This moves the inlines (which were previously declared as macros) back into
packet_history.h since the loss detection code needs to be able to read entries
from the RX history in order to create the relevant loss entries: it needs at
least tfrc_rx_hist_loss_prev() and tfrc_rx_hist_last_rcv(), which in turn
require the definition of the other inlines (macros).
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>
This separates RX/TX initialisation and puts all packet history / loss intervals
initialisation into tfrc.c.
The organisation is uniform: slab declaration -> {rx,tx}_init() -> {rx,tx}_exit()
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>
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>
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>
Moved up the comment "Receiver routines" above the first occurrence of
RX history routines.
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>
Credit here goes to Gerrit Renker, that provided the initial implementation for
this new codebase.
I modified it just to try to make it closer to the existing API, renaming some
functions, add namespacing and fix one bug where the tfrc_rx_hist_alloc was not
freeing the allocated ring entries on the error path.
Original changeset comment from Gerrit:
-----------
This provides a new, self-contained and generic RX history service for TFRC
based protocols.
Details:
* new data structure, initialisation and cleanup routines;
* allocation of dccp_rx_hist entries local to packet_history.c,
as a service exported by the dccp_tfrc_lib module.
* interface to automatically track highest-received seqno;
* receiver-based RTT estimation (needed for instance by RFC 3448, 6.3.1);
* a generic function to test for `data packets' as per RFC 4340, sec. 7.7.
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>
Only the sender sets window counters [RFC 4342, sections 5 and 8.1].
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 is in preparation for merging the new rx history code written by Gerrit Renker.
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>
This is in preparation for merging the new rx history code written by Gerrit Renker.
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>
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>
as per RFC 4340, sec. 7.7.
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 patch changes the tfrc_lib module in the following manner:
(1) a dedicated tfrc source file to call the packet history &
loss interval init/exit functions.
(2) a dedicated tfrc_pr_debug macro with toggle switch `tfrc_debug'.
Commiter note: renamed tfrc_module.c to tfrc.c, and made CONFIG_IP_DCCP_CCID3
select IP_DCCP_TFRC_LIB.
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>
Based on a previous patch by Gerrit Renker.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This removes a redundant test for unexpected packet types. In dccp_rcv_state_process
it is tested twice whether a DCCP-server has received a CloseReq (Step 7):
* first in the combined if-statement,
* then in the call to dccp_rcv_closereq().
The latter is necesssary since dccp_rcv_closereq() is also called from
__dccp_rcv_established().
This patch removes the duplicate test.
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>
This adds the necessary state transitions for the two forms of passive-close
* PASSIVE_CLOSE - which is entered when a host receives a Close;
* PASSIVE_CLOSEREQ - which is entered when a client receives a CloseReq.
Here is a detailed account of what the patch does in each state.
1) Receiving CloseReq
The pseudo-code in 8.5 says:
Step 13: Process CloseReq
If P.type == CloseReq and S.state < CLOSEREQ,
Generate Close
S.state := CLOSING
Set CLOSING timer.
This means we need to address what to do in CLOSED, LISTEN, REQUEST, RESPOND, PARTOPEN, and OPEN.
* CLOSED: silently ignore - it may be a late or duplicate CloseReq;
* LISTEN/RESPOND: will not appear, since Step 7 is performed first (we know we are the client);
* REQUEST: perform Step 13 directly (no need to enqueue packet);
* OPEN/PARTOPEN: enter PASSIVE_CLOSEREQ so that the application has a chance to process unread data.
When already in PASSIVE_CLOSEREQ, no second CloseReq is enqueued. In any other state, the CloseReq is ignored.
I think that this offers some robustness against rare and pathological cases: e.g. a simultaneous close where
the client sends a Close and the server a CloseReq. The client will then be retransmitting its Close until it
gets the Reset, so ignoring the CloseReq while in state CLOSING is sane.
2) Receiving Close
The code below from 8.5 is unconditional.
