set_primary_if is currently misused to update the mac address in vis
packets. This unneeded and introduces overhead due to other operations
which must be done when updating the primary interface.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
set_primary_if exchanges the current primary interfaces with a new one.
This is a new reference and thus we have to count it and decrease the
count of the old primary interface.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The return of get_batman_if_by_netdev and get_active_batman_if leaks a
pointer from the rcu protected list of interfaces. We must protect it to
prevent a too early release of the memory. Those functions must increase
the reference counter before rcu_read_unlock or it may be to late to
prevent a free.
hardif_add_interface must also increase the reference count for the
returned batman_if to make the behaviour consistent.
Reported-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
get_batman_if_by_netdev and get_active_batman_if may leak data from the
rcu protected list of interfaces. The rcu protected list of all gateway
nodes leaks the actual data outside the read-side critical area. This is
not valid as we may free the data using a call_rcu created callback
after we unlock using rcu_read_unlock. A workaround is to provide a
reference count to be sure that the memory isn't freed to early.
It is currently only to implement the already existing functionality and
doesn't provide the full tracking of all usage cases.
Additionally, we must hardif_hold inside the
rcu_read_lock()..rcu_read_unlock() before we attach to the structure
which "leaks" it. When another function now removed it from its usage
context (primary_if, usage on stack, ...) then we must hardif_put it. If
it is decremented to zero then we can issue the call_rcu to the freeing
function. So "put" is not allowed inside an rcu_read_lock.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
It is recommended [1] to use synchronize_rcu to simplify the code -
especially when otherwise extra locking is needed to protect other code
from picking stale elements. It also protects us for emitting to many
callbacks which may results in OOM conditions.
The only reason not to use it, would be in performance critical sections
or when we are not allowed to block.
[1] Documentation/RCU/checklist.txt
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Regions which do not use rcu functions don't need to protected by
rcu_read_lock. If we want to protect data from being freed than it must
be covered by the same read-side critical section or otherwise the grace
period may already ended and freed the memory before we called
rcu_read_lock again.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
receive_bat_packet is not called with rcu_read_lock so we must ensure by
ourself that we protect list_for_each_entry_rcu using the correct RCU
locks.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The update critical sections of if_list must be protected by a locking
primitive other than RCU. The iterator must also be protected by the
chosen locking mechanism.
The rtnl_lock in hardif_remove_interfaces must also be moved outside the
iterator primitive to ensure that we don't deadlock the kernel due to
differently nested locks in hardif_remove_interfaces and hard_if_event.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
synchronize_rcu respective synchronize_net only waits for the rcu grace
period to elapse and we may fail to finish the calls which were made to
call_rcu in that time. In result the module could be unloaded during the
execution of the RCU callbacks.
rcu_barrier[1] will now wait for all outstanding RCU callbacks to finish
before continuing.
[1] Documentation/RCU/rcubarrier.txt
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We don't allow to seek in the debugfs socket and log files. Thus we
should mark the file descriptor as nonseekable.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The official mailing list is run on lists.open-mesh.org and it should be
avoided to sent them to lists.open-mesh.net to reduce the number of
receipents and double posts.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
checkpatch now detects the start of a comment and warns about usage of
multiple spaces at the beginning of a line. We have to replace the ' '
in multiple lines comments by ' * ' to fix it.
Checkpatch also wants a comment after a definition of a spinlock_t which
describes what it protects. It is currently not possible to add it
before the actual struct which includes the spinlock.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
kmalloc() may fail, if so drop current packet.
Signed-off-by: Vasiliy Kulikov <segooon@gmail.com>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
[sven.eckelmann@gmx.de: Removed new introduced deadlock]
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Paul E. McKenney informed us that rcu is misused by leaking pointers to
rcu related elements outside read-side protected critical sections.
