2019-05-23 17:14:41 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0-or-later
|
2008-01-11 22:57:09 +08:00
|
|
|
/* SCTP kernel implementation
|
2005-04-17 06:20:36 +08:00
|
|
|
* Copyright (c) 1999-2000 Cisco, Inc.
|
|
|
|
* Copyright (c) 1999-2001 Motorola, Inc.
|
|
|
|
* Copyright (c) 2001-2003 International Business Machines Corp.
|
|
|
|
* Copyright (c) 2001 Intel Corp.
|
|
|
|
* Copyright (c) 2001 La Monte H.P. Yarroll
|
|
|
|
*
|
2008-01-11 22:57:09 +08:00
|
|
|
* This file is part of the SCTP kernel implementation
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
2020-11-23 02:07:04 +08:00
|
|
|
* This module provides the abstraction for an SCTP transport representing
|
2005-04-17 06:20:36 +08:00
|
|
|
* a remote transport address. For local transport addresses, we just use
|
|
|
|
* union sctp_addr.
|
|
|
|
*
|
|
|
|
* Please send any bug reports or fixes you make to the
|
|
|
|
* email address(es):
|
2013-07-23 20:51:47 +08:00
|
|
|
* lksctp developers <linux-sctp@vger.kernel.org>
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
|
|
|
* Written or modified by:
|
|
|
|
* La Monte H.P. Yarroll <piggy@acm.org>
|
|
|
|
* Karl Knutson <karl@athena.chicago.il.us>
|
|
|
|
* Jon Grimm <jgrimm@us.ibm.com>
|
|
|
|
* Xingang Guo <xingang.guo@intel.com>
|
|
|
|
* Hui Huang <hui.huang@nokia.com>
|
|
|
|
* Sridhar Samudrala <sri@us.ibm.com>
|
|
|
|
* Ardelle Fan <ardelle.fan@intel.com>
|
|
|
|
*/
|
|
|
|
|
2010-08-24 21:21:08 +08:00
|
|
|
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
|
|
|
|
|
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 16:04:11 +08:00
|
|
|
#include <linux/slab.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <linux/types.h>
|
2006-07-22 05:48:50 +08:00
|
|
|
#include <linux/random.h>
|
2005-04-17 06:20:36 +08:00
|
|
|
#include <net/sctp/sctp.h>
|
|
|
|
#include <net/sctp/sm.h>
|
|
|
|
|
|
|
|
/* 1st Level Abstractions. */
|
|
|
|
|
|
|
|
/* Initialize a new transport from provided memory. */
|
2012-08-07 15:26:14 +08:00
|
|
|
static struct sctp_transport *sctp_transport_init(struct net *net,
|
|
|
|
struct sctp_transport *peer,
|
2005-04-17 06:20:36 +08:00
|
|
|
const union sctp_addr *addr,
|
2005-10-07 14:46:04 +08:00
|
|
|
gfp_t gfp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
/* Copy in the address. */
|
|
|
|
peer->af_specific = sctp_get_af_specific(addr->sa.sa_family);
|
2019-07-30 20:38:19 +08:00
|
|
|
memcpy(&peer->ipaddr, addr, peer->af_specific->sockaddr_len);
|
2005-04-17 06:20:36 +08:00
|
|
|
memset(&peer->saddr, 0, sizeof(union sctp_addr));
|
|
|
|
|
2012-06-30 11:04:26 +08:00
|
|
|
peer->sack_generation = 0;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* From 6.3.1 RTO Calculation:
|
|
|
|
*
|
|
|
|
* C1) Until an RTT measurement has been made for a packet sent to the
|
|
|
|
* given destination transport address, set RTO to the protocol
|
|
|
|
* parameter 'RTO.Initial'.
