Rename SCTP_INC_STATS_BH() to __SCTP_INC_STATS()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Remove misleading _BH suffix.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename TCP_INC_STATS_BH() to __TCP_INC_STATS()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename UDP_INC_STATS_BH() to __UDP_INC_STATS(),
and UDP6_INC_STATS_BH() to __UDP6_INC_STATS()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename ICMP_INC_STATS_BH() to __ICMP_INC_STATS()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Rename DCCP_INC_STATS_BH() to __DCCP_INC_STATS()
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

In the old days (before linux-3.0), SNMP counters were duplicated,
one for user context, and one for BH context.

After commit 8f0ea0fe ("snmp: reduce percpu needs by 50%")
we have a single copy, and what really matters is preemption being
enabled or disabled, since we use this_cpu_inc() or __this_cpu_inc()
respectively.

We therefore kill SNMP_INC_STATS_USER(), SNMP_ADD_STATS_USER(),
NET_INC_STATS_USER(), NET_ADD_STATS_USER(), SCTP_INC_STATS_USER(),
SNMP_INC_STATS64_USER(), SNMP_ADD_STATS64_USER(), TCP_ADD_STATS_USER(),
UDP_INC_STATS_USER(), UDP6_INC_STATS_USER(), and XFRM_INC_STATS_USER()

Following patches will rename __BH helpers to make clear their
usage is not tied to BH being disabled.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Similar to 3bfd8472 ("net: Use passed in table for nexthop lookups")
for IPv4, if the route spec contains a table id use that to lookup the
next hop first and fall back to a full lookup if it fails (per the fix
4c9bcd11 ("net: Fix nexthop lookups")).

Example:

    root@kenny:~# ip -6 ro ls table red
    local 2100:1::1 dev lo  proto none  metric 0  pref medium
    2100:1::/120 dev eth1  proto kernel  metric 256  pref medium
    local 2100:2::1 dev lo  proto none  metric 0  pref medium
    2100:2::/120 dev eth2  proto kernel  metric 256  pref medium
    local fe80::e0:f9ff:fe09:3cac dev lo  proto none  metric 0  pref medium
    local fe80::e0:f9ff:fe1c:b974 dev lo  proto none  metric 0  pref medium
    fe80::/64 dev eth1  proto kernel  metric 256  pref medium
    fe80::/64 dev eth2  proto kernel  metric 256  pref medium
    ff00::/8 dev red  metric 256  pref medium
    ff00::/8 dev eth1  metric 256  pref medium
    ff00::/8 dev eth2  metric 256  pref medium
    unreachable default dev lo  metric 240  error -113 pref medium

    root@kenny:~# ip -6 ro add table red 2100:3::/64 via 2100:1::64
    RTNETLINK answers: No route to host

Route add fails even though 2100:1::64 is a reachable next hop:
    root@kenny:~# ping6 -I red  2100:1::64
    ping6: Warning: source address might be selected on device other than red.
    PING 2100:1::64(2100:1::64) from 2100:1::1 red: 56 data bytes
    64 bytes from 2100:1::64: icmp_seq=1 ttl=64 time=1.33 ms

With this patch:
    root@kenny:~# ip -6 ro add table red 2100:3::/64 via 2100:1::64
    root@kenny:~# ip -6 ro ls table red
    local 2100:1::1 dev lo  proto none  metric 0  pref medium
    2100:1::/120 dev eth1  proto kernel  metric 256  pref medium
    local 2100:2::1 dev lo  proto none  metric 0  pref medium
    2100:2::/120 dev eth2  proto kernel  metric 256  pref medium
    2100:3::/64 via 2100:1::64 dev eth1  metric 1024  pref medium
    local fe80::e0:f9ff:fe09:3cac dev lo  proto none  metric 0  pref medium
    local fe80::e0:f9ff:fe1c:b974 dev lo  proto none  metric 0  pref medium
    fe80::/64 dev eth1  proto kernel  metric 256  pref medium
    fe80::/64 dev eth2  proto kernel  metric 256  pref medium
    ff00::/8 dev red  metric 256  pref medium
    ff00::/8 dev eth1  metric 256  pref medium
    ff00::/8 dev eth2  metric 256  pref medium
    unreachable default dev lo  metric 240  error -113 pref medium
Signed-off-by: NDavid Ahern <dsa@cumulusnetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

No more users in the tree, remove NETDEV_TX_LOCKED support.
Adds another hole in softnet_stats struct, but better than keeping
the unused collision counter around.
Signed-off-by: NFlorian Westphal <fw@strlen.de>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

For sctp assoc, when rcvbuf_policy is set, it will has it's own
rmem_alloc, when we dump asoc info in sctp_diag, we should use that
value on RMEM_ALLOC as well, just like WMEM_ALLOC.
Signed-off-by: NXin Long <lucien.xin@gmail.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

I also fix commit 8b32ab9e6ef1: use nla_total_size_64bit() for
OVS_FLOW_ATTR_USED in ovs_flow_cmd_msg_size().

