linux-stable.git/net/core/rtnetlink.c, branch linux-2.6.26.y

net: Fix netdev_run_todo dead-lock

2008-11-10T19:18:00+00:00

[ Upstream commit 58ec3b4db9eb5a28e3aec5f407a54e28f7039c19 ]

Benjamin Thery tracked down a bug that explains many instances
of the error

unregister_netdevice: waiting for %s to become free. Usage count = %d

It turns out that netdev_run_todo can dead-lock with itself if
a second instance of it is run in a thread that will then free
a reference to the device waited on by the first instance.

The problem is really quite silly.  We were trying to create
parallelism where none was required.  As netdev_run_todo always
follows a RTNL section, and that todo tasks can only be added
with the RTNL held, by definition you should only need to wait
for the very ones that you've added and be done with it.

There is no need for a second mutex or spinlock.

This is exactly what the following patch does.

Signed-off-by: Herbert Xu 
Signed-off-by: David S. Miller 
Signed-off-by: Greg Kroah-Hartman

netlink: Improve returned error codes

2008-06-03T23:36:54+00:00

Make nlmsg_trim(), nlmsg_cancel(), genlmsg_cancel(), and
nla_nest_cancel() void functions.

Return -EMSGSIZE instead of -1 if the provided message buffer is not
big enough.

Signed-off-by: Thomas Graf 
Signed-off-by: David S. Miller

[RTNETLINK]: Fix bogus ASSERT_RTNL warning

2008-04-24T05:10:48+00:00

ASSERT_RTNL uses mutex_trylock to test whether the rtnl_mutex is
held. This bogus warnings when running in atomic context, which
f.e. happens when adding secondary unicast addresses through
macvlan or vlan or when synchronizing multicast addresses from
wireless devices.

Mid-term we might want to consider moving all address updates
to process context since the locking seems overly complicated,
for now just fix the bogus warning by changing ASSERT_RTNL to
use mutex_is_locked().

Signed-off-by: Patrick McHardy 
Signed-off-by: David S. Miller

[RTNL]: Introduce the rtnl_kill_links helper.

2008-04-16T07:46:52+00:00

This one is responsible for calling ->dellink on each net
device found in net to help with vlan net_exit hook in the
nearest future.

Signed-off-by: Pavel Emelyanov 
Acked-by: Patrick McHardy 
Signed-off-by: David S. Miller

[RTNL]: Relax for_each_netdev_safe in __rtnl_link_unregister.

2008-04-16T07:45:56+00:00

Each potential list_del (happening from inside a ->dellink call)
is followed by goto restart, so there's no need in _safe iteration.

Signed-off-by: Pavel Emelyanov 
Acked-by: Patrick McHardy 
Signed-off-by: David S. Miller

[NET] NETNS: Omit sock->sk_net without CONFIG_NET_NS.

2008-03-25T19:39:55+00:00

Introduce per-sock inlines: sock_net(), sock_net_set()
and per-inet_timewait_sock inlines: twsk_net(), twsk_net_set().
Without CONFIG_NET_NS, no namespace other than &init_net exists.
Let's explicitly define them to help compiler optimizations.

Signed-off-by: YOSHIFUJI Hideaki

[NET] NETNS: Omit net_device->nd_net without CONFIG_NET_NS.

2008-03-25T19:39:53+00:00

Introduce per-net_device inlines: dev_net(), dev_net_set().
Without CONFIG_NET_NS, no namespace other than &init_net exists.
Let's explicitly define them to help compiler optimizations.

Signed-off-by: YOSHIFUJI Hideaki

[RTNL]: Validate hardware and broadcast address attribute for RTM_NEWLINK

2008-02-24T03:54:36+00:00

RTM_NEWLINK allows for already existing links to be modified. For this
purpose do_setlink() is called which expects address attributes with a
payload length of at least dev->addr_len. This patch adds the necessary
validation for the RTM_NEWLINK case.

The address length for links to be created is not checked for now as the
actual attribute length is used when copying the address to the netdevice
structure. It might make sense to report an error if less than addr_len
bytes are provided but enforcing this might break drivers trying to be
smart with not transmitting all zero addresses.

Signed-off-by: Thomas Graf 
Signed-off-by: David S. Miller

[RTNL]: Add missing link netlink attribute policy definitions

2008-02-20T00:12:08+00:00

IFLA_LINK is no longer a write-only attribute on the kernel side and
must thus be validated. Same goes for the newly introduced
IFLA_LINKINFO.

Fixes undefined behaviour if either of the attributes are not well
formed.

Signed-off-by: Thomas Graf 
Signed-off-by: David S. Miller

Revert "[RTNETLINK]: Send a single notification on device state changes."

2008-02-18T02:35:07+00:00

This reverts commit 45b503548210fe6f23e92b856421c2a3f05fd034.

It break locking around dev->link_mode as well as cause
other bootup problems.

Signed-off-by: David S. Miller