bonding *bond_get_bond_by_slave (struct slave *slave)
Signature · 1 line
1.8.4
PATCH v2 7/27 net: add for_each iterators through neighbour lower link's private
In reply to Veaceslav FalicoAdd a possibility to iterate through netdev_adjacent's private, currently only for lower neighbours.
Add both RCU and RTNL/other locking variants of iterators.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
rename neighbour_dev_list/all_dev_list to adj_list/all_adj_list and rework the naming of functions - to make the use the pattern netdev_(all_)(lower|upper)_* . Also reword the commit message - I had to read it three times to understand :-/, and move it closer to the beginning of the patchset.
include/linux/netdevice.h | 17 ++++++++++++ net/core/dev.c | 66 +++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 83 insertions(+)
include/linux/netdevice.h +17 -0
@@ -2833,6 +2833,23 @@ extern struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dupdev ; \ updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter))) +extern void *netdev_lower_get_next_private (struct net_device *dev, +struct list_head **iter); +extern void *netdev_lower_get_next_private_rcu (struct net_device *dev, +struct list_head **iter); + +#define netdev_for_each_lower_private (dev, priv, iter) \ + for (iter = &(dev)->adj_list.lower, \ + priv = netdev_lower_get_next_private(dev, &(iter)); \ + priv; \ + priv = netdev_lower_get_next_private(dev, &(iter))) + +#define netdev_for_each_lower_private_rcu (dev, priv, iter) \ + for (iter = &(dev)->adj_list.lower, \ + priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \ + priv; \ + priv = netdev_lower_get_next_private_rcu(dev, &(iter))) +extern struct net_device *netdev_master_upper_dev_get (struct net_device *dev);extern struct net_device *netdev_master_upper_dev_get_rcu (struct net_device *dev); extern int netdev_upper_dev_link(struct net_device *dev,
net/core/dev.c +66 -0
72 lines not on this page · raw message
Signature · 1 line
1.8.4
Re: [PATCH v2 net-next 07/27] net: add for_each iterators through neighbour lower link's private
net/core/dev.c struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, · line 4588
Quoted 7 lines · Veaceslav Falico
net/core/dev.c +7
-0+void *netdev_lower_get_next_private_rcu (struct net_device *dev, +struct list_head **iter) +{ +struct netdev_adjacent *lower; + +WARN_ON_ONCE (!rcu_read_lock_held ()); +
Since this function has _rcu suffix, this warning is not useful at all.
Re: [PATCH v2 net-next 07/27] net: add for_each iterators through neighbour lower link's private
Quoted 10 lines · Cong Wang
net/core/dev.c struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev, · line 4588
Quoted 7 lines · Veaceslav Falico
net/core/dev.c +7
-0+void *netdev_lower_get_next_private_rcu (struct net_device *dev, +struct list_head **iter) +{ +struct netdev_adjacent *lower; + +WARN_ON_ONCE (!rcu_read_lock_held ()); +Since this function has
_rcusuffix, this warning is not useful at all.
I've been using it to catch offenders, just in case, like BUG_ON.
I can easily remove it, if needed, there are *tons* of cleanups that can be made for this patchset, tbh, I've just left it the way it is to be able to catch bug/review easier :).
p.s. Please, use reply-all when replying...
Signature · 2 lines
Best regards, Veaceslav Falico
PATCH v2 8/27 bonding: remove bond_for_each_slave_reverse()
In reply to Veaceslav FalicoWe only use it in rollback scenarios and can easily use the standart bond_for_each_dev() instead.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Move the patch right before we change bond_for_each_slave() conversion, so that we don't have to convert bond_for_each_slave_reverse() too.
drivers/net/bonding/bond_alb.c | 14 ++++++++------ drivers/net/bonding/bond_main.c | 34 ++++++++++++++++++++++------------ drivers/net/bonding/bonding.h | 10 ---------- 3 files changed, 30 insertions(+), 28 deletions(-)
drivers/net/bonding/bond_alb.c +8 -6
@@ -1246,9 +1246,9 @@ static int alb_handle_addr_collision_on_attach(struct bonding *bond, struct slav*/ static int alb_set_mac_address (struct bonding *bond,void *addr) { -char tmp_addr[ETH_ALEN ]; -struct slave *slave; +struct slave *slave, *rollback_slave;struct sockaddr sa; +char tmp_addr[ETH_ALEN ];int res;if (bond->alb_info.rlb_enabled)
@@ -1274,10 +1274,12 @@ unwind: sa.sa_family = bond->dev->type;/* unwind from head to the slave that failed */ -bond_for_each_slave_continue_reverse (bond, slave) { -memcpy (tmp_addr, slave->dev->dev_addr,ETH_ALEN ); -dev_set_mac_address (slave->dev, &sa); -memcpy (slave->dev->dev_addr, tmp_addr,ETH_ALEN ); +bond_for_each_slave (bond, rollback_slave) { +if (rollback_slave == slave) +break ; +memcpy (tmp_addr, rollback_slave->dev->dev_addr,ETH_ALEN ); +dev_set_mac_address (rollback_slave->dev, &sa); +memcpy (rollback_slave->dev->dev_addr, tmp_addr,ETH_ALEN ); }return res;
drivers/net/bonding/bond_main.c +22 -12
@@ -332,7 +332,7 @@ static int bond_vlan_rx_add_vid(struct net_device *bond_dev, __be16 proto, u16 vid) {struct bonding *bond =netdev_priv (bond_dev); -struct slave *slave; +struct slave *slave, *rollback_slave;int res;bond_for_each_slave (bond, slave) {
@@ -344,9 +344,13 @@ static int bond_vlan_rx_add_vid(struct net_device *bond_dev,return 0 ; unwind: -/* unwind from the slave that failed */ -bond_for_each_slave_continue_reverse (bond, slave) -vlan_vid_del (slave->dev, proto, vid); +/* unwind to the slave that failed */ +bond_for_each_slave (bond, rollback_slave) { +if (rollback_slave == slave) +break ; + +vlan_vid_del (rollback_slave->dev, proto, vid); + }return res; }
@@ -3468,7 +3472,7 @@ static int bond_neigh_setup(struct net_device *dev,static int bond_change_mtu (struct net_device *bond_dev,int new_mtu) {struct bonding *bond =netdev_priv (bond_dev); -struct slave *slave; +struct slave *slave, *rollback_slave;int res =0 ; pr_debug("bond=%p, name=%s, new_mtu=%d\n", bond,
@@ -3517,13 +3521,16 @@ static int bond_change_mtu(struct net_device *bond_dev, int new_mtu) unwind:/* unwind from head to the slave that failed */ -bond_for_each_slave_continue_reverse (bond, slave) { +bond_for_each_slave (bond, rollback_slave) {int tmp_res; - tmp_res =dev_set_mtu (slave->dev, bond_dev->mtu); +if (rollback_slave == slave); +break ; + + tmp_res =dev_set_mtu (rollback_slave->dev, bond_dev->mtu);if (tmp_res) {pr_debug ("unwind err %d dev %s\n" , - tmp_res, slave->dev->name); + tmp_res, rollback_slave->dev->name); } }
@@ -3540,8 +3547,8 @@ unwind:static int bond_set_mac_address (struct net_device *bond_dev,void *addr) {struct bonding *bond =netdev_priv (bond_dev); +struct slave *slave, *rollback_slave;struct sockaddr *sa = addr, tmp_sa; -struct slave *slave;int res =0 ;if (bond->params.mode ==BOND_MODE_ALB )
@@ -3607,13 +3614,16 @@ unwind: tmp_sa.sa_family = bond_dev->type;/* unwind from head to the slave that failed */ -bond_for_each_slave_continue_reverse (bond, slave) { +bond_for_each_slave (bond, rollback_slave) {int tmp_res; - tmp_res =dev_set_mac_address (slave->dev, &tmp_sa); +if (rollback_slave == slave) +break ; + + tmp_res =dev_set_mac_address (rollback_slave->dev, &tmp_sa);if (tmp_res) {pr_debug ("unwind err %d dev %s\n" , - tmp_res, slave->dev->name); + tmp_res, rollback_slave->dev->name); } }
drivers/net/bonding/bonding.h +0 -10
@@ -120,16 +120,6 @@#define bond_for_each_slave_rcu (bond, pos) \ list_for_each_entry_rcu(pos, &(bond)->slave_list, list) -/** -* bond_for_each_slave_reverse - iterate in reverse from a given position -* @bond: the bond holding this list -* @pos: slave to continue from -* -* Caller must hold bond->lock -*/ -#define bond_for_each_slave_continue_reverse (bond, pos) \ - list_for_each_entry_continue_reverse(pos, &(bond)->slave_list, list) -#ifdef CONFIG_NET_POLL_CONTROLLER extern atomic_t netpoll_block_tx;
Signature · 1 line
1.8.4
Re: [PATCH v2 net-next 08/27] bonding: remove bond_for_each_slave_reverse()
Quoted 3 lines · Veaceslav Falico
We only use it in rollback scenarios and can easily use the standart
bond_for_each_dev()instead.