He also recommended that it should be checked against the rcu checklist.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
gw_list_lock, gw_list and curr_gw are currently unused members of struct
bat_priv. They will be readded when gateway support is really
implemented.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
During the module shutdown procedure in batman_exit(), a rcu callback is
being scheduled (batman_exit -> hardif_remove_interfaces ->
hardif_remove_interfae -> call_rcu). However, when the kernel unloads
the module, the rcu callback might not have been executed yet, resulting
in a "unable to handle kernel paging request" in __rcu_process_callback
afterwards, causing the kernel to freeze.
The synchronize_net and synchronize_rcu in mesh_free are currently
called before the call_rcu in hardif_remove_interface and have no real
effect on it.
Therefore, we should always flush all rcu callback functions scheduled
during the shutdown procedure using synchronize_net. The call to
synchronize_rcu can be omitted because synchronize_net already calls it.
Signed-off-by: Linus Lüssing <linus.luessing@web.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We must reduce our own mtu when we reduce the mtu of any device we use
to transfer our packets. Otherwise we may accept to large packets which
gets dropped by the actual device.
Reported-by: Vasiliy Kulikov <segooon@gmail.com>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Information about dropped packets are usually only interesting for
debugging purposes and otherwise open the possibility to flood the logs
of the target machine with useless information.
pr_debug will not output those information on a nodebug kernel.
Reported-by: Vasiliy Kulikov <segooon@gmail.com>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
batman_if has the name of the net_dev as extra string in its own
structure, but also holds a reference to the actual net_device structure
which always has the current name of the device. This makes it
unneccessary and also more complex because we must update the name in
situations when we receive a NETDEV_CHANGENAME event.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch removes all remaining global variables and includes the
necessary bits into the bat_priv structure. It is the last
remaining piece to allow multiple concurrent mesh clouds on the
same device.
A few global variables have been rendered obsolete during the process
and have been removed entirely.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
[sven.eckelmann@gmx.de: Rework on top of current version]
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch replaces the static bat0 interface with a dynamic/abstracted
approach. It is now possible to create multiple batX interfaces by
assigning hard interfaces to them. Each batX interface acts as an
independent mesh network. A soft interface is removed once no hard
interface references it any longer.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
[sven.eckelmann@gmx.de: Rework on top of current version]
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We must ensure that all interesting data is linear and not paged out to
access all information in a header or a full batman-adv related packet.
Otherwise we may drop packets which have non-linear headers but which
hold valid data.
This doesn't affect non-linear skbs which have all headers in a linear
head unless we must process the whole packet like in ogms or vis
packets.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We can use skb_cow instead of a handwritten function to test and create
a writable skb buffer. This also allows us to pre-allocate headroom to
be able to send the data without re-allocating the buffer again to add
the ethernet header.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The vis information structure is used in a way that it can be transfered
directly as packet. It still had to be copied into a skb because of an
extra buffer used for the actual preparation of the data. This is
unnecessary and can be replaced by a simple clone instead of an full
copy before each send.
This makes also the send_raw_packet function obsolete.
Reported-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
All originator messages are send through aggregation buffers. Those
buffers can directly be allocated as skb to reduce the cost of
allocation an extra buffer and copying them to a new allocated skb
directly before it gets send.
Now only the skb must be cloned in case of send_packet_to_if as it gets
called by send_packet for each interface. Non-primary ogms must not
cloned at all because they will only send once and the forward_packet
structure is freed by send_outstanding_bat_packet afterwards.
Reported-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
batman-adv tries to resend broadcasts on all interfaces up to three
times. For each round and each interface it must provide a skb which
gets consumed by the sending function.
It is unnecessary to copy the data of each broadcast because the actual
data is either not shared or already copied by add_bcast_packet_to_list.
So it is enough to just copy the skb control data
Reported-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
my_skb_push provided an easy way to allocate enough headroom in
situation were we don't have enough space left and move the data pointer
to the new position, but we didn't checked wether we are allowed to
write to the new pushed header. This is for example a problem when the
skb was cloned and thus doesn't have a private data part.
my_skb_head_push now replaces my_skb_push by using skb_cow_head to
provide only a large enough, writable header without testing for the
rest of the (maybe shared) data. It will also move the data pointer
using skb_push when skb_cow_head doesn't fail.