Step 14: Process Close
If P.type == Close,
Generate Reset(Closed)
Tear down connection
Drop packet and return
Thus we need to consider all states:
* CLOSED: silently ignore, since this can happen when a retransmitted or late Close arrives;
* LISTEN: dccp_rcv_state_process() will generate a Reset ("No Connection");
* REQUEST: perform Step 14 directly (no need to enqueue packet);
* RESPOND: dccp_check_req() will generate a Reset ("Packet Error") -- left it at that;
* OPEN/PARTOPEN: enter PASSIVE_CLOSE so that application has a chance to process unread data;
* CLOSEREQ: server performed active-close -- perform Step 14;
* CLOSING: simultaneous-close: use a tie-breaker to avoid message ping-pong (see comment);
* PASSIVE_CLOSEREQ: ignore - the peer has a bug (sending first a CloseReq and now a Close);
* TIMEWAIT: packet is ignored.
Note that the condition of receiving a packet in state CLOSED here is different from the condition "there
is no socket for such a connection": the socket still exists, but its state indicates it is unusable.
Last, dccp_finish_passive_close sets either DCCP_CLOSED or DCCP_CLOSING = TCP_CLOSING, so that
sk_stream_wait_close() will wait for the final Reset (which will trigger CLOSING => CLOSED).
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>
This adds two auxiliary states to deal with passive closes:
* PASSIVE_CLOSE (reached from OPEN via reception of Close) and
* PASSIVE_CLOSEREQ (reached from OPEN via reception of CloseReq)
as internal intermediate states.
These states are used to allow a receiver to process unread data before
acknowledging the received connection-termination-request (the Close/CloseReq).
Without such support, it will happen that passively-closed sockets enter CLOSED
state while there is still unprocessed data in the queue; leading to unexpected
and erratic API behaviour.
PASSIVE_CLOSE has been mapped into TCPF_CLOSE_WAIT, so that the code will
seamlessly work with inet_accept() (which tests for this state).
The state names are thanks to Arnaldo, who suggested this naming scheme
following an earlier revision of this patch.
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>
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>
This patch was based on another made by Gerrit Renker, his changelog was:
------------------------------------------------------
The patch set migrates TFRC TX history to a singly-linked list.
The details are:
* use of a consistent naming scheme (all TFRC functions now begin with `tfrc_');
* allocation and cleanup are taken care of internally;
* provision of a lookup function, which is used by the CCID TX infrastructure
to determine the time a packet was sent (in turn used for RTT sampling);
* integration of the new interface with the present use in CCID3.
------------------------------------------------------
Simplifications I did:
. removing the tfrc_tx_hist_head that had a pointer to the list head and
another for the slabcache.
. No need for creating a slabcache for each CCID that wants to use the TFRC
tx history routines, create a single slabcache when the dccp_tfrc_lib module
init routine is called.
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 continues from the previous patch and adds support for actively aborting
a DCCP connection, using a Reset Code 2, "Aborted" to inform the peer of an
abortive release.
I have tried this in various client/server settings and it works as expected.
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 removes one FIXME with regard to close when there is still unread data.
The mechanism is implemented similar to TCP: with regard to DCCP-specifics,
a Reset with Code 2, "Aborted" is sent to the peer.
This corresponds in part to RFC 4340, 8.1.1 and 8.1.5.
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 removes a comment which identifies an `issue' with dccp_write_xmit() where there is none.
The comment assumes it is possible that a packet is sent between the calls to
ccid_hc_tx_send_packet(),
dccp_transmit_skb(),
ccid_hc_tx_packet_sent()
(in the above order) in dccp_write_xmit().
I think that this is impossible, since dccp_write_xmit() is always called under lock:
* when called as dccp_write_xmit(sk, 1) from dccp_send_close(), the socket is locked
(see code comment above dccp_send_close());
* when called as dccp_write_xmit(sk, 0) from dccp_send_msg(), it is after lock_sock() has been called;
* when called as dccp_write_xmit(sk, 0) from dccp_write_xmit_timer(), bh_lock_sock() has been called
and the if/else statement has made sure that sk_lock.owner is not set;
* there are no other places where dccp_write_xmit() is called.
