We currently allocate a fixed size (TCP_SYNQ_HSIZE=512) slots hash table for
each LISTEN socket, regardless of various parameters (listen backlog for
example)
On x86_64, this means order-1 allocations (might fail), even for 'small'
sockets, expecting few connections. On the contrary, a huge server wanting a
backlog of 50000 is slowed down a bit because of this fixed limit.
This patch makes the sizing of listen hash table a dynamic parameter,
depending of :
- net.core.somaxconn tunable (default is 128)
- net.ipv4.tcp_max_syn_backlog tunable (default : 256, 1024 or 128)
- backlog value given by user application (2nd parameter of listen())
For large allocations (bigger than PAGE_SIZE), we use vmalloc() instead of
kmalloc().
We still limit memory allocation with the two existing tunables (somaxconn &
tcp_max_syn_backlog). So for standard setups, this patch actually reduce RAM
usage.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The return value of kfifo_alloc() should be checked by IS_ERR().
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
TCP and RAW do not have this issue. Closes Bug #7432.
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix printk format warnings:
build2.out:net/dccp/ccids/ccid2.c:355: warning: long long unsigned int format, u64 arg (arg 3)
build2.out:net/dccp/ccids/ccid2.c:360: warning: long long unsigned int format, u64 arg (arg 3)
build2.out:net/dccp/ccids/ccid2.c:482: warning: long long unsigned int format, u64 arg (arg 5)
build2.out:net/dccp/ccids/ccid2.c:639: warning: long long unsigned int format, u64 arg (arg 3)
build2.out:net/dccp/ccids/ccid2.c:639: warning: long long unsigned int format, u64 arg (arg 4)
build2.out:net/dccp/ccids/ccid2.c:674: warning: long long unsigned int format, u64 arg (arg 3)
build2.out:net/dccp/ccids/ccid2.c:720: warning: long long unsigned int format, u64 arg (arg 3)
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Updates the references to spec documents throughout the code, taking into
account that
* the DCCP, CCID 2, and CCID 3 drafts all became RFCs in March this year
* RFC 1063 was obsoleted by RFC 1191
* draft-ietf-tcpimpl-pmtud-0x.txt was published as an Informational
RFC, RFC 2923 on 2000-09-22.
All references verified.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based upon a patch from Jesper Juhl. Try to match the
TCP IPv6 code this was copied from as much as possible,
so that it's easy to see where to add the ipv6 pktoptions
support code.
Signed-off-by: David S. Miller <davem@davemloft.net>
I think I got the cause for the Oops observed in
http://www.mail-archive.com/dccp@vger.kernel.org/msg00578.html
The problem is always with applications listening on PF_INET6 sockets. Apart
from the mentioned oops, I observed another one one, triggered at irregular
intervals via timer interrupt:
run_timer_softirq -> dccp_keepalive_timer
-> inet_csk_reqsk_queue_prune
-> reqsk_free
-> dccp_v6_reqsk_destructor
The latter function is the problem and is also the last function to be called
in said kernel panic.
In any case, there is a real problem with allocating the right request_sock
which is what this patch tackles.
It fixes the following problem:
- application listens on PF_INET6
- DCCPv4 packet comes in, is handed over to dccp_v4_do_rcv, from there
to dccp_v4_conn_request
Now: socket is PF_INET6, packet is IPv4. The following code then furnishes the
connection with IPv6 - request_sock operations:
req = reqsk_alloc(sk->sk_prot->rsk_prot);
The first problem is that all further incoming packets will get a Reset since
the connection can not be looked up.
The second problem is worse:
--> reqsk_alloc is called instead of inet6_reqsk_alloc
--> consequently inet6_rsk_offset is never set (dangling pointer)
--> the request_sock_ops are nevertheless still dccp6_request_ops
--> destructor is called via reqsk_free
--> dccp_v6_reqsk_destructor tries to free random memory location (inet6_rsk_offset not set)
--> panic
I have tested this for a while, DCCP sockets are now handled correctly in all
three scenarios (v4/v6 only/v4-mapped).
