Arnaldo and I agreed it could be applied now, because I have other
pending patches depending on this one (Thank you Arnaldo)
(The other important patch moves skc_refcnt in a separate cache line,
so that the SMP/NUMA performance doesnt suffer from cache line ping pongs)
1) First some performance data :
--------------------------------
tcp_v4_rcv() wastes a *lot* of time in __inet_lookup_established()
The most time critical code is :
sk_for_each(sk, node, &head->chain) {
if (INET_MATCH(sk, acookie, saddr, daddr, ports, dif))
goto hit; /* You sunk my battleship! */
}
The sk_for_each() does use prefetch() hints but only the begining of
"struct sock" is prefetched.
As INET_MATCH first comparison uses inet_sk(__sk)->daddr, wich is far
away from the begining of "struct sock", it has to bring into CPU
cache cold cache line. Each iteration has to use at least 2 cache
lines.
This can be problematic if some chains are very long.
2) The goal
-----------
The idea I had is to change things so that INET_MATCH() may return
FALSE in 99% of cases only using the data already in the CPU cache,
using one cache line per iteration.
3) Description of the patch
---------------------------
Adds a new 'unsigned int skc_hash' field in 'struct sock_common',
filling a 32 bits hole on 64 bits platform.
struct sock_common {
unsigned short skc_family;
volatile unsigned char skc_state;
unsigned char skc_reuse;
int skc_bound_dev_if;
struct hlist_node skc_node;
struct hlist_node skc_bind_node;
atomic_t skc_refcnt;
+ unsigned int skc_hash;
struct proto *skc_prot;
};
Store in this 32 bits field the full hash, not masked by (ehash_size -
1) Using this full hash as the first comparison done in INET_MATCH
permits us immediatly skip the element without touching a second cache
line in case of a miss.
Suppress the sk_hashent/tw_hashent fields since skc_hash (aliased to
sk_hash and tw_hash) already contains the slot number if we mask with
(ehash_size - 1)
File include/net/inet_hashtables.h
64 bits platforms :
#define INET_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
(((__sk)->sk_hash == (__hash))
((*((__u64 *)&(inet_sk(__sk)->daddr)))== (__cookie)) && \
((*((__u32 *)&(inet_sk(__sk)->dport))) == (__ports)) && \
(!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
32bits platforms:
#define TCP_IPV4_MATCH(__sk, __hash, __cookie, __saddr, __daddr, __ports, __dif)\
(((__sk)->sk_hash == (__hash)) && \
(inet_sk(__sk)->daddr == (__saddr)) && \
(inet_sk(__sk)->rcv_saddr == (__daddr)) && \
(!((__sk)->sk_bound_dev_if) || ((__sk)->sk_bound_dev_if == (__dif))))
- Adds a prefetch(head->chain.first) in
__inet_lookup_established()/__tcp_v4_check_established() and
__inet6_lookup_established()/__tcp_v6_check_established() and
__dccp_v4_check_established() to bring into cache the first element of the
list, before the {read|write}_lock(&head->lock);
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allocation for the optnames is similar to the DCCP options, with a
range for rx and tx half connection CCIDs.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Moving the TFRC sender and receiver variables to separate structs, so
that we can copy these structs to userspace thru getsockopt,
dccp_diag, etc.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Isolating it, that will be used when we introduce a CCID2 (TCP-Like)
implementation.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As discussed in the dccp@vger mailing list:
Now applications have to use setsockopt(DCCP_SOCKOPT_SERVICE, service[s]),
prior to calling listen() and connect().
An array of unsigned ints can be passed meaning that the listening sock accepts
connection requests for several services.
With this we can ditch struct sockaddr_dccp and use only sockaddr_in (and
sockaddr_in6 in the future).
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Moving the setting of DCCP_SKB_CB(skb)->dccpd_reset_code to the places
where events happen that trigger sending a RESET packet.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eliciting a SYNCACK in response, we were handling SYNC packets
only in the DCCP_OPEN state, in dccp_rcv_established.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
It is possible to receive more than one CLOSEREQ packet if the
CLOSE packet sent in response is somehow lost, change the state
to DCCP_CLOSING only on the first CLOSEREQ packet received.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Also use some BUG_ON where appropriate and use LIMIT_NETDEBUG for the unlikely
cases where we, at this stage, want to know about, that in my tests hasn't
appeared in the radar.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
To match more closely what is described in RFC 3448.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: Ian McDonald <iam4@cs.waikato.ac.nz>
To start the timestamps with 0.0ms, easing the integer maths in the CCIDs, this
probably will be reworked to use the to be introduced struct timeval_offset
infrastructure out of skb_get_timestamp, etc.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
The initialization of ccid3hcrx_rtt to 5ms is just a bandaid, I'll continue
auditing the CCID3 HC rx codebase to fix this properly, probably I'll add a
feedback timer as suggested in the CCID3 draft.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
We can get this value in an TIMESTAMP_ECHO and/or in an ELAPSED_TIME option, if
receiving both give precendence to the biggest one.
