In case of a client X roaming from a generic node A to another node B, it is
possible that a third node C gets A's OGM but not B's. At this point in time, if
C wants to send data to X it will send a unicast packet destined to A. The
packet header will contain A's last ttvn (C got A's OGM and so it knows it).
The packet will travel towards A without being intercepted because the ttvn
contained in its header is the newest for A.
Once A will receive the packet, A's state will not report to be in a "roaming
phase" (because, after a roaming, once A sends out its OGM, all the changes are
committed and the node is considered not to be in the roaming state anymore)
and it will match the ttvn carried by the packet. Therefore there is no reason
for A to try to alter the packet's route, thus dropping the packet because the
destination client is not there anymore.
However, C is well aware that it's routing information towards the client X is
outdated as it received an OGM from A saying that the client roamed away.
Thanks to this detail, this patch introduces a small change in behaviour: as
long as C is in the state of not knowing the new location of client X it will
forward the traffic to its last known location using ttvn-1 of the destination.
By using an older ttvn node A will be forced to re-route the packet.
Intermediate nodes are also allowed to update the packet's destination as long
as they have the information about the client's new location.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
translation_table.{c,h} have been heavily modified by another contributor and
for legal purposes it is better to include his name into the contributor list
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Before sending out a TT_Request packet we must convert the tt_crc field value
to network order (since it is 16bits long).
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
As the backbone gateways are connected to the same backbone, they
should announce the same clients on the backbone non-exclusively.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
as backbone gateways will all independently announce the same clients,
also the tt global table must be able to hold multiple originators per
client entry.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Acked-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Actually the TT_CLIENT_PENDING flag is never set in the tt_global_entry
structure, therefore this code is useless and can be removed.
Reported-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Each tt_local_set_pending is always followed by a bat_dbg invocation. This can
be simplified by moving the bat_dbg() call.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
In case of a new tt_local_entry, the TT_CLIENT_NEW flag has to be set before
adding such entry to the hash table. Otherwise, it opens a race condition
in which the entry can be found but the flag has not been set.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Acked-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
To increase readability the has_timed_out() functions has been introduced.
This patch converts existing time_after() calls to use this wrapper
function (if applicable).
This patch also converts all timeouts to miliseconds to be consistent.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Both translation tables and network coding use timeouts to do house
keeping, so we might as well share the function used to compare a
timestamp+timeout with current time.
For readability and simplicity, the function is renamed to
has_timed_out() and uses time_is_before_jiffies() instead of
time_after().
Signed-off-by: Martin Hundebøll <martin@hundeboll.net>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
All batman-adv packets have a common 3 byte header. It can be used to share
some code between different code paths, but it was never explicit stated that
this header has to be always the same for all packets. Therefore, new code
changes always have the problem that they may accidently introduce regressions
by moving some elements around.
A new structure is introduced that contains the common header and makes it
easier visible that these 3 bytes have to be the same for all on-wire packets.
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
(ttvn == 0) is currently used as initial condition. However this is not a good
idea because ttvn gets the vale zero each time after reaching the maximum value
(wrap around). For this reason a new flag is added in order to define whether a
node has an initialised table or not. Moreover, after invoking
tt_global_del_orig(), tt_initialised has to be set to false
Reported-by: Alexey Fisher <bug-track@fisher-privat.net>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Tested-by: Alexey Fisher <bug-track@fisher-privat.net>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
I didn't resolve the merge properly during the last pull of the net
tree into net-next.
The code in the final resolution should set flags to TT_CLIENT_ROAM
not TT_CLIENT_PENDING.
Signed-off-by: David S. Miller <davem@davemloft.net>
When receiving a DEL change for a client due to a roaming event (change is
marked with TT_CLIENT_ROAM), each node has to check if the client roamed
to itself or somewhere else.
In the latter case the global entry is kept to avoid having no route at all
otherwise we can safely delete the global entry
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
In case of a client roaming from node A to node B, the latter have to mark the
corresponding global entry with TT_CLIENT_ROAM (instead of TT_CLIENT_PENDING).
Marking a global entry with TT_CLIENT_PENDING will end up in keeping such entry
forever (because this flag is only meant to be used with local entries and it is
never checked on global ones).
