Do some simple changes to make congestion control API faster/cleaner.
* use ktime_t rather than timeval
* merge rtt sampling into existing ack callback
this means one indirect call versus two per ack.
* use flags bits to store options/settings
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Spring cleaning time...
There seems to be a lot of places in the network code that have
extra bogus semicolons after conditionals. Most commonly is a
bogus semicolon after: switch() { }
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
When a transmitted packet is looped back directly, CHECKSUM_PARTIAL
maps to the semantics of CHECKSUM_UNNECESSARY. Therefore we should
treat it as such in the stack.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the places where we need a pointer to the transport header, it is
still legal to touch skb->h.raw directly if just adding to,
subtracting from or setting it to another layer header.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the cases where the transport header is being set to a offset from
skb->data.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the cases where the network header is being set to a offset from skb->data.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the places where we need a pointer to the network header, it is still legal
to touch skb->nh.raw directly if just adding to, subtracting from or setting it
to another layer header.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
For the places where we need a pointer to the mac header, it is still legal to
touch skb->mac.raw directly if just adding to, subtracting from or setting it
to another layer header.
This one also converts some more cases to skb_reset_mac_header() that my
regex missed as it had no spaces before nor after '=', ugh.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
That is equal to skb->head before skb_reserve, to help in the layer header
changes.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add whitespace around keywords.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This allows the write queue implementation to be changed,
for example, to one which allows fast interval searching.
Signed-off-by: David S. Miller <davem@davemloft.net>
Where appropriate, convert references to xtime.tv_sec to the
get_seconds() helper function.
Signed-off-by: James Morris <jmorris@namei.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Undoing ssthresh is disabled in fastretrans_alert whenever
FLAG_ECE is set by clearing prior_ssthresh. The clearing does
not protect FRTO because FRTO operates before fastretrans_alert.
Moving the clearing of prior_ssthresh earlier seems to be a
suboptimal solution to the FRTO case because then FLAG_ECE will
cause a second ssthresh reduction in try_to_open (the first
occurred when FRTO was entered). So instead, FRTO falls back
immediately to the rate halving response, which switches TCP to
CA_CWR state preventing the latter reduction of ssthresh.
If the first ECE arrived before the ACK after which FRTO is able
to decide RTO as spurious, prior_ssthresh is already cleared.
Thus no undoing for ssthresh occurs. Besides, FLAG_ECE should be
set also in the following ACKs resulting in rate halving response
that sees TCP is already in CA_CWR, which again prevents an extra
ssthresh reduction on that round-trip.
If the first ECE arrived before RTO, ssthresh has already been
adapted and prior_ssthresh remains cleared on entry because TCP
is in CA_CWR (the same applies also to a case where FRTO is
entered more than once and ECE comes in the middle).
High_seq must not be touched after tcp_enter_cwr because CWR
round-trip calculation depends on it.
I believe that after this patch, FRTO should be ECN-safe and
even able to take advantage of synergy benefits.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
A local variable for icsk was created but this change was
missing. Spotted by Jarek Poplawski.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
New sysctl tcp_frto_response is added to select amongst these
responses:
- Rate halving based; reuses CA_CWR state (default)
- Very conservative; used to be the only one available (=1)
- Undo cwr; undoes ssthresh and cwnd reductions (=2)
The response with rate halving requires a new parameter to
tcp_enter_cwr because FRTO has already reduced ssthresh and
doing a second reduction there has to be prevented. In addition,
to keep things nice on 80 cols screen, a local variable was
added.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
The reordering detection must work also when FRTO has not been
used at all which was the original intention of mine, just the
expression of the idea was flawed.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implements the SACK-enhanced FRTO given in RFC4138 using the
variant given in Appendix B.
RFC4138, Appendix B:
"This means that in order to declare timeout spurious, the TCP
sender must receive an acknowledgment for non-retransmitted
segment between SND.UNA and RecoveryPoint in algorithm step 3.
RecoveryPoint is defined in conservative SACK-recovery
algorithm [RFC3517]"
The basic version of the FRTO algorithm can still be used also
when SACK is enabled. To enabled SACK-enhanced version, tcp_frto
sysctl is set to 2.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
To be honest, I'm not too sure how the reord stuff works in the
first place but this seems necessary.