Furthermore, the debug statement for printing the sequence number of the packet just sent has been
removed, since the entire list is being printed anyway and so the entry of that number appears last.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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>
The code used two different variables to count Acks, one of them redundant.
This patch reduces the number of Ack counters to one.
The type of the Ack counter has also been changed to u32 (twice the range of int);
and the variable has been renamed into `packets_acked' - for consistency with
RFC 3465 (and similarly named variables are used by TCP and SCTP).
Lastly, a slightly less aggressive `maxincr' increment is used (for even Ack Ratios,
maxincr was Ack Ratio/2 + 1 instead of Ack Ratio/2).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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 removes the synchronisation variable `ccid2hctx_sendwait', which is set to 1
when the CCID2 sender may send a new packet, and which is set to 0 otherwise
The variable is redundant, since it is only used in combination with the hc_tx_send_packet/
hc_tx_packet_sent function pair. Both functions are called under socket lock, so the
following happens when the CCID2 may send a new packet:
* it sets sendwait = 1 in tx_send_packet and returns 0;
* the subsequent call to tx_packet_sent clears the sendwait flag;
* since tx_send_packet returns 0 if and only if sendwait == 1, the BUG_ON condition
in tx_packet_sent is never satisfied, since that function is never called when
tx_send_packet returns a value different from 0 (cf. dccp_write_xmit);
* the call to tx_packet_sent clears the flag so that the condition "!sendwait" is
true the next time tx_packet_sent is called.
In other words, it is sufficient to just return 0 / not-0 to synchronise tx_send_packet
and tx_packet_sent -- which is what the patch does.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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 reduces the amount of redundant debugging messages:
* pipe/cwnd are printed in both tx_send_packet() and tx_packet_sent().
Both functions are called immediately after one another, so one occurrence is sufficient.
* Since tx_packet_sent() prints pipe/cwnd already, the second printk for pipe is redundant.
* In tx_packet_sent() the check_sanity function is called twice (at the begin and at the end).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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>
The function ccid2_change_pipe only does an assignment. This patch simplifies the code by
replacing the function with the assignment it performs.
Furthermore, the type of pipe is promoted from `signed' to unsigned (increasing the range).
As a result, a BUG_ON test for negative values now becomes obsolete (for safety not removed,
but replaced with a less annoying `DCCP_BUG').
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>
The current function ccid2_change_cwnd in effect makes only an assignment, as
the test whether cwnd has reached 0 is only required when cwnd is halved.
This patch simplifies the code by replacing the function with the assignment
it performs.
Furthermore, since ssthresh derives from cwnd and appears in many assignments and
comparisons, the type of ssthresh has also been changed to match that of cwnd.
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 replaces the field member `numdupack', which was used as a read-only
constant in the code, with a #define.
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 removes a variable `ccid2hctx_sent' which is incremented but
never referenced/read (i.e., dead code).
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 comments out a problematic section comprising a half-finished algorithm:
- The variable `ccid2hctx_ackloss' is never initialised to a value different from 0 and
hence in fact is a read-only constant.
- The `arsent' variable counts packets other than Acks (it is incremented for every packet),
and there is no test for Ack Loss.
- The concept of counting Acks as such leads to a complex calculation, and the calculation
at the moment is inconsistent with this concept.
The problem is that the number of Acks - rather than the number of windows - is counted,
which leads to a complex (cubic/quadratic) expression - this is not even implemented.
In its current state, the commented-out algorithm interfers with normal processing by
changing Ack Ratio incorrectly, and at the wrong times.
A new algorithm is necessary, which will not necessarily use the same variables as used by
the unfinished one; hence the old variables have been removed.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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>
RFC 4341, sec. 5 states that "The cwnd parameter is initialized to at most
four packets for new connections, following the rules from [RFC3390]", which
is implemented by this patch.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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 is because in the next patch CCID2 will assume that dccps_mss_cache is
non-zero.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch removes a bug in the current code. I agree with Andrea's comment
that there is a problem here but the way it is treated does not fix it.