Commiter note: I've added the dccp_request_sock_ops forward declaration to keep
the tree building and to reduce the size of the patch for 2.6.19,
later I'll move the functions to the top of the affected source
code to match what we have in the TCP counterpart, where this
problem hasn't existed in the first place, dumb me not to have
done the same thing on DCCP land 8)
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
annotated address arguments (port number left alone for now); ditto
for inferred net-endian variables in callers.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds DCCP probing shamelessly ripped off from TCP probes by Stephen
Hemminger.
I've put in here support for further CCID3 variables as well.
Andrea/Arnaldo might look to extend for CCID2.
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
With constants for CCID numbers this now uses them in some places.
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
This has been discussed on dccp@vger and removes the necessity for applications
to supply service codes in each and every case.
If an application does not want to provide a service code, that's fine, it will
be given 0. Otherwise, service codes can be set via socket options as before.
This patch has been tested using various client/server configurations
(including listening on multiple service codes).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Introduce methods which manipulate interesting congestion control
state such as pipe and rtt estimate. This is useful for people
wishing to monitor the variables of CCID and instrument the code
[perhaps using Kprobes]. Personally, I am a fan of
encapsulation---that justifies this change =D.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When multiple losses occur in one RTT, the window should be halved
only once [a single "congestion event"]. This is now implemented,
although not perfectly. Slightly changed the interface for changing
the cwnd: pass hctx instead of dp. This is required in order to allow
for change_cwnd to be called from _init().
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allocate more sequence state on demand. Each time a packet is sent
out by CCID2, a record of it needs to be kept. This list of records
grows proportionally to cwnd. Previously, the length of this list was
hardcored and therefore the cwnd could only grow to this value (of
128). Now, records are allocated on demand as necessary---cwnd may
grow as it wishes. The exceptional case of when memory is not
available is not handled gracefully. Perhaps, cwnd should be capped
at that point.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Allow the user to choose whether or not to enable CCID2 debugging via
Kconfig.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If not enough cwnd is available, tell the sender to check again as
soon as possible. This will increase CPU utilization (polling
frequently for cwnd) but will improve network performance. That is,
the sender will need to wait less before detecting the increase of
cwnd. A better architecture would be for the CCID to call-back (or
dequeue) from DCCP when it is able to transmit traffic -- not the
other way around as it currently occurs.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Initialize the slow-start threshold to infinity. This way, upon connection
initiation, slow-start will be exited only upon a packet loss. This patch will
allow connections to quickly gain speed.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Jiffies are now handled correctly (I hope) in CCID2. If they wrap, no
problem.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Get rid of unused variables in ackvector state.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix the way state is masked out. DCCP_ACKVEC_STATE_NOT_RECEIVED is
defined as appears in the packet, therefore bit shifting is not
required. This fix allows CCID2 to correctly detect losses.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Fix ackvector length calculation upon receiving an "ack-of-ack". This
patch avoids the ackvector from growing too large which causes it to
not be inserted into packets.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Function sk_filter() is called from tcp_v{4,6}_rcv() functions with arg
needlock = 0, while socket is not locked at that moment. In order to avoid
this and similar issues in the future, use rcu for sk->sk_filter field read
protection.
Signed-off-by: Dmitry Mishin <dim@openvz.org>
Signed-off-by: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Signed-off-by: Kirill Korotaev <dev@openvz.org>
As Arnaldo Carvalho de Melo points out I should be using list_entry in case
the structure changes in future. Current code functions but is reliant
on position and requires type cast.
Noticed when doing this that I have one more variable than I needed so
removing that also.
Signed off by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds transmit buffering to DCCP.
I have tested with CCID2/3 and with loss and rate limiting.
Signed off by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
This shifts further sysctls into feat.h. No change in
functionality - shifting code only.
Signed off by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
Based on MIPL2 kernel patch.