|
|
|
|
*/
|
2012-08-07 15:29:57 +08:00
|
|
|
peer->rto = msecs_to_jiffies(net->sctp.rto_initial);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2016-12-25 19:30:41 +08:00
|
|
|
peer->last_time_heard = 0;
|
2005-04-17 06:20:36 +08:00
|
|
|
peer->last_time_ecne_reduced = jiffies;
|
|
|
|
|
2005-12-23 03:36:46 +08:00
|
|
|
peer->param_flags = SPP_HB_DISABLE |
|
|
|
|
SPP_PMTUD_ENABLE |
|
|
|
|
SPP_SACKDELAY_ENABLE;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* Initialize the default path max_retrans. */
|
2012-08-07 15:29:57 +08:00
|
|
|
peer->pathmaxrxt = net->sctp.max_retrans_path;
|
|
|
|
peer->pf_retrans = net->sctp.pf_retrans;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
INIT_LIST_HEAD(&peer->transmitted);
|
|
|
|
INIT_LIST_HEAD(&peer->send_ready);
|
|
|
|
INIT_LIST_HEAD(&peer->transports);
|
|
|
|
|
2017-10-24 16:45:31 +08:00
|
|
|
timer_setup(&peer->T3_rtx_timer, sctp_generate_t3_rtx_event, 0);
|
|
|
|
timer_setup(&peer->hb_timer, sctp_generate_heartbeat_event, 0);
|
|
|
|
timer_setup(&peer->reconf_timer, sctp_generate_reconf_event, 0);
|
|
|
|
timer_setup(&peer->proto_unreach_timer,
|
|
|
|
sctp_generate_proto_unreach_event, 0);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2006-07-22 05:48:50 +08:00
|
|
|
/* Initialize the 64-bit random nonce sent with heartbeat. */
|
|
|
|
get_random_bytes(&peer->hb_nonce, sizeof(peer->hb_nonce));
|
|
|
|
|
2017-07-04 20:53:27 +08:00
|
|
|
refcount_set(&peer->refcnt, 1);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
return peer;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Allocate and initialize a new transport. */
|
2012-08-07 15:26:14 +08:00
|
|
|
struct sctp_transport *sctp_transport_new(struct net *net,
|
|
|
|
const union sctp_addr *addr,
|
2005-10-07 14:46:04 +08:00
|
|
|
gfp_t gfp)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2007-02-09 22:25:18 +08:00
|
|
|
struct sctp_transport *transport;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2013-06-17 17:40:04 +08:00
|
|
|
transport = kzalloc(sizeof(*transport), gfp);
|
2005-04-17 06:20:36 +08:00
|
|
|
if (!transport)
|
|
|
|
goto fail;
|
|
|
|
|
2012-08-07 15:26:14 +08:00
|
|
|
if (!sctp_transport_init(net, transport, addr, gfp))
|
2005-04-17 06:20:36 +08:00
|
|
|
goto fail_init;
|
|
|
|
|
|
|
|
SCTP_DBG_OBJCNT_INC(transport);
|
|
|
|
|
|
|
|
return transport;
|
|
|
|
|
|
|
|
fail_init:
|
|
|
|
kfree(transport);
|
|
|
|
|
|
|
|
fail:
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* This transport is no longer needed. Free up if possible, or
|
|
|
|
* delay until it last reference count.
|
|
|
|
*/
|
|
|
|
void sctp_transport_free(struct sctp_transport *transport)
|
|
|
|
{
|
|
|
|
/* Try to delete the heartbeat timer. */
|
|
|
|
if (del_timer(&transport->hb_timer))
|
|
|
|
sctp_transport_put(transport);
|
|
|
|
|
|
|
|
/* Delete the T3_rtx timer if it's active.
|
|
|
|
* There is no point in not doing this now and letting
|
|
|
|
* structure hang around in memory since we know
|
2020-11-23 02:07:04 +08:00
|
|
|
* the transport is going away.