In the worst case (all the clients roaming to the same node A) the local and the
global table will contain exactly the same clients. Batman-adv will continue to
work, but the memory usage is duplicated.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
if hash_add() fails, we should remove the structure to avoid memory
leaks.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Acked-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
The tt_local_reset_flags() is actually used for one use case only. It is not
generalised enough to be used indifferent situations. This patch make it general
enough in order to let other code use it whenever a flag set is requested over
the whole hash table (passed as parameter). The function is now called
tt_set_flags()
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Several functions in the translation table management code assume that the
tt_global_entry and tt_local_entry structures have the same initial fields such
as 'addr' and 'hash_entry'. To improve the code readability and to avoid
mistakes in later changes, a common substructure that substitute the shared
fields has been introduced (struct tt_common_entry).
Thanks to this modification, it has also been possible to slightly reduce the
code length by merging some functions like compare_ltt/gtt() and
tt_local/global_hash_find()
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
When the translation tables (global and local) are written for debugfs,
it is not neccesary to allocate a buffer, we can directly use
seq_printf() to print them out.
This might actually be safer if the table changes between size
calculation and traversal, and we can't estimate the required size
wrong.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
There are two reasons for this fix:
- the result of choose_orig() and vis_choose() is an index and therefore it can't
be negative. Hence it is correct to make the return type unsigned too.
- sizeof(int) may not be the same on ALL the architectures. Since we plan to use
choose_orig() as DHT hash function, we need to guarantee that, given the same
argument, the result is the same. Then it is correct to explicitly express
the size of the return type (and the second argument). Since the expected
length is currently 4, uint32_t is the most convenient choice.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
get_orig_node() tries to retrieve an orig_node object based on a mac address
and creates it if not present. This is not the wanted behaviour in the
translation table code as we don't want to create new orig_code objects but
expect a NULL pointer if the object does not exist.
Reported-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
When deleting the entries, tt_global_del_orig() has to print the message passed
as argument instead of a static one.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
After removing the batman-adv module, the hash may be already gone
when tt_global_del_orig() tries to clean the hash. This patch adds
a sanity check to avoid this.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Tested-by: Alexey Fisher <bug-track@fisher-privat.net>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
struct tt_global_entry holds a reference to an orig_node which must be
decremented before deallocating the structure.
Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Tested-by: Alexey Fisher <bug-track@fisher-privat.net>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
In the TT_RESPONSE packet, the number of carried entries is not correctly set.
This leads to a wrong interpretation of the packet payload on the receiver side
causing random entries to be added to the global translation table. Therefore
the latter gets always corrupted, triggering a table recovery all the time.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Currently the counter of tt_local_entry structures (tt_local_num) is incremented
each time the tt_local_reset_flags() is invoked causing the node to send wrong
TT_REPONSE packets containing a copy of non-initialised memory thus corrupting
other nodes global translation table and making higher level communication
impossible.
Reported-by: Junkeun Song <jun361@gmail.com>
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Acked-by: Junkeun Song <jun361@gmail.com>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Since clients can have several flags on or off, this patches make them
appear in the local/global transtable output so that they can be checked
for debugging purposes.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
If a node has to send a packet issued by a WIFI client to another WIFI client,
the packet is dropped.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
When a node receives a unicast packet it checks if the source and the
destination client can communicate or not due to the AP isolation
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Clients connected through a 802.11 device are now marked with the
TT_CLIENT_WIFI flag. This flag is also advertised with the tt
announcement.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Several typos have been corrected and some sentences have been rephrased
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
To keep consistency of other originator tables, new clients detected as
roamed, are kept in the global table but are marked as TT_CLIENT_PENDING
They are purged only when the new ttvn is received by the corresponding
originator. Moreover they need to be considered as removed in case of global
transtable lookup.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
To keep transtable consistency among all the nodes, an originator must
not send not yet announced clients within a full table TT_RESPONSE.
Instead, deleted client have to be kept in the table in order to be sent
within an immediate TT_RESPONSE. In this way all the nodes in the
network will always provide the same response for the same request.
All the modification are committed at the next ttvn increment event.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
now tt_local_event() takes a flags argument instead of a sequence of
boolean values which would grow up with the time.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
The tt_local_entry structure now has a 'flags' field. This helps to
unify the flags format to all the client related structures (tt_global_entry
and tt_change). The 'never_purge' field is now encoded in the 'flags' one.
To optimise the usage of this field, its length has been increased to 16bit
in order to use the eight leading bits (from 0 to 7) to store flags that
have to be sent on the wire, while the eight ending ones are used for local
computation only.
