The kernel's implementation of notifier chains is unsafe.  There is no
protection against entries being added to or removed from a chain while the
chain is in use.  The issues were discussed in this thread:

    http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2

We noticed that notifier chains in the kernel fall into two basic usage
classes:

	"Blocking" chains are always called from a process context
	and the callout routines are allowed to sleep;

	"Atomic" chains can be called from an atomic context and
	the callout routines are not allowed to sleep.

We decided to codify this distinction and make it part of the API.  Therefore
this set of patches introduces three new, parallel APIs: one for blocking
notifiers, one for atomic notifiers, and one for "raw" notifiers (which is
really just the old API under a new name).  New kinds of data structures are
used for the heads of the chains, and new routines are defined for
registration, unregistration, and calling a chain.  The three APIs are
explained in include/linux/notifier.h and their implementation is in
kernel/sys.c.

With atomic and blocking chains, the implementation guarantees that the chain
links will not be corrupted and that chain callers will not get messed up by
entries being added or removed.  For raw chains the implementation provides no
guarantees at all; users of this API must provide their own protections.  (The
idea was that situations may come up where the assumptions of the atomic and
blocking APIs are not appropriate, so it should be possible for users to
handle these things in their own way.)

There are some limitations, which should not be too hard to live with.  For
atomic/blocking chains, registration and unregistration must always be done in
a process context since the chain is protected by a mutex/rwsem.  Also, a
callout routine for a non-raw chain must not try to register or unregister
entries on its own chain.  (This did happen in a couple of places and the code
had to be changed to avoid it.)

Since atomic chains may be called from within an NMI handler, they cannot use
spinlocks for synchronization.  Instead we use RCU.  The overhead falls almost
entirely in the unregister routine, which is okay since unregistration is much
less frequent that calling a chain.

Here is the list of chains that we adjusted and their classifications.  None
of them use the raw API, so for the moment it is only a placeholder.

  ATOMIC CHAINS
  -------------
arch/i386/kernel/traps.c:		i386die_chain
arch/ia64/kernel/traps.c:		ia64die_chain
arch/powerpc/kernel/traps.c:		powerpc_die_chain
arch/sparc64/kernel/traps.c:		sparc64die_chain
arch/x86_64/kernel/traps.c:		die_chain
drivers/char/ipmi/ipmi_si_intf.c:	xaction_notifier_list
kernel/panic.c:				panic_notifier_list
kernel/profile.c:			task_free_notifier
net/bluetooth/hci_core.c:		hci_notifier
net/ipv4/netfilter/ip_conntrack_core.c:	ip_conntrack_chain
net/ipv4/netfilter/ip_conntrack_core.c:	ip_conntrack_expect_chain
net/ipv6/addrconf.c:			inet6addr_chain
net/netfilter/nf_conntrack_core.c:	nf_conntrack_chain
net/netfilter/nf_conntrack_core.c:	nf_conntrack_expect_chain
net/netlink/af_netlink.c:		netlink_chain

  BLOCKING CHAINS
  ---------------
arch/powerpc/platforms/pseries/reconfig.c:	pSeries_reconfig_chain
arch/s390/kernel/process.c:		idle_chain
arch/x86_64/kernel/process.c		idle_notifier
drivers/base/memory.c:			memory_chain
drivers/cpufreq/cpufreq.c		cpufreq_policy_notifier_list
drivers/cpufreq/cpufreq.c		cpufreq_transition_notifier_list
drivers/macintosh/adb.c:		adb_client_list
drivers/macintosh/via-pmu.c		sleep_notifier_list
drivers/macintosh/via-pmu68k.c		sleep_notifier_list
drivers/macintosh/windfarm_core.c	wf_client_list
drivers/usb/core/notify.c		usb_notifier_list
drivers/video/fbmem.c			fb_notifier_list
kernel/cpu.c				cpu_chain
kernel/module.c				module_notify_list
kernel/profile.c			munmap_notifier
kernel/profile.c			task_exit_notifier
kernel/sys.c				reboot_notifier_list
net/core/dev.c				netdev_chain
net/decnet/dn_dev.c:			dnaddr_chain
net/ipv4/devinet.c:			inetaddr_chain

It's possible that some of these classifications are wrong.  If they are,
please let us know or submit a patch to fix them.  Note that any chain that
gets called very frequently should be atomic, because the rwsem read-locking
used for blocking chains is very likely to incur cache misses on SMP systems.
(However, if the chain's callout routines may sleep then the chain cannot be
atomic.)

The patch set was written by Alan Stern and Chandra Seetharaman, incorporating
material written by Keith Owens and suggestions from Paul McKenney and Andrew
Morton.

