Age | Commit message (Collapse) | Author |
|
CCID-2's cwnd increases like TCP during slow-start, which has implications for
* the local Sequence Window value (should be > cwnd),
* the Ack Ratio value.
Hence an exponential growth, if it does not reflect the actual network
conditions, can quickly lead to instability.
This patch adds congestion-window validation (RFC2861) to CCID-2:
* cwnd is constrained if the sender is application limited;
* cwnd is reduced after a long idle period, as suggested in the '90 paper
by Van Jacobson, in RFC 2581 (sec. 4.1);
* cwnd is never reduced below the RFC 3390 initial window.
As marked in the comments, the code is actually almost a direct copy of the
TCP congestion-window-validation algorithms. By continuing this work, it may
in future be possible to use the TCP code (not possible at the moment).
The mechanism can be turned off using a module parameter. Sampling of the
currently-used window (moving-maximum) is however done constantly; this is
used to determine the expected window, which can be exploited to regulate
DCCP's Sequence Window value.
This patch also sets slow-start-after-idle (RFC 4341, 5.1), i.e. it behaves like
TCP when net.ipv4.tcp_slow_start_after_idle = 1.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
|
|
This replaces a switch statement with a test, using the equivalent
function dccp_data_packet(skb). It also doubles the range of the field
`rx_num_data_pkts' by changing the type from `int' to `u32', avoiding
signed/unsigned comparison with the u16 field `dccps_r_ack_ratio'.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
|
|
This moves CCID-2's initial window function into the header file, since several
parts throughout the CCID-2 code need to call it (CCID-2 still uses RFC 3390).
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Acked-by: Leandro Melo de Sales <leandro@ic.ufal.br>
|
|
Change the CCID (de)activation message to start with the
protocol name, as 'CCID' is already in there.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
|
|
Realising the following call pattern,
* first dccp_entail() is called to enqueue a new skb and
* then skb_clone() is called to transmit a clone of that skb,
this patch integrates both into the same function.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
|
|
This patch rearranges the order of statements of the slow-path input processing
(i.e. any other state than OPEN), to resolve the following issues.
1. Dependencies: the order of statements now better matches RFC 4340, 8.5, i.e.
step 7 is before step 9 (previously 9 was before 7), and parsing options in
step 8 (which may consume resources) now comes after step 7.
2. Sequence number checks are omitted if in state LISTEN/REQUEST, due to the
note underneath the table in RFC 4340, 7.5.3.
As a result, CCID processing is now indeed confined to OPEN/PARTOPEN states,
i.e. congestion control is performed only on the flow of data packets. This
avoids pathological cases of doing congestion control on those messages
which set up and terminate the connection.
3. Packets are now passed on to Ack Vector / CCID processing only after
- step 7 (receive unexpected packets),
- step 9 (receive Reset),
- step 13 (receive CloseReq),
- step 14 (receive Close)
and only if the state is PARTOPEN. This simplifies CCID processing:
- in LISTEN/CLOSED the CCIDs are non-existent;
- in RESPOND/REQUEST the CCIDs have not yet been negotiated;
- in CLOSEREQ and active-CLOSING the node has already closed this socket;
- in passive-CLOSING the client is waiting for its Reset.
In the last case, RFC 4340, 8.3 leaves it open to ignore further incoming
data, which is the approach taken here.
Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
|
|
Use pr_fmt() without KBUILD_MODNAME to allow AUN and econet prefixes.
Convert printks with KERN_DEBUG to pr_debug.
Hoist assigns from if.
80 column wrapping.
Move open braces to end of line.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Too trivial to live.
cc: WANG Cong <amwang@redhat.com>
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Unused symbols waste space.
Commit 0e34e93177fb
"(netpoll: add generic support for bridge and bonding devices)"
added the symbol more than a year ago with the promise of "future use".
Because it is so far unused, remove it for now.
It can be easily readded if or when it actually needs to be used.
cc: WANG Cong <amwang@redhat.com>
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git2.kernel.org/pub/scm/linux/kernel/git/lowpan/lowpan
|
|
Correct the syntax so that both array and pointer are const.
Signed-off-by: Greg Dietsche <Gregory.Dietsche@cuw.edu>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows 80 column line reflowing.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows 80 column line reflowing.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows useless break;s removed after returns
and a comment added to an unnecessary default: break;
because of a dubious gcc warning.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows code on same line as case moved
to separate lines and a "/* Fallthrough */" comment
added where appropriate.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows 80 column reflowing.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows 80 column reflowing,
removal of a useless break after return, and moving
open brace after case instead of separate line.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows no difference.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows miscellaneous 80 column wrapping,
comment reflowing and a comment for a useless gcc
warning for an otherwise unused default: case.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows miscellaneous 80 column wrapping.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
Simplify a default case / return (unreached)
Remove unnecessary break after return.
