The earlier version was just very basic one which is "playing
safe" by always clearing the hints. However, clearing of a hint
is extremely costly operation with large windows, so it must be
avoided at all cost whenever possible, there is a way with
shifting too achieve not-clearing.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
During SACK processing, most of the benefits of TSO are eaten by
the SACK blocks that one-by-one fragment SKBs to MSS sized chunks.
Then we're in problems when cleanup work for them has to be done
when a large cumulative ACK comes. Try to return back to pre-split
state already while more and more SACK info gets discovered by
combining newly discovered SACK areas with the previous skb if
that's SACKed as well.
This approach has a number of benefits:
1) The processing overhead is spread more equally over the RTT
2) Write queue has less skbs to process (affect everything
which has to walk in the queue past the sacked areas)
3) Write queue is consistent whole the time, so no other parts
of TCP has to be aware of this (this was not the case with
some other approach that was, well, quite intrusive all
around).
4) Clean_rtx_queue can release most of the pages using single
put_page instead of previous PAGE_SIZE/mss+1 calls
In case a hole is fully filled by the new SACK block, we attempt
to combine the next skb too which allows construction of skbs
that are even larger than what tso split them to and it handles
hole per on every nth patterns that often occur during slow start
overshoot pretty nicely. Though this to be really useful also
a retransmission would have to get lost since cumulative ACKs
advance one hole at a time in the most typical case.
TODO: handle upwards only merging. That should be rather easy
when segment is fully sacked but I'm leaving that as future
work item (it won't make very large difference anyway since
this current approach already covers quite a lot of normal
cases).
I was earlier thinking of some sophisticated way of tracking
timestamps of the first and the last segment but later on
realized that it won't be that necessary at all to store the
timestamp of the last segment. The cases that can occur are
basically either:
1) ambiguous => no sensible measurement can be taken anyway
2) non-ambiguous is due to reordering => having the timestamp
of the last segment there is just skewing things more off
than does some good since the ack got triggered by one of
the holes (besides some substle issues that would make
determining right hole/skb even harder problem). Anyway,
it has nothing to do with this change then.
I choose to route some abnormal looking cases with goto noop,
some could be handled differently (eg., by stopping the
walking at that skb but again). In general, they either
shouldn't happen at all or are rare enough to make no difference
in practice.
In theory this change (as whole) could cause some macroscale
regression (global) because of cache misses that are taken over
the round-trip time but it gets very likely better because of much
less (local) cache misses per other write queue walkers and the
big recovery clearing cumulative ack.
Worth to note that these benefits would be very easy to get also
without TSO/GSO being on as long as the data is in pages so that
we can merge them. Currently I won't let that happen because
DSACK splitting at fragment that would mess up pcounts due to
sk_can_gso in tcp_set_skb_tso_segs. Once DSACKs fragments gets
avoided, we have some conditions that can be made less strict.
TODO: I will probably have to convert the excessive pointer
passing to struct sacktag_state... :-)
My testing revealed that considerable amount of skbs couldn't
be shifted because they were cloned (most likely still awaiting
tx reclaim)...
[The rest is considering future work instead since I got
repeatably EFAULT to tcpdump's recvfrom when I added
pskb_expand_head to deal with clones, so I separated that
into another, later patch]
...To counter that, I gave up on the fifth advantage:
5) When growing previous SACK block, less allocs for new skbs
are done, basically a new alloc is needed only when new hole
is detected and when the previous skb runs out of frags space
...which now only happens of if reclaim is fast enough to dispose
the clone before the SACK block comes in (the window is RTT long),
otherwise we'll have to alloc some.