What you remove actually is bond_for_each_slave_continue_reverse()... $subject needs to be fixed.
drivers/net/bonding/bond_alb.c static int alb_handle_addr_collision_on_attach(struct bonding *bond, struct slav · line 1251
Quoted 13 lines
drivers/net/bonding/bond_alb.c +2
-2@@ -1246,9 +1246,9 @@ static int alb_handle_addr_collision_on_attach(struct bonding *bond, struct slav*/ static int alb_set_mac_address (struct bonding *bond,void *addr) { -char tmp_addr[ETH_ALEN ]; -struct slave *slave; +struct slave *slave, *rollback_slave;struct sockaddr sa; +char tmp_addr[ETH_ALEN ];
Why are you moving tmp_addr[]?
Re: [PATCH v2 net-next 08/27] bonding: remove bond_for_each_slave_reverse()
Quoted 7 lines · Cong Wang
Quoted 3 lines · Veaceslav Falico
We only use it in rollback scenarios and can easily use the standart
bond_for_each_dev()instead.What you remove actually is
bond_for_each_slave_continue_reverse()... $subject needs to be fixed.
Indeed, thank you, will fix :).
Quoted 17 lines
Quoted 7 lines
drivers/net/bonding/bond_alb.c +0
-0drivers/net/bonding/bond_alb.c static int alb_handle_addr_collision_on_attach(struct bonding *bond, struct slav · line 1251
Quoted 8 lines
drivers/net/bonding/bond_alb.c +2
-2@@ -1246,9 +1246,9 @@ static int alb_handle_addr_collision_on_attach(struct bonding *bond, struct slav*/ static int alb_set_mac_address (struct bonding *bond,void *addr) { -char tmp_addr[ETH_ALEN ]; -struct slave *slave; +struct slave *slave, *rollback_slave;struct sockaddr sa; +char tmp_addr[ETH_ALEN ];Why are you moving tmp_addr[]?
It's more readable, IMO, if structs go first.
Signature · 2 lines
Best regards, Veaceslav Falico
PATCH v2 9/27 bonding: make bond_for_each_slave() use lower neighbour's private
In reply to Veaceslav FalicoIt needs a list_head *iter, so add it wherever needed. Use both non-rcu and rcu variants.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Move the patch right after we're ready to use it - for easier review. Follow renaming strategy.
Add cxgb4 driver - it also began to use bond_for_each_slave().
drivers/net/bonding/bond_3ad.c | 6 +- drivers/net/bonding/bond_alb.c | 18 ++++-- drivers/net/bonding/bond_main.c | 86 ++++++++++++++++--------- drivers/net/bonding/bond_procfs.c | 5 +- drivers/net/bonding/bond_sysfs.c | 23 ++++--- drivers/net/bonding/bonding.h | 12 ++-- drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c | 3 +- 7 files changed, 98 insertions(+), 55 deletions(-)
drivers/net/bonding/bond_3ad.c +4 -2
@@ -2419,6 +2419,7 @@ int bond_3ad_xmit_xor(struct sk_buff *skb, struct net_device *dev) {struct slave *slave, *start_at;struct bonding *bond =netdev_priv (dev); +struct list_head *iter;int slave_agg_no;int slaves_in_agg;int agg_id;
@@ -2444,7 +2445,7 @@ int bond_3ad_xmit_xor(struct sk_buff *skb, struct net_device *dev) slave_agg_no = bond->xmit_hash_policy (skb, slaves_in_agg); -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {struct aggregator *agg =SLAVE_AD_INFO (slave).port.aggregator;if (agg && (agg->aggregator_identifier == agg_id)) {
@@ -2515,11 +2516,12 @@ int bond_3ad_lacpdu_recv(const struct sk_buff *skb, struct bonding *bond,void bond_3ad_update_lacp_rate (struct bonding *bond) {struct port *port =NULL ; +struct list_head *iter;struct slave *slave;int lacp_fast; lacp_fast = bond->params.lacp_fast; -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) { port = &(SLAVE_AD_INFO (slave).port);__get_state_machine_lock (port);if (lacp_fast)
drivers/net/bonding/bond_alb.c +11 -7
@@ -223,13 +223,14 @@ static long long compute_gap(struct slave *slave)static struct slave *tlb_get_least_loaded_slave (struct bonding *bond) {struct slave *slave, *least_loaded; +struct list_head *iter;long long max_gap; least_loaded =NULL ; max_gap =LLONG_MIN ;/* Find the slave with the largest gap */ -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {if (SLAVE_IS_OK (slave)) {long long gap =compute_gap (slave);
@@ -1172,8 +1173,9 @@ static void alb_change_hw_addr_on_detach(struct bonding *bond, struct slave *sla*/ static int alb_handle_addr_collision_on_attach (struct bonding *bond,struct slave *slave) { -struct slave *tmp_slave1, *free_mac_slave =NULL ;struct slave *has_bond_addr = bond->curr_active_slave; +struct slave *tmp_slave1, *free_mac_slave =NULL ; +struct list_head *iter;if (list_empty (&bond->slave_list)) {/* this is the first slave */
@@ -1196,7 +1198,7 @@ static int alb_handle_addr_collision_on_attach(struct bonding *bond, struct slav/* The slave's address is equal to the address of the bond. * Search for a spare address in the bond for this slave. */ -bond_for_each_slave (bond, tmp_slave1) { +bond_for_each_slave (bond, tmp_slave1, iter) {if (!bond_slave_has_mac (bond, tmp_slave1->perm_hwaddr)) {/* no slave has tmp_slave1's perm addr * as its curr addr
@@ -1247,6 +1249,7 @@ static int alb_handle_addr_collision_on_attach(struct bonding *bond, struct slavstatic int alb_set_mac_address (struct bonding *bond,void *addr) {struct slave *slave, *rollback_slave; +struct list_head *iter;struct sockaddr sa;char tmp_addr[ETH_ALEN ];int res;
@@ -1254,7 +1257,7 @@ static int alb_set_mac_address(struct bonding *bond, void *addr)if (bond->alb_info.rlb_enabled)return 0 ; -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {/* save net_device's current hw address */ memcpy (tmp_addr, slave->dev->dev_addr,ETH_ALEN );
@@ -1274,7 +1277,7 @@ unwind: sa.sa_family = bond->dev->type;/* unwind from head to the slave that failed */ -bond_for_each_slave (bond, rollback_slave) { +bond_for_each_slave (bond, rollback_slave, iter) {if (rollback_slave == slave)break ;memcpy (tmp_addr, rollback_slave->dev->dev_addr,ETH_ALEN );
@@ -1460,6 +1463,7 @@ void bond_alb_monitor(struct work_struct *work) struct bonding *bond = container_of(work, struct bonding, alb_work.work);struct alb_bond_info *bond_info = &(BOND_ALB_INFO (bond)); +struct list_head *iter;struct slave *slave;read_lock (&bond->lock);
@@ -1482,7 +1486,7 @@ void bond_alb_monitor(struct work_struct *work)*/ read_lock (&bond->curr_slave_lock); -bond_for_each_slave (bond, slave) +bond_for_each_slave (bond, slave, iter)alb_send_learning_packets (slave, slave->dev->dev_addr);read_unlock (&bond->curr_slave_lock);
@@ -1495,7 +1499,7 @@ void bond_alb_monitor(struct work_struct *work)read_lock (&bond->curr_slave_lock); -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {tlb_clear_slave (bond, slave,1 );if (slave == bond->curr_active_slave) { SLAVE_TLB_INFO(slave).load =
drivers/net/bonding/bond_main.c +56 -30
364 lines not on this page · raw message
drivers/net/bonding/bond_procfs.c +3 -2
@@ -10,8 +10,9 @@ static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos) __acquires(&bond->lock) {struct bonding *bond = seq->private; -loff_t off =0 ; +struct list_head *iter;struct slave *slave; +loff_t off =0 ;/* make sure the bond won't be taken away */ rcu_read_lock ();
@@ -20,7 +21,7 @@ static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos)if (*pos ==0 )return SEQ_START_TOKEN ; -bond_for_each_slave (bond, slave) +bond_for_each_slave (bond, slave, iter)if (++off == *pos)return slave;
drivers/net/bonding/bond_sysfs.c +15 -8
