Back in 2006, commit 1a2449a87b ("[I/OAT]: TCP recv offload to I/OAT")
added support for receive offloading to IOAT dma engine if available.
The code in tcp_rcv_established() tries to perform early DMA copy if
applicable. It however does so without checking whether the userspace
task is actually expecting the data in the buffer.
This is not a problem under normal circumstances, but there is a corner
case where this doesn't work -- and that's when MSG_TRUNC flag to
recvmsg() is used.
If the IOAT dma engine is not used, the code properly checks whether
there is a valid ucopy.task and the socket is owned by userspace, but
misses the check in the dmaengine case.
This problem can be observed in real trivially -- for example 'tbench' is a
good reproducer, as it makes a heavy use of MSG_TRUNC. On systems utilizing
IOAT, you will soon find tbench waiting indefinitely in sk_wait_data(), as they
have been already early-copied in tcp_rcv_established() using dma engine.
This patch introduces the same check we are performing in the simple
iovec copy case to the IOAT case as well. It fixes the indefinite
recvmsg(MSG_TRUNC) hangs.
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit 92101b3b2e (ipv4: Prepare for change of rt->rt_iif encoding.)
invalidated TCP early demux, because rx_dst_ifindex is not properly
initialized and checked.
Also remove the use of inet_iif(skb) in favor or skb->skb_iif
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP_USER_TIMEOUT is a TCP level socket option that takes an unsigned int. But
patch "tcp: Add TCP_USER_TIMEOUT socket option"(dca43c75) didn't check the negative
values. If a user assign -1 to it, the socket will set successfully and wait
for 4294967295 miliseconds. This patch add a negative value check to avoid
this issue.
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the IPv6 missing bits for infrastructure added in commit
41063e9dd1 (ipv4: Early TCP socket demux.)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With the routing cache removal we lost the "noref" code paths on
input, and this can kill some routing workloads.
Reinstate the noref path when we hit a cached route in the FIB
nexthops.
With help from Eric Dumazet.
Reported-by: Alexander Duyck <alexander.duyck@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
commit d2d68ba9fe (ipv4: Cache input routes in fib_info nexthops.)
introduced rt_cache_valid() helper. It unfortunately doesn't check if
route is expired before caching it.
I noticed sk_setup_caps() was constantly called on a tcp workload.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1) Remove a non needed pskb_may_pull() in tcp_v4_early_demux()
and fix a potential bug if skb->head was reallocated
(iph & th pointers were not reloaded)
TCP stack will pull/check headers anyway.
2) must reload iph in ip_rcv_finish() after early_demux()
call since skb->head might have changed.
3) skb->dev->ifindex can be now replaced by skb->skb_iif
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On input packet processing, rt->rt_iif will be zero if we should
use skb->dev->ifindex.
Since we access rt->rt_iif consistently via inet_iif(), that is
the only spot whose interpretation have to adjust.
Signed-off-by: David S. Miller <davem@davemloft.net>
Use inet_iif() consistently, and for TCP record the input interface of
cached RX dst in inet sock.
rt->rt_iif is going to be encoded differently, so that we can
legitimately cache input routes in the FIB info more aggressively.
When the input interface is "use SKB device index" the rt->rt_iif will
be set to zero.
This forces us to move the TCP RX dst cache installation into the ipv4
specific code, and as well it should since doing the route caching for
ipv6 is pointless at the moment since it is not inspected in the ipv6
input paths yet.
Also, remove the unlikely on dst->obsolete, all ipv4 dsts have
obsolete set to a non-zero value to force invocation of the check
callback.
Signed-off-by: David S. Miller <davem@davemloft.net>
Alexey removed kernel side support for requests, and the
only thing we do for replies is log a message if something
doesn't look right.
As Alexey's comment indicates, this belongs in userspace (if
anywhere), and thus we can safely just get rid of this code.
Signed-off-by: David S. Miller <davem@davemloft.net>
With CONFIG_SPARSE_RCU_POINTER=y sparse identified references which did not
specificy __rcu in ip_vti.c
Signed-off-by: Saurabh Mohan <saurabh.mohan@vyatta.com>
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It is redundant to set no_addr and accept_local to 0 and then set them
with other values just after that.
Signed-off-by: Lin Ming <mlin@ss.pku.edu.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
ICMP messages generated in output path if frame length is bigger than
mtu are actually lost because socket is owned by user (doing the xmit)
One example is the ipgre_tunnel_xmit() calling
icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, htonl(mtu));
We had a similar case fixed in commit a34a101e1e (ipv6: disable GSO on
sockets hitting dst_allfrag).
Problem of such fix is that it relied on retransmit timers, so short tcp
sessions paid a too big latency increase price.