In my tests they are very close if not equal at all times, so we may well think
about removing the code in CCID3 that inserts this option and leaving this to
the core, and perhaps even use just TIMESTAMP_ECHO including the elapsed time.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
This makes the send rate calculations behave way more closely to what
is specified, with the jitter previously seen on x and x_recv
disappearing completely on non lossy setups.
This resembles the tcp_data_snd_check code, that possibly we'll end up
using in DCCP as well, perhaps moving this code to
inet_connection_sock.
For now I'm doing the simplest implementation tho.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
So that applications can set dccp_sock->dccps_pkt_size, that in turn
is used in the CCID3 half connection init routines to set
ccid3hc[tr]x_s and use it in its rate calculations.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Renaming it to dccp_rx_hist_detect_loss.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Renaming it to dccp_rx_hist_add_packet.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
I'll now take a look at the other proposed TFRC DCCP CCIDs to find
more code that is now in ccid3.c and move to this module, the loss
event rate, calc_X, etc most probably will be moved there.
The main goal of these changes is to pave the way for the
implementation of more TFRC based DCCP CCIDs and to shrink ccid3.c,
reducing its complexity and helping in getting it rock solid.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
And put this into net/dccp/ccids/lib/, where packet_history.[ch] will also be
moved and then we'll have a tfrc_lib.ko module that will be used by
dccp_ccid3.ko and other CCIDs that are variations of TFRC (RFC 3448).
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To avoid open coding this all over the place.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Introducing functions to add to or subtract from a timeval variable
and renaming now_delta to timeval_new_delta that calls do_gettimeofday
and then timeval_delta, that should be used when there are several
deltas made relative to the current time or setting variables to it,
so as to avoid calling do_gettimeofday excessively.
I'm leaving these "timeval_" prefixed funcions internal to DCCP for a
while till we're sure there are no subtle bugs in it.
It also is more correct as it checks if the number of usecs added to
or subtracted from a tv_usec field is more than 2 seconds.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is not quite what I think we should have long term but improves
performance for now, so lets use it till we get CCID3 working well,
then we can think about using sk_write_queue, perhaps using some ideas
from Juwen Lai's old stack for 2.4.20.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch puts mostly read only data in the right section
(read_mostly), to help sharing of these data between CPUS without
memory ping pongs.
On one of my production machine, tcp_statistics was sitting in a
heavily modified cache line, so *every* SNMP update had to force a
reload.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Tested with a patched netcat, no horror stories so far 8)
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
And also hc_tx and hc_rx get_info functions for the CCIDs to fill in
information that is specific to them.
For now reusing struct tcp_info, later I'll try to figure out a better
solution, for now its really nice to get this kind of info:
[root@qemu ~]# ./ss -danemi
State Recv-Q Send-Q Local Addr:Port Peer Addr:Port
LISTEN 0 0 *:5001 *:* ino:628 sk:c1340040
mem:(r0,w0,f0,t0) cwnd:0 ssthresh:0
ESTAB 0 0 172.20.0.2:5001 172.20.0.1:32785 ino:629 sk:c13409a0
mem:(r0,w0,f0,t0) ts rto:1000 rtt:0.004/0 cwnd:0 ssthresh:0 rcv_rtt:61.377
This, for instance, shows that we're not congestion controlling ACKs,
as the above output is in the ttcp receiving host, and ttcp is a one
way app, i.e. the received never calls sendmsg, so
ccid_hc_tx_send_packet is never called, so the TX half connection
stays in TFRC_SSTATE_NO_SENT state and hctx_rtt is never calculated,
stays with the value set in ccid3_hc_tx_init, 4us, as show above in
milliseconds (0.004ms), upcoming patches will fix this.
rcv_rtt seems sane tho, matching ping results :-)
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using TIMESTAMP_ECHO and ELAPSED_TIME options received.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
And don't insert a TIMESTAMP option in all packets, leave the decision
to the CCIDs.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Just like kfree, etc it will just not call the CCID exit
routines when the private data area is set to NULL.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
So that we retransmit CLOSE/CLOSEREQ packets till they elicit an
answer or we hit a timeout.
Most of the machinery uses TCP approaches, this code has to be
polished & audited, but this is better than we had before.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This is from a first audit, more eyeballs are more than welcome.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
CCID3 keeps this variable in usecs, inet_connection_socks in jiffies,
so to avoid Mars orbiter losses lets reintroduce ccid3hctx_t_rto 8)
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This changes timestamp, timestamp echo, and elapsed time to use units of 10
usecs as per DCCP spec. This has been tested to verify that times are correct.
Also fixed up length and used hton/ntoh more.
Still to add in later patches:
- actually use elapsed time to adjust RTT
(commented out as was prior to this patch)
- send options at times more closely following the spec
(content is now correct)
Signed-off-by: Ian McDonald <iam4@cs.waikato.ac.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The elapsed time can be two bytes or four bytes only.