@@ -210,11 +210,12 @@ static ssize_t bonding_show_slaves(struct device *d, struct device_attribute *attr, char *buf) { struct bonding *bond = to_bond(d); +struct list_head *iter;struct slave *slave;int res =0 ;read_lock (&bond->lock); -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {if (res > (PAGE_SIZE -IFNAMSIZ )) {/* not enough space for another interface name */ if ((PAGE_SIZE - res) >10 )
@@ -637,6 +638,7 @@ static ssize_t bonding_store_arp_targets(struct device *d, const char *buf, size_t count) {struct bonding *bond =to_bond (d); +struct list_head *iter;struct slave *slave;__be32 newtarget, *targets;unsigned long *targets_rx;
@@ -669,7 +671,7 @@ static ssize_t bonding_store_arp_targets(struct device *d, &newtarget);/* not to race with bond_arp_rcv */ write_lock_bh (&bond->lock); -bond_for_each_slave (bond, slave) +bond_for_each_slave (bond, slave, iter) slave->target_last_arp_rx[ind] = jiffies; targets[ind] = newtarget;write_unlock_bh (&bond->lock);
@@ -695,7 +697,7 @@ static ssize_t bonding_store_arp_targets(struct device *d, &newtarget);write_lock_bh (&bond->lock); -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) { targets_rx = slave->target_last_arp_rx; j = ind;for (; (j <BOND_MAX_ARP_TARGETS -1 ) && targets[j+1 ]; j++)
@@ -1092,6 +1094,7 @@ static ssize_t bonding_store_primary(struct device *d, const char *buf, size_t count) {struct bonding *bond =to_bond (d); +struct list_head *iter;char ifname[IFNAMSIZ ];struct slave *slave;
@@ -1119,7 +1122,7 @@ static ssize_t bonding_store_primary(struct device *d, goto out; } -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {if (strncmp (slave->dev->name, ifname,IFNAMSIZ ) ==0 ) {pr_info ("%s: Setting %s as primary slave.\n" , bond->dev->name, slave->dev->name);
@@ -1267,6 +1270,7 @@ static ssize_t bonding_store_active_slave(struct device *d, {struct slave *slave, *old_active, *new_active;struct bonding *bond =to_bond (d); +struct list_head *iter;char ifname[IFNAMSIZ ];if (!rtnl_trylock ())
@@ -1294,7 +1298,7 @@ static ssize_t bonding_store_active_slave(struct device *d, goto out; } -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {if (strncmp (slave->dev->name, ifname,IFNAMSIZ ) ==0 ) { old_active = bond->curr_active_slave; new_active = slave;
@@ -1474,6 +1478,7 @@ static ssize_t bonding_show_queue_id(struct device *d, char *buf) {struct bonding *bond =to_bond (d); +struct list_head *iter;struct slave *slave;int res =0 ;
@@ -1481,7 +1486,7 @@ static ssize_t bonding_show_queue_id(struct device *d,return restart_syscall ();read_lock (&bond->lock); -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {if (res > (PAGE_SIZE -IFNAMSIZ -6 )) {/* not enough space for another interface_name:queue_id pair */ if ((PAGE_SIZE - res) >10 )
@@ -1510,6 +1515,7 @@ static ssize_t bonding_store_queue_id(struct device *d, {struct slave *slave, *update_slave;struct bonding *bond =to_bond (d); +struct list_head *iter;u16 qid;int ret = count;char *delim;
@@ -1546,7 +1552,7 @@ static ssize_t bonding_store_queue_id(struct device *d,/* Search for thes slave and check for duplicate qids */ update_slave =NULL ; -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {if (sdev == slave->dev)/* * We don't need to check the matching
@@ -1600,6 +1606,7 @@ static ssize_t bonding_store_slaves_active(struct device *d, {struct bonding *bond =to_bond (d);int new_value, ret = count; +struct list_head *iter;struct slave *slave;if (sscanf (buf,"%d" , &new_value) !=1 ) {
@@ -1622,7 +1629,7 @@ static ssize_t bonding_store_slaves_active(struct device *d, }read_lock (&bond->lock); -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {if (!bond_is_active_slave (slave)) {if (new_value) slave->inactive =0 ;
drivers/net/bonding/bonding.h +7 -5
@@ -110,15 +110,16 @@* bond_for_each_slave - iterate over all slaves * @bond: the bond holding this list * @pos: current slave +* @iter: list_head * iterator * * Caller must hold bond->lock */ -#define bond_for_each_slave (bond, pos) \ - list_for_each_entry(pos, &(bond)->slave_list, list) +#define bond_for_each_slave (bond, pos, iter) \ + netdev_for_each_lower_private((bond)->dev, pos, iter)/* Caller must have rcu_read_lock */ -#define bond_for_each_slave_rcu (bond, pos) \ - list_for_each_entry_rcu(pos, &(bond)->slave_list, list) +#define bond_for_each_slave_rcu (bond, pos, iter) \ + netdev_for_each_lower_private_rcu((bond)->dev, pos, iter)#ifdef CONFIG_NET_POLL_CONTROLLER extern atomic_t netpoll_block_tx;
@@ -475,9 +476,10 @@ static inline void bond_destroy_proc_dir(struct bond_net *bn)static inline struct slave *bond_slave_has_mac (struct bonding *bond,const u8 *mac) { +struct list_head *iter;struct slave *tmp; -bond_for_each_slave (bond, tmp) +bond_for_each_slave (bond, tmp, iter)if (ether_addr_equal_64bits (mac, tmp->dev->dev_addr))return tmp;
drivers/net/ethernet/chelsio/cxgb4/cxgb4_main.c +2 -1
@@ -3983,6 +3983,7 @@ static int cxgb4_inet6addr_handler(struct notifier_block *this,struct net_device *event_dev;int ret =NOTIFY_DONE ;struct bonding *bond =netdev_priv (ifa->idev->dev); +struct list_head *iter;struct slave *slave;struct pci_dev *first_pdev =NULL ;
@@ -3995,7 +3996,7 @@ static int cxgb4_inet6addr_handler(struct notifier_block *this,* in all of them only once. */ read_lock (&bond->lock); -bond_for_each_slave (bond, slave) { +bond_for_each_slave (bond, slave, iter) {if (!first_pdev) { ret =clip_add (slave->dev, ifa, event);/* If clip_add is success then only initialize
Signature · 1 line
1.8.4
PATCH v2 10/27 bonding: use bond_for_each_slave() in bond_uninit()
In reply to Veaceslav FalicoWe're safe agains removal there, cause we use neighbours primitives.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Move the patch rigth after we start using neighbour lists for
bond_for_each_slave().drivers/net/bonding/bond_main.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
drivers/net/bonding/bond_main.c +3 -2
@@ -4090,12 +4090,13 @@ static void bond_setup(struct net_device *bond_dev)static void bond_uninit (struct net_device *bond_dev) {struct bonding *bond =netdev_priv (bond_dev); -struct slave *slave, *tmp_slave; +struct list_head *iter; +struct slave *slave;bond_netpoll_cleanup (bond_dev);/* Release the bonded slaves */ -list_for_each_entry_safe (slave, tmp_slave, &bond->slave_list, list) +bond_for_each_slave (bond, slave, iter)__bond_release_one (bond_dev, slave->dev, true);pr_info ("%s: released all slaves\n" , bond_dev->name);
Signature · 1 line
1.8.4
PATCH v2 11/27 bonding: rework bond_3ad_xmit_xor() to use bond_for_each_slave() only
In reply to Veaceslav FalicoCurrently, there are two loops - first we find the first slave in an aggregator after the xmit_hash_policy() returned number, and after that we loop from that slave, over bonding head, and till that slave to find any suitable slave to send the packet through.
Replace it by just one bond_for_each_slave() loop, which first loops through the requested number of slaves, saving the first suitable one, and after that we've hit the requested number of slaves to skip - search for any up slave to send the packet through. If we don't find such kind of slave - then just send the packet through the first suitable slave found.