This should give us enough flexibility in situation were skbs will be
queued by underlying layers and still doesn't unnecessarily copy the
data in situations when the skb was consumed right away during
dev_queue_xmit.
Reported-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Batman-adv globally registered the batman-adv packet type and installed
a hook to batman_skb_recv(). Each interface receiving a packet with that
type would end up in this function which then had to loop through all
batman-adv internal interface structures to find the its meta data. The
more interfaces a system had the longer the loops might take. Each and
every packet goes through this function making it a performance critical
loop.
This patch installs the hook for each activated interface. The called
batman_skb_recv() can distinguish these calls, therefore avoiding the
loop through the interface structures.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
[sven.eckelmann@gmx.de: Rework on top of current version]
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
It is unnecessary to generate an icmp packet in an extra memory region
and than copying it to a new allocated skb.
This also resolved the problem that we do inform the user that we
couldn't send the packet because we couldn't allocate the socket buffer.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch implements a simple layer2 fragmentation to allow traffic
exchange over network interfaces with a MTU smaller than 1500 bytes. The
fragmentation splits the big packets into two parts and marks the frames
accordingly. The receiving end buffers the packets to reassemble the
orignal packet before passing it to the higher layers. This feature
makes it necessary to modify the batman-adv encapsulation for unicast
packets by adding a sequence number, flags and the originator address.
This modifcation is part of a seperate packet type for fragemented
packets to keep the original overhead as low as possible. This patch
enables the feature by default to ensure the data traffic can travel
through the network. But it also prints a warning to notify the user
about the performance implications.
Note: Fragmentation should be avoided at all costs since it has a
dramatic impact on the performance, especially when it comes wifi
networks. Instead of a single packet, 2 packets have to be sent! Not
only valuable airtime is wasted but also packetloss decreases the
throughput. A link with 50% packetloss and fragmentation enabled is
pretty much unusable.
Signed-off-by: Andreas Langer <an.langer@gmx.de>
[sven.eckelmann@gmx.de: Rework on top of current version]
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
[sven.eckelmann@gmx.de: Rework on top of current version]
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
to support multiple mesh devices later, we need to move global variables
like the queues into corresponding private structs bat_priv of the soft
devices.
Note that this patch still has a lot of FIXMEs and depends on the global
soft_device variable. This should be resolved later, e.g. by referencing
the parent soft device in batman_if.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
[sven.eckelmann@gmx.de: Rework on top of current version]
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The Kernighan algorithm is not able to calculate the number of set bits
in parallel and the compiler cannot replace it with optimized
instructions.
The kernel provides specialised functions for each cpu which can either
use a software implementation or hardware instruction depending on the
target cpu.
Reported-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The Ethernet header is counted when transmitting a packet, so it should also
be counted when receiving a packet. With this patch, the rx_bytes and tx_bytes
statistics behave like an ordinary Ethernet interface.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Version 2010.1.0 of the extra kernel module was released and thus the
documentation should be updated and everything prepared for the the
upcoming patchset.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The changelog is only generated on standalone releases. Thus it has no
real value for the in-kernel version of batman-adv.
Reported-by: Abraham Arce <abraham.arce.moreno@gmail.com>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This reverts commit 96d592ed59.
The netfilter hook seems to be misused and may leak skbs in situations
when NF_HOOK returns NF_STOLEN. It may not filter everything as
expected. Also the ethernet bridge tables are not yet capable to
understand batman-adv packet correctly.
It was only added for testing purposes and can be removed again.
Reported-by: Vasiliy Kulikov <segooon@gmail.com>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Each net_device in a system will automatically managed as a possible
batman_if and holds different informations like a buffer with a prepared
originator messages. To reduce the memory usage, the packet_buff will
only be allocated when the interface is really added/enabled for
batman-adv.