The problem is that whenever Ack Ratio > cwnd, starvation/deadlock occurs:
* the receiver will not send an Ack until (Ack Ratio - cwnd) data packets
have arrived;
* the sender will not send any data packet before the receipt of an Ack
advances the send window.
The only way that the connection then progresses was via RTO timeout. In one
extreme case (bulk transfer), it was observed that this happened for every single
packet; i.e. hundreds of packets, each a RTO timeout of 1..3 seconds apart:
a transfer which normally would take a fraction of a second thus grew to
several minutes.
The solution taken by this approach is to observe the relation
"Ack Ratio <= cwnd"
by using the constraint (1) from RFC 4341, 6.1.2; i.e. set
Ack Ratio = ceil(cwnd / 2)
and update it whenever either Ack Ratio or cwnd change. This ensures that
the deadlock problem can not arise.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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>
Since it makes not sense to assign negative values to Ack Ratio, this
patch disallows this possibility.
As a consequence, a Bug test for negative Ack Ratio values becomes obsolete.
Furthermore, a check against overflow (as Ack Ratio may not exceed 2 bytes,
due to RFC 4340, 11.3) has been added.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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 replaces use of normal subtraction with modulo-48 subtraction.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-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>
In CCID2 the receiver-history is sorted in ascending order of sequence number,
but the processing of received Ack Vectors requires the list traversal in the
opposite direction.
The current code has a bug in this regard: the list traversal is upwards. As a
consequence, only Ack Vectors with a run length of 1 will pass, in all other
Ack Vectors the remaining (acked) sequence numbers are missed, and may later
falsely be identified as lost.
Note: This bug is only visible when Ack Ratio > 1, since otherwise the run
lengths of Ack Vectors are 0.
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 is reduces the length of the struct ackvec/ackvec_record fields. It is
a purely text-based replacement:
s#dccpavr_#avr_#g;
s#dccpav_#av_#g;
and increases readability somewhat.
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>
Small update with regard to RFC 4340 (references added as documentation):
on Requests, Ack Vectors / Elapsed Time should be ignored.
Length handling of Elapsed Time also simplified.
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 cleans up the consequences of an earlier patch which
introduced the `if IP_DCCP' clause into net/dccp/Kconfig.
The CCID Kconfig menu is sourced within this clause; as a
consequence, all tests of type `depends on IP_DCCP' are now
redundant.
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 patch addresses the following problems:
1. DCCP relies for its proper functioning on having at least one CCID module
enabled (as in TCP plugable congestion control). Currently it is possible to
disable both CCIDs and thus leave the DCCP module in a compiled, but entirely
non-functional state: no sockets can be created when no CCID is available.
Furthermore, the protocol is (again like TCP) not intended to be used without
CCIDs. Last, a non-empty CCID list is needed for doing CCID feature negotiation.
2. Internally the default CCID that is advertised by the Linux host is set to CCID2
(DCCPF_INITIAL_CCID in include/linux/dccp.h). Disabling CCID2 in the Kconfig
menu without changing the defaults leads to a failure `module not found' when
trying to load the dccp module (which internally tries to load the default CCID).
3. The specification (RFC 4340, sec. 10) treats CCID2 somewhat like a
`minimum common denominator'; the specification says that:
* "New connections start with CCID 2 for both endpoints"
* "A DCCP implementation intended for general use, such as an implementation in a
general-purpose operating system kernel, SHOULD implement at least CCID 2.
The intent is to make CCID 2 broadly available for interoperability [...]"
Providing CCID2 as minimum-required CCID (like Reno/Cubic in TCP) thus seems reasonable.
Hence this patch automatically selects CCID2 when DCCP is enabled. Documentation also added.
Discussions with Ian McDonald on this subject are gratefully acknowledged.