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: Ville Nuorvala <vnuorval@tcs.hut.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>
Right now most inet_lookup_* functions take a host-order hnum instead
of a network-order dport because that's how it is represented
internally.
This means that users of these functions have to be careful about
using the right byte-order. To add more confusion, inet_lookup takes
a network-order dport unlike all other functions.
So this patch changes all visible inet_lookup functions to take a
dport and move all dport->hnum conversion inside them.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This automatically labels the TCP, Unix stream, and dccp child sockets
as well as openreqs to be at the same MLS level as the peer. This will
result in the selection of appropriately labeled IPSec Security
Associations.
This also uses the sock's sid (as opposed to the isec sid) in SELinux
enforcement of secmark in rcv_skb and postroute_last hooks.
Signed-off-by: Venkat Yekkirala <vyekkirala@TrustedCS.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This labels the flows that could utilize IPSec xfrms at the points the
flows are defined so that IPSec policy and SAs at the right label can
be used.
The following protos are currently not handled, but they should
continue to be able to use single-labeled IPSec like they currently
do.
ipmr
ip_gre
ipip
igmp
sit
sctp
ip6_tunnel (IPv6 over IPv6 tunnel device)
decnet
Signed-off-by: Venkat Yekkirala <vyekkirala@TrustedCS.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixes CCID3 to give much closer performance to RFC4342.
CCID3 is meant to alter sending rate based on RTT and loss.
The performance was verified against:
http://wand.net.nz/~perry/max_download.php
For example I tested with netem and had the following parameters:
Delayed Acks 1, MSS 256 bytes, RTT 105 ms, packet loss 5%.
This gives a theoretical speed of 71.9 Kbits/s. I measured across three
runs with this patch set and got 70.1 Kbits/s. Without this patchset the
average was 232 Kbits/s which means Linux can't be used for CCID3 research
properly.
I also tested with netem turned off so box just acting as router with 1.2
msec RTT. The performance with this is the same with or without the patch
at around 30 Mbit/s.
Signed off by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds a new function dccp_rx_hist_find_entry.
Signed off by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds a new function to see if two sequence numbers follow each
other.
Signed off by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixes a small typo in net/dccp/libs/packet_history.c
Signed off by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
The current users of ip6_dst_lookup can be divided into two classes:
1) The caller holds no locks and is in user-context (UDP).
2) The caller does not want to lookup the dst cache at all.
The second class covers everyone except UDP because most people do
the cache lookup directly before calling ip6_dst_lookup. This patch
adds ip6_sk_dst_lookup for the first class.
Similarly ip6_dst_store users can be divded into those that need to
take the socket dst lock and those that don't. This patch adds
__ip6_dst_store for those (everyone except UDP/datagram) that don't
need an extra lock.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
When using the default sequence window size (100) I got the following in
my logs:
Jun 22 14:24:09 localhost kernel: [ 1492.114775] DCCP: Step 6 failed for
DATA packet, (LSWL(6279674225) <= P.seqno(6279674749) <=
S.SWH(6279674324)) and (P.ackno doesn't exist or LAWL(18798206530) <=
P.ackno(1125899906842620) <= S.AWH(18798206548), sending SYNC...
Jun 22 14:24:09 localhost kernel: [ 1492.115147] DCCP: Step 6 failed for
DATA packet, (LSWL(6279674225) <= P.seqno(6279674750) <=
S.SWH(6279674324)) and (P.ackno doesn't exist or LAWL(18798206530) <=
P.ackno(1125899906842620) <= S.AWH(18798206549), sending SYNC...
I went to alter the default sysctl and it didn't take for new sockets.
Below patch fixes this.
I think the default is too low but it is what the DCCP spec specifies.
As a side effect of this my rx speed using iperf goes from about 2.8 Mbits/sec
to 3.5. This is still far too slow but it is a step in the right direction.
Compile tested only for IPv6 but not particularly complex change.