This patch uses the tcp_release_cb() infrastructure so that MTU
reduction messages (ICMP messages) are not lost, and no extra delay
is added in TCP transmits.
Reported-by: Maciej Żenczykowski <maze@google.com>
Diagnosed-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Tore Anderson <tore@fud.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
In tcp_tw_remember_stamp we incorrectly checked tw
instead of tm, it can lead to oops if the cached entry is
not found.
tcpm_stamp was not updated in tcpm_check_stamp when
tcpm_suck_dst was called, move the update into tcpm_suck_dst,
so that we do not call it infinitely on every next cache hit
after TCP_METRICS_TIMEOUT.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
The ipv4 routing cache is non-deterministic, performance wise, and is
subject to reasonably easy to launch denial of service attacks.
The routing cache works great for well behaved traffic, and the world
was a much friendlier place when the tradeoffs that led to the routing
cache's design were considered.
What it boils down to is that the performance of the routing cache is
a product of the traffic patterns seen by a system rather than being a
product of the contents of the routing tables. The former of which is
controllable by external entitites.
Even for "well behaved" legitimate traffic, high volume sites can see
hit rates in the routing cache of only ~%10.
The general flow of this patch series is that first the routing cache
is removed. We build a completely new rtable entry every lookup
request.
Next we make some simplifications due to the fact that removing the
routing cache causes several members of struct rtable to become no
longer necessary.
Then we need to make some amends such that we can legally cache
pre-constructed routes in the FIB nexthops. Firstly, we need to
invalidate routes which are hit with nexthop exceptions. Secondly we
have to change the semantics of rt->rt_gateway such that zero means
that the destination is on-link and non-zero otherwise.
Now that the preparations are ready, we start caching precomputed
routes in the FIB nexthops. Output and input routes need different
kinds of care when determining if we can legally do such caching or
not. The details are in the commit log messages for those changes.
The patch series then winds down with some more struct rtable
simplifications and other tidy ups that remove unnecessary overhead.
On a SPARC-T3 output route lookups are ~876 cycles. Input route
lookups are ~1169 cycles with rpfilter disabled, and about ~1468
cycles with rpfilter enabled.
These measurements were taken with the kbench_mod test module in the
net_test_tools GIT tree:
git://git.kernel.org/pub/scm/linux/kernel/git/davem/net_test_tools.git
That GIT tree also includes a udpflood tester tool and stresses
route lookups on packet output.
For example, on the same SPARC-T3 system we can run:
time ./udpflood -l 10000000 10.2.2.11
with routing cache:
real 1m21.955s user 0m6.530s sys 1m15.390s
without routing cache:
real 1m31.678s user 0m6.520s sys 1m25.140s
Performance undoubtedly can easily be improved further.
For example fib_table_lookup() performs a lot of excessive
computations with all the masking and shifting, some of it
conditionalized to deal with edge cases.
Also, Eric's no-ref optimization for input route lookups can be
re-instated for the FIB nexthop caching code path. I would be really
pleased if someone would work on that.
In fact anyone suitable motivated can just fire up perf on the loading
of the test net_test_tools benchmark kernel module. I spend much of
my time going:
bash# perf record insmod ./kbench_mod.ko dst=172.30.42.22 src=74.128.0.1 iif=2
bash# perf report
Thanks to helpful feedback from Joe Perches, Eric Dumazet, Ben
Hutchings, and others.
Signed-off-by: David S. Miller <davem@davemloft.net>
Set unicast_sock uc_ttl to -1 so that we select the right ttl,
instead of sending packets with a 0 ttl.
Bug added in commit be9f4a44e7 (ipv4: tcp: remove per net tcp_sock)
Signed-off-by: Hiroaki SHIMODA <shimoda.hiroaki@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It's not really needed.
We only grabbed a reference to the fib_info for the sake of fib_info
local metrics.
However, fib_info objects are freed using RCU, as are therefore their
private metrics (if any).
We would have triggered a route cache flush if we eliminated a
reference to a fib_info object in the routing tables.
Therefore, any existing cached routes will first check and see that
they have been invalidated before an errant reference to these
metric values would occur.
Signed-off-by: David S. Miller <davem@davemloft.net>
That is this value's only use, as a boolean to indicate whether
a route is an input route or not.
So implement it that way, using a u16 gap present in the struct
already.
Signed-off-by: David S. Miller <davem@davemloft.net>
Never actually used.
It was being set on output routes to the original OIF specified in the
flow key used for the lookup.
Adjust the only user, ipmr_rt_fib_lookup(), for greater correctness of
the flowi4_oif and flowi4_iif values, thanks to feedback from Julian
Anastasov.