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2013-02-04 04:32:57 +08:00
|
|
|
if (del_timer(&transport->T3_rtx_timer))
|
2005-04-17 06:20:36 +08:00
|
|
|
sctp_transport_put(transport);
|
|
|
|
|
2017-01-18 00:44:43 +08:00
|
|
|
if (del_timer(&transport->reconf_timer))
|
|
|
|
sctp_transport_put(transport);
|
|
|
|
|
2010-05-10 00:56:07 +08:00
|
|
|
/* Delete the ICMP proto unreachable timer if it's active. */
|
2013-02-04 04:32:57 +08:00
|
|
|
if (del_timer(&transport->proto_unreach_timer))
|
2020-11-14 13:22:53 +08:00
|
|
|
sctp_transport_put(transport);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
sctp_transport_put(transport);
|
|
|
|
}
|
|
|
|
|
2012-12-06 17:25:05 +08:00
|
|
|
static void sctp_transport_destroy_rcu(struct rcu_head *head)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2012-12-06 17:25:05 +08:00
|
|
|
struct sctp_transport *transport;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2012-12-06 17:25:05 +08:00
|
|
|
transport = container_of(head, struct sctp_transport, rcu);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
dst_release(transport->dst);
|
|
|
|
kfree(transport);
|
|
|
|
SCTP_DBG_OBJCNT_DEC(transport);
|
|
|
|
}
|
|
|
|
|
2012-12-06 17:25:05 +08:00
|
|
|
/* Destroy the transport data structure.
|
|
|
|
* Assumes there are no more users of this structure.
|
|
|
|
*/
|
|
|
|
static void sctp_transport_destroy(struct sctp_transport *transport)
|
|
|
|
{
|
2017-07-04 20:53:27 +08:00
|
|
|
if (unlikely(refcount_read(&transport->refcnt))) {
|
net: sctp: rework debugging framework to use pr_debug and friends
We should get rid of all own SCTP debug printk macros and use the ones
that the kernel offers anyway instead. This makes the code more readable
and conform to the kernel code, and offers all the features of dynamic
debbuging that pr_debug() et al has, such as only turning on/off portions
of debug messages at runtime through debugfs. The runtime cost of having
CONFIG_DYNAMIC_DEBUG enabled, but none of the debug statements printing,
is negligible [1]. If kernel debugging is completly turned off, then these
statements will also compile into "empty" functions.
While we're at it, we also need to change the Kconfig option as it /now/
only refers to the ifdef'ed code portions in outqueue.c that enable further
debugging/tracing of SCTP transaction fields. Also, since SCTP_ASSERT code
was enabled with this Kconfig option and has now been removed, we
transform those code parts into WARNs resp. where appropriate BUG_ONs so
that those bugs can be more easily detected as probably not many people
have SCTP debugging permanently turned on.
To turn on all SCTP debugging, the following steps are needed:
# mount -t debugfs none /sys/kernel/debug
# echo -n 'module sctp +p' > /sys/kernel/debug/dynamic_debug/control
This can be done more fine-grained on a per file, per line basis and others
as described in [2].
[1] https://www.kernel.org/doc/ols/2009/ols2009-pages-39-46.pdf
[2] Documentation/dynamic-debug-howto.txt
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-29 01:49:40 +08:00
|
|
|
WARN(1, "Attempt to destroy undead transport %p!\n", transport);
|
|
|
|
return;
|
|
|
|
}
|
2012-12-06 17:25:05 +08:00
|
|
|
|
sctp: sctp_close: fix release of bindings for deferred call_rcu's
It seems due to RCU usage, i.e. within SCTP's address binding list,
a, say, ``behavioral change'' was introduced which does actually
not conform to the RFC anymore. In particular consider the following
(fictional) scenario to demonstrate this:
do:
Two SOCK_SEQPACKET-style sockets are opened (S1, S2)
S1 is bound to 127.0.0.1, port 1024 [server]
S2 is bound to 127.0.0.1, port 1025 [client]
listen(2) is invoked on S1
From S2 we call one sendmsg(2) with msg.msg_name and
msg.msg_namelen parameters set to the server's
address
S1, S2 are closed
goto do
The first pass of this loop passes successful, while the second round
fails during binding of S1 (address still in use). What is happening?