[jes@sgi.com: restructure the notifier chain initialization macros]
Signed-off-by: NAlan Stern <stern@rowland.harvard.edu>
Signed-off-by: NChandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: NJes Sorensen <jes@sgi.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

	Originally submitted by Kenzo Iwami; his original description is:

The current bonding driver receives duplicate packets when broadcast/
multicast packets are sent by other devices or packets are flooded by the
switch. In this patch, new flags are added in priv_flags of net_device
structure to let the bonding driver discard duplicate packets in
dev.c:skb_bond().

	Modified by Jay Vosburgh to change a define name, update some
comments, rearrange the new skb_bond() for clarity, clear all bonding
priv_flags on slave release, and update the driver version.
Signed-off-by: NKenzo Iwami <k-iwami@cj.jp.nec.com>
Signed-off-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJeff Garzik <jeff@garzik.org>

bond_release returns EINVAL without releasing the bond lock if the
slave device is not being bonded by the bond.  The following patch
ensures that the lock is released in this case.
Signed-off-by: NStephen J. Bevan <stephen@dino.dnsalias.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

drivers/net/bonding/bond_sysfs.c:263:27: warning: Using plain integer as NULL pointer
drivers/net/bonding/bond_sysfs.c:998:26: warning: Using plain integer as NULL pointer
drivers/net/bonding/bond_sysfs.c:1126:26: warning: Using plain integer as NULL pointer
Signed-off-by: NLuiz Capitulino <lcapitulino@mandriva.com.br>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Add NETIF_F_TSO (NETIF_F_UFO) to BOND_INTERSECT_FEATURES so that it can
be used by a bonding device iff all its slave devices support TSO (UFO).
Signed-off-by: NArthur Kepner <akepner@sgi.com>
Signed-off-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Since get_settings() returns a signed int and it gets checked
for < 0 to catch an error, res should be a signed int too.
Signed-off-by: NEric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: NAlexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

	I believe I see the race Michael refers to (tlb_choose_channel
may set head, which tlb_init_slave clears), although I was not able to
reproduce it.  I have updated his patch for the current netdev-2.6.git
tree and added a version update.  His original comment follows:

Our systems have been crashing during testing of PCI HotPlug
support in the various networking components.  We've faulted in
the bonding driver due to a bug in bond_alb.c:tlb_clear_slave()

In that routine, the last modification to the TLB hash table is
made without protection of the lock, allowing a race that can lead
tlb_choose_channel() to select an invalid table element.

	-J
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

"extern inline" doesn't make much sense.
Signed-off-by: NAdrian Bunk <bunk@stusta.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Garzik <jgarzik@pobox.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

No need to record this information in source code, its all in the git
repository, and kernel archives.

Bonding source files still have changelogs in the comments.  This, then,
is an update to that changelog.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Minor spelling and whitespace corrections.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Update the version number for the bonding module.  Since we've just
added a significant new feature (sysfs support), bump the major number.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This large patch adds sysfs functionality to the channel bonding module.
Bonds can be added, removed, and reconfigured at runtime without having
to reload the module.  Multiple bonds with different configurations are
easily configured, and ifenslave is no longer required to configure bonds.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Make the /proc files show which ARP targets are in use by each bond.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

With the sysfs interface, the user can remove entries from the ARP table
at runtime.  The ARP monitor code now allows for empty entries in the
table.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The sysfs interface can create bonds at runtime, and __init code goes away
after module init.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The sysfs interface can create bonds at runtime, so we need a separate
function to do this, instead of just doing it in the module init code.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The sysfs code needs access these functions, so make them
not static, and move the protos to the header file.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The sysfs code needs to know what these structs look like, so make them
not static, and move the definition to the header.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Explicitly clear RLB flag during ALB init.  This is needed for sysfs
support, since the bond mode can be changed at runtime via sysfs.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Move memory allocations out of the spinlock during ALB init.  This gets
rid of a sleeping-inside-spinlock warning and accompanying stack dump.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Take the primary slave name shown in /proc from the actual slave dev
instead of from the command-line parameter, which won't be present
if the bond is created via sysfs.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Adds information about the recently-added transmit policy setting to each
bond's /proc file.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Expand and correct the parameter descriptions shown by modinfo.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Add the bond name to all error messages so we can tell which one is
complaining.  Also reformats some error messages to be more consistent.
Signed-off-by: NMitch Williams <mitch.a.williams@intel.com>
Acked-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This should resolve http://bugzilla.kernel.org/show_bug.cgi?id=5519

The current feature computation loses bits that it doesn't know about,
resulting in an inability to add VLANs and possibly other havoc.
Rewrote function to preserve bits it doesn't know about, remove an
unneeded state variable, and simplify the code.
Signed-off-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Expand comment explaining MAC address selection for replicated IGMP
frames transmitted in bonding mode 1 (active-backup).  Also, a small
whitespace cleanup.
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

 - added typedef unsigned int __nocast gfp_t;

 - replaced __nocast uses for gfp flags with gfp_t - it gives exactly
   the same warnings as far as sparse is concerned, doesn't change
   generated code (from gcc point of view we replaced unsigned int with
   typedef) and documents what's going on far better.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

Fix implicit nocast warnings in bonding code:
drivers/net/bonding/bond_main.c:1302:49: warning: implicit cast to nocast type
Signed-off-by: NRandy Dunlap <rdunlap@xenotime.net>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Replicate IGMP frames across all slaves in activebackup mode. This
ensures fail-over is rapid for multicast traffic as well. Otherwise,
multicast traffic will be lost until the next IGMP membership report
poll timeout.