When FRTO has been active, the one and only retransmission could
be unnecessary but the state and sending order might not be what
the sacktag code expects it to be (to work correctly).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP without FRTO would be in Loss state with small cwnd. FRTO,
however, leaves cwnd (typically) to a larger value which causes
ssthresh to become too large in case RTO is triggered again
compared to what conventional recovery would do. Because
consecutive RTOs result in only a single ssthresh reduction,
RTO+cumulative ACK+RTO pattern is required to trigger this
event.
A large comment is included for congestion control module writers
trying to figure out what CA_EVENT_FRTO handler should do because
there exists a remote possibility of incompatibility between
FRTO and module defined ssthresh functions.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Previously RETRANS bits were cleared on the entry to FRTO. We
postpone that into tcp_enter_frto_loss, which is really the
place were the clearing should be done anyway. This allows
simplification of the logic from a clearing loop to the head skb
clearing only.
Besides, the other changes made in the previous patches to
tcp_use_frto made it impossible for the non-SACKed FRTO to be
entered if other than the head has been rexmitted.
With SACK-enhanced FRTO (and Appendix B), however, there can be
a number retransmissions in flight when RTO expires (same thing
could happen before this patchset also with non-SACK FRTO). To
not introduce any jumpiness into the packet counting during FRTO,
instead of clearing RETRANS bits from skbs during entry, do it
later on.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
This interpretation comes from RFC4138:
"If the sender implements some loss recovery algorithm other
than Reno or NewReno [FHG04], the F-RTO algorithm SHOULD
NOT be entered when earlier fast recovery is underway."
I think the RFC means to say (especially in the light of
Appendix B) that ...recovery is underway (not just fast recovery)
or was underway when it was interrupted by an earlier (F-)RTO
that hasn't yet been resolved (snd_una has not advanced enough).
Thus, my interpretation is that whenever TCP has ever
retransmitted other than head, basic version cannot be used
because then the order assumptions which are used as FRTO basis
do not hold.
NewReno has only the head segment retransmitted at a time.
Therefore, walk up to the segment that has not been SACKed, if
that segment is not retransmitted nor anything before it, we know
for sure, that nothing after the non-SACKed segment should be
either. This assumption is valid because TCPCB_EVER_RETRANS does
not leave holes but each non-SACKed segment is rexmitted
in-order.
Check for retrans_out > 1 avoids more expensive walk through the
skb list, as we can know the result beforehand: F-RTO will not be
allowed.
SACKed skb can turn into non-SACked only in the extremely rare
case of SACK reneging, in this case we might fail to detect
retransmissions if there were them for any other than head. To
get rid of that feature, whole rexmit queue would have to be
walked (always) or FRTO should be prevented when SACK reneging
happens. Of course RTO should still trigger after reneging which
makes this issue even less likely to show up. And as long as the
response is as conservative as it's now, nothing bad happens even
then.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
FRTO controls cwnd when it still processes the ACK input or it
has just reverted back to conventional RTO recovery; the normal
rules apply when FRTO has reverted to standard congestion
control.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Because TCP is not in Loss state during FRTO recovery, fast
recovery could be triggered by accident. Non-SACK FRTO is more
robust than not yet included SACK-enhanced version (that can
receiver high number of duplicate ACKs with SACK blocks during
FRTO), at least with unidirectional transfers, but under
extraordinary patterns fast recovery can be incorrectly
triggered, e.g., Data loss+ACK losses => cumulative ACK with
enough SACK blocks to meet sacked_out >= dupthresh condition).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Since purpose is to reduce CWND, we prevent immediate growth. This
is not a major issue nor is "the correct way" specified anywhere.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
The FRTO detection did not care how ACK pattern affects to cwnd
calculation of the conventional recovery. This caused incorrect
setting of cwnd when the fallback becames necessary. The
knowledge tcp_process_frto() has about the incoming ACK is now
passed on to tcp_enter_frto_loss() in allowed_segments parameter
that gives the number of segments that must be added to
packets-in-flight while calculating the new cwnd.
