24 Feb, 2011
33 commits
-
Signed-off-by: David S. Miller
-
Enhance TIPC to skip unnecessary (and, in some cases, redundant)
preparation work when sending a broadcast link NACK message, since this
preparation is only required for broadcast messages that are sent in a
reliable manner. This change also fixes a bug that caused NACK messages
to be improperly counted as "TX packets" in TIPC's broadcast link
statistics.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Eliminates support for the "number of requested links" field in a neighbor
discovery message. This field was never used and has been removed from
the TIPC 2.0 protocol specification.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Eliminates TIPC's prototype support for message sequence numbering
on routable connections (i.e. connections requiring more than one hop).
This capability isn't currently used, and can be removed since TIPC
only supports systems in which all inter-node communication can be
achieved in a single hop.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Ensure that the routine that starts up processing on a newly created
link endpoint takes the spinlock of the node object that owns the link,
to prevent possible conflicts with processing involving other links
owned by that node object.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Modifies TIPC's congestion control between a connected port and its
peer so that it works as documented. The following changes have been
made:1) The counter of the number of messages sent by a port now starts
at zero, rather than one. This prevents the port from reporting port
congestion one message earlier than it was supposed to.2) The counter of the number of messages sent by a port is now
incremented only if a non-empty message is sent successfully.
This prevents the port from becoming permanently congested if
too many send attempts are unsuccessful because of congestion
(or other reasons). It also removes the risk that empty hand-
shaking messages used during connection setup might cause the
port to report congestion earlier than it was supposed to.3) The counter of the number of unacknowledged messages received by
a port controlled by an internal TIPC service is now incremented
only if the message is non-empty, in order to be consistent with
the aforementioned changes.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Eliminates a local iovec structure containing no data, which was
previously used during the establishment of a topology service connection,
since the same effect can be achieved by passing in a NULL pointer and
an iovec length of zero.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Ensures that a link reset or activate message has a "probe" field
of zero. (This field is currently unused in these messages, but this
could potentially change in future versions of TIPC.)Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Enhances TIPC's unicast and broadcast link code to update the transmit
queue maximum size counter in a single place, namely the routine that
adds messages to the queue. This ensures that the maximum size statistic
reported for unicast links is completely accurate, rather than being
partially based on statistical sampling.The changes to link.h are just documenting the roles of the variables.
Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Modifies the initialization code for TIPC's topology service to
avoid taking the spinlock protecting the subscriber list, since
there is no need to do this.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Allows the broadcast link to track the node that is requesting a retransmit
in a new field dedicated to that purpose. This replaces the existing
mechanism that (ab)uses an existing node structure linked list field to do
the tracking.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Remove all instances of legacy or proposed-but-not-implemented code
that lives within an #if 0 ... #endif block. If some of it is needed
in the future it can recovered out of history, but there is no need
for it to clutter up the active code base.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Corrects print statements that use %x to print pointer values to use
%p instead, so that 64-bit pointer values are displayed correctly.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Enhances TIPC link code to ignore an invalid link tolerance value
contained in an incoming LINK_PROTOCOL message, rather than
processing the value and potentially causing a divide-by-zero error.Also add a compile-time check that catches attempts to redefine
TIPC's minimum link tolerance value in a manner that might result
in the same divide-by-zero error at run-time.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Reject TIPC configuration service messages without a full message
header. Previously, an application that sent a message to the
configuration service that was too short could cause the validation
code to access an uninitialized field in the msghdr structure,
resulting in a memory access exception.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Eliminates a global variable that was previously used by TIPC's user
registry to track the number of distinct applications using TIPC. Due to
the recent elimination of the user registry this variable no longer serves
any purpose and can be removed.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Combines two distinct structures containing information about a TIPC bearer
into a single structure. The structures were previously kept separate so
that public information about a bearer could be made available to plug-in
media types using TIPC's native API, while the remaining information was
kept private for use by TIPC itself. However, now that the native API has
been removed there is no longer any need for this arrangement.Since one of the structures was already embedded within the other, the
change largely involves replacing instances of "publ.foo" with "foo".
The changes do not otherwise alter the operation of TIPC bearers.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Merge two distinct structures containing information about a TIPC port
into a single structure. The structures were previously kept separate
so that public information about a port could be made available to
applications using TIPC's native API, while the remaining information
was kept private for use by TIPC itself. However, now that the native
API has been removed there is no longer any need for this somewhat
confusing arrangement.Since one of the structures was already embedded within the other, the
change largely involves replacing instances of "publ.foo" with "foo".
The changes do not otherwise alter the operation of TIPC ports.Signed-off-by: Allan Stephens
Signed-off-by: Paul Gortmaker -
Max rx queues and tx queues are governed by fimware.
So driver should not validate these values.Signed-off-by: Rajesh Borundia
Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
o Remove checks of 1 for auto_fw_reset module parameter.
auto_fw_reset is of type int and can have value > 1.
o Remove unnecessary #define for 1Signed-off-by: Amit Kumar Salecha
Signed-off-by: David S. Miller -
Use discrete setting ops for not updated drivers. This will not make
them conform to full G/SFEATURES semantics, though.Signed-off-by: Michał Mirosław
Signed-off-by: David S. Miller -
Implement getting rx checksum state for not updated drivers.