Logic remains unchainged, and we skip two loops. Also, refactor it a bit for readability.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
New patch.drivers/net/bonding/bond_3ad.c | 46 ++++++++++++++++++++---------------------- 1 file changed, 22 insertions(+), 24 deletions(-)
drivers/net/bonding/bond_3ad.c +22 -24
@@ -2417,15 +2417,15 @@ int bond_3ad_get_active_agg_info(struct bonding *bond, struct ad_info *ad_info)int bond_3ad_xmit_xor (struct sk_buff *skb,struct net_device *dev) { -struct slave *slave, *start_at;struct bonding *bond =netdev_priv (dev); +struct slave *slave, *first_ok_slave; +struct aggregator *agg; +struct ad_info ad_info;struct list_head *iter; -int slave_agg_no;int slaves_in_agg; -int agg_id; -int i; -struct ad_info ad_info; +int slave_agg_no;int res =1 ; +int agg_id;read_lock (&bond->lock);if (__bond_3ad_get_active_agg_info (bond, &ad_info)) {
@@ -2438,20 +2438,28 @@ int bond_3ad_xmit_xor(struct sk_buff *skb, struct net_device *dev) agg_id = ad_info.aggregator_id;if (slaves_in_agg ==0 ) { -/*the aggregator is empty*/ pr_debug ("%s: Error: active aggregator is empty\n" , dev->name); goto out; } slave_agg_no = bond->xmit_hash_policy (skb, slaves_in_agg); + first_ok_slave =NULL ;bond_for_each_slave (bond, slave, iter) { -struct aggregator *agg =SLAVE_AD_INFO (slave).port.aggregator; + agg =SLAVE_AD_INFO (slave).port.aggregator; +if (!agg || agg->aggregator_identifier != agg_id) +continue ; -if (agg && (agg->aggregator_identifier == agg_id)) { +if (slave_agg_no >=0 ) { +if (!first_ok_slave &&SLAVE_IS_OK (slave)) + first_ok_slave = slave; slave_agg_no--; -if (slave_agg_no <0 ) -break ; +continue ; + } + +if (SLAVE_IS_OK (slave)) { + res =bond_dev_queue_xmit (bond, skb, slave->dev); + goto out; } }
@@ -2461,20 +2469,10 @@ int bond_3ad_xmit_xor(struct sk_buff *skb, struct net_device *dev) goto out; } - start_at = slave; - -bond_for_each_slave_from (bond, slave, i, start_at) { -int slave_agg_id =0 ; -struct aggregator *agg =SLAVE_AD_INFO (slave).port.aggregator; - -if (agg) - slave_agg_id = agg->aggregator_identifier; - -if (SLAVE_IS_OK (slave) && agg && (slave_agg_id == agg_id)) { - res =bond_dev_queue_xmit (bond, skb, slave->dev); -break ; - } - } +/* we couldn't find any suitable slave after the agg_no, so use the +* first suitable found, if found. */ +if (first_ok_slave) + res =bond_dev_queue_xmit (bond, skb, first_ok_slave->dev); out:read_unlock (&bond->lock);
Signature · 1 line
1.8.4
PATCH v2 12/27 bonding: rework rlb_next_rx_slave() to use bond_for_each_slave()
In reply to Veaceslav FalicoCurrently, we're using bond_for_each_slave_from(), which is really hard to implement under RCU and/or neighbour list.
Remove it and use bond_for_each_slave() instead, taking care of the last used slave.
Also, rename next_rx_slave to rx_slave and store the current (last) rx_slave.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
New patch.drivers/net/bonding/bond_alb.c | 41 +++++++++++++++++++++-------------------- drivers/net/bonding/bond_alb.h | 4 +--- 2 files changed, 22 insertions(+), 23 deletions(-)
drivers/net/bonding/bond_alb.c +21 -20
@@ -383,30 +383,31 @@ out:static struct slave *rlb_next_rx_slave (struct bonding *bond) {struct alb_bond_info *bond_info = &(BOND_ALB_INFO (bond)); -struct slave *rx_slave, *slave, *start_at; -int i =0 ; - -if (bond_info->next_rx_slave) - start_at = bond_info->next_rx_slave; -else - start_at =bond_first_slave (bond); - - rx_slave =NULL ; +struct slave *before =NULL , *rx_slave =NULL , *slave; +struct list_head *iter; +bool found = false; -bond_for_each_slave_from (bond, slave, i, start_at) { -if (SLAVE_IS_OK (slave)) { -if (!rx_slave) { - rx_slave = slave; - }else if (slave->speed > rx_slave->speed) { +bond_for_each_slave (bond, slave, iter) { +if (!SLAVE_IS_OK (slave)) +continue ; +if (!found) { +if (!before || before->speed < slave->speed) + before = slave; + }else { +if (!rx_slave || rx_slave->speed < slave->speed) rx_slave = slave; - } } +if (slave == bond_info->rx_slave) + found = true; } +/* we didn't find anything after the current or we have something +* better before and up to the current slave +*/ +if (!rx_slave || (before && rx_slave->speed < before->speed)) + rx_slave = before; -if (rx_slave) { - slave =bond_next_slave (bond, rx_slave); - bond_info->next_rx_slave = slave; - } +if (rx_slave) + bond_info->rx_slave = rx_slave;return rx_slave; }
@@ -1611,7 +1612,7 @@ void bond_alb_deinit_slave(struct bonding *bond, struct slave *slave)tlb_clear_slave (bond, slave,0 );if (bond->alb_info.rlb_enabled) { - bond->alb_info.next_rx_slave =NULL ; + bond->alb_info.rx_slave =NULL ;rlb_clear_slave (bond, slave); } }
drivers/net/bonding/bond_alb.h +1 -3
@@ -153,9 +153,7 @@ struct alb_bond_info {u8 rx_ntt;/* flag - need to transmit * to all rx clients */ -struct slave *next_rx_slave;/* next slave to be assigned -* to a new rx client for -*/ +struct slave *rx_slave;/* last slave to xmit from */ u8 primary_is_promisc;/* boolean */ u32 rlb_promisc_timeout_counter;/* counts primary * promiscuity time
Signature · 1 line
1.8.4
PATCH v2 13/27 bonding: rework bond_find_best_slave() to use bond_for_each_slave()
In reply to Veaceslav Falicobond_find_best_slave() does not have to be balanced - i.e. return the slave that is *after* some other slave, but rather return the best slave that suits, except of bond->primary_slave - in which case we just return it if it's suitable.
After that we just look through all the slaves and return either first up slave or the slave whose link came back earliest.
We also don't care about curr_active_slave lock cause we use it in bond_should_change_active() only and there we take it right away - i.e. it won't go away.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
New patch.drivers/net/bonding/bond_main.c | 43 ++++++++++++----------------------------- 1 file changed, 12 insertions(+), 31 deletions(-)
drivers/net/bonding/bond_main.c +12 -31
@@ -785,43 +785,24 @@ static bool bond_should_change_active(struct bonding *bond)/** * find_best_interface - select the best available slave to be the active one * @bond: our bonding struct -* -* Warning: Caller must hold curr_slave_lock for writing. */ static struct slave *bond_find_best_slave (struct bonding *bond) { -struct slave *new_active, *old_active; -struct slave *bestslave =NULL ; +struct slave *slave, *bestslave =NULL ; +struct list_head *iter;int mintime = bond->params.updelay; -int i; - new_active = bond->curr_active_slave; - -if (!new_active) {/* there were no active slaves left */ - new_active =bond_first_slave (bond); -if (!new_active) -return NULL ;/* still no slave, return NULL */ - } +if (bond->primary_slave && bond->primary_slave->link ==BOND_LINK_UP && +bond_should_change_active (bond)) +return bond->primary_slave; -if ((bond->primary_slave) && - bond->primary_slave->link ==BOND_LINK_UP && -bond_should_change_active (bond)) { - new_active = bond->primary_slave; - } - -/* remember where to stop iterating over the slaves */ - old_active = new_active; - -bond_for_each_slave_from (bond, new_active, i, old_active) { -if (new_active->link ==BOND_LINK_UP ) { -return new_active; - }else if (new_active->link ==BOND_LINK_BACK && -IS_UP (new_active->dev)) { -/* link up, but waiting for stabilization */ -if (new_active->delay < mintime) { - mintime = new_active->delay; - bestslave = new_active; - } +bond_for_each_slave (bond, slave, iter) { +if (slave->link ==BOND_LINK_UP ) +return slave; +if (slave->link ==BOND_LINK_BACK &&IS_UP (slave->dev) && + slave->delay < mintime) { + mintime = slave->delay; + bestslave = slave; } }
Signature · 1 line
1.8.4
PATCH v2 14/27 bonding: rework bond_ab_arp_probe() to use bond_for_each_slave()
In reply to Veaceslav FalicoCurrently it uses the hard-to-rcuify bond_for_each_slave_from(), and also it doesn't check every slave for disrepencies between the actual IS_UP(slave) and the slave->link == BOND_LINK_UP, but only till we find the next suitable slave.
Fix this by using bond_for_each_slave() and storing the first good slave in *before till we find the current_arp_slave, after that we store the first good slave in new_slave. If new_slave is empty - use the slave stored in before, and if it's also empty - then we didn't find any suitable slave.