Instead of snd_una we use FLAG_DATA_ACKED in duplicate ACK
detection because RFC4138 states (in Section 2.2):
If the first acknowledgment after the RTO retransmission
does not acknowledge all of the data that was retransmitted
in step 1, the TCP sender reverts to the conventional RTO
recovery. Otherwise, a malicious receiver acknowledging
partial segments could cause the sender to declare the
timeout spurious in a case where data was lost.
If the next ACK after RTO is duplicate, we do not retransmit
anything, which is equal to what conservative conventional
recovery does in such case.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Handles RFC4138 shortcoming (in step 2); it should also have case
c) which ignores ACKs that are not duplicates nor advance window
(opposite dir data, winupdate).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Retransmission counter assumptions are to be changed. Forcing
reason to do this exist: Using sysctl in check would be racy
as soon as FRTO starts to ignore some ACKs (doing that in the
following patches). Userspace may disable it at any moment
giving nice oops if timing is right. frto_counter would be
inaccessible from userspace, but with SACK enhanced FRTO
retrans_out can include other than head, and possibly leaving
it non-zero after spurious RTO, boom again.
Luckily, solution seems rather simple: never go directly to Open
state but use Disorder instead. This does not really change much,
since TCP could anyway change its state to Disorder during FRTO
using path tcp_fastretrans_alert -> tcp_try_to_open (e.g., when
a SACK block makes ACK dubious). Besides, Disorder seems to be
the state where TCP should be if not recovering (in Recovery or
Loss state) while having some retransmissions in-flight (see
tcp_try_to_open), which is exactly what happens with FRTO.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
In case a latency spike causes more than one RTO, the later should not
cause the already reduced ssthresh to propagate into the prior_ssthresh
since FRTO declares all such RTOs spurious at once or none of them. In
treating of ssthresh, we mimic what tcp_enter_loss() does.
The previous state (in frto_counter) must be available until we have
checked it in tcp_enter_frto(), and also ACK information flag in
process_frto().
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Moved comments out from the body of process_frto() to the head
(preferred way; see Documentation/CodingStyle). Bonus: it's much
easier to read in this compacted form.
FRTO algorithm and implementation is described in greater detail.
For interested reader, more information is available in RFC4138.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
FRTO spurious RTO detection algorithm (RFC4138) does not include response
to a detected spurious RTO but can use different response algorithms.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
FRTO was slightly too brave... Should only clear
TCPCB_SACKED_RETRANS bit.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
We clear the unused parts of the SACK cache, This prevents us from mistakenly
taking the cache data if the old data in the SACK cache is the same as the data
in the SACK block. This assumes that we never receive an empty SACK block with
start and end both at zero.
Signed-off-by: Baruch Even <baruch@ev-en.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move DSACK code outside the SACK fast-path checking code. If the DSACK
determined that the information was too old we stayed with a partial cache
copied. Most likely this matters very little since the next packet will not be
DSACK and we will find it in the cache. but it's still not good form and there
is little reason to couple the two checks.
Since the SACK receive cache doesn't need the data to be in host order we also
remove the ntohl in the checking loop.
Signed-off-by: Baruch Even <baruch@ev-en.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only advance the SACK fast-path pointer for the first block, the
fast-path assumes that only the first block advances next time so we
should not move the cached skb for the next sack blocks.
Signed-off-by: Baruch Even <baruch@ev-en.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
The sorting of SACK blocks actually munges them rather than sort,
causing the TCP stack to ignore some SACK information and breaking the
assumption of ordered SACK blocks after sorting.
The sort takes the data from a second buffer which isn't moved causing
subsequent data moves to occur from the wrong location. The fix is to
use a temporary buffer as a normal sort does.
Signed-off-By: Baruch Even <baruch@ev-en.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
I encountered a kernel panic with my test program, which is a very
simple IPv6 client-server program.
The server side sets IPV6_RECVPKTINFO on a listening socket, and the
client side just sends a message to the server. Then the kernel panic
occurs on the server. (If you need the test program, please let me
know. I can provide it.)
This problem happens because a skb is forcibly freed in
tcp_rcv_state_process().