Fixes: 8b32ab9e6ef1 ("ovs: use nla_put_u64_64bit()")
Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

I also fix the value of INET_DIAG_MAX. It's wrong since commit 8f840e47
which is only in net-next right now, thus I didn't make a separate patch.

Fixes: 8f840e47 ("sctp: add the sctp_diag.c file")
Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It was a simple idea -- save IPv6 configured addresses on a link down
so that IPv6 behaves similar to IPv4. As always the devil is in the
details and the IPv6 stack as too many behavioral differences from IPv4
making the simple idea more complicated than it needs to be.

The current implementation for keeping IPv6 addresses can panic or spit
out a warning in one of many paths:

1. IPv6 route gets an IPv4 route as its 'next' which causes a panic in
   rt6_fill_node while handling a route dump request.

2. rt->dst.obsolete is set to DST_OBSOLETE_DEAD hitting the WARN_ON in
   fib6_del

3. Panic in fib6_purge_rt because rt6i_ref count is not 1.

The root cause of all these is references related to the host route for
an address that is retained.

So, this patch deletes the host route every time the ifdown loop runs.
Since the host route is deleted and will be re-generated an up there is
no longer a need for the l3mdev fix up. On the 'admin up' side move
addrconf_permanent_addr into the NETDEV_UP event handling so that it
runs only once versus on UP and CHANGE events.

All of the current panics and warnings appear to be related to
addresses on the loopback device, but given the catastrophic nature when
a bug is triggered this patch takes the conservative approach and evicts
all host routes rather than trying to determine when it can be re-used
and when it can not. That can be a later optimizaton if desired.
Signed-off-by: NDavid Ahern <dsa@cumulusnetworks.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This reverts commit 841645b5.

Ok, this puts the feature back.  I've decided to apply David A.'s
bug fix and run with that rather than make everyone wait another
whole release for this feature.
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Support checksum neutral ILA as described in the ILA draft. The low
order 16 bits of the identifier are used to contain the checksum
adjustment value.

The csum-mode parameter is added to described checksum processing. There
are three values:
 - adjust transport checksum (previous behavior)
 - do checksum neutral mapping
 - do nothing

On output the csum-mode in the ila_params is checked and acted on. If
mode is checksum neutral mapping then to mapping and set C-bit.

On input, C-bit is checked. If it is set checksum-netural mapping is
done (regardless of csum-mode in ila params) and C-bit will be cleared.
If it is not set then action in csum-mode is taken.
Signed-off-by: NTom Herbert <tom@herbertland.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Change model of xlat to be used only for input where lookup is done on
the locator part of an address (comparing to locator_match as key
in rhashtable). This is needed for checksum neutral translation
which obfuscates the low order 16 bits of the identifier. It also
permits hosts to be in muliple ILA domains (each locator can map
to a different SIR address). A check is also added to disallow
translating non-ILA addresses (check of type in identifier).
Signed-off-by: NTom Herbert <tom@herbertland.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Add structures for identifiers, locators, and an ila address which
is composed of a locator and identifier and in6_addr can be cast to
it. This includes a three bit type field and enums for the types defined
in ILA I-D.

In ILA lwt don't allow user to set a translation for a non-ILA
address (type of identifier is zero meaning it is an IID). This also
requires that the destination prefix is at least 65 bytes (64
bit locator and first byte of identifier).
Signed-off-by: NTom Herbert <tom@herbertland.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

The memcpy of ipv6 header destination address to the skb control block
(sbk->cb) in header_create() results in currupted memory when bt_xmit()
is issued. The skb->cb is "released" in the return of header_create()
making room for lower layer to minipulate the skb->cb.

The value retrieved in bt_xmit is not persistent across header creation
and sending, and the lower layer will overwrite portions of skb->cb,
making the copied destination address wrong.

The memory corruption will lead to non-working multicast as the first 4
bytes of the copied destination address is replaced by a value that
resolves into a non-multicast prefix.