Signed-off-by: Ian McDonald <iam4@cs.waikato.ac.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Also fix step 6 when receiving SYNC or SYNCACK packets, i.e. we were not using
the updated swl.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
may be a false warning if there always is something on ccid3hcrx_hist:
net/dccp/ccids/ccid3.c: In function 'ccid3_hc_rx_packet_recv':
net/dccp/ccids/ccid3.c:1634: warning: 'tstamp.tv_usec' may be used uninitialized in this function
net/dccp/ccids/ccid3.c:1634: warning: 'tstamp.tv_sec' may be used uninitialized in this function
const on inline functions doesn't have any effect:
net/dccp/dccp.h:64: warning: type qualifiers ignored on function return type
net/dccp/dccp.h:70: warning: type qualifiers ignored on function return type
net/dccp/dccp.h:76: warning: type qualifiers ignored on function return type
Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To avoid holding TIMEWAIT state for sockets in the LISTEN state.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Only available if CONFIG_DEBUG_KERNEL is enabled in the "Kernel
Hacking" Menu.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based on discussions with Nishida-san.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As requested by Ian.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: Ian McDonald <iam4@cs.waikato.ac.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
With this the previous setup is back, i.e. tcp_diag can be built as a module,
as dccp_diag and both share the infrastructure available in inet_diag.
If one selects CONFIG_INET_DIAG as module CONFIG_INET_TCP_DIAG will also be
built as a module, as will CONFIG_INET_DCCP_DIAG, if CONFIG_IP_DCCP was
selected static or as a module, if CONFIG_INET_DIAG is y, being statically
linked CONFIG_INET_TCP_DIAG will follow suit and CONFIG_INET_DCCP_DIAG will be
built in the same manner as CONFIG_IP_DCCP.
Now to aim at UDP, converting it to use inet_hashinfo, so that we can use
iproute2 for UDP sockets as well.
Ah, just to show an example of this new infrastructure working for DCCP :-)
[root@qemu ~]# ./ss -dane
State Recv-Q Send-Q Local Address:Port Peer Address:Port
LISTEN 0 0 *:5001 *:* ino:942 sk:cfd503a0
ESTAB 0 0 127.0.0.1:5001 127.0.0.1:32770 ino:943 sk:cfd50a60
ESTAB 0 0 127.0.0.1:32770 127.0.0.1:5001 ino:947 sk:cfd50700
TIME-WAIT 0 0 127.0.0.1:32769 127.0.0.1:5001 timer:(timewait,3.430ms,0) ino:0 sk:cf209620
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Next changeset will introduce net/ipv4/tcp_diag.c, moving the code that was put
transitioanlly in inet_diag.c.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Next changeset will rename tcp_diag.[ch] to inet_diag.[ch].
I'm taking this longer route so as to easy review, making clear the changes
made all along the way.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Next changeset will rename tcp_diag to inet_diag and move the tcp_diag code out
of it and into a new tcp_diag.c, similar to the net/dccp/diag.c introduced in
this changeset, completing the transition to a generic inet_diag
infrastructure.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This also changes the list_for_each_entry_safe_continue behaviour to match its
kerneldoc comment, that is, to start after the pos passed.
Also adds several helper functions from previously open coded fragments, making
the code more clear.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
With ugly ifdefs, etc, but this actually:
1. keeps the existing ABI, i.e. no need to recompile the iproute2
utilities if not interested in DCCP.
2. Provides all the tcp_diag functionality in DCCP, with just a
small patch that makes iproute2 support DCCP.
Of course I'll get this cleaned-up in time, but for now I think its
OK to be this way to quickly get this functionality.
iproute2-ss050808 patch at:
http://vger.kernel.org/~acme/iproute2-ss050808.dccp.patch
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Using most of the infrastructure TCP uses, with a dccp_death_row,
etc. As per my current interpretation of the draft what we have with
this changeset seems to be all we need (or very close to it 8)).
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fixes nasty bug related to the retransmit timer (yeah, DCCP does
retransmits) firing too early.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This way it gets closer to the TCP flow, where congestion window
checks are done, it seems we can map ccid_hc_tx_send_packet in
dccp_write_xmit to tcp_snd_wnd_test in tcp_write_xmit, a CCID2
decision should just fit in here as well...
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Syntax is net-pf-PROTOCOL_FAMILY-PROTOCOL-SOCK_TYPE and if this
fails net-pf-PROTOCOL_FAMILY-PROTOCOL.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Yoshifumi Nishida <nishida@csl.sony.co.jp>
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
This also improves reqsk_queue_prune and renames it to
inet_csk_reqsk_queue_prune, as it deals with both inet_connection_sock
and inet_request_sock objects, not just with request_sock ones thus
belonging to inet_request_sock.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Development to this point was done on a subversion repository at:
http://oops.ghostprotocols.net:81/cgi-bin/viewcvs.cgi/dccp-2.6/
This repository will be kept at this site for the foreseable future,
so that interested parties can see the history of this code,
attributions, etc.
If I ever decide to take this offline I'll provide the full history at
some other suitable place.
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