Also, in the meanwhile, check for each slave status.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
New patch.drivers/net/bonding/bond_main.c | 38 ++++++++++++++++++++++++-------------- 1 file changed, 24 insertions(+), 14 deletions(-)
drivers/net/bonding/bond_main.c +24 -14
@@ -2761,8 +2761,9 @@ do_failover:*/ static void bond_ab_arp_probe (struct bonding *bond) { -struct slave *slave, *next_slave; -int i; +struct slave *slave, *before =NULL , *new_slave =NULL ; +struct list_head *iter; +bool found = false;read_lock (&bond->curr_slave_lock);
@@ -2792,18 +2793,12 @@ static void bond_ab_arp_probe(struct bonding *bond)bond_set_slave_inactive_flags (bond->current_arp_slave); -/* search for next candidate */ - next_slave =bond_next_slave (bond, bond->current_arp_slave); -bond_for_each_slave_from (bond, slave, i, next_slave) { -if (IS_UP (slave->dev)) { - slave->link =BOND_LINK_BACK ; -bond_set_slave_active_flags (slave); -bond_arp_send_all (bond, slave); - slave->jiffies = jiffies; - bond->current_arp_slave = slave; -break ; - } +bond_for_each_slave (bond, slave, iter) { +if (!found && !before &&IS_UP (slave->dev)) + before = slave; +if (found && !new_slave &&IS_UP (slave->dev)) + new_slave = slave;/* if the link state is up at this point, we * mark it down - this can happen if we have * simultaneous link failures and
@@ -2811,7 +2806,7 @@ static void bond_ab_arp_probe(struct bonding *bond)* one the current slave so it is still marked * up when it is actually down */ -if (slave->link ==BOND_LINK_UP ) { +if (!IS_UP (slave->dev) && slave->link ==BOND_LINK_UP ) { slave->link =BOND_LINK_DOWN ;if (slave->link_failure_count <UINT_MAX ) slave->link_failure_count++;
@@ -2821,7 +2816,22 @@ static void bond_ab_arp_probe(struct bonding *bond)pr_info ("%s: backup interface %s is now down.\n" , bond->dev->name, slave->dev->name); } +if (slave == bond->current_arp_slave) + found = true; } + +if (!new_slave && before) + new_slave = before; + +if (!new_slave) +return ; + + new_slave->link =BOND_LINK_BACK ; +bond_set_slave_active_flags (new_slave); +bond_arp_send_all (bond, new_slave); + new_slave->jiffies = jiffies; + bond->current_arp_slave = new_slave; + }void bond_activebackup_arp_mon (struct work_struct *work)
Signature · 1 line
1.8.4
PATCH v2 15/27 bonding: remove unused bond_for_each_slave_from()
In reply to Veaceslav FalicoIt has no users, so we can remove it.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
New patch.drivers/net/bonding/bonding.h | 13 ------------- 1 file changed, 13 deletions(-)
drivers/net/bonding/bonding.h +0 -13
@@ -94,19 +94,6 @@ bond_to_slave((pos)->list.prev))/** -* bond_for_each_slave_from - iterate the slaves list from a starting point -* @bond: the bond holding this list. -* @pos: current slave. -* @cnt: counter for max number of moves -* @start: starting point. -* -* Caller must hold bond->lock -*/ -#define bond_for_each_slave_from (bond, pos, cnt, start) \ - for (cnt = 0, pos = start; pos && cnt < (bond)->slave_cnt; \ - cnt++, pos = bond_next_slave(bond, pos)) - -/** * bond_for_each_slave - iterate over all slaves * @bond: the bond holding this list * @pos: current slave
Signature · 1 line
1.8.4
PATCH v2 16/27 bonding: add bond_has_slaves() and use it
In reply to Veaceslav FalicoCurrently we verify if we have slaves by checking if bond->slave_list is empty. Create a define bond_has_slaves() and use it, a bit more readable and easier to change in the future.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
No change.drivers/net/bonding/bond_3ad.c | 2 +- drivers/net/bonding/bond_alb.c | 8 ++++---- drivers/net/bonding/bond_main.c | 32 ++++++++++++++++---------------- drivers/net/bonding/bond_sysfs.c | 4 ++-- drivers/net/bonding/bonding.h | 2 ++ 5 files changed, 25 insertions(+), 23 deletions(-)
drivers/net/bonding/bond_3ad.c +1 -1
@@ -2117,7 +2117,7 @@ void bond_3ad_state_machine_handler(struct work_struct *work)read_lock (&bond->lock);//check if there are any slaves -if (list_empty (&bond->slave_list)) +if (!bond_has_slaves (bond)) goto re_arm;// check if agg_select_timer timer after initialize is timed out
drivers/net/bonding/bond_alb.c +4 -4
@@ -1178,7 +1178,7 @@ static int alb_handle_addr_collision_on_attach(struct bonding *bond, struct slavstruct slave *tmp_slave1, *free_mac_slave =NULL ;struct list_head *iter; -if (list_empty (&bond->slave_list)) { +if (!bond_has_slaves (bond)) {/* this is the first slave */ return 0 ; }
@@ -1469,7 +1469,7 @@ void bond_alb_monitor(struct work_struct *work)read_lock (&bond->lock); -if (list_empty (&bond->slave_list)) { +if (!bond_has_slaves (bond)) { bond_info->tx_rebalance_counter =0 ; bond_info->lp_counter =0 ; goto re_arm;
@@ -1606,7 +1606,7 @@ int bond_alb_init_slave(struct bonding *bond, struct slave *slave)*/ void bond_alb_deinit_slave (struct bonding *bond,struct slave *slave) { -if (!list_empty (&bond->slave_list)) +if (bond_has_slaves (bond))alb_change_hw_addr_on_detach (bond, slave);tlb_clear_slave (bond, slave,0 );
@@ -1676,7 +1676,7 @@ void bond_alb_handle_active_change(struct bonding *bond, struct slave *new_slave swap_slave = bond->curr_active_slave;rcu_assign_pointer (bond->curr_active_slave, new_slave); -if (!new_slave ||list_empty (&bond->slave_list)) +if (!new_slave || !bond_has_slaves (bond))return ;/* set the new curr_active_slave to the bonds mac address
drivers/net/bonding/bond_main.c +16 -16
@@ -391,7 +391,7 @@ static int bond_set_carrier(struct bonding *bond)struct list_head *iter;struct slave *slave; -if (list_empty (&bond->slave_list)) +if (!bond_has_slaves (bond)) goto down;if (bond->params.mode ==BOND_MODE_8023AD )
@@ -1085,7 +1085,7 @@ static netdev_features_t bond_fix_features(struct net_device *dev,netdev_features_t mask;struct slave *slave; -if (list_empty (&bond->slave_list)) { +if (!bond_has_slaves (bond)) {/* Disable adding VLANs to empty bond. But why? --mq */ features |=NETIF_F_VLAN_CHALLENGED ;return features;
@@ -1120,7 +1120,7 @@ static void bond_compute_features(struct bonding *bond)unsigned int gso_max_size =GSO_MAX_SIZE ;u16 gso_max_segs =GSO_MAX_SEGS ; -if (list_empty (&bond->slave_list)) +if (!bond_has_slaves (bond)) goto done;bond_for_each_slave (bond, slave, iter) {
@@ -1310,7 +1310,7 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)* bond ether type mutual exclusion - don't allow slaves of dissimilar * ether type (eg ARPHRD_ETHER and ARPHRD_INFINIBAND) share the same bond */ -if (list_empty (&bond->slave_list)) { +if (!bond_has_slaves (bond)) {if (bond_dev->type != slave_dev->type) { pr_debug("%s: change device type from %d to %d\n", bond_dev->name,
@@ -1349,7 +1349,7 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) }if (slave_ops->ndo_set_mac_address ==NULL ) { -if (list_empty (&bond->slave_list)) { +if (!bond_has_slaves (bond)) {pr_warning ("%s: Warning: The first slave device specified does not support setting the MAC address. Setting fail_over_mac to active." , bond_dev->name); bond->params.fail_over_mac =BOND_FOM_ACTIVE ;
@@ -1365,7 +1365,7 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)/* If this is the first slave, then we need to set the master's hardware * address to be the same as the slave's. */ -if (list_empty (&bond->slave_list) && +if (!bond_has_slaves (bond) && bond->dev->addr_assign_type ==NET_ADDR_RANDOM )bond_set_dev_addr (bond->dev, slave_dev);
@@ -1696,7 +1696,7 @@ err_free: err_undo_flags:bond_compute_features (bond);/* Enslave of first slave has failed and we need to fix master's mac */ -if (list_empty (&bond->slave_list) && +if (!bond_has_slaves (bond) &ðer_addr_equal (bond_dev->dev_addr, slave_dev->dev_addr))eth_hw_addr_random (bond_dev);
@@ -1776,7 +1776,7 @@ static int __bond_release_one(struct net_device *bond_dev,if (!all && !bond->params.fail_over_mac) {if (ether_addr_equal (bond_dev->dev_addr, slave->perm_hwaddr) && - !list_empty (&bond->slave_list)) +bond_has_slaves (bond)) pr_warn("%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", bond_dev->name, slave_dev->name, slave->perm_hwaddr,
@@ -1819,7 +1819,7 @@ static int __bond_release_one(struct net_device *bond_dev,write_lock_bh (&bond->lock); } -if (list_empty (&bond->slave_list)) { +if (!bond_has_slaves (bond)) {bond_set_carrier (bond);eth_hw_addr_random (bond_dev);