Moreover 'enum tt_change_flags' is now called 'enum tt_client_flags' and the
defined values apply to the tt_local_entry, tt_global_entry and the tt_change
'flags' field.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
The local and the global translation-tables are now lock free and rcu
protected.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Acked-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
With the current client announcement implementation, in case of roaming,
an update is triggered on the new AP serving the client. At that point
the new information is spread around by means of the OGM broadcasting
mechanism. Until this operations is not executed, no node is able to
correctly route traffic towards the client. This obviously causes packet
drops and introduces a delay in the time needed by the client to recover
its connections.
A new packet type called ROAMING_ADVERTISEMENT is added to account this
issue.
This message is sent in case of roaming from the new AP serving the
client to the old one and will contain the client MAC address. In this
way an out-of-OGM update is immediately committed, so that the old node
can update its global translation table. Traffic reaching this node will
then be redirected to the correct destination utilising the fresher
information. Thus reducing the packet drops and the connection recovery
delay.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
The client announcement mechanism informs every mesh node in the network
of any connected non-mesh client, in order to find the path towards that
client from any given point in the mesh.
The old implementation was based on the simple idea of appending a data
buffer to each OGM containing all the client MAC addresses the node is
serving. All other nodes can populate their global translation tables
(table which links client MAC addresses to node addresses) using this
MAC address buffer and linking it to the node's address contained in the
OGM. A node that wants to contact a client has to lookup the node the
client is connected to and its address in the global translation table.
It is easy to understand that this implementation suffers from several
issues:
- big overhead (each and every OGM contains the entire list of
connected clients)
- high latencies for client route updates due to long OGM trip time and
OGM losses
The new implementation addresses these issues by appending client
changes (new client joined or a client left) to the OGM instead of
filling it with all the client addresses each time. In this way nodes
can modify their global tables by means of "updates", thus reducing the
overhead within the OGMs.
To keep the entire network in sync each node maintains a translation
table version number (ttvn) and a translation table checksum. These
values are spread with the OGM to allow all the network participants to
determine whether or not they need to update their translation table
information.
When a translation table lookup is performed in order to send a packet
to a client attached to another node, the destination's ttvn is added to
the payload packet. Forwarding nodes can compare the packet's ttvn with
their destination's ttvn (this node could have a fresher information
than the source) and re-route the packet if necessary. This greatly
reduces the packet loss of clients roaming from one AP to the next.
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Documentation/CodingStyle recommends to use the form
p = kmalloc(sizeof(*p), ...);
to calculate the size of a struct and not the version where the struct
name is spelled out to prevent bugs when the type of p changes. This
also seems appropriate for manipulation of buffers when they are
directly associated with p.
Signed-off-by: Sven Eckelmann <sven@narfation.org>
It is not necessary to cast a void* to the pointer type when we just
store it and don't want to do pointer arithmetic before the actual
assignment.
Signed-off-by: Sven Eckelmann <sven@narfation.org>
batman-adv uses pointers which are marked as const and should not
violate that type qualifier by passing it to functions which force a
cast to the non-const version.
Signed-off-by: Sven Eckelmann <sven@narfation.org>
To be coherent, all the functions/variables/constants have been renamed
to the TranslationTable style
Signed-off-by: Antonio Quartulli <ordex@autistici.org>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
The rcu protected macros rcu_dereference() and rcu_assign_pointer()
for the bat_priv->primary_if need to be used, as well as spin/rcu locking.
Otherwise we might end up using a primary_if pointer pointing to already
freed memory.
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
It might be possible that 2 threads access the same data in the same
rcu grace period. The first thread calls call_rcu() to decrement the
refcount and free the data while the second thread increases the
refcount to use the data. To avoid this race condition all refcount
operations have to be atomic.
Reported-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: Marek Lindner <lindner_marek@yahoo.de>
types.h is included by main.h, which is included at the beginning of any
other c-file anyway. Therefore this commit removes those duplicate
inclussions.
Signed-off-by: Linus Lüssing <linus.luessing@ascom.ch>
Signed-off-by: Sven Eckelmann <sven@narfation.org>
hna_local_fill_buffer must return the number of added hna entries and
not the last checked hash bucket.
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
B.A.T.M.A.N. (better approach to mobile ad-hoc networking) is a routing
protocol for multi-hop ad-hoc mesh networks. The networks may be wired or
wireless. See http://www.open-mesh.org/ for more information and user space
tools.
Signed-off-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
