People are using bonding over Infiniband IPoIB connections, and who knows
what else. Infiniband has a hardware address length of 20 octets
(INFINIBAND_ALEN), and the network core defines a MAX_ADDR_LEN of 32.
Various places in the bonding code are currently hard-wired to 6 octets
(ETH_ALEN), such as the 3ad code, which I've left untouched here. Besides,
only alb is currently possible on Infiniband links right now anyway, due
to commit 1533e773, so the alb code is where most of the changes are.

One major component of this change is the addition of a bond_hw_addr_copy
function that takes a length argument, instead of using ether_addr_copy
everywhere that hardware addresses need to be copied about. The other
major component of this change is converting the bonding code from using
struct sockaddr for address storage to struct sockaddr_storage, as the
former has an address storage space of only 14, while the latter is 128
minus a few, which is necessary to support bonding over device with up to
MAX_ADDR_LEN octet hardware addresses. Additionally, this probably fixes
up some memory corruption issues with the current code, where it's
possible to write an infiniband hardware address into a sockaddr declared
on the stack.

Lightly tested on a dual mlx4 IPoIB setup, which properly shows a 20-octet
hardware address now:

$ cat /proc/net/bonding/bond0
Ethernet Channel Bonding Driver: v3.7.1 (April 27, 2011)

Bonding Mode: fault-tolerance (active-backup) (fail_over_mac active)
Primary Slave: mlx4_ib0 (primary_reselect always)
Currently Active Slave: mlx4_ib0
MII Status: up
MII Polling Interval (ms): 100
Up Delay (ms): 100
Down Delay (ms): 100

Slave Interface: mlx4_ib0
MII Status: up
Speed: Unknown
Duplex: Unknown
Link Failure Count: 0
Permanent HW addr:
80:00:02:08:fe:80:00:00:00:00:00:00:e4:1d:2d:03:00:1d:67:01
Slave queue ID: 0

Slave Interface: mlx4_ib1
MII Status: up
Speed: Unknown
Duplex: Unknown
Link Failure Count: 0
Permanent HW addr:
80:00:02:09:fe:80:00:00:00:00:00:01:e4:1d:2d:03:00:1d:67:02
Slave queue ID: 0

Also tested with a standard 1Gbps NIC bonding setup (with a mix of
e1000 and e1000e cards), running LNST's bonding tests.

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
CC: netdev@vger.kernel.org
Signed-off-by: NJarod Wilson <jarod@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Convert alb_send_learning_packets and bond_has_this_ip to use the new
netdev_walk_all_upper_dev_rcu API. In both cases this is just a code
conversion; no functional change is intended.

v2
- removed typecast of data and simplified bond_upper_dev_walk
Signed-off-by: NDavid Ahern <dsa@cumulusnetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

ether_addr_equal_64bits() requires some care about its arguments,
namely that 8 bytes might be read, even if last 2 byte values are not
used.

KASan detected a violation with null_mac_addr and lacpdu_mcast_addr
in bond_3ad.c

Same problem with mac_bcast[] and mac_v6_allmcast[] in bond_alb.c :
Although the 8-byte alignment was there, KASan would detect out
of bound accesses.

Fixes: 815117ad ("bonding: use ether_addr_equal_unaligned for bond addr compare")
Fixes: bb54e589 ("bonding: Verify RX LACPDU has proper dest mac-addr")
Fixes: 885a136c ("bonding: use compare_ether_addr_64bits() in ALB")
Signed-off-by: NEric Dumazet <edumazet@google.com>
Reported-by: NDmitry Vyukov <dvyukov@google.com>
Acked-by: NDmitry Vyukov <dvyukov@google.com>
Acked-by: NNikolay Aleksandrov <nikolay@cumulusnetworks.com>
Acked-by: NDing Tianhong <dingtianhong@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The return value of kzalloc on failure of allocation of memory should
be -ENOMEM and not -1.