@@ -1835,7 +1835,7 @@ static int __bond_release_one(struct net_device *bond_dev,unblock_netpoll_tx ();synchronize_rcu (); -if (list_empty (&bond->slave_list)) { +if (!bond_has_slaves (bond)) {call_netdevice_notifiers (NETDEV_CHANGEADDR , bond->dev);call_netdevice_notifiers (NETDEV_RELEASE , bond->dev); }
@@ -1904,7 +1904,7 @@ static int bond_release_and_destroy(struct net_device *bond_dev,int ret; ret =bond_release (bond_dev, slave_dev); -if (ret ==0 &&list_empty (&bond->slave_list)) { +if (ret ==0 && !bond_has_slaves (bond)) { bond_dev->priv_flags |=IFF_DISABLE_NETPOLL ;pr_info ("%s: destroying bond %s.\n" , bond_dev->name, bond_dev->name);
@@ -2219,7 +2219,7 @@ void bond_mii_monitor(struct work_struct *work) delay =msecs_to_jiffies (bond->params.miimon); -if (list_empty (&bond->slave_list)) +if (!bond_has_slaves (bond)) goto re_arm; should_notify_peers =bond_should_notify_peers (bond);
@@ -2513,7 +2513,7 @@ void bond_loadbalance_arp_mon(struct work_struct *work)read_lock (&bond->lock); -if (list_empty (&bond->slave_list)) +if (!bond_has_slaves (bond)) goto re_arm; oldcurrent = bond->curr_active_slave;
@@ -2845,7 +2845,7 @@ void bond_activebackup_arp_mon(struct work_struct *work) delta_in_ticks =msecs_to_jiffies (bond->params.arp_interval); -if (list_empty (&bond->slave_list)) +if (!bond_has_slaves (bond)) goto re_arm; should_notify_peers =bond_should_notify_peers (bond);
@@ -3169,7 +3169,7 @@ static int bond_open(struct net_device *bond_dev)/* reset slave->backup and slave->inactive */ read_lock (&bond->lock); -if (!list_empty (&bond->slave_list)) { +if (bond_has_slaves (bond)) {read_lock (&bond->curr_slave_lock);bond_for_each_slave (bond, slave, iter) { if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP)
@@ -3892,7 +3892,7 @@ static netdev_tx_t bond_start_xmit(struct sk_buff *skb, struct net_device *dev)return NETDEV_TX_BUSY ;rcu_read_lock (); -if (!list_empty (&bond->slave_list)) +if (bond_has_slaves (bond)) ret =__bond_start_xmit (skb, dev);else kfree_skb (skb);
drivers/net/bonding/bond_sysfs.c +2 -2
@@ -327,7 +327,7 @@ static ssize_t bonding_store_mode(struct device *d, goto out; } -if (!list_empty (&bond->slave_list)) { +if (bond_has_slaves (bond)) {pr_err ("unable to update mode of %s because it has slaves.\n" , bond->dev->name); ret = -EPERM ;
@@ -509,7 +509,7 @@ static ssize_t bonding_store_fail_over_mac(struct device *d,if (!rtnl_trylock ())return restart_syscall (); -if (!list_empty (&bond->slave_list)) { +if (bond_has_slaves (bond)) {pr_err ("%s: Can't alter fail_over_mac with slaves in bond.\n" , bond->dev->name); ret = -EPERM ;
drivers/net/bonding/bonding.h +2 -0
@@ -72,6 +72,8 @@ res; })/* slave list primitives */ +#define bond_has_slaves (bond) !list_empty(&(bond)->slave_list) +#define bond_to_slave (ptr) list_entry(ptr, struct slave, list)/* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */
Signature · 1 line
1.8.4
PATCH v2 17/27 bonding: convert bond_has_slaves() to use the neighbour list
In reply to Veaceslav FalicoThe same way as it was used for its own slave_list.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Use the renamed adj_list instead of neighbour_dev_list.drivers/net/bonding/bonding.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/net/bonding/bonding.h +1 -1
@@ -72,7 +72,7 @@ res; })/* slave list primitives */ -#define bond_has_slaves (bond) !list_empty(&(bond)->slave_list) +#define bond_has_slaves (bond) !list_empty(&(bond)->dev->adj_list.lower)#define bond_to_slave (ptr) list_entry(ptr, struct slave, list)
Signature · 1 line
1.8.4
PATCH v2 18/27 net: add a possibility to get private from netdev_adjacent->list
In reply to Veaceslav FalicoIt will be useful to get first/last element.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
No changes.include/linux/netdevice.h | 1 + net/core/dev.c | 10 ++++++++++ 2 files changed, 11 insertions(+)
include/linux/netdevice.h +1 -0
@@ -2850,6 +2850,7 @@ extern void *netdev_lower_get_next_private_rcu(struct net_device *dev,priv ; \ priv = netdev_lower_get_next_private_rcu(dev, &(iter))) +extern void *netdev_adjacent_get_private (struct list_head *adj_list);extern struct net_device *netdev_master_upper_dev_get (struct net_device *dev);extern struct net_device *netdev_master_upper_dev_get_rcu (struct net_device *dev); extern int netdev_upper_dev_link(struct net_device *dev,
net/core/dev.c +10 -0
@@ -4512,6 +4512,16 @@ struct net_device *netdev_master_upper_dev_get(struct net_device *dev) }EXPORT_SYMBOL (netdev_master_upper_dev_get); +void *netdev_adjacent_get_private (struct list_head *adj_list) +{ +struct netdev_adjacent *adj; + + adj = list_entry(adj_list, struct netdev_adjacent, list); + +return adj->private; +} +EXPORT_SYMBOL (netdev_adjacent_get_private); +/* netdev_all_upper_get_next_dev_rcu - Get the next dev from upper list * @dev: device * @iter: list_head ** of the current position
Signature · 1 line
1.8.4
PATCH v2 19/27 bonding: convert first/last slave logic to use neighbours
In reply to Veaceslav FalicoFor that, use netdev_adjacent_get_private(list_head) on bond's lower neighbour list members. Also, add a small macro - bond_slave_list(bond), which returns the bond list via neighbour list.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Rename neieghour_dev_list to adj_list.drivers/net/bonding/bonding.h | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-)
drivers/net/bonding/bonding.h +11 -6
@@ -72,19 +72,24 @@ res; })/* slave list primitives */ -#define bond_has_slaves (bond) !list_empty(&(bond)->dev->adj_list.lower) +#define bond_slave_list (bond) (&(bond)->dev->adj_list.lower) + +#define bond_has_slaves (bond) !list_empty(bond_slave_list(bond))#define bond_to_slave (ptr) list_entry(ptr, struct slave, list)/* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */ #define bond_first_slave (bond) \ - list_first_entry_or_null(&(bond)->slave_list, struct slave, list) + (bond_has_slaves(bond) ? \ + netdev_adjacent_get_private(bond_slave_list(bond)->next) : \ + NULL)#define bond_last_slave (bond) \ - (list_empty(&(bond)->slave_list) ? NULL : \ - bond_to_slave((bond)->slave_list.prev)) + (bond_has_slaves(bond) ? \ + netdev_adjacent_get_private(bond_slave_list(bond)->prev) : \ + NULL) -#define bond_is_first_slave (bond, pos) ((pos)->list.prev == &(bond)->slave_list) -#define bond_is_last_slave (bond, pos) ((pos)->list.next == &(bond)->slave_list) +#define bond_is_first_slave (bond, pos) (pos == bond_first_slave(bond)) +#define bond_is_last_slave (bond, pos) (pos == bond_last_slave(bond))/* Since bond_first/last_slave can return NULL, these can return NULL too */ #define bond_next_slave (bond, pos) \
Signature · 1 line
1.8.4
PATCH v2 20/27 bonding: remove bond_prev_slave()
In reply to Veaceslav FalicoWe don't really need it, and it's really hard to RCUify the list->prev.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
New patch.drivers/net/bonding/bond_main.c | 9 +++------ drivers/net/bonding/bonding.h | 4 ---- 2 files changed, 3 insertions(+), 10 deletions(-)
drivers/net/bonding/bond_main.c +3 -6
@@ -1255,7 +1255,7 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) {struct bonding *bond =netdev_priv (bond_dev);const struct net_device_ops *slave_ops = slave_dev->netdev_ops; -struct slave *new_slave =NULL ; +struct slave *new_slave =NULL , *prev_slave;struct sockaddr addr;int link_reporting;int res =0 , i;
@@ -1472,6 +1472,7 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)write_lock_bh (&bond->lock); + prev_slave =bond_last_slave (bond);bond_attach_slave (bond, new_slave); new_slave->delay =0 ;
@@ -1566,9 +1567,6 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)*/ bond_3ad_initialize (bond,1000 /AD_TIMER_INTERVAL ); }else { -struct slave *prev_slave; - - prev_slave =bond_prev_slave (bond, new_slave);SLAVE_AD_INFO (new_slave).id =SLAVE_AD_INFO (prev_slave).id +1 ; }
@@ -3506,9 +3504,8 @@ static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)*/ bond_for_each_slave (bond, slave, iter) { -pr_debug ("s %p s->p %p c_m %p\n" , +pr_debug ("s %p c_m %p\n" , slave, -bond_prev_slave (bond, slave), slave->dev->netdev_ops->ndo_change_mtu); res =dev_set_mtu (slave->dev, new_mtu);
