Jiri mentioned that flowi6_tos of struct flowi6 is never used/read
anywhere. In fact, rest of the kernel uses the flowi6's flowlabel,
where the traffic class _and_ the flowlabel (aka flowinfo) is encoded.
For example, for policy routing, fib6_rule_match() uses ip6_tclass()
that is applied on the flowlabel member for matching on tclass. Similar
fix is needed for geneve, where flowi6_tos is set as well. Installing
a v6 blackhole rule that f.e. matches on tos is now working with vxlan.
Fixes: 1400615d64 ("vxlan: allow setting ipv6 traffic class")
Reported-by: Jiri Benc <jbenc@redhat.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
follows up commit 45f6fad84c ("ipv6: add complete rcu protection around
np->opt") which added mixed rcu/refcount protection to np->opt.
Given the current implementation of rcu_pointer_handoff(), this has no
effect at runtime.
Signed-off-by: Benjamin Poirier <bpoirier@suse.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds a static inline function ipv6_addr_prefix_copy which
copies a ipv6 address prefix(argument pfx) into the ipv6 address prefix.
The prefix len is given by plen as bits. This function mainly based on
ipv6_addr_prefix which copies one address prefix from address into a new
ipv6 address destination and zero all other address bits.
The difference is that ipv6_addr_prefix_copy don't get a prefix from an
ipv6 address, it sets a prefix to an ipv6 address with keeping other
address bits. The use case is for context based address compression
inside 6LoWPAN IPHC header which keeping ipv6 prefixes inside a context
table to lookup address-bits without sending them.
Cc: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Cc: James Morris <jmorris@namei.org>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: Patrick McHardy <kaber@trash.net>
Acked-by: Łukasz Duda <lukasz.duda@nordicsemi.no>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Acked-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Stefan Schmidt <stefan@osg.samsung.com>
Signed-off-by: Alexander Aring <alex.aring@gmail.com>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
This patch addresses multiple problems :
UDP/RAW sendmsg() need to get a stable struct ipv6_txoptions
while socket is not locked : Other threads can change np->opt
concurrently. Dmitry posted a syzkaller
(http://github.com/google/syzkaller) program desmonstrating
use-after-free.
Starting with TCP/DCCP lockless listeners, tcp_v6_syn_recv_sock()
and dccp_v6_request_recv_sock() also need to use RCU protection
to dereference np->opt once (before calling ipv6_dup_options())
This patch adds full RCU protection to np->opt
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
If a fragmented multicast packet is received on an ethernet device which
has an active macvlan on top of it, each fragment is duplicated and
received both on the underlying device and the macvlan. If some
fragments for macvlan are processed before the whole packet for the
underlying device is reassembled, the "overlapping fragments" test in
ip6_frag_queue() discards the whole fragment queue.
To resolve this, add device ifindex to the search key and require it to
match reassembling multicast packets and packets to link-local
addresses.
Note: similar patch has been already submitted by Yoshifuji Hideaki in
http://patchwork.ozlabs.org/patch/220979/
but got lost and forgotten for some reason.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
The network namespace is already passed into dst_output pass it into
dst->output lwt->output and friends.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Stop hidding the sk parameter with an inline helper function and make
all of the callers pass it, so that it is clear what the function is
doing.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only __ip6_local_out_sk has callers so rename __ip6_local_out_sk
__ip6_local_out and remove the previous __ip6_local_out.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For consistency with the other similar methods in the kernel pass a
struct sock into the dst_ops .local_out method.
Simplifying the socket passing case is needed a prequel to passing a
struct net reference into .local_out.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is to document that socket lock might not be held at this point.
skb_set_owner_w() and ipv6_local_error() are using proper atomic ops
or spinlocks, so we promote the socket to non const when calling them.
netfilter hooks should never assume socket lock is held,
we also promote the socket to non const.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ip6_dst_lookup_flow() and ip6_dst_lookup_tail() do not touch
socket, lets add a const qualifier.
This will permit the same change in inet6_csk_route_req()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is immediately motivated by the bridge code that chains functions that
call into netfilter. Without passing net into the okfns the bridge code would
need to guess about the best expression for the network namespace to process
packets in.
As net is frequently one of the first things computed in continuation functions
after netfilter has done it's job passing in the desired network namespace is in
many cases a code simplification.