In the first round, the initial handshake is being done, and, at the
time close(2) is called on S1, a non-graceful shutdown is performed via
ABORT since in S1's receive queue an unprocessed packet is present,
thus stating an error condition. This can be considered as a correct
behavior.
During close also all bound addresses are freed, thus nothing *must*
be active anymore. In reference to RFC2960:
After checking the Verification Tag, the receiving endpoint shall
remove the association from its record, and shall report the
termination to its upper layer. (9.1 Abort of an Association)
Also, no half-open states are supported, thus after an ungraceful
shutdown, we leave nothing behind. However, this seems not to be
happening though. In a real-world scenario, this is exactly where
it breaks the lksctp-tools functional test suite, *for instance*:
./test_sockopt
test_sockopt.c 1 PASS : getsockopt(SCTP_STATUS) on a socket with no assoc
test_sockopt.c 2 PASS : getsockopt(SCTP_STATUS)
test_sockopt.c 3 PASS : getsockopt(SCTP_STATUS) with invalid associd
test_sockopt.c 4 PASS : getsockopt(SCTP_STATUS) with NULL associd
test_sockopt.c 5 BROK : bind: Address already in use
The underlying problem is that sctp_endpoint_destroy() hasn't been
triggered yet while the next bind attempt is being done. It will be
triggered eventually (but too late) by sctp_transport_destroy_rcu()
after one RCU grace period:
sctp_transport_destroy()
sctp_transport_destroy_rcu() ----.
sctp_association_put() [*] <--+--> sctp_packet_free()
sctp_association_destroy() [...]
sctp_endpoint_put() skb->destructor
sctp_endpoint_destroy() sctp_wfree()
sctp_bind_addr_free() sctp_association_put() [*]
Thus, we move out the condition with sctp_association_put() as well as
the sctp_packet_free() invocation and the issue can be solved. We also
better free the SCTP chunks first before putting the ref of the association.
With this patch, the example above (which simulates a similar scenario
as in the implementation of this test case) and therefore also the test
suite run successfully through. Tested by myself.
Cc: Vlad Yasevich <vyasevich@gmail.com>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Acked-by: Vlad Yasevich <vyasevich@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-02-01 12:37:43 +08:00
|
|
|
sctp_packet_free(&transport->packet);
|
|
|
|
|
|
|
|
if (transport->asoc)
|
|
|
|
sctp_association_put(transport->asoc);
|
2013-08-09 22:25:21 +08:00
|
|
|
|
|
|
|
call_rcu(&transport->rcu, sctp_transport_destroy_rcu);
|
2012-12-06 17:25:05 +08:00
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* Start T3_rtx timer if it is not already running and update the heartbeat
|
|
|
|
* timer. This routine is called every time a DATA chunk is sent.
|
|
|
|
*/
|
sctp: avoid refreshing heartbeat timer too often
Currently on high rate SCTP streams the heartbeat timer refresh can
consume quite a lot of resources as timer updates are costly and it
contains a random factor, which a) is also costly and b) invalidates
mod_timer() optimization for not editing a timer to the same value.
It may even cause the timer to be slightly advanced, for no good reason.
As suggested by David Laight this patch now removes this timer update
from hot path by leaving the timer on and re-evaluating upon its
expiration if the heartbeat is still needed or not, similarly to what is
done for TCP. If it's not needed anymore the timer is re-scheduled to
the new timeout, considering the time already elapsed.
For this, we now record the last tx timestamp per transport, updated in
the same spots as hb timer was restarted on tx. Also split up
sctp_transport_reset_timers into sctp_transport_reset_t3_rtx and
sctp_transport_reset_hb_timer, so we can re-arm T3 without re-arming the
heartbeat one.