With clones being handled I got these numbers (will be somewhat
worse without that), taken with fine-grained mibs:
TCPSackShifted 398
TCPSackMerged 877
TCPSackShiftFallback 320
TCPSACKCOLLAPSEFALLBACKGSO 0
TCPSACKCOLLAPSEFALLBACKSKBBITS 0
TCPSACKCOLLAPSEFALLBACKSKBDATA 0
TCPSACKCOLLAPSEFALLBACKBELOW 0
TCPSACKCOLLAPSEFALLBACKFIRST 1
TCPSACKCOLLAPSEFALLBACKPREVBITS 318
TCPSACKCOLLAPSEFALLBACKMSS 1
TCPSACKCOLLAPSEFALLBACKNOHEAD 0
TCPSACKCOLLAPSEFALLBACKSHIFT 0
TCPSACKCOLLAPSENOOPSEQ 0
TCPSACKCOLLAPSENOOPSMALLPCOUNT 0
TCPSACKCOLLAPSENOOPSMALLLEN 0
TCPSACKCOLLAPSEHOLE 12
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is preparatory work for SACK combiner patch which may
have to count TCP state changes for only a part of the skb
because it will intentionally avoids splitting skb to SACKed
and not sacked parts.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Sadly enough, this adds possible divide though we try to avoid
it by checking one mss as common case.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
I knew already when rewriting the sacktag that this condition
was too conservative, change it now since it prevent lot of
useless work (especially in the sack shifter decision code
that is being added by a later patch). This shouldn't change
anything really, just save some processing regardless of the
shifter.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
I always had thought that collapsing up to two at a time was
intentional decision to avoid excessive processing if 1 byte
sized skbs are to be combined for a full mtu, and consecutive
retransmissions would make the size of the retransmittee
double each round anyway, but some recent discussion made me
to understand that was not the case. Thus make collapse work
more and wait less.
It would be possible to take advantage of the shifting
machinery (added in the later patch) in the case of paged
data but that can be implemented on top of this change.
tcp_skb_is_last check is now provided by the loop.
I tested a bit (ss-after-idle-off, fill 4096x4096B xfer,
10s sleep + 4096 x 1byte writes while dropping them for
some a while with netem):
. 16774097:16775545(1448) ack 1 win 46
. 16775545:16776993(1448) ack 1 win 46
. ack 16759617 win 2399
P 16776993:16777217(224) ack 1 win 46
. ack 16762513 win 2399
. ack 16765409 win 2399
. ack 16768305 win 2399
. ack 16771201 win 2399
. ack 16774097 win 2399
. ack 16776993 win 2399
. ack 16777217 win 2399
P 16777217:16777257(40) ack 1 win 46
. ack 16777257 win 2399
P 16777257:16778705(1448) ack 1 win 46
P 16778705:16780153(1448) ack 1 win 46
FP 16780153:16781313(1160) ack 1 win 46
. ack 16778705 win 2399
. ack 16780153 win 2399
F 1:1(0) ack 16781314 win 2399
While without drop-all period I get this:
. 16773585:16775033(1448) ack 1 win 46
. ack 16764897 win 9367
. ack 16767793 win 9367
. ack 16770689 win 9367
. ack 16773585 win 9367
. 16775033:16776481(1448) ack 1 win 46
P 16776481:16777217(736) ack 1 win 46
. ack 16776481 win 9367
. ack 16777217 win 9367
P 16777217:16777218(1) ack 1 win 46
P 16777218:16777219(1) ack 1 win 46
P 16777219:16777220(1) ack 1 win 46
...
P 16777247:16777248(1) ack 1 win 46
. ack 16777218 win 9367
. ack 16777219 win 9367
...
. ack 16777233 win 9367
. ack 16777248 win 9367
P 16777248:16778696(1448) ack 1 win 46
P 16778696:16780144(1448) ack 1 win 46
FP 16780144:16781313(1169) ack 1 win 46
. ack 16780144 win 9367
F 1:1(0) ack 16781314 win 9367
The window seems to be 30-40 segments, which were successfully
combined into: P 16777217:16777257(40) ack 1 win 46
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can reduce pressure on dst entry refcount that slowdown UDP transmit
path on SMP machines. This pressure is visible on RTP servers when
delivering content to mediagateways, especially big ones, handling
thousand of streams. Several cpus send UDP frames to the same
destination, hence use the same dst entry.
This patch makes ip_push_pending_frames() steal the refcount its
callers had to take when filling inet->cork.dst.
This doesnt avoid all refcounting, but still gives speedups on SMP,
on UDP/RAW transmit path.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can reduce pressure on dst entry refcount that slowdown UDP transmit
path on SMP machines. This pressure is visible on RTP servers when
delivering content to mediagateways, especially big ones, handling
thousand of streams. Several cpus send UDP frames to the same
destination, hence use the same dst entry.
