Adding memory barrier after the poll_wait function, paired with
receive callbacks. Adding fuctions sock_poll_wait and sk_has_sleeper
to wrap the memory barrier.
Without the memory barrier, following race can happen.
The race fires, when following code paths meet, and the tp->rcv_nxt
and __add_wait_queue updates stay in CPU caches.
CPU1 CPU2
sys_select receive packet
... ...
__add_wait_queue update tp->rcv_nxt
... ...
tp->rcv_nxt check sock_def_readable
... {
schedule ...
if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
wake_up_interruptible(sk->sk_sleep)
...
}
If there was no cache the code would work ok, since the wait_queue and
rcv_nxt are opposit to each other.
Meaning that once tp->rcv_nxt is updated by CPU2, the CPU1 either already
passed the tp->rcv_nxt check and sleeps, or will get the new value for
tp->rcv_nxt and will return with new data mask.
In both cases the process (CPU1) is being added to the wait queue, so the
waitqueue_active (CPU2) call cannot miss and will wake up CPU1.
The bad case is when the __add_wait_queue changes done by CPU1 stay in its
cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 will then
endup calling schedule and sleep forever if there are no more data on the
socket.
Calls to poll_wait in following modules were ommited:
net/bluetooth/af_bluetooth.c
net/irda/af_irda.c
net/irda/irnet/irnet_ppp.c
net/mac80211/rc80211_pid_debugfs.c
net/phonet/socket.c
net/rds/af_rds.c
net/rfkill/core.c
net/sunrpc/cache.c
net/sunrpc/rpc_pipe.c
net/tipc/socket.c
Signed-off-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Define three accessors to get/set dst attached to a skb
struct dst_entry *skb_dst(const struct sk_buff *skb)
void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
void skb_dst_drop(struct sk_buff *skb)
This one should replace occurrences of :
dst_release(skb->dst)
skb->dst = NULL;
Delete skb->dst field
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixes a problem caused by the overlap of the connection-setup and
established-state phases of DCCP connections.
During connection setup, the client retransmits Confirm Feature-Negotiation
options until a response from the server signals that it can move from the
half-established PARTOPEN into the OPEN state, whereupon the connection is
fully established on both ends (RFC 4340, 8.1.5).
However, since the client may already send data while it is in the PARTOPEN
state, consequences arise for the Maximum Packet Size: the problem is that the
initial option overhead is much higher than for the subsequent established
phase, as it involves potentially many variable-length list-type options
(server-priority options, RFC 4340, 6.4).
Applying the standard MPS is insufficient here: especially with larger
payloads this can lead to annoying, counter-intuitive EMSGSIZE errors.
On the other hand, reducing the MPS available for the established phase by
the added initial overhead is highly wasteful and inefficient.
The solution chosen therefore is a two-phase strategy:
If the payload length of the DataAck in PARTOPEN is too large, an Ack is sent
to carry the options, and the feature-negotiation list is then flushed.
This means that the server gets two Acks for one Response. If both Acks get
lost, it is probably better to restart the connection anyway and devising yet
another special-case does not seem worth the extra complexity.
The result is a higher utilisation of the available packet space for the data
transmission phase (established state) of a connection.
The patch (over-)estimates the initial overhead to be 32*4 bytes -- commonly
seen values were around 90 bytes for initial feature-negotiation options.
It uses sizeof(u32) to mean "aligned units of 4 bytes".
For consistency, another use of 4-byte alignment is adapted.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch resolves a long-standing FIXME to dynamically update the Maximum
Packet Size depending on actual options usage.
It uses the flags set by the feature-negotiation infrastructure to compute
the required header option size.
Most options are fixed-size, a notable exception are Ack Vectors (required
currently only by CCID-2). These can have any length between 3 and 1020
bytes. As a result of testing, 16 bytes (2 bytes for type/length plus 14 Ack
Vector cells) have been found to be sufficient for loss-free situations.