On loopback with MTU of 65535 and data chunks with 1636, so that we
have a considerable amount of chunks without stressing system calls,
netperf -t SCTP_STREAM -l 30, perf looked like this before:
Samples: 103K of event 'cpu-clock', Event count (approx.): 25833000000
Overhead Command Shared Object Symbol
+ 6,15% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,43% netperf [kernel.vmlinux] [k] _raw_write_unlock_irqrestore
- _raw_write_unlock_irqrestore
- 96,54% _raw_spin_unlock_irqrestore
- 36,14% mod_timer
+ 97,24% sctp_transport_reset_timers
+ 2,76% sctp_do_sm
+ 33,65% __wake_up_sync_key
+ 28,77% sctp_ulpq_tail_event
+ 1,40% del_timer
- 1,84% mod_timer
+ 99,03% sctp_transport_reset_timers
+ 0,97% sctp_do_sm
+ 1,50% sctp_ulpq_tail_event
And after this patch, now with netperf -l 60:
Samples: 230K of event 'cpu-clock', Event count (approx.): 57707250000
Overhead Command Shared Object Symbol
+ 5,65% netperf [kernel.vmlinux] [k] memcpy_erms
+ 5,59% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,05% netperf [kernel.vmlinux] [k] _raw_spin_unlock_irqrestore
- _raw_spin_unlock_irqrestore
+ 49,89% __wake_up_sync_key
+ 45,68% sctp_ulpq_tail_event
- 2,85% mod_timer
+ 76,51% sctp_transport_reset_t3_rtx
+ 23,49% sctp_do_sm
+ 1,55% del_timer
+ 2,50% netperf [sctp] [k] sctp_datamsg_from_user
+ 2,26% netperf [sctp] [k] sctp_sendmsg
Throughput-wise, from 6800mbps without the patch to 7050mbps with it,
~3.7%.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-04-07 02:15:19 +08:00
|
|
|
void sctp_transport_reset_t3_rtx(struct sctp_transport *transport)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
|
|
|
/* RFC 2960 6.3.2 Retransmission Timer Rules
|
|
|
|
*
|
|
|
|
* R1) Every time a DATA chunk is sent to any address(including a
|
|
|
|
* retransmission), if the T3-rtx timer of that address is not running
|
|
|
|
* start it running so that it will expire after the RTO of that
|
|
|
|
* address.
|
|
|
|
*/
|
|
|
|
|
2010-05-01 10:41:09 +08:00
|
|
|
if (!timer_pending(&transport->T3_rtx_timer))
|
2005-04-17 06:20:36 +08:00
|
|
|
if (!mod_timer(&transport->T3_rtx_timer,
|
|
|
|
jiffies + transport->rto))
|
|
|
|
sctp_transport_hold(transport);
|
sctp: avoid refreshing heartbeat timer too often
Currently on high rate SCTP streams the heartbeat timer refresh can
consume quite a lot of resources as timer updates are costly and it
contains a random factor, which a) is also costly and b) invalidates
mod_timer() optimization for not editing a timer to the same value.
It may even cause the timer to be slightly advanced, for no good reason.
As suggested by David Laight this patch now removes this timer update
from hot path by leaving the timer on and re-evaluating upon its
expiration if the heartbeat is still needed or not, similarly to what is
done for TCP. If it's not needed anymore the timer is re-scheduled to
the new timeout, considering the time already elapsed.
For this, we now record the last tx timestamp per transport, updated in
the same spots as hb timer was restarted on tx. Also split up
sctp_transport_reset_timers into sctp_transport_reset_t3_rtx and
sctp_transport_reset_hb_timer, so we can re-arm T3 without re-arming the
heartbeat one.