Found using Coccinelle. A simplified version of the semantic patch
used is:

//<smpl>
@@
expression *e;
@@

e = kzalloc(...);
if (e == NULL) {
...
return
- -1
+ -ENOMEM
;
}
//</smpl>

The single call site only checks that the return value is not 0,
hence no change is required at the call site.
Signed-off-by: NAmitoj Kaur Chawla <amitoj1606@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Since both tx and rx paths work with skb->vlan_tci, there's no need for
this function anymore. Switch users directly to __vlan_hwaccel_put_tag.
Signed-off-by: NJiri Pirko <jiri@resnulli.us>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This ways drivers like cxgb4 don't need to do ugly relative includes.
Reported-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Because bonding stats are usually sum of slave stats, it was
not easy to account for tx drops at bonding layer.

We can use dev->tx_dropped for this, as this counter is later
added to the device stats (in dev_get_stats())

This extends the idea we had in commit ee637714 ("bonding: Simplify
the xmit function for modes that use xmit_hash") for bond_3ad_xor_xmit()
to other bonding modes.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Cc: Mahesh Bandewar <maheshb@google.com>
Reviewed-by: NNikolay Aleksandrov <nikolay@redhat.com>
Acked-by: NMahesh Bandewar <maheshb@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Earlier change to use usable slave array for TLB mode had an additional
performance advantage. So extending the same logic to all other modes
that use xmit-hash for slave selection (viz 802.3AD, and XOR modes).
Also consolidating this with the earlier TLB change.

The main idea is to build the usable slaves array in the control path
and use that array for slave selection during xmit operation.

Measured performance in a setup with a bond of 4x1G NICs with 200
instances of netperf for the modes involved (3ad, xor, tlb)
cmd: netperf -t TCP_RR -H <TargetHost> -l 60 -s 5

Mode        TPS-Before   TPS-After

802.3ad   : 468,694      493,101
TLB (lb=0): 392,583      392,965
XOR       : 475,696      484,517
Signed-off-by: NMahesh Bandewar <maheshb@google.com>
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

First adjust a couple of locking comments that were left inaccurate,
then adjust comments to use the netdev styling and remove extra new
lines where necessary and add a couple of new lines between declarations
and code. These are all trivial styling changes, no functional change.
Also removed a couple of outdated or obvious comments.
This patch is by no means a complete fix of all netdev style violations
but it gets the bonding closer.
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

__rlb_next_rx_slave() is a copy of rlb_next_rx_slave() with the
difference that it uses rcu primitives to walk the slave list. We don't
need the two functions and can make rlb_next_rx_slave() a wrapper for
callers which hold RTNL.
So add a comment and ASSERT_RTNL() to make sure what is intended.
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Now that locks have been removed, remove some unnecessary comments and
adjust others to reflect reality. Also add a comment to "mode_lock" to
describe its current users and give a brief summary why they need it.
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The ALB/TLB specific spinlocks are no longer necessary as we now have
bond->mode_lock for this purpose, so convert them and remove them from
struct alb_bond_info.
Also remove the unneeded lock/unlock functions and use spin_lock/unlock
directly.
Suggested-by: NJay Vosburgh <jay.vosburgh@canonical.com>
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Mostly all users of curr_slave_lock already have RTNL as we've discussed
previously so there's no point in using it, the one case where the lock
must stay is the 3ad code, in fact it's the only one.
It's okay to remove it from bond_do_fail_over_mac() as it's called with
RTNL and drops the curr_slave_lock anyway.
bond_change_active_slave() is one of the main places where
curr_slave_lock was used, it's okay to remove it as all callers use RTNL
these days before calling it, that's why we move the ASSERT_RTNL() in
the beginning to catch any potential offenders to this rule.
The RTNL argument actually applies to all of the places where
curr_slave_lock has been removed from in this patch.
Also remove the unnecessary bond_deref_active_protected() macro and use
rtnl_dereference() instead.
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

First in rlb_teach_disabled_mac_on_primary() it's okay to remove
curr_slave_lock as all callers except bond_alb_monitor() already hold
RTNL, and in case bond_alb_monitor() is executing we can at most have a
period with bad throughput (very unlikely though).
In bond_alb_monitor() it's okay to remove the read_lock as the slave
list is walked with RCU and the worst that could happen is another
transmitter at the same time and thus for a period which currently is 10
seconds (bond_alb.h: BOND_ALB_LP_TICKS).
And bond_alb_handle_active_change() is okay because it's always called
with RTNL. Removed the ASSERT_RTNL() because it'll be inserted in the
parent function in a following patch.
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We can remove the lock/unlock as it's no longer necessary since
RTNL should be held while calling bond_alb_set_mac_address().
Signed-off-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This "rcu_dereference()" call is used directly in a condition.
Since its return value is never dereferenced it is recommended to use
"rcu_access_pointer()" instead of "rcu_dereference()".
Therefore, this patch makes this replacement.