There are currently no CCID-specific header options which may appear on data
packets, thus it is not necessary to define a corresponding CCID field as
suggested in the old comment.
Further changes:
----------------
Adjusted the type of 'cur_mps' to match the unsigned return type of the
function.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds a hook to resolve features whose value depends on the choice of
CCID. It is done at the server since it can only be done after the CCID
values have been negotiated; i.e. the client will add its CCID preference
list on the Change options sent in the Request, which will be reconciled
with the local preference list of the server.
The concept is documented on
http://www.erg.abdn.ac.uk/users/gerrit/dccp/notes/feature_negotiation/\
implementation_notes.html#ccid_dependencies
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
This provides a missing link in the code chain, as several features implicitly
depend and/or rely on the choice of CCID. Most notably, this is the Send Ack Vector
feature, but also Ack Ratio and Send Loss Event Rate (also taken care of).
For Send Ack Vector, the situation is as follows:
* since CCID2 mandates the use of Ack Vectors, there is no point in allowing
endpoints which use CCID2 to disable Ack Vector features such a connection;
* a peer with a TX CCID of CCID2 will always expect Ack Vectors, and a peer
with a RX CCID of CCID2 must always send Ack Vectors (RFC 4341, sec. 4);
* for all other CCIDs, the use of (Send) Ack Vector is optional and thus
negotiable. However, this implies that the code negotiating the use of Ack
Vectors also supports it (i.e. is able to supply and to either parse or
ignore received Ack Vectors). Since this is not the case (CCID-3 has no Ack
Vector support), the use of Ack Vectors is here disabled, with a comment
in the source code.
An analogous consideration arises for the Send Loss Event Rate feature,
since the CCID-3 implementation does not support the loss interval options
of RFC 4342. To make such use explicit, corresponding feature-negotiation
options are inserted which signal the use of the loss event rate option,
as it is used by the CCID3 code.
Lastly, the values of the Ack Ratio feature are matched to the choice of CCID.
The patch implements this as a function which is called after the user has
made all other registrations for changing default values of features.
The table is variable-length, the reserved (and hence for feature-negotiation
invalid, confirmed by considering section 19.4 of RFC 4340) feature number `0'
is used to mark the end of the table.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit a3116ac5c2 from 1st October ("tcp: Port
redirection support for TCP") broke DCCP skb lookup by changing inet_csk_clone,
which is used by DCCP to generate the child socket after the handshake.
This patch updates DCCP to use 'loc_port' instead of 'sport', which fixes the
problem, and thus inheriting port redirection support via the new interface.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: KOVACS Krisztian <hidden@sch.bme.hu>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The AWL lower Ack validity window advances in proportion to GSS, the greatest
sequence number sent. Updating AWL other than at connection setup (in the
DCCP-Request sent by dccp_v{4,6}_connect()) was missing in the DCCP code.
This bug lead to syslog messages such as
"kernel: dccp_check_seqno: DCCP: Step 6 failed for DATAACK packet, [...]
P.ackno exists or LAWL(82947089) <= P.ackno(82948208)
<= S.AWH(82948728), sending SYNC..."
The difference between AWL/AWH here is 1639 packets, while the expected value
(the Sequence Window) would have been 100 (the default). A closer look showed
that LAWL = AWL = 82947089 equalled the ISS on the Response.
The patch now updates AWL with each increase of GSS.
Further changes:
----------------
The patch also enforces more stringent checks on the ISS sequence number:
* AWL is initialised to ISS at connection setup and remains at this value;
* AWH is then always set to GSS (via dccp_update_gss());
* so on the first Request: AWL = AWH = ISS,
and on the n-th Request: AWL = ISS, AWH = ISS + n.
As a consequence, only Response packets that refer to Requests sent by this
host will pass, all others are discarded. This is the intention and in effect
implements the initial adjustments for AWL as specified in RFC 4340, 7.5.1.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
This patch allows the sender to distinguish original and retransmitted packets,
which is in particular needed for the retransmission of DCCP-Requests:
* the first Request uses ISS (generated in net/dccp/ip*.c), and sets GSS = ISS;
* all retransmitted Requests use GSS' = GSS + 1, so that the n-th retransmitted
Request has sequence number ISS + n (mod 48).