To support this change the function dst_output_okfn is introduced to
simplify passing dst_output as an okfn. For the moment dst_output_okfn
just silently drops the struct net.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Initialize auto_flowlabels to one. This enables automatic flow labels,
individual socket may disable them using the IPV6_AUTOFLOWLABEL socket
option.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change the meaning of net.ipv6.auto_flowlabels to provide a mode for
automatic flow labels generation. There are four modes:
0: flow labels are disabled
1: flow labels are enabled, sockets can opt-out
2: flow labels are allowed, sockets can opt-in
3: flow labels are enabled and enforced, no opt-out for sockets
np->autoflowlabel is initialized according to the sysctl value.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can't call skb_get_hash here since the packet is not complete to do
flow_dissector. Create hash based on flowi6 instead.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds net argument to ipv6_stub_impl.ipv6_dst_lookup
for use cases where sk is not available (like mpls).
sk appears to be needed to get the namespace 'net' and is optional
otherwise. This patch series changes ipv6_stub_impl.ipv6_dst_lookup
to take net argument. sk remains optional.
All callers of ipv6_stub_impl.ipv6_dst_lookup have been modified
to pass net. I have modified them to use already available
'net' in the scope of the call. I can change them to
sock_net(sk) to avoid any unintended change in behaviour if sock
namespace is different. They dont seem to be from code inspection.
Signed-off-by: Roopa Prabhu <roopa@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch creates sk_set_txhash and eliminates protocol specific
inet_set_txhash and ip6_set_txhash. sk_set_txhash simply sets a
random number instead of performing flow dissection. sk_set_txash
is also allowed to be called multiple times for the same socket,
we'll need this when redoing the hash for negative routing advice.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds full IPv6 addresses into flow_keys and uses them as
input to the flow hash function. The implementation supports either
IPv4 or IPv6 addresses in a union, and selector is used to determine
how may words to input to jhash2.
We also add flow_get_u32_dst and flow_get_u32_src functions which are
used to get a u32 representation of the source and destination
addresses. For IPv6, ipv6_addr_hash is called. These functions retain
getting the legacy values of src and dst in flow_keys.
With this patch, Ethertype and IP protocol are now included in the
flow hash input.
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch changes flow hashing to use jhash2 over the flow_keys
structure instead just doing jhash_3words over src, dst, and ports.
This method will allow us take more input into the hashing function
so that we can include full IPv6 addresses, VLAN, flow labels etc.
without needing to resort to xor'ing which makes for a poor hash.
Acked-by: Jiri Pirko <jiri@resnulli.us>
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
ipv6_select_ident() returns a 32bit value in network order.
Fixes: 286c2349f6 ("ipv6: Clean up ipv6_select_ident() and ip6_fragment()")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes the assumptions that the returned rt is always
a RTF_CACHE entry with the rt6i_dst and rt6i_src containing the
destination and source address. The dst and src can be recovered from
the calling site.
We may consider to rename (rt6i_dst, rt6i_src) to
(rt6i_key_dst, rt6i_key_src) later.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Reviewed-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch changes the ipv6_select_ident() signature to return a
fragment id instead of taking a whole frag_hdr as a param to
only set the frag_hdr->identification.
It also cleans up ip6_fragment() to obtain the fragment id at the
beginning instead of using multiple "if" later to check fragment id
has been generated or not.
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Cc: Hannes Frederic Sowa <hannes@stressinduktion.org>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch divides the IPv6 flow label space into two ranges:
0-7ffff is reserved for flow label manager, 80000-fffff will be
used for creating auto flow labels (per RFC6438). This only affects how
labels are set on transmit, it does not affect receive. This range split
can be disbaled by systcl.
Background:
IPv6 flow labels have been an unmitigated disappointment thus far
in the lifetime of IPv6. Support in HW devices to use them for ECMP
is lacking, and OSes don't turn them on by default. If we had these
we could get much better hashing in IPv6 networks without resorting
to DPI, possibly eliminating some of the motivations to to define new
encaps in UDP just for getting ECMP.
Unfortunately, the initial specfications of IPv6 did not clarify
how they are to be used. There has always been a vague concept that
these can be used for ECMP, flow hashing, etc. and we do now have a
good standard how to this in RFC6438. The problem is that flow labels
can be either stateful or stateless (as in RFC6438), and we are
presented with the possibility that a stateless label may collide
with a stateful one. Attempts to split the flow label space were
rejected in IETF. When we added support in Linux for RFC6438, we
could not turn on flow labels by default due to this conflict.