When a socket in listening state(TCP_LISTEN) receives a syn packet,
then tcp_v6_conn_request() will be called from
tcp_rcv_state_process(). If the tcp_v6_conn_request() successfully
returns, the skb would be discarded by __kfree_skb().
However, in case of a listening socket which was already set
IPV6_RECVPKTINFO, an address of the skb will be stored in
treq->pktopts and a ref count of the skb will be incremented in
tcp_v6_conn_request(). But, even if the skb is still in use, the skb
will be freed. Then someone still using the freed skb will cause the
kernel panic.
I suggest to use kfree_skb() instead of __kfree_skb().
Signed-off-by: Masayuki Nakagawa <nakagawa.msy@ncos.nec.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
I believe all the below memory barriers only matter on SMP so
therefore the smp_* variant of the barrier should be used.
I'm wondering if the barrier in net/ipv4/inet_timewait_sock.c should be
dropped entirely. schedule_work's implementation currently implies a
memory barrier and I think sane semantics of schedule_work() should imply
a memory barrier, as needed so the caller shouldn't have to worry.
It's not quite obvious why the barrier in net/packet/af_packet.c is
needed; maybe it should be implied through flush_dcache_page?
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix SO_PEERSEC for tcp sockets to return the security context of
the peer (as represented by the SA from the peer) as opposed to the
SA used by the local/source socket.
Signed-off-by: Venkat Yekkirala <vyekkirala@TrustedCS.com>
Signed-off-by: James Morris <jmorris@namei.org>
GCC can't tell we always initialize 'tv' in all the cases
we actually use it, so explicitly set it up with zeros.
Signed-off-by: David S. Miller <davem@davemloft.net>
This changes the microsecond RTT sampling so that samples are taken in
the same way that RTT samples are taken for the RTO calculator: on the
last segment acknowledged, and only when the segment hasn't been
retransmitted.
Signed-off-by: John Heffner <jheffner@psc.edu>
Acked-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some of the instances of tcp_sack_block are host-endian, some - net-endian.
Define struct tcp_sack_block_wire identical to struct tcp_sack_block
with u32 replaced with __be32; annotate uses of tcp_sack_block replacing
net-endian ones with tcp_sack_block_wire. Change is obviously safe since
for cc(1) __be32 is typedefed to u32.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
It does not affect either mss-sized connections (obviously) or
connections controlled by Nagle (because there is only one small
segment in flight).
The idea is to record the fact that a small segment arrives on a
connection, where one small segment has already been received and
still not-ACKed. In this case ACK is forced after tcp_recvmsg() drains
receive buffer.
In other words, it is a "soft" each-2nd-segment ACK, which is enough
to preserve ACK clock even when ABC is enabled.
Signed-off-by: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
By passing a Linux-generated TSO packet straight back into Linux, Xen
becomes our first LRO user :) Unfortunately, there is at least one spot
in our stack that needs to be changed to cope with this.
The receive MSS estimate is computed from the raw packet size. This is
broken if the packet is GSO/LRO. Fortunately the real MSS can be found
in gso_size so we simply need to use that if it is non-zero.
Real LRO NICs should of course set the gso_size field in future.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Change net/core, ipv4 and ipv6 sysctl variables to __read_mostly.
Couldn't actually measure any performance increase while testing (.3%
I consider noise), but seems like the right thing to do.
Signed-off-by: Brian Haley <brian.haley@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Turn Appropriate Byte Count off by default because it unfairly
penalizes applications that do small writes. Add better documentation
to describe what it is so users will understand why they might want to
turn it on.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
1) fix slow start after retransmit timeout
2) fix case of L=2*SMSS acked bytes comparison
Signed-off-by: Daikichi Osuga <osugad@s1.nttdocomo.co.jp>
Signed-off-by: David S. Miller <davem@davemloft.net>
Whenever a transfer is application limited, we are allowed at least
initial window worth of data per window unless cwnd is previously
less than that.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the current TSO implementation, NETIF_F_TSO and ECN cannot be
turned on together in a TCP connection. The problem is that most
hardware that supports TSO does not handle CWR correctly if it is set
in the TSO packet. Correct handling requires CWR to be set in the
first packet only if it is set in the TSO header.