To add generic support, the patch reorganises existing code so that:
* icsk_retransmits == 0 for the original packet and
* icsk_retransmits = n > 0 for the n-th retransmitted packet
at the time dccp_transmit_skb() is called, via dccp_retransmit_skb().
Thanks to Wei Yongjun for pointing this problem out.
Further changes:
----------------
* removed the `skb' argument from dccp_retransmit_skb(), since sk_send_head
is used for all retransmissions (the exception is client-Acks in PARTOPEN
state, but these do not use sk_send_head);
* since sk_send_head always contains the original skb (via dccp_entail()),
skb_cloned() never evaluated to true and thus pskb_copy() was never used.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
This patch fixes the following sparse warnings:
* nested min(max()) expression:
net/dccp/ccids/ccid3.c:91:21: warning: symbol '__x' shadows an earlier one
net/dccp/ccids/ccid3.c:91:21: warning: symbol '__y' shadows an earlier one
* Declaration of function prototypes in .c instead of .h file, resulting in
"should it be static?" warnings.
* Declared "struct dccpw" static (local to dccp_probe).
* Disabled dccp_delayed_ack() - not fully removed due to RFC 4340, 11.3
("Receivers SHOULD implement delayed acknowledgement timers ...").
* Used a different local variable name to avoid
net/dccp/ackvec.c:293:13: warning: symbol 'state' shadows an earlier one
net/dccp/ackvec.c:238:33: originally declared here
* Removed unused functions `dccp_ackvector_print' and `dccp_ackvec_print'.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
dev_queue_xmit() and the other IP output functions expect to get a skb
with clear or properly initialized skb->cb. Unlike TCP and UDP, the
dccp_skb_cb doesn't contain a struct inet_skb_parm at the beginning,
so the DCCP-specific data is interpreted by the IP output functions.
This can cause false negatives for the conditional POST_ROUTING hook
invocation, making the packet bypass the hook.
Add a inet_skb_parm/inet6_skb_parm union to the beginning of
dccp_skb_cb to avoid clashes. Also add a BUILD_BUG_ON to make
sure it fits in the cb.
[ Combined with patch from Gerrit Renker to remove two now unnecessary
memsets of IPCB(skb)->opt ]
Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Replace dccp_v(4|6)_ctl_socket with sock to unify a code with TCP/ICMP.
Signed-off-by: Denis V. Lunev <den@openvz.org>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This introduces a CCMPS field for setting a CCID-specific upper bound on the application payload
size, as is defined in RFC 4340, section 14.
Only the TX CCID is considered in setting this limit, since the RX CCID generates comparatively
small (DCCP-Ack) feedback packets. The CCMPS field includes network and transport layer header
lengths. The only current CCMPS customer is CCID4 (via RFC 4828).
A wrapper is used to allow querying the CCMPS even at times where the CCID modules may not have
been fully negotiated yet.
In dccp_sync_mss() the variable `mss_now' has been renamed into `cur_mps', to reflect that we are
dealing with an MPS, but not an MSS.
Since the DCCP code closely follows the TCP code, the identifiers `dccp_sync_mss' and
`dccps_mss_cache' have been kept, as they have direct TCP counterparts.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This provides a separate routine to insert options during the initial handshake.
The main purpose is to conduct feature negotiation, for the moment the only user
is the timestamp echo needed for the (CCID3) handshake RTT sample.
Padding of options has been put into a small separate routine, to be shared among
the two functions. This could also be used as a generic routine to finish inserting
options.
Also removed an `XXX' comment since its content was obvious.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This adds a socket option and signalling support for the case where the server
holds timewait state on closing the connection, as described in RFC 4340, 8.3.
Since holding timewait state at the server is the non-usual case, it is enabled
via a socket option. Documentation for this socket option has been added.