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
git diff -w shows differences for line wrapping.
(fit multiple lines to 80 columns, join where possible)
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the case labels the same indent as the switch.
(git diff -w net/appletalk shows no difference)
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This patch adds a change sequence counter to each net namespace
which is bumped whenever a netdevice is added or removed from
the list. If such a change occurred while a link dump took place,
the dump will have the NLM_F_DUMP_INTR flag set in the first
message which has been interrupted and in all subsequent messages
of the same dump.
Note that links may still be modified or renamed while a dump is
taking place but we can guarantee for userspace to receive a
complete list of links and not miss any.
Testing:
I have added 500 VLAN netdevices to make sure the dump is split
over multiple messages. Then while continuously dumping links in
one process I also continuously deleted and re-added a dummy
netdevice in another process. Multiple dumps per seconds have
had the NLM_F_DUMP_INTR flag set.
I guess we can wait for Johannes patch to hit net-next via the
wireless tree. I just wanted to give this some testing right away.
Signed-off-by: Thomas Graf <tgraf@infradead.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Signed-off-by: Alexander Smirnov <alex.bluesman.smirnov@gmail.com>
Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
|
|
Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
|
|
Signed-off-by: Dmitry Eremin-Solenikov <dbaryshkov@gmail.com>
|
|
This patch enables the 6131 family of chips to forward DSA
packets to other switch chips. This is needed if multiple
DSA chips are used in a device. Without this patch the
chip will drop any DSA packets not destined for it.
This patch only enables the forwarding of DSA packets if
multiple chips are used in the switch configuration.
Signed-off-by: Barry Grussling <barry@grussling.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Results on dummy device can be seen in my netconf 2011
slides. These results are for a 10Gige IXGBE intel
nic - on another i5 machine, very similar specs to
the one used in the netconf2011 results.
It turns out - this is a hell lot worse than dummy
and so this patch is even more beneficial for 10G.
Test setup:
----------
System under test sending packets out.
Additional box connected directly dropping packets.
Installed prio qdisc on the eth device and default
netdev default length of 1000 used as is.
The 3 prio bands each were set to 100 (didnt factor in
the results).
5 packet runs were made and the middle 3 picked.
results
-------
The "cpu" column indicates the which cpu the sample
was taken on,
The "Pkt runx" carries the number of packets a cpu
dequeued when forced to be in the "dequeuer" role.
The "avg" for each run is the number of times each
cpu should be a "dequeuer" if the system was fair.
3.0-rc4 (plain)
cpu Pkt run1 Pkt run2 Pkt run3
================================================
cpu0 21853354 21598183 22199900
cpu1 431058 473476 393159
cpu2 481975 477529 458466
cpu3 23261406 23412299 22894315
avg 11506948 11490372 11486460
3.0-rc4 with patch and default weight 64
cpu Pkt run1 Pkt run2 Pkt run3
================================================
cpu0 13205312 13109359 13132333
cpu1 10189914 10159127 10122270
cpu2 10213871 10124367 10168722
cpu3 13165760 13164767 13096705
avg 11693714 11639405 11630008
As you can see the system is still not perfect but
is a lot better than what it was before...
At the moment we use the old backlog weight, weight_p
which is 64 packets. It seems to be reasonably fine
with that value.
The system could be made more fair if we reduce the
weight_p (as per my presentation), but we are going
to affect the shared backlog weight. Unless deemed
necessary, I think the default value is fine. If not
we could add yet another knob.
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
File net/TUNABLE has never be updated since git age.
For some tunable parameters which user can control with proc file-system,
They are all in ip-sysctl.txt doc.
For tunable parameters that only at compile time, no meaning to note them.
Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/paulg/net-next-2.6
|
|
Simplifies the creation of connection protocol messages by eliminating
the passing of information that is no longer required, is constant,
or is contained within the port structure that is issuing the message.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Modifies the logic that creates a connection termination payload
message so that it no longer (mis)uses a routine that creates a
connection protocol message. The revised code is now more easily
understood, and avoids setting several fields that are either not
present in payload messages or were being set more than once.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Restructures the logic used in tipc_port_recv_proto_msg() to ensure
that incoming connection protocol messages are handled properly. The
routine now uses a two-stage process that first ensures the message
applies on an existing connection and then processes the request.