The function to update the hw address information inside the packet_buff
just assumes that the packet_buff is always initialised and thus the
kernel will just oops when we try to change the hw address of a not
already fully enabled interface.
We must always check if the packet_buff is allocated before we try to
change information inside of it.
Reported-by: Tim Glaremin <Tim.Glaremin@web.de>
Reported-by: Kazuki Shimada <zukky@bb.banban.jp>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
dev_put allows a device to be freed when all its references are dropped.
After that we are not allowed to access that information anymore. Access
to the data structure of a net_device must be surrounded a dev_hold
and ended using dev_put.
batman-adv adds a device to its own management structure in
hardif_add_interface and will release it in hardif_remove_interface.
Thus it must hold a reference all the time between those functions to
prevent any access to the already released net_device structure.
Reported-by: Tim Glaremin <Tim.Glaremin@web.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We try to get all events for all net_devices to be able to add special
sysfs folders for the batman-adv configuration. This also includes such
events like NETDEV_POST_INIT which has no valid kobject according to
v2.6.32-rc3-13-g7ffbe3f. This would create an oops in that situation.
It is enough to create the batman_if only on NETDEV_REGISTER events
because we will also receive those events for devices which already
existed when we registered the notifier call.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Earlier batman-adv versions would only create a batman_if struct after
a corresponding interface had been activated by a user. Now each
existing system interface has a batman_if struct and has to be checked
by verifying the IF_ACTIVE flag.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When receiving an batman icmp echo request or in case of a time-to-live
exceeded batman would reply with the mac address of the outgoing
interface which might be a secondary interface. Because secondary
interfaces are not globally known this might lead to confusion.
Now, replies are sent with the mac address of the primary interface.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
If a batman icmp packet had to be routed over a secondary interface
at the first hop, the mac address of that secondary interface would
be written in the 'orig' field of the icmp packet. A node which is
more than one hop away is not aware of the mac address because
secondary interfaces are not flooded through the whole mesh and
therefore can't send a reply.
This patch always sends the mac address of the primary interface
in the 'orig' field of the icmp packet.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The orig_hash_lock spinlock always has to be locked with IRQs being
disabled to avoid deadlocks between code that is being executed in
IRQ context and code that is being executed in non-IRQ context.
Reported-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Greg Kroah-Hartman merged Linus 2.6.36 tree in
e9563355ac with his staging tree.
Different parts of the merge conflicts were resolved incorrectly and may
result in an abnormal behavior.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6: (1443 commits)
phy/marvell: add 88ec048 support
igb: Program MDICNFG register prior to PHY init
e1000e: correct MAC-PHY interconnect register offset for 82579
hso: Add new product ID
can: Add driver for esd CAN-USB/2 device
l2tp: fix export of header file for userspace
can-raw: Fix skb_orphan_try handling
Revert "net: remove zap_completion_queue"
net: cleanup inclusion
phy/marvell: add 88e1121 interface mode support
u32: negative offset fix
net: Fix a typo from "dev" to "ndev"
igb: Use irq_synchronize per vector when using MSI-X
ixgbevf: fix null pointer dereference due to filter being set for VLAN 0
e1000e: Fix irq_synchronize in MSI-X case
e1000e: register pm_qos request on hardware activation
ip_fragment: fix subtracting PPPOE_SES_HLEN from mtu twice
net: Add getsockopt support for TCP thin-streams
cxgb4: update driver version
cxgb4: add new PCI IDs
...
Manually fix up conflicts in:
- drivers/net/e1000e/netdev.c: due to pm_qos registration
infrastructure changes
- drivers/net/phy/marvell.c: conflict between adding 88ec048 support
and cleaning up the IDs
- drivers/net/wireless/ipw2x00/ipw2100.c: trivial ipw2100_pm_qos_req
conflict (registration change vs marking it static)
David S. Miller provided some new ideas and found problems in his review
20100719.212625.255369607.davem@davemloft.net. These issues must be
resolved before it can be merged into net.
Signed-off-by: Sven Eckelmann <sven.eckelmann@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