This patch makes ip_append_data() eventually steal the refcount its
callers had to take on the dst entry.
This doesnt avoid all refcounting, but still gives speedups on SMP,
on UDP/RAW transmit path
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The rule of calling sock_prot_inuse_add() is that BHs must
be disabled. Some new calls were added where this was not
true and this tiggers warnings as reported by Ilpo.
Fix this by adding explicit BH disabling around those call sites,
or moving sock_prot_inuse_add() call inside an existing BH disabled
section.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is the last step to be able to perform full RCU lookups
in __inet_lookup() : After established/timewait tables, we
add RCU lookups to listening hash table.
The only trick here is that a socket of a given type (TCP ipv4,
TCP ipv6, ...) can now flight between two different tables
(established and listening) during a RCU grace period, so we
must use different 'nulls' end-of-chain values for two tables.
We define a large value :
#define LISTENING_NULLS_BASE (1U << 29)
So that slots in listening table are guaranteed to have different
end-of-chain values than slots in established table. A reader can
still detect it finished its lookup in the right chain.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1) Use eq_net() in inet_netns_ok() to speedup socket creation if
!CONFIG_NET_NS
2) Reorder the tests about inet_ehash_secret generation (once only)
Use the unlikely() macro when testing if inet_ehash_secret already
generated.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now TCP & DCCP use RCU lookups, we can convert ehash rwlocks to spinlocks.
/proc/net/tcp and other seq_file 'readers' can safely be converted to 'writers'.
This should speedup writers, since spin_lock()/spin_unlock()
only use one atomic operation instead of two for write_lock()/write_unlock()
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert to network device ops. Needed to change to directly call
the init routine since two sides share same ops. In the process
found by inspection a device ref count leak if register_netdevice failed.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert to new network device ops interface.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes sparse warnings:
net/ipv4/ip_sockglue.c:146:15: warning: incorrect type in assignment (different base types)
net/ipv4/ip_sockglue.c:146:15: expected restricted __be16 [assigned] [usertype] sin_port
net/ipv4/ip_sockglue.c:146:15: got unsigned short [unsigned] [short] [usertype] <noident>
net/ipv4/ip_sockglue.c:130:6: warning: symbol 'ip_cmsg_recv_dstaddr' was not declared. Should it be static?
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
inet_sk_rebuild_header() does a new route lookup if the dst_entry
associated with a socket becomes stale. However inet_sk_rebuild_header()
didn't use struct flowi->flags, causing the route lookup to
fail for foreign-bound IP_TRANSPARENT sockets, causing an error
state to be set for the sockets in question.
Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
udp_sendmsg() didn't fill struct flowi->flags, which means that
the route lookup would fail for non-local IPs even if the
IP_TRANSPARENT sockopt was set.
This prevents sendto() to work properly for UDP sockets, whereas
bind(foreign-ip) + connect() + send() worked fine.
Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch prepares RCU migration of listening_hash table for
TCP/DCCP protocols.
listening_hash table being small (32 slots per protocol), we add
a spinlock for each slot, instead of a single rwlock for whole table.
This should reduce hold time of readers, and writers concurrency.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Convert to net_device_ops function table pointer for ioctl.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The first argument to csum_partial is const void *
casts to char/u8 * are not necessary
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Similarly to IPv6 ip6_mr_init() (fixed last week), the order of cleanup
operations in the error/exit section of ip_mr_init() is completely
inversed. It should be the other way around.
Also a del_timer() is missing in the error path.
I should have guessed last week that this same error existed in ipmr.c
too, as ip6mr.c is largely inspired by ipmr.c.
Signed-off-by: Benjamin Thery <benjamin.thery@bull.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
RCU was added to UDP lookups, using a fast infrastructure :
- sockets kmem_cache use SLAB_DESTROY_BY_RCU and dont pay the
price of call_rcu() at freeing time.
- hlist_nulls permits to use few memory barriers.
This patch uses same infrastructure for TCP/DCCP established
and timewait sockets.