Signed-off-by: Michał Mirosław
Signed-off-by: David S. Miller -
Avoid "Features changed" message and ndo_set_features call on device
registration caused by automatic enabling of GSO and GRO. Driver should
have enabled hardware offloads it set in features, so the ndo_set_features()
is not needed at registration time.Signed-off-by: Michał Mirosław
Signed-off-by: David S. Miller -
Fix netdev_update_features() messages on register time by moving
the call further in register_netdevice(). When
netdev->reg_state != NETREG_REGISTERED, netdev_name() returns
"(unregistered netdevice)" even if the dev's name is already filled.Signed-off-by: Michał Mirosław
Signed-off-by: David S. Miller -
Signed-off-by: Florian Fainelli
Signed-off-by: David S. Miller -
The original code does not work well when the number of mulitcast
address to handle is greater than MCAST_MAX. It only enable promiscous
mode instead of multicast hash table mode, so the hash table function
will not be activated and all multicast frames will be recieved in this
condition.This patch fixes the following issues with the r6040 NIC operating in
multicast:1) When the IFF_ALLMULTI flag is set, we should write 0xffff to the NIC
hash table registers to make it process multicast traffic.2) When the number of multicast address to handle is smaller than
MCAST_MAX, we should use the NIC multicast registers MID1_{L,M,H}.3) The hashing of the address was not correct, due to an invalid
substraction (15 - (crc & 0x0f)) instead of (crc & 0x0f) and an
incorrect crc algorithm (ether_crc_le) instead of (ether_crc).4) If necessary, we should set HASH_EN flag in MCR0 to enable multicast
hash table function.Reported-by: Marc Leclerc
Tested-by: Marc Leclerc
Signed-off-by: Shawn Lin
Signed-off-by: Albert Chen
Signed-off-by: Florian Fainelli
Signed-off-by: David S. Miller -
gfp_t needs to be cast to integer.
Signed-off-by: Stephen Hemminger
Signed-off-by: David S. Miller -
* make qdisc_ops local
* add sparse annotation about expected unlock/unlock in dump_class_stats
* fix indentationSigned-off-by: Stephen Hemminger
Signed-off-by: David S. Miller -
Signed-off-by: Stephen Hemminger
Signed-off-by: David S. Miller -
The dmaw_block is an enum and max_pay_load is u32. Therefore
sparse gives warning about comparison of unsigned and signed value.
Resolve by using min_t to force cast.Signed-off-by: Stephen Hemminger
Signed-off-by: David S. Miller -
Use __force to quiet sparse warnings for cases where the code
is simulating user space pointers.Signed-off-by: Stephen Hemminger
Signed-off-by: David S. Miller -
This is the Stochastic Fair Blue scheduler, based on work from :
W. Feng, D. Kandlur, D. Saha, K. Shin. Blue: A New Class of Active Queue
Management Algorithms. U. Michigan CSE-TR-387-99, April 1999.http://www.thefengs.com/wuchang/blue/CSE-TR-387-99.pdf
This implementation is based on work done by Juliusz Chroboczek
General SFB algorithm can be found in figure 14, page 15:
B[l][n] : L x N array of bins (L levels, N bins per level)
enqueue()
Calculate hash function values h{0}, h{1}, .. h{L-1}
Update bins at each level
for i = 0 to L - 1
if (B[i][h{i}].qlen > bin_size)
B[i][h{i}].p_mark += p_increment;
else if (B[i][h{i}].qlen == 0)
B[i][h{i}].p_mark -= p_decrement;
p_min = min(B[0][h{0}].p_mark ... B[L-1][h{L-1}].p_mark);
if (p_min == 1.0)
ratelimit();
else
mark/drop with probabilty p_min;I did the adaptation of Juliusz code to meet current kernel standards,
and various changes to address previous comments :http://thread.gmane.org/gmane.linux.network/90225
http://thread.gmane.org/gmane.linux.network/90375Default flow classifier is the rxhash introduced by RPS in 2.6.35, but
we can use an external flow classifier if wanted.tc qdisc add dev $DEV parent 1:11 handle 11: \
est 0.5sec 2sec sfb limit 128tc filter add dev $DEV protocol ip parent 11: handle 3 \
flow hash keys dst divisor 1024Notes:
1) SFB default child qdisc is pfifo_fast. It can be changed by another
qdisc but a child qdisc MUST not drop a packet previously queued. This
is because SFB needs to handle a dequeued packet in order to maintain
its virtual queue states. pfifo_head_drop or CHOKe should not be used.2) ECN is enabled by default, unlike RED/CHOKe/GRED
With help from Patrick McHardy & Andi Kleen
Signed-off-by: Eric Dumazet
CC: Juliusz Chroboczek
CC: Stephen Hemminger
CC: Patrick McHardy
CC: Andi Kleen
CC: John W. Linville
Signed-off-by: David S. Miller
23 Feb, 2011
7 commits
-
Signed-off-by: David S. Miller
-
Signed-off-by: David S. Miller
-
Signed-off-by: David S. Miller
-
Signed-off-by: David S. Miller
-
Signed-off-by: David S. Miller
-
Signed-off-by: David S. Miller
-
Signed-off-by: David S. Miller