On loopback with MTU of 65535 and data chunks with 1636, so that we
have a considerable amount of chunks without stressing system calls,
netperf -t SCTP_STREAM -l 30, perf looked like this before:
Samples: 103K of event 'cpu-clock', Event count (approx.): 25833000000
Overhead Command Shared Object Symbol
+ 6,15% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,43% netperf [kernel.vmlinux] [k] _raw_write_unlock_irqrestore
- _raw_write_unlock_irqrestore
- 96,54% _raw_spin_unlock_irqrestore
- 36,14% mod_timer
+ 97,24% sctp_transport_reset_timers
+ 2,76% sctp_do_sm
+ 33,65% __wake_up_sync_key
+ 28,77% sctp_ulpq_tail_event
+ 1,40% del_timer
- 1,84% mod_timer
+ 99,03% sctp_transport_reset_timers
+ 0,97% sctp_do_sm
+ 1,50% sctp_ulpq_tail_event
And after this patch, now with netperf -l 60:
Samples: 230K of event 'cpu-clock', Event count (approx.): 57707250000
Overhead Command Shared Object Symbol
+ 5,65% netperf [kernel.vmlinux] [k] memcpy_erms
+ 5,59% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,05% netperf [kernel.vmlinux] [k] _raw_spin_unlock_irqrestore
- _raw_spin_unlock_irqrestore
+ 49,89% __wake_up_sync_key
+ 45,68% sctp_ulpq_tail_event
- 2,85% mod_timer
+ 76,51% sctp_transport_reset_t3_rtx
+ 23,49% sctp_do_sm
+ 1,55% del_timer
+ 2,50% netperf [sctp] [k] sctp_datamsg_from_user
+ 2,26% netperf [sctp] [k] sctp_sendmsg
Throughput-wise, from 6800mbps without the patch to 7050mbps with it,
~3.7%.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-04-07 02:15:19 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
void sctp_transport_reset_hb_timer(struct sctp_transport *transport)
|
|
|
|
{
|
|
|
|
unsigned long expires;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
/* When a data chunk is sent, reset the heartbeat interval. */
|
sctp: avoid refreshing heartbeat timer too often
Currently on high rate SCTP streams the heartbeat timer refresh can
consume quite a lot of resources as timer updates are costly and it
contains a random factor, which a) is also costly and b) invalidates
mod_timer() optimization for not editing a timer to the same value.
It may even cause the timer to be slightly advanced, for no good reason.
As suggested by David Laight this patch now removes this timer update
from hot path by leaving the timer on and re-evaluating upon its
expiration if the heartbeat is still needed or not, similarly to what is
done for TCP. If it's not needed anymore the timer is re-scheduled to
the new timeout, considering the time already elapsed.
For this, we now record the last tx timestamp per transport, updated in
the same spots as hb timer was restarted on tx. Also split up
sctp_transport_reset_timers into sctp_transport_reset_t3_rtx and
sctp_transport_reset_hb_timer, so we can re-arm T3 without re-arming the
heartbeat one.
On loopback with MTU of 65535 and data chunks with 1636, so that we
have a considerable amount of chunks without stressing system calls,
netperf -t SCTP_STREAM -l 30, perf looked like this before:
Samples: 103K of event 'cpu-clock', Event count (approx.): 25833000000
Overhead Command Shared Object Symbol
+ 6,15% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,43% netperf [kernel.vmlinux] [k] _raw_write_unlock_irqrestore
- _raw_write_unlock_irqrestore
- 96,54% _raw_spin_unlock_irqrestore
- 36,14% mod_timer
+ 97,24% sctp_transport_reset_timers
+ 2,76% sctp_do_sm
+ 33,65% __wake_up_sync_key
+ 28,77% sctp_ulpq_tail_event
+ 1,40% del_timer
- 1,84% mod_timer
+ 99,03% sctp_transport_reset_timers
+ 0,97% sctp_do_sm
+ 1,50% sctp_ulpq_tail_event
And after this patch, now with netperf -l 60:
Samples: 230K of event 'cpu-clock', Event count (approx.): 57707250000
Overhead Command Shared Object Symbol
+ 5,65% netperf [kernel.vmlinux] [k] memcpy_erms
+ 5,59% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,05% netperf [kernel.vmlinux] [k] _raw_spin_unlock_irqrestore
- _raw_spin_unlock_irqrestore
+ 49,89% __wake_up_sync_key
+ 45,68% sctp_ulpq_tail_event
- 2,85% mod_timer
+ 76,51% sctp_transport_reset_t3_rtx
+ 23,49% sctp_do_sm
+ 1,55% del_timer
+ 2,50% netperf [sctp] [k] sctp_datamsg_from_user
+ 2,26% netperf [sctp] [k] sctp_sendmsg
Throughput-wise, from 6800mbps without the patch to 7050mbps with it,
~3.7%.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-04-07 02:15:19 +08:00
|
|
|
expires = jiffies + sctp_transport_timeout(transport);
|
2019-02-22 16:45:26 +08:00
|
|
|
if ((time_before(transport->hb_timer.expires, expires) ||
|
|
|
|
!timer_pending(&transport->hb_timer)) &&
|
sctp: avoid refreshing heartbeat timer too often
Currently on high rate SCTP streams the heartbeat timer refresh can
consume quite a lot of resources as timer updates are costly and it
contains a random factor, which a) is also costly and b) invalidates
mod_timer() optimization for not editing a timer to the same value.