The following Coccinelle semantic patch was used for solving it:
@@
@@

(
 if(
 (<+...
- rcu_dereference
+ rcu_access_pointer
  (...)
 ...+>)) {...}
|
 while(
 (<+...
- rcu_dereference
+ rcu_access_pointer
  (...)
  ...+>)) {...}
)
Signed-off-by: NAndreea-Cristina Bernat <bernat.ada@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently we're limited by a constant level of vlan nestings, and fail to
find anything beyound that level (currently 2).

To fix this - remove the limit of nestings when going through device tree,
and when the end device is found - allocate the needed amount of vlan tags
and return them, instead of found/not found.

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NVeaceslav Falico <vfalico@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In TLB mode if tlb_dynamic_lb is NOT set, slaves from the bond
group are selected based on the hash distribution. This does not
exclude dead links which are part of the bond. Also if there is a
temporary link event which brings down the interface, packets
hashed on that interface would be dropped too.

This patch fixes these issues and distributes flows across the
UP links only. Also the array construction of links which are
capable of sending packets happen in the control path leaving
only link-selection during the data-path.

One possible side effect of this is - at a link event; all
flows will be shuffled to get good distribution. But impact of
this should be minimum with the assumption that a member or
members of the bond group are not available is a very temporary
situation.
Signed-off-by: NMahesh Bandewar <maheshb@google.com>
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To maintain the same message structure as netdev_* functions print.

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NVeaceslav Falico <vfalico@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NVeaceslav Falico <vfalico@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

RCU was added to bonding in linux-3.12 but lacked proper sparse annotations.

Using __rcu annotation actually helps to spot all accesses to bond->curr_active_slave
are correctly protected, with LOCKDEP support.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Acked-by: NVeaceslav Falico <vfalico@gmail.com>
Reviewed-by: NNikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

To make TLB mode work, the patch allows learning packets
to be sent using mac addresses assigned to macvlan devices,
also taking into an account vlans that may be between the
bond and macvlan device.

To make RLB work, all we have to do is accept ARP packets
for addresses added to the bond dev->uc list.  Since RLB
mode will take care to update the peers directly with
correct mac addresses, learning packets for these addresses
do not have be send to switch.
Signed-off-by: NVlad Yasevich <vyasevic@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

ALB learning packets are currentlyalways sent using the slave mac
address for all vlans configured on top of bond.   This is not always
correct, as vlans may change their mac address.
This patch introduced a concept of strict matching where the
source of learning packets can either strictly match the address
passed in, or it can determine a more correct address to use.

There are 3 casese to consider:
  1) Switchover.  In this case, we have a new active slave and we need
     tell the switch about all addresses available on the slave.
  2) Monitor.  We'll periodically refresh learning info for all slaves.
     In this case, we refresh all addresses for current active, and just
     the slave address for other slaves.
  3) Teaching of disabled adddress.  This happens as part of the
     failover and in this case, we alwyas to use just the address
     provided.

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NVlad Yasevich <vyasevic@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

TLB/ALB learning packets always assume 802.1Q vlan protocol, but
that is no longer the case since we now have support for Q-in-Q
on top of bonding.  Pass the vlan protocol to alb_send_lp_vid()
so that the packets are properly tagged.

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Veaceslav Falico <vfalico@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NVlad Yasevich <vyasevic@redhat.com>
Acked-by: NVeaceslav Falico <vfalico@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

ALB/TLB learning packets use all vlans configured on top
of the bond.  This ends up being incorrect if we have a stack
of vlans on top of the bond.  ALB/TLB should only use
first level/outer most vlans in its announcements.
Signed-off-by: NVlad Yasevich <vyasevic@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

They're verifying the same thing (except of IFF_UP, which is implied for
netif_running(), which is also a prerequisite).