The setsockopt statement has been made resilient against different possible cases
of expressing boolean `true' values using a suggestion by Ian McDonald.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When performing active close, RFC 4340, 8.3. requires to retransmit the
Close/CloseReq with a backoff-retransmit timer starting at intially 2 RTTs.
This patch shifts the existing code for active-close retransmit timer
into output.c, so that the retransmit timer is started when the first
Close/CloseReq is sent. Previously, the timer was started when, after
releasing the socket in dccp_close(), the actively-closing side had not yet
reached the CLOSED/TIMEWAIT state.
The patch further reduces the initial timeout from 3 seconds to the required
2 RTTs, where - in absence of a known RTT - the fallback value specified in
RFC 4340, 3.4 is used.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixes a nasty bug: dccp_send_reset() is called by both DCCPv4 and DCCPv6, but uses
inet_sk_rebuild_header() in each case. This leads to unpredictable and weird behaviour:
under some conditions, DCCPv6 Resets were sent, in other not.
The fix is to use the AF-independent rebuild_header routine.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The sock_wake_async() performs a bit different actions
depending on "how" argument. Unfortunately this argument
ony has numerical magic values.
I propose to give names to their constants to help people
reading this function callers understand what's going on
without looking into this function all the time.
I suppose this is 2.6.25 material, but if it's not (or the
naming seems poor/bad/awful), I can rework it against the
current net-2.6 tree.
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
This performs a minor optimisation: when ccid_hc_tx_send_packet
returns a value greater zero, then the same call previously was done
again at the begin of the while loop in dccp_wait_for_ccid.
This patch exploits the available information and schedule-timeouts
directly instead.
Documentation also added.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This factors code common to dccp_v{4,6}_ctl_send_reset into a separate function,
and adds support for filling in the Data 1 ... Data 3 fields from RFC 4340, 5.6.
It is useful to have this separate, since the following Reset codes will always
be generated from the control socket rather than via dccp_send_reset:
* Code 3, "No Connection", cf. 8.3.1;
* Code 4, "Packet Error" (identification for Data 1 added);
* Code 5, "Option Error" (identification for Data 1..3 added, will be used later);
* Code 6, "Mandatory Error" (same as Option Error);
* Code 7, "Connection Refused" (what on Earth is the difference to "No Connection"?);
* Code 8, "Bad Service Code";
* Code 9, "Too Busy";
* Code 10, "Bad Init Cookie" (not used).
Code 0 is not recommended by the RFC, the following codes would be used in
dccp_send_reset() instead, since they all relate to an established DCCP connection:
* Code 1, "Closed";
* Code 2, "Aborted";
* Code 11, "Aggression Penalty" (12.3).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
In this patch, duplicated code is removed for the case when a Reset packet is
sent from a connected socket. This code duplication is between dccp_make_reset
and dccp_transmit_skb, which already contained an (up to now entirely unused)
switch statement to fill in the reset code from the DCCP_SKB_CB.
The only thing that has been removed is the call to dst_clone(dst), since
the queue_xmit functions use sk_dst_cache anyway.
I wasn't sure which purpose inet_sk_rebuild_header served, so I left it in.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
This adds a FIXME to signal that the function dccp_send_delayed_ack is nowhere
used in the entire DCCP/CCID code.
Using a delayed Ack timer is suggested in 11.3 of RFC 4340, but it has also
rather subtle implications for the Ack-Ratio-accounting.
CCID2 does not use this (maybe it should).
I think leaving the function in is good, in case someone wants to implement
this.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
The parameter `seq' of dccp_send_sync() is in fact an acknowledgement number
and not a sequence number - thus renamed by this patch into `ackno'.
Secondly, a `critical' warning is added when a Sync/SyncAck could not be sent.
Sanity: I have checked all other functions that are called in dccp_transmit_skb,
there are no clashes with the use of dccpd_ack_seq; no other function is
using this slot at the same time.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Ian McDonald <ian.mcdonald@jandi.co.nz>
Signed-off-by: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