This corrects a loophole that allowed a connection probe request to
be processed if it was sent to an unconnected port that had no names
bound to it.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Speeds up the creation of the FIN message that terminates a TIPC
connection. The typical peer termination message is now created by
duplicating the terminating port's standard payload message header
and adjusting the message size, importance, and error code fields,
rather than building all fields of the message from scratch. A FIN
message that is directed to the port itself is created the same way.
but also requires swapping the origin and destination address fields.
In addition to reducing the work required to create FIN messages,
these changes eliminate several instances of duplicated code,
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Performs cosmetic cleanup of the symbolic names used to specify TIPC
payload message header sizes. The revised names now more accurately
reflect the payload messages in which they can appear. In addition,
several places where these payload message symbol names were being used
to create non-payload messages have been updated to use the proper
internal message symbolic name.
No functional changes are introduced by this rework.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Gets rid of code that allows tipc_msg_init() to create a short
payload message header. This optimization is possible because
there are no longer any callers who require this capability.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Eliminates a pair of #include statements for files that are brought in
automatically by including core.h.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Gets rid of counter that records the number of times a bearer has
resumed after congestion or blocking, since the value is never
referenced anywhere.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Fixes a minor error in the title of one of the message size profiling
values printed as part of TIPC's link statistics.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Gets rid of a pair of checks to see if a name sequence entry in
TIPC's name table has an empty zone list. These checks are pointless
since the zone list can never be empty (i.e. as soon as the list
becomes empty the associated name sequence entry is deleted).
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Modifies the main circular linked lists of publications used in TIPC's
name table to use the standard kernel linked list type. This change
simplifies the deletion of an existing publication by eliminating
the need to search up to three lists to locate the publication.
The use of standard list routines also helps improve the readability
of the name table code by make it clearer what each list operation
being performed is actually doing.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Modifies the name table array structure that contains the name
sequence instances for a given name type so that the publication
lists associated with a given instance are stored in a dynamically
allocated structure, rather than being embedded within the array
entry itself. This change is being done for several reasons:
1) It reduces the amount of data that needs to be copied whenever
a given array is expanded or contracted to accommodate the first
publication of a new name sequence or the removal of the last
publication of an existing name sequence.
2) It reduces the amount of memory associated with array entries that
are currently unused.
3) It facilitates the upcoming conversion of the publication lists
from TIPC-specific circular lists to standard kernel lists. (Standard
lists cannot be used with the former array structure because the
relocation of array entries during array expansion and contraction
would corrupt the lists.)
Note that, aside from introducing a small amount of code to dynamically
allocate and free the structure that now holds publication list info,
this change is largely a simple renaming exercise that replaces
references to "sseq->LIST" with "sseq->info->LIST" (or "info->LIST").
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Gets rid of unnecessary masking in two routines that set TIPC message
header fields. (The msg_set_bits() routine already takes care of
masking the new value to the correct size.)
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Gets rid of a pair of routines that provide support for temporarily
caching the destination node for a message in the associated message
buffer's application handle, since this capability is no longer used.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Optimizes the creation of a returned payload message by duplicating
the original message and then updating the small number of fields
that need to be adjusted, rather than building the new message header
from scratch. In addition, certain operations that are not always
required are relocated so that they are only done if needed.
These optimizations also have the effect of addressing other issues
that were present previously:
1) Fixes a bug that caused the socket send routines to return the
size of the returned message, rather than the size of the sent
message, when a returnable payload message was sent to a non-existent
destination port.
2) The message header of the returned message now matches that of
the original message more closely. The header is now always the same
size as the original header, and some message header fields that
weren't being initialized in the returned message header are now
populated correctly -- namely the "d" and "s" bits, and the upper
bound of a multicast name instance (where present).
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Reduces the work involved in transmitting a returned payload message
by doing only the work necessary to route such a message directly to
the specified destination port, rather than invoking the code used
to route an arbitrary message to an arbitrary destination.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Introduces an internal sanity check to ensure that the only undeliverable
messages TIPC attempts to return to their origin are application payload
messages.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Modifies the routine that handles the rejection of payload messages
so that it has a single exit point that frees up the rejected message,
thereby eliminating some duplicated code.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|
|
Eliminates a TIPC-specific assert() macro that is no longer used.
Signed-off-by: Allan Stephens <allan.stephens@windriver.com>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
|