Signed-off-by: David S. Miller <davem@davemloft.net>
Don't bother incrementing dst->__use and setting dst->lastuse,
they are completely pointless and just slow things down.
Signed-off-by: David S. Miller <davem@davemloft.net>
Caching input routes is slightly simpler than output routes, since we
don't need to be concerned with nexthop exceptions. (locally
destined, and routed packets, never trigger PMTU events or redirects
that will be processed by us).
However, we have to elide caching for the DIRECTSRC and non-zero itag
cases.
Signed-off-by: David S. Miller <davem@davemloft.net>
If we have an output route that lacks nexthop exceptions, we can cache
it in the FIB info nexthop.
Such routes will have DST_HOST cleared because such routes refer to a
family of destinations, rather than just one.
The sequence of the handling of exceptions during route lookup is
adjusted to make the logic work properly.
Before we allocate the route, we lookup the exception.
Then we know if we will cache this route or not, and therefore whether
DST_HOST should be set on the allocated route.
Then we use DST_HOST to key off whether we should store the resulting
route, during rt_set_nexthop(), in the FIB nexthop cache.
With help from Eric Dumazet.
Signed-off-by: David S. Miller <davem@davemloft.net>
Add a big comment explaining how the field works, and use defines
instead of magic constants for the values assigned to it.
Suggested by Joe Perches.
Signed-off-by: David S. Miller <davem@davemloft.net>
In order to allow prefixed routes, we have to adjust how rt_gateway
is set and interpreted.
The new interpretation is:
1) rt_gateway == 0, destination is on-link, nexthop is iph->daddr
2) rt_gateway != 0, destination requires a nexthop gateway
Abstract the fetching of the proper nexthop value using a new
inline helper, rt_nexthop(), as suggested by Joe Perches.
Signed-off-by: David S. Miller <davem@davemloft.net>
Tested-by: Vijay Subramanian <subramanian.vijay@gmail.com>
They are always used in contexts where they can be reconstituted,
or where the finally resolved rt->rt_{src,dst} is semantically
equivalent.
Signed-off-by: David S. Miller <davem@davemloft.net>
The "noref" argument to ip_route_input_common() is now always ignored
because we do not cache routes, and in that case we must always grab
a reference to the resulting 'dst'.
Signed-off-by: David S. Miller <davem@davemloft.net>
The ipv4 routing cache is non-deterministic, performance wise, and is
subject to reasonably easy to launch denial of service attacks.
The routing cache works great for well behaved traffic, and the world
was a much friendlier place when the tradeoffs that led to the routing
cache's design were considered.
What it boils down to is that the performance of the routing cache is
a product of the traffic patterns seen by a system rather than being a
product of the contents of the routing tables. The former of which is
controllable by external entitites.
Even for "well behaved" legitimate traffic, high volume sites can see
hit rates in the routing cache of only ~%10.
Signed-off-by: David S. Miller <davem@davemloft.net>
Modern TCP stack highly depends on tcp_write_timer() having a small
latency, but current implementation doesn't exactly meet the
expectations.
When a timer fires but finds the socket is owned by the user, it rearms
itself for an additional delay hoping next run will be more
successful.
tcp_write_timer() for example uses a 50ms delay for next try, and it
defeats many attempts to get predictable TCP behavior in term of
latencies.
Use the recently introduced tcp_release_cb(), so that the user owning
the socket will call various handlers right before socket release.
This will permit us to post a followup patch to address the
tcp_tso_should_defer() syndrome (some deferred packets have to wait
RTO timer to be transmitted, while cwnd should allow us to send them
sooner)
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Cc: H.K. Jerry Chu <hkchu@google.com>
Cc: John Heffner <johnwheffner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When/if sysctl_tcp_abc > 1, we expect to increase cwnd by 2 if the
received ACK acknowledges more than 2*MSS bytes, in tcp_slow_start()
Problem is this RFC 3465 statement is not correctly coded, as
the while () loop increases snd_cwnd one by one.
Add a new variable to avoid this off-by one error.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tom Herbert <therbert@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Nandita Dukkipati <nanditad@google.com>
Cc: John Heffner <johnwheffner@gmail.com>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix a missing roundup_pow_of_two(), since tcpmhash_entries is not
guaranteed to be a power of two.
Uses hash_32() instead of custom hash.
tcpmhash_entries should be an unsigned int.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Applied to a set of static inline functions in tcp_input.c
Signed-off-by: Vijay Subramanian <subramanian.vijay@gmail.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix again the diff value in rt_bind_exception
after collision of two latest patches, my original commit
actually fixed the same problem.