This patch adds the ability to turn on NETIF_F_TSO and ECN using
GSO if necessary to handle TSO packets with CWR set. Hardware
that handles CWR correctly can turn on NETIF_F_TSO_ECN in the dev->
features flag.
All TSO packets with CWR set will have the SKB_GSO_TCPV4_ECN set. If
the output device does not have the NETIF_F_TSO_ECN feature set, GSO
will split the packet up correctly with CWR only set in the first
segment.
With help from Herbert Xu <herbert@gondor.apana.org.au>.
Since ECN can always be enabled with TSO, the SOCK_NO_LARGESEND sock
flag is completely removed.
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Having separate fields in sk_buff for TSO/UFO (tso_size/ufo_size) is not
going to scale if we add any more segmentation methods (e.g., DCCP). So
let's merge them.
They were used to tell the protocol of a packet. This function has been
subsumed by the new gso_type field. This is essentially a set of netdev
feature bits (shifted by 16 bits) that are required to process a specific
skb. As such it's easy to tell whether a given device can process a GSO
skb: you just have to and the gso_type field and the netdev's features
field.
I've made gso_type a conjunction. The idea is that you have a base type
(e.g., SKB_GSO_TCPV4) that can be modified further to support new features.
For example, if we add a hardware TSO type that supports ECN, they would
declare NETIF_F_TSO | NETIF_F_TSO_ECN. All TSO packets with CWR set would
have a gso_type of SKB_GSO_TCPV4 | SKB_GSO_TCPV4_ECN while all other TSO
packets would be SKB_GSO_TCPV4. This means that only the CWR packets need
to be emulated in software.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Many of the TCP congestion methods all just use ssthresh
as the minimum congestion window on decrease. Rather than
duplicating the code, just have that be the default if that
handle in the ops structure is not set.
Minor behaviour change to TCP compound. It probably wants
to use this (ssthresh) as lower bound, rather than ssthresh/2
because the latter causes undershoot on loss.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
We only want to take receive RTT mesaurements for data
bearing frames, here in the header prediction fast path
for a pure-sender, we know that we have a pure-ACK and
thus the checks in tcp_rcv_rtt_mesaure_ts() will not pass.
Signed-off-by: David S. Miller <davem@davemloft.net>
Locks down user pages and sets up for DMA in tcp_recvmsg, then calls
dma_async_try_early_copy in tcp_v4_do_rcv
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Aki M Nyrhinen <anyrhine@cs.helsinki.fi>
IMHO the current fix to the problem (in_flight underflow in reno)
is incorrect. it treats the symptons but ignores the problem. the
problem is timing out packets other than the head packet when we
don't have sack. i try to explain (sorry if explaining the obvious).
with sack, scanning the retransmit queue for timed out packets is
fine because we know which packets in our retransmit queue have been
acked by the receiver.
without sack, we know only how many packets in our retransmit queue the
receiver has acknowledged, but no idea which packets.
think of a "typical" slow-start overshoot case, where for example
every third packet in a window get lost because a router buffer gets
full.
with sack, we check for timeouts on those every third packet (as the
rest have been sacked). the packet counting works out and if there
is no reordering, we'll retransmit exactly the packets that were
lost.
without sack, however, we check for timeout on every packet and end up
retransmitting consecutive packets in the retransmit queue. in our
slow-start example, 2/3 of those retransmissions are unnecessary. these
unnecessary retransmissions eat the congestion window and evetually
prevent fast recovery from continuing, if enough packets were lost.
Signed-off-by: David S. Miller <davem@davemloft.net>
From: "Angelo P. Castellani" <angelo.castellani+lkml@gmail.com>
Using NewReno, if a sk_buff is timed out and is accounted as lost_out,
it should also be removed from the sacked_out.
This is necessary because recovery using NewReno fast retransmit could
take up to a lot RTTs and the sk_buff RTO can expire without actually
being really lost.
left_out = sacked_out + lost_out
in_flight = packets_out - left_out + retrans_out
Using NewReno without this patch, on very large network losses,
left_out becames bigger than packets_out + retrans_out (!!).
