On Enhanced Retransmission Mode and Streaming Mode a entity can send, on
a successful Conf RSP, new values for the RFC fields. For example, the
entity can send txWindow and MPS values less than the value received on
a Conf REQ.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Tx Window value shall not be used with Streaming Mode and the receiver
of the config Request shall ignore its value.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

We ack I-frames on each txWindow/5 I-frames received, but if the sender
stop to send I-frames and it's not a txWindow multiple we can leave some
frames unacked.
So I added a timer to ack I-frames on this case. The timer expires in
200ms.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Now the code handles the case under SREJ_SENT state.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

tx_seq var already has the value of __get_reqseq().
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

It also fix a bug: we weren't acknowledging I-frames when P=1.
Note that when F=1 we are acknowledging packets before setting
RemoteBusy to False. The spec says we should do that in the opposite
order, but acknowledment of packets doesn't care about RemoteBusy flag
so we can do that in the order we want.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

We weren't handling the receipt under SREJ_SENT state table.
It also introduce l2cap_send_srejtail(). It will be used in the nexts
commits too.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Create a function for each type fo S-frame and avoid a lot of nested
code.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

This finishes the implementation of Recv RR (P=0)(F=0) for the Enhanced
Retransmission Mode on L2CAP.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Abstract the send of of P-bit and avoids code duplication like we
did with the setting of F-bit.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Shall be used to ack received frames, It must decide type of
acknowledgment between a RR frame, a RNR frame or transmission of
pending I-frames.
It also modifies l2cap_ertm_send() to report the number of frames sent.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

After reassembly the SDU we need to check his size. It can't overflow
the MTU size.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

All packets with size fewer than the minimum specified is dropped.
Note that the size of the l2cap basic header, FCS and SAR fields are
already subtracted of len at the moment of the size check.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

On receipt of a F=1 under WAIT_F state ERTM shall stop monitor timer and
start retransmission timer (if there are unacked frames).
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

After receive a RR with P bit set ERTM shall use this funcion to choose
what type of frame to reply with F bit = 1.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Trivial clean up.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

ERTM and Streaming Modes was having problems when the ACL MTU is lower
than MPS. The 'minus 10' is to take in account the header and fcs
lenghts.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

All operation related to the txWindow should be modulo 64.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

l2cap_data_channel do not free the S-frame, so we free it here.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Inside "case L2CAP_MODE_BASIC:" we don't need to check for sk_type and
L2CAP mode. So only the length check is fine.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Remove extra braces and labels, break over column 80 lines, etc
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

It also removes an unneeded check for the MTU. The check is done before
on sco_send_frame()
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Reviewed-by: NJoão Paulo Rechi Vita <jprvita@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

Set the proper error(ENOMEM), instead of just return 0.
Signed-off-by: NGustavo F. Padovan <padovan@profusion.mobi>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>

A while back there was a discussion regarding the rt_secret_interval timer.
Given that we've had the ability to do emergency route cache rebuilds for awhile
now, based on a statistical analysis of the various hash chain lengths in the
cache, the use of the flush timer is somewhat redundant.  This patch removes the
rt_secret_interval sysctl, allowing us to rely solely on the statistical
analysis mechanism to determine the need for route cache flushes.
Signed-off-by: NNeil Horman <nhorman@tuxdriver.com>
Acked-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This reverts commit 39e0786d.

On request from microblaze developers, they are going to handle
this differently.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add prefetches of the skb and the next rx descriptor to speed up rx path.

Use prefetchw() for the skb [suggested by Eric Dumazet].

The rx descriptor is in skb->data which is mapped for streaming mode DMA.
Eric Dumazet pointed out that we should not prefetch the data before
dma_sync.  So we prefetch only if dma_sync is no_op on the system.
Signed-off-by: NMichael Chan <mchan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

And turn on NETIF_F_GRO by default [requested by DaveM].
Signed-off-by: NMichael Chan <mchan@broadcom.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set the napi member it to 0 explicitely.
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memory for the private data is allocated using kzalloc in
alloc_etherdev (or alloc_netdev_mq respectively) so there is no need to
set it to 0 again.
Signed-off-by: NTobias Klauser <tklauser@distanz.ch>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Introduce ____napi_schedule() helper for callers in irq disabled
contexts. rps_trigger_softirq() becomes a leaf function.

Use container_of() in process_backlog() instead of accessing per_cpu
address.

Use a custom inlined version of __napi_complete() in process_backlog()
to avoid one locked instruction :

 only current cpu owns and manipulates this napi,
 and NAPI_STATE_SCHED is the only possible flag set on backlog.
 we can use a plain write instead of clear_bit(),
 and we dont need an smp_mb() memory barrier, since RPS is on,
 backlog is protected by a spinlock.
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

NET_IP_ALIGN defaults to 2, no need to override.

NET_SKB_PAD is now 64, which is much larger than microblaze's
L1_CACHE_SIZE so no need to override that either.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

eth_type_trans() & get_rps_cpus() currently need two 64bytes cache
lines in packet to compute rxhash.

Increasing NET_SKB_PAD from 32 to 64 reduces the need to one cache
line only, and makes RPS faster.

NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Signed-off-by: NEric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>