This patch splits the flow label space and should give us
a path to enabling auto flow labels by default for all IPv6 packets.
This is an API change so we need to consider compatibility with
existing deployment. The stateful range is chosen to be the lower
values in hopes that most uses would have chosen small numbers.
Once we resolve the stateless/stateful issue, we can proceed to
look at enabling RFC6438 flow labels by default (starting with
scaled testing).
Signed-off-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
That was we can make sure the output path of ipv4/ipv6 operate on
the UDP socket rather than whatever random thing happens to be in
skb->sk.
Based upon a patch by Jiri Pirko.
Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
On the output paths in particular, we have to sometimes deal with two
socket contexts. First, and usually skb->sk, is the local socket that
generated the frame.
And second, is potentially the socket used to control a tunneling
socket, such as one the encapsulates using UDP.
We do not want to disassociate skb->sk when encapsulating in order
to fix this, because that would break socket memory accounting.
The most extreme case where this can cause huge problems is an
AF_PACKET socket transmitting over a vxlan device. We hit code
paths doing checks that assume they are dealing with an ipv4
socket, but are actually operating upon the AF_PACKET one.
Signed-off-by: David S. Miller <davem@davemloft.net>
As namespaces are sometimes used with overlapping ip address ranges,
we should also use the namespace as input to the hash to select the ip
fragmentation counter bucket.
Cc: Eric Dumazet <edumazet@google.com>
Cc: Flavio Leitner <fbl@redhat.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
in favor of their inner __ ones, which doesn't grab rtnl.
As these functions need to operate on a locked socket, we can't be
grabbing rtnl by then. It's too late and doing so causes reversed
locking.
So this patch:
- move rtnl handling to callers instead while already fixing some
reversed locking situations, like on vxlan and ipvs code.
- renames __ ones to not have the __ mark:
__ip_mc_{join,leave}_group -> ip_mc_{join,leave}_group
__ipv6_sock_mc_{join,drop} -> ipv6_sock_mc_{join,drop}
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based on the igmp v4 changes from Eric Dumazet.
959d10f6bbf6("igmp: add __ip_mc_{join|leave}_group()")
These changes are needed to perform igmp v6 join/leave while
RTNL is held.
Make ipv6_sock_mc_join and ipv6_sock_mc_drop wrappers around
__ipv6_sock_mc_join and __ipv6_sock_mc_drop to avoid
proliferation of work queues.
Signed-off-by: Madhu Challa <challa@noironetworks.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make __ipv6_select_ident() static as it isn't used outside
the file.
Fixes: 0508c07f5e (ipv6: Select fragment id during UFO segmentation if not set.)
Signed-off-by: Vladislav Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts:
drivers/net/vxlan.c
drivers/vhost/net.c
include/linux/if_vlan.h
net/core/dev.c
The net/core/dev.c conflict was the overlap of one commit marking an
existing function static whilst another was adding a new function.
In the include/linux/if_vlan.h case, the type used for a local
variable was changed in 'net', whereas the function got rewritten
to fix a stacked vlan bug in 'net-next'.
In drivers/vhost/net.c, Al Viro's iov_iter conversions in 'net-next'
overlapped with an endainness fix for VHOST 1.0 in 'net'.
In drivers/net/vxlan.c, vxlan_find_vni() added a 'flags' parameter
in 'net-next' whereas in 'net' there was a bug fix to pass in the
correct network namespace pointer in calls to this function.
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/ipv6.h:713:22: warning: incorrect type in assignment (different base types)
include/net/ipv6.h:713:22: expected restricted __be32 [usertype] hash
include/net/ipv6.h:713:22: got unsigned int
include/net/ipv6.h:719:25: warning: restricted __be32 degrades to integer
include/net/ipv6.h:719:22: warning: invalid assignment: ^=
include/net/ipv6.h:719:22: left side has type restricted __be32
include/net/ipv6.h:719:22: right side has type unsigned int
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the IPv6 fragment id has not been set and we perform
fragmentation due to UFO, select a new fragment id.
We now consider a fragment id of 0 as unset and if id selection
process returns 0 (after all the pertrubations), we set it to
0x80000000, thus giving us ample space not to create collisions
with the next packet we may have to fragment.
When doing UFO integrity checking, we also select the
fragment id if it has not be set yet. This is stored into
the skb_shinfo() thus allowing UFO to function correclty.