For this reason unsigned integer in_flight overflows to 2^32 - something.
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch contains the following possible cleanups:
- make the following needlessly global function static:
- arp.c: arp_rcv()
- remove the following unused EXPORT_SYMBOL's:
- devinet.c: devinet_ioctl
- fib_frontend.c: ip_rt_ioctl
- inet_hashtables.c: inet_bind_bucket_create
- inet_hashtables.c: inet_bind_hash
- tcp_input.c: sysctl_tcp_abc
- tcp_ipv4.c: sysctl_tcp_tw_reuse
- tcp_output.c: sysctl_tcp_mtu_probing
- tcp_output.c: sysctl_tcp_base_mss
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
This moves some TCP-specific MTU probing state out of
inet_connection_sock back to tcp_sock.
Signed-off-by: John Heffner <jheffner@psc.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implementation of packetization layer path mtu discovery for TCP, based on
the internet-draft currently found at
<http://www.ietf.org/internet-drafts/draft-ietf-pmtud-method-05.txt>.
Signed-off-by: John Heffner <jheffner@psc.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
This changes some simple "if (x) BUG();" statements to "BUG_ON(x);"
Signed-off-by: Kris Katterjohn <kjak@users.sourceforge.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP inline usage cleanup:
* get rid of inline in several places
* replace __inline__ with inline where possible
* move functions used in one file out of tcp.h
* let compiler decide on used once cases
On x86_64:
text data bss dec hex filename
3594701 648348 567400 4810449 4966d1 vmlinux.orig
3593133 648580 567400 4809113 496199 vmlinux
On sparc64:
text data bss dec hex filename
2538278 406152 530392 3474822 350586 vmlinux.ORIG
2536382 406384 530392 3473158 34ff06 vmlinux
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
As DCCP needs to be called in the same spots.
Now we have a member in inet_sock (is_icsk), set at sock creation time from
struct inet_protosw->flags (if INET_PROTOSW_ICSK is set, like for TCP and
DCCP) to see if a struct sock instance is a inet_connection_sock for places
like the ones in ip_sockglue.c (v4 and v6) where we previously were looking if
sk_type was SOCK_STREAM, that is insufficient because we now use the same code
for DCCP, that has sk_type SOCK_DCCP.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
And move it to struct inet_connection_sock. DCCP will use it in the
upcoming changesets.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Use "hints" to speed up the SACK processing. Various forms
of this have been used by TCP developers (Web100, STCP, BIC)
to avoid the 2x linear search of outstanding segments.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is a patch for discussion addressing some receive buffer growing issues.
This is partially related to the thread "Possible BUG in IPv4 TCP window
handling..." last week.
Specifically it addresses the problem of an interaction between rcvbuf
moderation (receiver autotuning) and rcv_ssthresh. The problem occurs when
sending small packets to a receiver with a larger MTU. (A very common case I
have is a host with a 1500 byte MTU sending to a host with a 9k MTU.) In
such a case, the rcv_ssthresh code is targeting a window size corresponding
to filling up the current rcvbuf, not taking into account that the new rcvbuf
moderation may increase the rcvbuf size.
One hunk makes rcv_ssthresh use tcp_rmem[2] as the size target rather than
rcvbuf. The other changes the behavior when it overflows its memory bounds
with in-order data so that it tries to grow rcvbuf (the same as with
out-of-order data).
These changes should help my problem of mixed MTUs, and should also help the
case from last week's thread I think. (In both cases though you still need
tcp_rmem[2] to be set much larger than the TCP window.) One question is if
this is too aggressive at trying to increase rcvbuf if it's under memory
stress.
Orignally-from: John Heffner <jheffner@psc.edu>
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is an updated version of the RFC3465 ABC patch originally
for Linux 2.6.11-rc4 by Yee-Ting Li. ABC is a way of counting
bytes ack'd rather than packets when updating congestion control.
The orignal ABC described in the RFC applied to a Reno style
algorithm. For advanced congestion control there is little
change after leaving slow start.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Simplify the code that comuputes microsecond rtt estimate used
by TCP Vegas. Move the callback out of the RTT sampler and into
the end of the ack cleanup.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
This bug is responsible for causing the infamous "Treason uncloaked"
messages that's been popping up everywhere since the printk was added.