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
In trusted networks, e.g., intranet, data-center, the client does not
need to use Fast Open cookie to mitigate DoS attacks. In cookie-less
mode, sendmsg() with MSG_FASTOPEN flag will send SYN-data regardless
of cookie availability.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On paths with firewalls dropping SYN with data or experimental TCP options,
Fast Open connections will have experience SYN timeout and bad performance.
The solution is to track such incidents in the cookie cache and disables
Fast Open temporarily.
Since only the original SYN includes data and/or Fast Open option, the
SYN-ACK has some tell-tale sign (tcp_rcv_fastopen_synack()) to detect
such drops. If a path has recurring Fast Open SYN drops, Fast Open is
disabled for 2^(recurring_losses) minutes starting from four minutes up to
roughly one and half day. sendmsg with MSG_FASTOPEN flag will succeed but
it behaves as connect() then write().
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sendmsg() (or sendto()) with MSG_FASTOPEN is a combo of connect(2)
and write(2). The application should replace connect() with it to
send data in the opening SYN packet.
For blocking socket, sendmsg() blocks until all the data are buffered
locally and the handshake is completed like connect() call. It
returns similar errno like connect() if the TCP handshake fails.
For non-blocking socket, it returns the number of bytes queued (and
transmitted in the SYN-data packet) if cookie is available. If cookie
is not available, it transmits a data-less SYN packet with Fast Open
cookie request option and returns -EINPROGRESS like connect().
Using MSG_FASTOPEN on connecting or connected socket will result in
simlar errno like repeating connect() calls. Therefore the application
should only use this flag on new sockets.
The buffer size of sendmsg() is independent of the MSS of the connection.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
On receiving the SYN-ACK after SYN-data, the client needs to
a) update the cached MSS and cookie (if included in SYN-ACK)
b) retransmit the data not yet acknowledged by the SYN-ACK in the final ACK of
the handshake.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch implements sending SYN-data in tcp_connect(). The data is
from tcp_sendmsg() with flag MSG_FASTOPEN (implemented in a later patch).
The length of the cookie in tcp_fastopen_req, init'd to 0, controls the
type of the SYN. If the cookie is not cached (len==0), the host sends
data-less SYN with Fast Open cookie request option to solicit a cookie
from the remote. If cookie is not available (len > 0), the host sends
a SYN-data with Fast Open cookie option. If cookie length is negative,
the SYN will not include any Fast Open option (for fall back operations).
To deal with middleboxes that may drop SYN with data or experimental TCP
option, the SYN-data is only sent once. SYN retransmits do not include
data or Fast Open options. The connection will fall back to regular TCP
handshake.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With help from Eric Dumazet, add Fast Open metrics in tcp metrics cache.
The basic ones are MSS and the cookies. Later patch will cache more to
handle unfriendly middleboxes.
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch impelements the common code for both the client and server.
1. TCP Fast Open option processing. Since Fast Open does not have an
option number assigned by IANA yet, it shares the experiment option
code 254 by implementing draft-ietf-tcpm-experimental-options
with a 16 bits magic number 0xF989. This enables global experiments
without clashing the scarce(2) experimental options available for TCP.
When the draft status becomes standard (maybe), the client should
switch to the new option number assigned while the server supports
both numbers for transistion.
2. The new sysctl tcp_fastopen
3. A place holder init function
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_v4_send_reset() and tcp_v4_send_ack() use a single socket
per network namespace.
This leads to bad behavior on multiqueue NICS, because many cpus
contend for the socket lock and once socket lock is acquired, extra
false sharing on various socket fields slow down the operations.
To better resist to attacks, we use a percpu socket. Each cpu can
run without contention, using appropriate memory (local node)
Additional features :
1) We also mirror the queue_mapping of the incoming skb, so that
answers use the same queue if possible.
2) Setting SOCK_USE_WRITE_QUEUE socket flag speedup sock_wfree()
3) We now limit the number of in-flight RST/ACK [1] packets
per cpu, instead of per namespace, and we honor the sysctl_wmem_default
limit dynamically. (Prior to this patch, sysctl_wmem_default value was
copied at boot time, so any further change would not affect tcp_sock
limit)
[1] These packets are only generated when no socket was matched for
the incoming packet.
Reported-by: Bill Sommerfeld <wsommerfeld@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use global seqlock for the nh_exceptions. Call
fnhe_oldest with the right hash chain. Correct the diff
value for dst_set_expires.
v2: after suggestions from Eric Dumazet:
* get rid of spin lock fnhe_lock, rearrange update_or_create_fnhe
* continue daddr search in rt_bind_exception
v3:
* remove the daddr check before seqlock in rt_bind_exception
* restart lookup in rt_bind_exception on detected seqlock change,
as suggested by David Miller
Signed-off-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: David S. Miller <davem@davemloft.net>
Andrew Morton