Signed off by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
No actual bugs that I can see just a couple of unmarked casts
getting annoying in my debug log files.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Default values for boolean and tristate options can only be 'y', 'm' or 'n'.
This patch removes wrong default for IP_DCCP_ACKVEC.
Signed-off-by: Jean-Luc Leger <jean-luc.leger@dspnet.fr.eu.org>
Cc: Arnaldo Carvalho de Melo <acme@conectiva.com.br>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add an extra argument to sk_eat_skb, and make it move early copied
packets to the async_wait_queue instead of freeing them.
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
A soft lockup existed in the handling of ack vector records.
Specifically, when a tail of the list of ack vector records was
removed, it was possible to end up iterating infinitely on an element
of the tail.
Signed-off-by: Andrea Bittau <a.bittau@cs.ucl.ac.uk>
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
Calling sock_orphan inside bh_lock_sock in dccp_close can lead to dead
locks. For example, the inet_diag code holds sk_callback_lock without
disabling BH. If an inbound packet arrives during that admittedly tiny
window, it will cause a dead lock on bh_lock_sock. Another possible
path would be through sock_wfree if the network device driver frees the
tx skb in process context with BH enabled.
We can fix this by moving sock_orphan out of bh_lock_sock.
The tricky bit is to work out when we need to destroy the socket
ourselves and when it has already been destroyed by someone else.
By moving sock_orphan before the release_sock we can solve this
problem. This is because as long as we own the socket lock its
state cannot change.
So we simply record the socket state before the release_sock
and then check the state again after we regain the socket lock.
If the socket state has transitioned to DCCP_CLOSED in the time being,
we know that the socket has been destroyed. Otherwise the socket is
still ours to keep.
This problem was discoverd by Ingo Molnar using his lock validator.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
we dont free req if we cant parse the options.
This fixes coverity bug id #1046
Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
From: Randy Dunlap <rdunlap@xenotime.net>
Use NULL instead of 0 for pointers.
Fix these sparse warnings:
net/dccp/feat.c:207:20: warning: Using plain integer as NULL pointer
net/dccp/feat.c:325:21: warning: Using plain integer as NULL pointer
net/dccp/feat.c:526:20: warning: Using plain integer as NULL pointer
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implement the half-closed devices notifiation, by adding a new POLLRDHUP
(and its alias EPOLLRDHUP) bit to the existing poll/select sets. Since the
existing POLLHUP handling, that does not report correctly half-closed
devices, was feared to be changed, this implementation leaves the current
POLLHUP reporting unchanged and simply add a new bit that is set in the few
places where it makes sense. The same thing was discussed and conceptually
agreed quite some time ago:
http://lkml.org/lkml/2003/7/12/116
Since this new event bit is added to the existing Linux poll infrastruture,
even the existing poll/select system calls will be able to use it. As far
as the existing POLLHUP handling, the patch leaves it as is. The
pollrdhup-2.6.16.rc5-0.10.diff defines the POLLRDHUP for all the existing
archs and sets the bit in the six relevant files. The other attached diff
is the simple change required to sys/epoll.h to add the EPOLLRDHUP
definition.
There is "a stupid program" to test POLLRDHUP delivery here:
http://www.xmailserver.org/pollrdhup-test.c
It tests poll(2), but since the delivery is same epoll(2) will work equally.
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Michael Kerrisk <mtk-manpages@gmx.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This is in preparation for having a dccp_minisock embedded into
dccp_request_sock so that feature negotiation can be done prior to
creating the full blown dccp_sock.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This will later be included in struct dccp_request_sock so that we can
have per connection feature negotiation state while in the 3way
handshake, when we clone the DCCP_ROLE_LISTEN socket (in
dccp_create_openreq_child) we'll just copy this state from
dreq_minisock to dccps_minisock.
Also the feature negotiation and option parsing code will mostly touch
dccps_minisock, which will simplify some stuff.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
No code changes, just tidying up, in some cases moving EXPORT_SYMBOLs
to just after the function exported, etc.
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