Thanks to SLAB_DESTROY_BY_RCU, no slowdown for applications
using short lived TCP connections. A followup patch, converting
rwlocks to spinlocks will even speedup this case.
__inet_lookup_established() is pretty fast now we dont have to
dirty a contended cache line (read_lock/read_unlock)
Only established and timewait hashtable are converted to RCU
(bind table and listen table are still using traditional locking)
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a straightforward patch, using hlist_nulls infrastructure.
RCUification already done on UDP two weeks ago.
Using hlist_nulls permits us to avoid some memory barriers, both
at lookup time and delete time.
Patch is large because it adds new macros to include/net/sock.h.
These macros will be used by TCP & DCCP in next patch.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case UDP traffic is redirected to a local UDP socket,
the originally addressed destination address/port
cannot be recovered with the in-kernel tproxy.
This patch adds an IP_RECVORIGDSTADDR sockopt that enables
a IP_ORIGDSTADDR ancillary message in recvmsg(). This
ancillary message contains the original destination address/port
of the packet being received.
Signed-off-by: Balazs Scheidler <bazsi@balabit.hu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Ben Greear wrote:
> I have 500 mac-vlans on a system talking to 500 other
> mac-vlans. My problem is that the arp-table gets extremely
> huge because every time an arp-request comes in on all mac-vlans,
> a stale arp entry is added for each mac-vlan. I have filtering
> turned on, but that doesn't help because the neigh_event_ns call
> below will cause a stale neighbor entry to be created regardless
> of whether a replay will be sent or not.
> Maybe the neigh_event code should be below the checks for dont_send,
> and only create check neigh_event_ns if we are !dont_send?
The attached patch makes it work much better for me. The patch
will cause the code to NOT create a stale neighbor entry if we
are not going to respond to the ARP request. The old code
*would* create a stale entry even if we are not going to respond.
Signed-off-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Failure to pass netns_ok check is SILENT, except some MIB counter is
incremented somewhere.
And adding "netns_ok = 1" (after long head-scratching session) is
usually the last step in making some protocol netns-ready...
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch fixes a minor bug in tcp_htcp.c which has been
highlighted by Lachlan Andrew and Lawrence Stewart. Currently, the
time since the last congestion event, which is stored in variable
last_cong, is reset whenever there is a state change into
TCP_CA_Open. This includes transitions of the type
TCP_CA_Open->TCP_CA_Disorder->TCP_CA_Open which are not associated
with backoff of cwnd. The patch changes last_cong to be updated
only on transitions into TCP_CA_Open that occur after experiencing
the congestion-related states TCP_CA_Loss, TCP_CA_Recovery,
TCP_CA_CWR.
Signed-off-by: Doug Leith <doug.leith@nuim.ie>
Signed-off-by: David S. Miller <davem@davemloft.net>
We can shrink size of "struct inet_bind_bucket" by 50%, using
read_pnet() and write_pnet()
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
icmpmsg_put() can happily corrupt kernel memory, using a static
table and forgetting to reset an array index in a loop.
Remove the static array since its not safe without proper locking.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Vito Caputo noticed that tcp_recvmsg() returns immediately from
partial reads when MSG_PEEK is used. In particular, this means that
SO_RCVLOWAT is not respected.
Simply remove the test. And this matches the behavior of several
other systems, including BSD.
Signed-off-by: David S. Miller <davem@davemloft.net>
While adding MIGRATE support to strongSwan, Andreas Steffen noticed that
the selectors provided in XFRM_MSG_ACQUIRE have their family field
uninitialized (those in MIGRATE do have their family set).
Looking at the code, this is because the af-specific init_tempsel()
(called via afinfo->init_tempsel() in xfrm_init_tempsel()) do not set
the value.
Reported-by: Andreas Steffen <andreas.steffen@strongswan.org>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Arnaud Ebalard <arno@natisbad.org>
I want to compile out proc_* and sysctl_* handlers totally and
stub them to NULL depending on config options, however usage of &
will prevent this, since taking adress of NULL pointer will break
compilation.
So, drop & in front of every ->proc_handler and every ->strategy
handler, it was never needed in fact.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