This is conceptually similar to the treatment of IGMP traffic in
bond_alb_xmit. In that case, IGMP traffic transmitted on any slave
is re-routed to the active slave in order to ensure that multicast
traffic continues to be directed to the active receiver.
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

David S. Miller <davem@davemloft.net> wrote:
>I think removing support for older ifenslave binaries is
>the least painful solution to this problem.

	This patch removes backwards compatibility for old ifenslave
binaries (ifenslave prior to verison 1.0.0).

	I did not similarly modify ifenslave itself; with sysfs on the
horizon, I don't see that as being worthwhile.
Signed-off-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

The following patch renames __in_dev_get() to __in_dev_get_rtnl() and
introduces __in_dev_get_rcu() to cover the second case.

1) RCU with refcnt should use in_dev_get().
2) RCU without refcnt should use __in_dev_get_rcu().
3) All others must hold RTNL and use __in_dev_get_rtnl().

There is one exception in net/ipv4/route.c which is in fact a pre-existing
race condition.  I've marked it as such so that we remember to fix it.

This patch is based on suggestions and prior work by Suzanne Wood and
Paul McKenney.
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Fix bond_enslave link monitoring warning to check use_carrier status
and ethtool_ops in addition to do_ioctl.  This version checks ethtool_ops
as well as do_ioctl, and also uses the per-bond params.use_carrier
instead of the global use_carrier.
Signed-off-by: NJason R. Martin <nsxfreddy@gmail.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

From: Florin Malita <fmalita@gmail.com>

bond_init() is not releasing rtnl_sem after register_netdevice() and before
calling unregister_netdevice() (from bond_free_all()) in the exception
path.  As the device registration is not completed (dev->reg_state ==
NETREG_REGISTERING), the call to unregister_netdevice() triggers
BUG_ON(dev->reg_state != NETREG_REGISTERED).
Signed-off-by: NFlorin Malita <fmalita@gmail.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

	Bonding leaks route structures when the ARP monitor is
configured to send probes over VLANs.

	Originally reported by Ian Abel <ian.abel@mxtelecom.com>; his
original fix was modified by Jay Vosburgh to correct coding style and to
close a leak it missed.
Signed-off-by: NJay Vosburgh <fubar@us.ibm.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Bonding just wants the device before the skb_bond()
decapsulation occurs, so simply pass that original
device into packet_type->func() as an argument.

It remains to be seen whether we can use this same
exact thing to get rid of skb->input_dev as well.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This change allows a bonding device to inherit the "zero-copy"
features of its slave devices.

It was inspired by a couple of previous postings on this topic:
http://marc.theaimsgroup.com/?l=bonding-devel&m=111924607327794&w=2
http://marc.theaimsgroup.com/?l=bonding-devel&m=111925242706297&w=2
and it's largely a combination of the patches that appear in those
emails.
Signed-off-by: NArthur Kepner <akepner@sgi.com>

In ALB mode, allow new slave to use bond's MAC address if the new
slave's MAC address is being used within the bond and no other slave
is using the bond's MAC address.
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>
Signed-off-by: NJeff Garzik <jgarzik@pobox.com>

Add support for alternate slave selection algorithms to bonding
balance-xor and 802.3ad modes.  Default mode (what we have now: xor of
MAC addresses) is "layer2", new choice is "layer3+4", using IP and port
information for hashing to select peer.

Originally submitted by Jason Gabler for balance-xor mode;
modified by Jay Vosburgh to additionally support 802.3ad mode.  Jason's
original comment is as follows:

The attached patch to the Linux Etherchannel Bonding driver modifies the
driver's "balance-xor" mode as follows:

      - alternate hashing policy support for mode 2
        * Added kernel parameter "xmit_policy" to allow the specification
          of different hashing policies for mode 2.  The original mode 2
          policy is the default, now found in xmit_hash_policy_layer2().
        * Added xmit_hash_policy_layer34()

This patch was inspired by hashing policies implemented by Cisco,
Foundry and IBM, which are explained in
Foundry documentation found at:
http://www.foundrynet.com/services/documentation/sribcg/Trunking.html#112750Signed-off-by: NJason Gabler <jygabler@lbl.gov>
Signed-off-by: NJay Vosburgh <fubar@us.ibm.com>