It has usually been blamed on foreign operating systems. However,
some of those reports implicate Linux as both systems are running
Linux or the TCP connection is going across the loopback interface.
In fact, there really is a bug in the Linux TCP header prediction code
that's been there since at least 2.1.8. This bug was tracked down with
help from Dale Blount.
The effect of this bug ranges from harmless "Treason uncloaked"
messages to hung/aborted TCP connections. The details of the bug
and fix is as follows.
When snd_wnd is updated, we only update pred_flags if
tcp_fast_path_check succeeds. When it fails (for example,
when our rcvbuf is used up), we will leave pred_flags with
an out-of-date snd_wnd value.
When the out-of-date pred_flags happens to match the next incoming
packet we will again hit the fast path and use the current snd_wnd
which will be wrong.
In the case of the treason messages, it just happens that the snd_wnd
cached in pred_flags is zero while tp->snd_wnd is non-zero. Therefore
when a zero-window packet comes in we incorrectly conclude that the
window is non-zero.
In fact if the peer continues to send us zero-window pure ACKs we
will continue making the same mistake. It's only when the peer
transmits a zero-window packet with data attached that we get a
chance to snap out of it. This is what triggers the treason
message at the next retransmit timeout.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
From: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Handle better the case where the sender sends full sized
frames initially, then moves to a mode where it trickles
out small amounts of data at a time.
This known problem is even mentioned in the comments
above tcp_grow_window() in tcp_input.c, specifically:
...
* The scheme does not work when sender sends good segments opening
* window and then starts to feed us spagetti. But it should work
* in common situations. Otherwise, we have to rely on queue collapsing.
...
When the sender gives full sized frames, the "struct sk_buff" overhead
from each packet is small. So we'll advertize a larger window.
If the sender moves to a mode where small segments are sent, this
ratio becomes tilted to the other extreme and we start overrunning
the socket buffer space.
tcp_clamp_window() tries to address this, but it's clamping of
tp->window_clamp is a wee bit too aggressive for this particular case.
Fix confirmed by Ion Badulescu.
Signed-off-by: David S. Miller <davem@davemloft.net>
The problem is that the SACK fragmenting code may incorrectly call
tcp_fragment() with a length larger than the skb->len. This happens
when the skb on the transmit queue completely falls to the LHS of the
SACK.
And add a BUG() check to tcp_fragment() so we can spot this kind of
error more quickly in the future.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
All we need to do is resegment the queue so that
we record SACK information accurately. The edges
of the SACK blocks guide our resegmenting decisions.
With help from Herbert Xu.
Signed-off-by: David S. Miller <davem@davemloft.net>
Of this type, mostly:
CHECK net/ipv6/netfilter.c
net/ipv6/netfilter.c:96:12: warning: symbol 'ipv6_netfilter_init' was not declared. Should it be static?
net/ipv6/netfilter.c:101:6: warning: symbol 'ipv6_netfilter_fini' was not declared. Should it be static?
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This changeset basically moves tcp_sk()->{ca_ops,ca_state,etc} to inet_csk(),
minimal renaming/moving done in this changeset to ease review.
Most of it is just changes of struct tcp_sock * to struct sock * parameters.
With this we move to a state closer to two interesting goals:
1. Generalisation of net/ipv4/tcp_diag.c, becoming inet_diag.c, being used
for any INET transport protocol that has struct inet_hashinfo and are
derived from struct inet_connection_sock. Keeps the userspace API, that will
just not display DCCP sockets, while newer versions of tools can support
DCCP.
2. INET generic transport pluggable Congestion Avoidance infrastructure, using
the current TCP CA infrastructure with DCCP.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With this we're very close to getting all of the current TCP
refactorings in my dccp-2.6 tree merged, next changeset will export
some functions needed by the current DCCP code and then dccp-2.6.git
will be born!