drivers/net/bonding/bonding.h +0 -4
@@ -96,10 +96,6 @@ (bond_is_last_slave (bond, pos) ?bond_first_slave (bond) : \bond_to_slave ((pos)->list.next)) -#define bond_prev_slave (bond, pos) \ - (bond_is_first_slave(bond, pos) ? bond_last_slave(bond) : \ - bond_to_slave((pos)->list.prev)) -/** * bond_for_each_slave - iterate over all slaves * @bond: the bond holding this list
Signature · 1 line
1.8.4
PATCH v2 21/27 net: add a function to get the next private
In reply to Veaceslav FalicoIt searches for the provided private and returns the next one. If private is not found or next list element is list head - returns NULL.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Drop the bool switch - it only goes to ->next now, and rework it so
that it doesn't go over the head, but rather returns NULL.include/linux/netdevice.h | 2 ++ net/core/dev.c | 27 +++++++++++++++++++++++++++ 2 files changed, 29 insertions(+)
include/linux/netdevice.h +2 -0
@@ -2866,6 +2866,8 @@ extern void *netdev_lower_dev_get_private_rcu(struct net_device *dev, struct net_device *lower_dev);extern void *netdev_lower_dev_get_private (struct net_device *dev,struct net_device *lower_dev); +extern void *netdev_lower_dev_get_next_private (struct net_device *dev, +void *private);extern int skb_checksum_help (struct sk_buff *skb);extern struct sk_buff *__skb_gso_segment (struct sk_buff *skb,netdev_features_t features,bool tx_path);
net/core/dev.c +27 -0
@@ -5055,6 +5055,33 @@ void *netdev_lower_dev_get_private(struct net_device *dev, }EXPORT_SYMBOL (netdev_lower_dev_get_private); +/* netdev_lower_dev_get_next_private - return the ->private of the list +* element whos ->private == private. +* @dev - device to search +* @private - private pointer to search for. +* +* Returns the next ->private pointer, if ->next is not head and private is +* found. +*/ +extern void *netdev_lower_dev_get_next_private (struct net_device *dev, +void *private) +{ +struct netdev_adjacent *lower; + +list_for_each_entry (lower, &dev->adj_list.lower, list) { +if (lower->private == private) { + lower =list_entry (lower->list.next, + struct netdev_adjacent, list); +if (&lower->list == &dev->adj_list.lower) +return NULL ; +return lower->private; + } + } + +return NULL ; +} +EXPORT_SYMBOL (netdev_lower_dev_get_next_private); +static void dev_change_rx_flags (struct net_device *dev,int flags) {const struct net_device_ops *ops = dev->netdev_ops;
Signature · 1 line
1.8.4
PATCH v2 22/27 bonding: use neighbours for bond_next_slave()
In reply to Veaceslav FalicoIf it's the last slave - return the first slave, otherwise - return the next slave via netdev_lower_dev_get_next_private().
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Change to standard logic - if it's the last slave - return the first one, otherwise - return the next via netdev_lower_dev_get_next_private(). Also, bond_prev_slave() was dropped - so we don't need it.
drivers/net/bonding/bonding.h | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-)
drivers/net/bonding/bonding.h +3 -4
@@ -76,8 +76,6 @@#define bond_has_slaves (bond) !list_empty(bond_slave_list(bond)) -#define bond_to_slave (ptr) list_entry(ptr, struct slave, list) -/* IMPORTANT: bond_first/last_slave can return NULL in case of an empty list */ #define bond_first_slave (bond) \ (bond_has_slaves(bond) ? \
@@ -93,8 +91,9 @@/* Since bond_first/last_slave can return NULL, these can return NULL too */ #define bond_next_slave (bond, pos) \ - (bond_is_last_slave(bond, pos) ? bond_first_slave(bond) : \ - bond_to_slave((pos)->list.next)) + (bond_is_last_slave(bond, pos) ? \ + bond_first_slave(bond) : \ + netdev_lower_dev_get_next_private((bond)->dev, pos))/** * bond_for_each_slave - iterate over all slaves
Signature · 1 line
1.8.4
PATCH v2 23/27 bonding: remove slave lists
In reply to Veaceslav FalicoAnd all the initialization.
---
Notes:
RFC -> v2:
New patch.drivers/net/bonding/bond_main.c | 4 ---- drivers/net/bonding/bonding.h | 2 -- 2 files changed, 6 deletions(-)
drivers/net/bonding/bond_main.c +0 -4
@@ -972,7 +972,6 @@ void bond_select_active_slave(struct bonding *bond)*/ static void bond_attach_slave (struct bonding *bond,struct slave *new_slave) { -list_add_tail_rcu (&new_slave->list, &bond->slave_list); bond->slave_cnt++; }
@@ -988,7 +987,6 @@ static void bond_attach_slave(struct bonding *bond, struct slave *new_slave)*/ static void bond_detach_slave (struct bonding *bond,struct slave *slave) { -list_del_rcu (&slave->list); bond->slave_cnt--; }
@@ -1374,7 +1372,6 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) res = -ENOMEM ; goto err_undo_flags; } -INIT_LIST_HEAD (&new_slave->list);/* * Set the new_slave's queue_id to be zero. Queue ID mapping * is set via sysfs or module option if desired.
@@ -4022,7 +4019,6 @@ static void bond_setup(struct net_device *bond_dev)/* initialize rwlocks */ rwlock_init (&bond->lock);rwlock_init (&bond->curr_slave_lock); -INIT_LIST_HEAD (&bond->slave_list); bond->params = bonding_defaults;/* Initialize pointers */
drivers/net/bonding/bonding.h +0 -2
@@ -167,7 +167,6 @@ struct bond_parm_tbl {struct slave {struct net_device *dev;/* first - useful for panic debug */ -struct list_head list;struct bonding *bond;/* our master */ int delay;unsigned long jiffies;
@@ -207,7 +206,6 @@ struct slave {*/ struct bonding {struct net_device *dev;/* first - useful for panic debug */ -struct list_head slave_list;struct slave *curr_active_slave;struct slave *current_arp_slave;struct slave *primary_slave;
Signature · 1 line
1.8.4
PATCH v2 24/27 vlan: link the upper neighbour only after registering
In reply to Veaceslav FalicoOtherwise users might access it without being fully registered, as per sysfs - it only inits in register_netdevice(), so is unusable till it is called.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
No changes.net/8021q/vlan.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
net/8021q/vlan.c +7 -7
@@ -169,13 +169,13 @@ int register_vlan_dev(struct net_device *dev)if (err <0 ) goto out_uninit_mvrp; - err =netdev_upper_dev_link (real_dev, dev); -if (err) - goto out_uninit_mvrp; - err =register_netdevice (dev);if (err <0 ) - goto out_upper_dev_unlink; + goto out_uninit_mvrp; + + err =netdev_upper_dev_link (real_dev, dev); +if (err) + goto out_unregister_netdev;/* Account for reference in struct vlan_dev_priv */ dev_hold (real_dev);
@@ -191,8 +191,8 @@ int register_vlan_dev(struct net_device *dev)return 0 ; -out_upper_dev_unlink: -netdev_upper_dev_unlink (real_dev, dev); +out_unregister_netdev: +unregister_netdevice (dev); out_uninit_mvrp:if (grp->nr_vlan_devs ==0 )vlan_mvrp_uninit_applicant (real_dev);
Signature · 1 line
1.8.4
PATCH v2 25/27 vlan: unlink the upper neighbour before unregistering
In reply to Veaceslav FalicoOn netdev unregister we're removing also all of its sysfs-associated stuff, including the sysfs symlinks that are controlled by netdev neighbour code. Also, it's a subtle race condition - cause we can still access it after unregistering.
Move the unlinking right before the unregistering to fix both.
---
Notes:
v1 -> v2:
No changes.
RFC -> v2:
New patch.net/8021q/vlan.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
net/8021q/vlan.c +2 -2
@@ -98,14 +98,14 @@ void unregister_vlan_dev(struct net_device *dev, struct list_head *head)vlan_gvrp_request_leave (dev);vlan_group_set_device (grp, vlan->vlan_proto, vlan_id,NULL ); + +netdev_upper_dev_unlink (real_dev, dev);/* Because unregister_netdevice_queue() makes sure at least one rcu * grace period is respected before device freeing, * we dont need to call synchronize_net() here. */ unregister_netdevice_queue (dev, head); -netdev_upper_dev_unlink (real_dev, dev); -if (grp->nr_vlan_devs ==0 ) {vlan_mvrp_uninit_applicant (real_dev);vlan_gvrp_uninit_applicant (real_dev);
Signature · 1 line
1.8.4
PATCH v2 26/27 net: expose the master link to sysfs, and remove it from bond
In reply to Veaceslav FalicoCurrently, we can have only one master upper neighbour, so it would be useful to create a symlink to it in the sysfs device directory, the way that bonding now does it, for every device. Lower devices from bridge/team/etc will automagically get it, so we could rely on it.