This patch also removes duplicate fragment id generation code
and moves ipv6_select_ident() into the header as it may be
used during GSO.
Signed-off-by: Vladislav Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is very similar to
commit 1c32c5ad6f
Author: Herbert Xu <herbert@gondor.apana.org.au>
Date: Tue Mar 1 02:36:47 2011 +0000
inet: Add ip_make_skb and ip_finish_skb
It adds IPv6 version of the helpers ip6_make_skb and ip6_finish_skb.
The job of ip6_make_skb is to collect messages into an ipv6 packet
and poplulate ipv6 eader. The job of ip6_finish_skb is to transmit
the generated skb. Together they replicated the job of
ip6_push_pending_frames() while also provide the capability to be
called independently. This will be needed to add lockless UDP sendmsg
support.
Signed-off-by: Vladislav Yasevich <vyasevic@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
UFO is now disabled on all drivers that work with virtio net headers,
but userland may try to send UFO/IPv6 packets anyway. Instead of
sending with ID=0, we should select identifiers on their behalf (as we
used to).
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Fixes: 916e4cf46d ("ipv6: reuse ip6_frag_id from ip6_ufo_append_data")
Signed-off-by: David S. Miller <davem@davemloft.net>
ipv6_opt_accepted() assumes IP6CB(skb) holds the struct inet6_skb_parm
that it needs. Lets not assume this, as TCP stack might use a different
place.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds a new sysctl_mld_qrv knob to configure the mldv1/v2 query
robustness variable. It specifies how many retransmit of unsolicited mld
retransmit should happen. Admins might want to tune this on lossy links.
Also reset mld state on interface down/up, so we pick up new sysctl
settings during interface up event.
IPv6 certification requests this knob to be available.
I didn't make this knob netns specific, as it is mostly a setting in a
physical environment and should be per host.
Cc: Flavio Leitner <fbl@redhat.com>
Signed-off-by: Hannes Frederic Sowa <hannes@stressinduktion.org>
Acked-by: Flavio Leitner <fbl@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The 'nqueues' counter is protected by the lru list lock,
once thats removed this needs to be converted to atomic
counter. Given this isn't used for anything except for
reporting it to userspace via /proc, just remove it.
We still report the memory currently used by fragment
reassembly queues.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
sparse is throwing warnings when building sunrpc modules due to some
endianness shenanigans in ipv6.h. Specifically:
CHECK net/sunrpc/addr.c
include/net/ipv6.h:573:17: warning: restricted __be64 degrades to integer
include/net/ipv6.h:577:34: warning: restricted __be32 degrades to integer
include/net/ipv6.h:573:17: warning: restricted __be64 degrades to integer
include/net/ipv6.h:577:34: warning: restricted __be32 degrades to integer
Sprinkle some endianness fixups to silence them. These should all get
fixed up at compile time, so I don't think this will add any extra work
to be done at runtime.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit cb1ce2ef38 ("ipv6: Implement automatic flow label generation
on transmit") introduced ip6_make_flowlabel, while commit b73c3d0e4f
("net: Save TX flow hash in sock and set in skbuf on xmit") introduced
ip6_set_txhash.
ip6_set_tx_hash() uses sk_v6_daddr which references
__sk_common.skc_v6_daddr from struct sock_common, which is gated with
IS_ENABLED(CONFIG_IPV6).
ip6_make_flowlabel() uses the ipv6 member from struct net which is
also gated with IS_ENABLED(CONFIG_IPV6).
When CONFIG_IPV6 is disabled, we will hit a build failure that looks
like this when the compiler attempts inlining these functions:
CC [M] drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.o
In file included from include/net/inet_sock.h:27:0,
from include/net/ip.h:30,
from drivers/net/ethernet/broadcom/cnic.c:37:
include/net/ipv6.h: In function 'ip6_set_txhash':
include/net/sock.h:327:33: error: 'struct sock_common' has no member named 'skc_v6_daddr'
#define sk_v6_daddr __sk_common.skc_v6_daddr
^
include/net/ipv6.h:696:49: note: in expansion of macro 'sk_v6_daddr'
keys.dst = (__force __be32)ipv6_addr_hash(&sk->sk_v6_daddr);
^
In file included from include/net/inetpeer.h:15:0,
from include/net/route.h:28,
from include/net/ip.h:31,
from drivers/net/ethernet/broadcom/cnic.c:37:
include/net/ipv6.h: In function 'ip6_make_flowlabel':
include/net/ipv6.h:706:37: error: 'struct net' has no member named 'ipv6'
if (!flowlabel && (autolabel || net->ipv6.sysctl.auto_flowlabels)) {
^
Fixes: cb1ce2ef38 ("ipv6: Implement automatic flow label generation on transmit")
Fixes: b73c3d0e4f ("net: Save TX flow hash in sock and set in skbuf on xmit")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Automatically generate flow labels for IPv6 packets on transmit.