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Completing the previous changeset, this also generalises tcp_v4_synq_add,
renaming it to inet_csk_reqsk_queue_hash_add, already geing used in the
DCCP tree, which I plan to merge RSN.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This creates struct inet_connection_sock, moving members out of struct
tcp_sock that are shareable with other INET connection oriented
protocols, such as DCCP, that in my private tree already uses most of
these members.
The functions that operate on these members were renamed, using a
inet_csk_ prefix while not being moved yet to a new file, so as to
ease the review of these changes.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove the "list" member of struct sk_buff, as it is entirely
redundant. All SKB list removal callers know which list the
SKB is on, so storing this in sk_buff does nothing other than
taking up some space.
Two tricky bits were SCTP, which I took care of, and two ATM
drivers which Francois Romieu <romieu@fr.zoreil.com> fixed
up.
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Francois Romieu <romieu@fr.zoreil.com>
This is part of the grand scheme to eliminate the qlen
member of skb_queue_head, and subsequently remove the
'list' member of sk_buff.
Most users of skb_queue_len() want to know if the queue is
empty or not, and that's trivially done with skb_queue_empty()
which doesn't use the skb_queue_head->qlen member and instead
uses the queue list emptyness as the test.
Signed-off-by: David S. Miller <davem@davemloft.net>
Make TSO segment transmit size decisions at send time not earlier.
The basic scheme is that we try to build as large a TSO frame as
possible when pulling in the user data, but the size of the TSO frame
output to the card is determined at transmit time.
This is guided by tp->xmit_size_goal. It is always set to a multiple
of MSS and tells sendmsg/sendpage how large an SKB to try and build.
Later, tcp_write_xmit() and tcp_push_one() chop up the packet if
necessary and conditions warrant. These routines can also decide to
"defer" in order to wait for more ACKs to arrive and thus allow larger
TSO frames to be emitted.
A general observation is that TSO elongates the pipe, thus requiring a
larger congestion window and larger buffering especially at the sender
side. Therefore, it is important that applications 1) get a large
enough socket send buffer (this is accomplished by our dynamic send
buffer expansion code) 2) do large enough writes.
Signed-off-by: David S. Miller <davem@davemloft.net>
In tcp_clean_rtx_queue(), if the TSO packet is not even partially
acked, do not waste time calling tcp_tso_acked().
Signed-off-by: David S. Miller <davem@davemloft.net>
Everything stated there is out of data. tcp_trim_skb()
does adjust the available socket send buffer space and
skb->truesize now.
Signed-off-by: David S. Miller <davem@davemloft.net>
'nonagle' should be passed to the tcp_snd_test() function
as 'TCP_NAGLE_PUSH' if we are checking an SKB not at the
tail of the write_queue. This is because Nagle does not
apply to such frames since we cannot possibly tack more
data onto them.
However, while doing this __tcp_push_pending_frames() makes
all of the packets in the write_queue use this modified
'nonagle' value.
Fix the bug and simplify this function by just calling
tcp_write_xmit() directly if sk_send_head is non-NULL.
As a result, we can now make tcp_data_snd_check() just call
tcp_push_pending_frames() instead of the specialized
__tcp_data_snd_check().
Signed-off-by: David S. Miller <davem@davemloft.net>
It reimplements portions of tcp_snd_check(), so it
we move it to tcp_output.c we can consolidate it's
logic much easier in a later change.
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow TCP to have multiple pluggable congestion control algorithms.
Algorithms are defined by a set of operations and can be built in
or modules. The legacy "new RENO" algorithm is used as a starting
point and fallback.
Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
When we are doing ucopy, we try to defer the ACK generation to
cleanup_rbuf(). This works most of the time very well, but if the
ucopy prequeue is large, this ACKing behavior kills performance.
With TSO, it is possible to fill the prequeue so large that by the
time the ACK is sent and gets back to the sender, most of the window
has emptied of data and performance suffers significantly.
This behavior does help in some cases, so we should think about
re-enabling this trick in the future, using some kind of limit in
order to avoid the bug case.
Signed-off-by: David S. Miller <davem@davemloft.net>
Ross moved. Remove the bad email address so people will find the correct
one in ./CREDITS.
Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This patch removes a superfluous intialization from tcp_data_queue().
Signed-off-by: James Morris <jmorris@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!