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 extends the DCCP socket API by honouring any shutdown(2) option set by the user.
The behaviour is, as much as possible, made consistent with the API for TCP's shutdown.
This patch exploits the information provided by the user via the socket API to reduce
processing costs:
* if the read end is closed (SHUT_RD), it is not necessary to deliver to input CCID;
* if the write end is closed (SHUT_WR), the same idea applies, but with a difference -
as long as the TX queue has not been drained, we need to receive feedback to keep
congestion-control rates up to date. Hence SHUT_WR is honoured only after the last
packet (under congestion control) has been sent;
* although SHUT_RDWR seems nonsensical, it is nevertheless supported in the same manner
as for TCP (and agrees with test for SHUTDOWN_MASK in dccp_poll() in net/dccp/proto.c).
Furthermore, most of the code already honours the sk_shutdown flags (dccp_recvmsg() for
instance sets the read length to 0 if SHUT_RD had been called); CCID handling is now added
to this by the present patch.
There will also no longer be any delivery when the socket is in the final stages, i.e. when
one of dccp_close(), dccp_fin(), or dccp_done() has been called - which is fine since at
that stage the connection is its final stages.
Motivation and background are on http://www.erg.abdn.ac.uk/users/gerrit/dccp/notes/shutdown
A FIXME has been added to notify the other end if SHUT_RD has been set (RFC 4340, 11.7).
Note: There is a comment in inet_shutdown() in net/ipv4/af_inet.c which asks to "make
sure the socket is a TCP socket". This should probably be extended to mean
`TCP or DCCP socket' (the code is also used by UDP and raw sockets).
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>
The moving average computation occurs so frequently in the CCID 3 code that
it merits an inline function of its own. This is uses a suggestion by
Arnaldo as per http://www.mail-archive.com/dccp@vger.kernel.org/msg01662.html
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/updates the handling of idle and application-limited periods in CCID3,
which currently is broken: there is no detection as to how long a sender has been
idle - there is only one flag which is toggled in between function calls.
Being obsolete now, the `idle' flag is removed.
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 patch fixes a previously undiscovered bug; the problem is in computing
the elapsed time as the time between `receiving' the packet (i.e. skb enters
CCID module) and sending feedback:
- there is no layer-processing, queueing, or delay involved,
- hence the elapsed time is in the order of 1 function call
- this is in the dimension of maximally 50..100usec
- which renders the use of elapsed time almost entirely useless.
The fix is simply to ignore such trivial amounts of elapsed time.
As a further advantage, the now useless elapsed_time field can be removed from
the socket, which reduces the socket structure by another four bytes.
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 updates the CCID3 code with regard to two instances of using `MSS' in place of `s':
1. The RFC3390-based initial rate: both rfc3448bis as well as the Faster Restart
draft now consistently use `s' instead of MSS.
2. Now agrees with section 4.2 of rfc3448bis: "If the sender is ready to send data when
it does not yet have a round trip sample, the value of X is set to s bytes per
second, for segment size s [...]"
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>
Many-many code in the kernel initialized the timer->function
and timer->data together with calling init_timer(timer). There
is already a helper for this. Use it for networking code.
The patch is HUGE, but makes the code 130 lines shorter
(98 insertions(+), 228 deletions(-)).
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
As done two years ago on IP route cache table (commit
22c047ccbc) , we can avoid using one
lock per hash bucket for the huge TCP/DCCP hash tables.
On a typical x86_64 platform, this saves about 2MB or 4MB of ram, for
litle performance differences. (we hit a different cache line for the
rwlock, but then the bucket cache line have a better sharing factor
among cpus, since we dirty it less often). For netstat or ss commands
that want a full scan of hash table, we perform fewer memory accesses.
Using a 'small' table of hashed rwlocks should be more than enough to
provide correct SMP concurrency between different buckets, without
using too much memory. Sizing of this table depends on
num_possible_cpus() and various CONFIG settings.
This patch provides some locking abstraction that may ease a future
work using a different model for TCP/DCCP table.
Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