The flow label is computed based on skb_get_hash. The flow label will
only automatically be set when it is zero otherwise (i.e. flow label
manager hasn't set one). This supports the transmit side functionality
of RFC 6438.
Added an IPv6 sysctl auto_flowlabels to enable/disable this behavior
system wide, and added IPV6_AUTOFLOWLABEL socket option to enable this
functionality per socket.
By default, auto flowlabels are disabled to avoid possible conflicts
with flow label manager, however if this feature proves useful we
may want to enable it by default.
It should also be noted that FreeBSD has already implemented automatic
flow labels (including the sysctl and socket option). In FreeBSD,
automatic flow labels default to enabled.
Performance impact:
Running super_netperf with 200 flows for TCP_RR and UDP_RR for
IPv6. Note that in UDP case, __skb_get_hash will be called for
every packet with explains slight regression. In the TCP case
the hash is saved in the socket so there is no regression.
Automatic flow labels disabled:
TCP_RR:
86.53% CPU utilization
127/195/322 90/95/99% latencies
1.40498e+06 tps
UDP_RR:
90.70% CPU utilization
118/168/243 90/95/99% latencies
1.50309e+06 tps
Automatic flow labels enabled:
TCP_RR:
85.90% CPU utilization
128/199/337 90/95/99% latencies
1.40051e+06
UDP_RR
92.61% CPU utilization
115/164/236 90/95/99% latencies
1.4687e+06
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For a connected socket we can precompute the flow hash for setting
in skb->hash on output. This is a performance advantage over
calculating the skb->hash for every packet on the connection. The
computation is done using the common hash algorithm to be consistent
with computations done for packets of the connection in other states
where thers is no socket (e.g. time-wait, syn-recv, syn-cookies).
This patch adds sk_txhash to the sock structure. inet_set_txhash and
ip6_set_txhash functions are added which are called from points in
TCP and UDP where socket moves to established state.
skb_set_hash_from_sk is a function which sets skb->hash from the
sock txhash value. This is called in UDP and TCP transmit path when
transmitting within the context of a socket.
Tested: ran super_netperf with 200 TCP_RR streams over a vxlan
interface (in this case skb_get_hash called on every TX packet to
create a UDP source port).
Before fix:
95.02% CPU utilization
154/256/505 90/95/99% latencies
1.13042e+06 tps
Time in functions:
0.28% skb_flow_dissect
0.21% __skb_get_hash
After fix:
94.95% CPU utilization
156/254/485 90/95/99% latencies
1.15447e+06
Neither __skb_get_hash nor skb_flow_dissect appear in perf
Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ideally, we would need to generate IP ID using a per destination IP
generator.
linux kernels used inet_peer cache for this purpose, but this had a huge
cost on servers disabling MTU discovery.
1) each inet_peer struct consumes 192 bytes
2) inetpeer cache uses a binary tree of inet_peer structs,
with a nominal size of ~66000 elements under load.
3) lookups in this tree are hitting a lot of cache lines, as tree depth
is about 20.
4) If server deals with many tcp flows, we have a high probability of
not finding the inet_peer, allocating a fresh one, inserting it in
the tree with same initial ip_id_count, (cf secure_ip_id())
5) We garbage collect inet_peer aggressively.
IP ID generation do not have to be 'perfect'
Goal is trying to avoid duplicates in a short period of time,
so that reassembly units have a chance to complete reassembly of
fragments belonging to one message before receiving other fragments
with a recycled ID.
We simply use an array of generators, and a Jenkin hash using the dst IP
as a key.
ipv6_select_ident() is put back into net/ipv6/ip6_output.c where it
belongs (it is only used from this file)
secure_ip_id() and secure_ipv6_id() no longer are needed.
Rename ip_select_ident_more() to ip_select_ident_segs() to avoid
unnecessary decrement/increment of the number of segments.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daniel Borkmann