This fix removes the dependency on the skb control block between header
creation and send, by moving the destination address memcpy to the send
function path (setup_create, which is called from bt_xmit).
Signed-off-by: NGlenn Ruben Bakke <glenn.ruben.bakke@nordicsemi.no>
Acked-by: NJukka Rissanen <jukka.rissanen@linux.intel.com>
Signed-off-by: NMarcel Holtmann <marcel@holtmann.org>
Cc: stable@vger.kernel.org # 4.5+

rds-stress experiments with request size 256 bytes, 8K acks,
using 16 threads show a 40% improvment when pskb_extract()
replaces the {skb_clone(..); pskb_pull(..); pskb_trim(..);}
pattern in the Rx path, so we leverage the perf gain with
this commit.
Signed-off-by: NSowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

A pattern of skb usage seen in modules such as RDS-TCP is to
extract `to_copy' bytes from the received TCP segment, starting
at some offset `off' into a new skb `clone'. This is done in
the ->data_ready callback, where the clone skb is queued up for rx on
the PF_RDS socket, while the parent TCP segment is returned unchanged
back to the TCP engine.

The existing code uses the sequence
	clone = skb_clone(..);
	pskb_pull(clone, off, ..);
	pskb_trim(clone, to_copy, ..);
with the intention of discarding the first `off' bytes. However,
skb_clone() + pskb_pull() implies pksb_expand_head(), which ends
up doing a redundant memcpy of bytes that will then get discarded
in __pskb_pull_tail().

To avoid this inefficiency, this commit adds pskb_extract() that
creates the clone, and memcpy's only the relevant header/frag/frag_list
to the start of `clone'. pskb_trim() is then invoked to trim clone
down to the requested to_copy bytes.
Signed-off-by: NSowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

It was impossible to include codel.h for the
purpose of having access to codel_params or
codel_vars structure definitions and using them
for embedding in other more complex structures.

This splits allows codel.h itself to be treated
like any other header file while codel_qdisc.h and
codel_impl.h contain function definitions with
logic that was previously in codel.h.

This copies over copyrights and doesn't involve
code changes other than adding a few additional
include directives to net/sched/sch*codel.c.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This strips out qdisc specific bits from the code
and makes it slightly more reusable. Codel will be
used by wireless/mac80211 in the future.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We should call consume_skb(skb) when skb is properly consumed,
or kfree_skb(skb) when skb must be dropped in error case.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

We now have proper per-listener but also per network namespace counters
for SYN packets that might be dropped.

We replace the kfree_skb() by consume_skb() to be drop monitor [1]
friendly, and remove an obsolete comment.
FastOpen SYN packets can carry payload in them just fine.

[1] perf record -a -g -e skb:kfree_skb sleep 1; perf report
Signed-off-by: NEric Dumazet <edumazet@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

This reverts the following three commits:

70af921d
799977d9
f1705ec1

The feature was ill conceived, has terrible semantics, and has added
nothing but regressions to the already fragile ipv6 stack.

Fixes: f1705ec1 ("net: ipv6: Make address flushing on ifdown optional")
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Signed-off-by: NNicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

d894ba18 ("soreuseport: fix ordering for mixed v4/v6 sockets")
was merged as a bug fix to the net tree.  Two conflicting changes
were committed to net-next before the above fix was merged back to
net-next:
ca065d0c ("udp: no longer use SLAB_DESTROY_BY_RCU")
3b24d854 ("tcp/dccp: do not touch listener sk_refcnt under synflood")

These changes switched the datastructure used for TCP and UDP sockets
from hlist_nulls to hlist.  This patch applies the necessary parts
of the net tree fix to net-next which were not automatic as part of the
merge.

Fixes: 1602f49b ("Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net")
Signed-off-by: NCraig Gallek <kraig@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

After commit fbd40ea0 ("ipv4: Don't do expensive useless work
during inetdev destroy.") when deleting an interface,
fib_del_ifaddr() can be executed without any primary address
present on the dead interface.

The above is safe, but triggers some "bug: prim == NULL" warnings.

This commit avoids warning if the in_dev is dead
Signed-off-by: NPaolo Abeni <pabeni@redhat.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>

Linux TCP stack painfully segments all TSO/GSO packets before retransmits.

This was fine back in the days when TSO/GSO were emerging, with their
bugs, but we believe the dark age is over.

Keeping big packets in write queues, but also in stack traversal
has a lot of benefits.
 - Less memory overhead, because write queues have less skbs
 - Less cpu overhead at ACK processing.
 - Better SACK processing, as lot of studies mentioned how
   awful linux was at this ;)
 - Less cpu overhead to send the rtx packets
   (IP stack traversal, netfilter traversal, drivers...)
 - Better latencies in presence of losses.
 - Smaller spikes in fq like packet schedulers, as retransmits
   are not constrained by TCP Small Queues.

1 % packet losses are common today, and at 100Gbit speeds, this
translates to ~80,000 losses per second.
Losses are often correlated, and we see many retransmit events
leading to 1-MSS train of packets, at the time hosts are already
under stress.
Signed-off-by: NEric Dumazet <edumazet@google.com>
Acked-by: NYuchung Cheng <ycheng@google.com>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>