Also, remove the same functionality from bonding.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
No changes.drivers/net/bonding/bond_sysfs.c | 12 +----------- net/core/dev.c | 9 +++++++++ 2 files changed, 10 insertions(+), 11 deletions(-)
drivers/net/bonding/bond_sysfs.c +1 -11
@@ -174,20 +174,11 @@ int bond_create_slave_symlinks(struct net_device *master,char linkname[IFNAMSIZ +7 ];int ret =0 ; -/* first, create a link from the slave back to the master */ - ret =sysfs_create_link (&(slave->dev.kobj), &(master->dev.kobj), -"master" ); -if (ret) -return ret; -/* next, create a link from the master to the slave */ +/* create a link from the master to the slave */ sprintf (linkname,"slave_%s" , slave->name); ret =sysfs_create_link (&(master->dev.kobj), &(slave->dev.kobj), linkname); -/* free the master link created earlier in case of error */ -if (ret) -sysfs_remove_link (&(slave->dev.kobj),"master" ); -return ret; }
@@ -197,7 +188,6 @@ void bond_destroy_slave_symlinks(struct net_device *master, {char linkname[IFNAMSIZ +7 ]; -sysfs_remove_link (&(slave->dev.kobj),"master" );sprintf (linkname,"slave_%s" , slave->name);sysfs_remove_link (&(master->dev.kobj), linkname); }
net/core/dev.c +9 -0
@@ -4638,6 +4638,7 @@ static int __netdev_adjacent_dev_insert(struct net_device *dev, bool upper, void *private) { struct netdev_adjacent *adj, *neigh = NULL; +int ret; adj =__netdev_find_adj (dev, adj_dev, upper, false);
@@ -4693,6 +4694,12 @@ static int __netdev_adjacent_dev_insert(struct net_device *dev,/* Ensure that master upper link is always the first item in list. */ if (master) { + ret =sysfs_create_link (&(dev->dev.kobj), &(adj_dev->dev.kobj),"master" ); +if (ret) { +kfree (neigh); +kfree (adj); +return ret; + }if (neigh)list_add_rcu (&neigh->list, &dev->adj_list.upper);
@@ -4749,6 +4756,8 @@ void __netdev_adjacent_dev_remove(struct net_device *dev,pr_debug ("dev_put for %s, because of %s link removed from %s to %s (neighbour)\n" , adj_dev->name, upper ?"upper" :"lower" , dev->name, adj_dev->name); +if (neighbour->master && upper) +sysfs_remove_link (&(dev->dev.kobj),"master" );list_del_rcu (&neighbour->list);dev_put (adj_dev);kfree_rcu (neighbour, rcu);
Signature · 1 line
1.8.4
PATCH v2 27/27 net: create sysfs symlinks for neighbour devices
In reply to Veaceslav FalicoAlso, remove the same functionality from bonding - it will be already done for any device that links to its lower/upper neighbour.
The links will be created for dev's kobject, and will look like lower_eth0 for lower device eth0 and upper_bridge0 for upper device bridge0.
---
Notes:
v1 -> v2:
No changes.
RFC -> v1:
Rename lower devices from slave_eth0 to lower_eth0, so that it meets the lower/upper naming. I've searched for any active users of bonding's slave_eth0, and seems that nobody's actively using it or relying on it.
drivers/net/bonding/bond_main.c | 11 +---------- drivers/net/bonding/bond_sysfs.c | 25 ------------------------- drivers/net/bonding/bonding.h | 2 -- net/core/dev.c | 29 +++++++++++++++++++++++++++++ 4 files changed, 30 insertions(+), 37 deletions(-)
drivers/net/bonding/bond_main.c +1 -10
@@ -1612,15 +1612,11 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)read_unlock (&bond->lock); - res =bond_create_slave_symlinks (bond_dev, slave_dev); -if (res) - goto err_detach; - res =netdev_rx_handler_register (slave_dev, bond_handle_frame, new_slave);if (res) {pr_debug ("Error %d calling netdev_rx_handler_register\n" , res); - goto err_dest_symlinks; + goto err_detach; } res =bond_master_upper_dev_link (bond_dev, slave_dev, new_slave);
@@ -1642,9 +1638,6 @@ int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev) err_unregister:netdev_rx_handler_unregister (slave_dev); -err_dest_symlinks: -bond_destroy_slave_symlinks (bond_dev, slave_dev); - err_detach:if (!USES_PRIMARY (bond->params.mode))bond_hw_addr_flush (bond_dev, slave_dev);
@@ -1842,8 +1835,6 @@ static int __bond_release_one(struct net_device *bond_dev, bond_dev->name, slave_dev->name, bond_dev->name);/* must do this from outside any spinlocks */ -bond_destroy_slave_symlinks (bond_dev, slave_dev); -vlan_vids_del_by_dev (slave_dev, bond_dev);/* If the mode USES_PRIMARY, then this cases was handled above by
drivers/net/bonding/bond_sysfs.c +0 -25
@@ -168,31 +168,6 @@ static const struct class_attribute class_attr_bonding_masters = { .namespace = bonding_namespace, }; -int bond_create_slave_symlinks (struct net_device *master, -struct net_device *slave) -{ -char linkname[IFNAMSIZ +7 ]; -int ret =0 ; - -/* create a link from the master to the slave */ -sprintf (linkname,"slave_%s" , slave->name); - ret =sysfs_create_link (&(master->dev.kobj), &(slave->dev.kobj), - linkname); - -return ret; - -} - -void bond_destroy_slave_symlinks (struct net_device *master, -struct net_device *slave) -{ -char linkname[IFNAMSIZ +7 ]; - -sprintf (linkname,"slave_%s" , slave->name); -sysfs_remove_link (&(master->dev.kobj), linkname); -} - -/* * Show the slaves in the current bond. */
drivers/net/bonding/bonding.h +0 -2
@@ -410,8 +410,6 @@ int bond_create(struct net *net, const char *name);int bond_create_sysfs (struct bond_net *net);void bond_destroy_sysfs (struct bond_net *net);void bond_prepare_sysfs_group (struct bonding *bond); -int bond_create_slave_symlinks (struct net_device *master,struct net_device *slave); -void bond_destroy_slave_symlinks (struct net_device *master,struct net_device *slave);int bond_enslave (struct net_device *bond_dev,struct net_device *slave_dev);int bond_release (struct net_device *bond_dev,struct net_device *slave_dev);void bond_mii_monitor (struct work_struct *);
net/core/dev.c +29 -0
@@ -4638,6 +4638,7 @@ static int __netdev_adjacent_dev_insert(struct net_device *dev, bool upper, void *private) { struct netdev_adjacent *adj, *neigh = NULL; +char linkname[IFNAMSIZ +7 ];int ret; adj =__netdev_find_adj (dev, adj_dev, upper, false);
@@ -4685,6 +4686,16 @@ static int __netdev_adjacent_dev_insert(struct net_device *dev,*/ if (master) neigh->private = private; + +sprintf (linkname,"lower_%s" , adj_dev->name); + ret =sysfs_create_link (&(dev->dev.kobj), + &(adj_dev->dev.kobj), + linkname); +if (ret) { +kfree (neigh); +kfree (adj); +return ret; + }list_add_tail_rcu (&neigh->list, &dev->adj_list.lower); }
@@ -4692,10 +4703,24 @@ static int __netdev_adjacent_dev_insert(struct net_device *dev,return 0 ; } +/* it's upper neighbour, we don't care if it's master or not */ +if (neigh) { +sprintf (linkname,"upper_%s" , adj_dev->name); + ret =sysfs_create_link (&(dev->dev.kobj), &(adj_dev->dev.kobj), + linkname); +if (ret) { +kfree (neigh); +kfree (adj); +return ret; + } + } +/* Ensure that master upper link is always the first item in list. */ if (master) { ret =sysfs_create_link (&(dev->dev.kobj), &(adj_dev->dev.kobj),"master" );if (ret) { +if (neigh) +sysfs_remove_link (&(dev->dev.kobj), linkname);kfree (neigh);kfree (adj);return ret;
@@ -4736,6 +4761,7 @@ void __netdev_adjacent_dev_remove(struct net_device *dev, bool is_neigh) {struct netdev_adjacent *adj, *neighbour; +char linkname[IFNAMSIZ +7 ];if (upper) { adj =__netdev_all_upper_find (dev, adj_dev);
@@ -4758,6 +4784,9 @@ void __netdev_adjacent_dev_remove(struct net_device *dev, adj_dev->name);if (neighbour->master && upper)sysfs_remove_link (&(dev->dev.kobj),"master" ); +sprintf (linkname,"%s_%s" , upper ?"upper" :"lower" , + adj_dev->name); +sysfs_remove_link (&(dev->dev.kobj), linkname);list_del_rcu (&neighbour->list);dev_put (adj_dev);kfree_rcu (neighbour, rcu);
Signature · 1 line
1.8.4