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NVeaceslav Falico <vfalico@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

CC: Jay Vosburgh <j.vosburgh@gmail.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NVeaceslav Falico <vfalico@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The aggresive load balancing causes packet re-ordering as active
flows are moved from a slave to another within the group. Sometime
this aggresive lb is not necessary if the preference is for less
re-ordering. This parameter if used with value "0" disables
this dynamic flow shuffling minimizing packet re-ordering. Of course
the side effect is that it has to live with the static load balancing
that the hashing distribution provides. This impact is less severe if
the correct xmit-hashing-policy is used for the tlb setup.

The default value of the parameter is set to "1" mimicing the earlier
behavior.

Ran the netperf test with 200 stream for 1 min between two hosts with
4x1G trunk (xmit-lb mode with xmit-policy L3+4) before and after these
changes. Following was the command used for those 200 instances -

    netperf -t TCP_RR -l 60 -s 5 -H <host> -- -r81920,81920

Transactions per second:
    Before change: 1,367.11
    After  change: 1,470.65

Change-Id: Ie3f75c77282cf602e83a6e833c6eb164e72a0990
Signed-off-by: NMahesh Bandewar <maheshb@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Re-organized the xmit function for the lb mode separating tlb xmit
from the alb mode. This will enable use of the hashing policies
like 802.3ad mode. Also extended use of xmit-hash-policy to tlb mode.

Now the tlb-mode defaults to BOND_XMIT_POLICY_LAYER2 if the xmit policy
module parameter is not set (just like 802.3ad, or Xor mode).

Change-Id: I140257403d272df75f477b380207338d0f04963e
Signed-off-by: NMahesh Bandewar <maheshb@google.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Separating the actual xmit part from the function in a separate
function that can be used in the tlb_xmit in the next patch. Also
there is no reason do_tx_balance to be an int so changing it to
bool type.

Change-Id: I9c48ff30487810f68587e621a191db616f49bd3b
Signed-off-by: NMahesh Bandewar <maheshb@google.com>
Acked-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Currently it's using the wrong ETH_P_LOOP type, which is sometimes treated
as packet length instead of ether type (because it's 0x0060).

Use the new ETH_P_LOOPBACK type.

CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NVeaceslav Falico <vfalico@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The commit d3ab3ffd
(bonding: use rlb_client_info->vlan_id instead of ->tag)
remove the rlb_client_info->tag, but occur some issues,
The vlan_get_tag() will return 0 for success and -EINVAL for
error, so the client_info->vlan_id always be set to 0 if the
vlan_get_tag return 0 for success, so the client_info would
never get a correct vlan id.

We should only set the vlan id to 0 when the vlan_get_tag return error.

Fixes: d3ab3ffd (bonding: use rlb_client_info->vlan_id instead of ->tag)

CC: Ding Tianhong <dingtianhong@huawei.com>
CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: NDing Tianhong <dingtianhong@huawei.com>
Acked-by: NVeaceslav Falico <vfalico@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replace kfree_skb with dev_kfree_skb_any in functions that can
be called in hard irq and other contexts.
Signed-off-by: N"Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It's unnecessary and less readable after a clause ending in a goto.
Signed-off-by: NJoe Perches <joe@perches.com>
Reviewed-by: NDing Tianhong <dingtianhong@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It's smaller and faster for some architectures.
Signed-off-by: NJoe Perches <joe@perches.com>
Reviewed-by: NDing Tianhong <dingtianhong@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

ether_addr_copy is smaller and faster for some architectures.

This relies on a stack frame being at least __aligned(2)
for one use of an Ethernet address on the stack.
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Use more current logging style.

Coalesce formats, realign arguments, drop unnecessary periods.
Signed-off-by: NJoe Perches <joe@perches.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NWang Yufen <wangyufen@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NWang Yufen <wangyufen@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NWang Yufen <wangyufen@huawei.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>