It may even cause the timer to be slightly advanced, for no good reason.
As suggested by David Laight this patch now removes this timer update
from hot path by leaving the timer on and re-evaluating upon its
expiration if the heartbeat is still needed or not, similarly to what is
done for TCP. If it's not needed anymore the timer is re-scheduled to
the new timeout, considering the time already elapsed.
For this, we now record the last tx timestamp per transport, updated in
the same spots as hb timer was restarted on tx. Also split up
sctp_transport_reset_timers into sctp_transport_reset_t3_rtx and
sctp_transport_reset_hb_timer, so we can re-arm T3 without re-arming the
heartbeat one.
On loopback with MTU of 65535 and data chunks with 1636, so that we
have a considerable amount of chunks without stressing system calls,
netperf -t SCTP_STREAM -l 30, perf looked like this before:
Samples: 103K of event 'cpu-clock', Event count (approx.): 25833000000
Overhead Command Shared Object Symbol
+ 6,15% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,43% netperf [kernel.vmlinux] [k] _raw_write_unlock_irqrestore
- _raw_write_unlock_irqrestore
- 96,54% _raw_spin_unlock_irqrestore
- 36,14% mod_timer
+ 97,24% sctp_transport_reset_timers
+ 2,76% sctp_do_sm
+ 33,65% __wake_up_sync_key
+ 28,77% sctp_ulpq_tail_event
+ 1,40% del_timer
- 1,84% mod_timer
+ 99,03% sctp_transport_reset_timers
+ 0,97% sctp_do_sm
+ 1,50% sctp_ulpq_tail_event
And after this patch, now with netperf -l 60:
Samples: 230K of event 'cpu-clock', Event count (approx.): 57707250000
Overhead Command Shared Object Symbol
+ 5,65% netperf [kernel.vmlinux] [k] memcpy_erms
+ 5,59% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,05% netperf [kernel.vmlinux] [k] _raw_spin_unlock_irqrestore
- _raw_spin_unlock_irqrestore
+ 49,89% __wake_up_sync_key
+ 45,68% sctp_ulpq_tail_event
- 2,85% mod_timer
+ 76,51% sctp_transport_reset_t3_rtx
+ 23,49% sctp_do_sm
+ 1,55% del_timer
+ 2,50% netperf [sctp] [k] sctp_datamsg_from_user
+ 2,26% netperf [sctp] [k] sctp_sendmsg
Throughput-wise, from 6800mbps without the patch to 7050mbps with it,
~3.7%.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-04-07 02:15:19 +08:00
|
|
|
!mod_timer(&transport->hb_timer,
|
|
|
|
expires + prandom_u32_max(transport->rto)))
|
|
|
|
sctp_transport_hold(transport);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2017-01-18 00:44:43 +08:00
|
|
|
void sctp_transport_reset_reconf_timer(struct sctp_transport *transport)
|
|
|
|
{
|
|
|
|
if (!timer_pending(&transport->reconf_timer))
|
|
|
|
if (!mod_timer(&transport->reconf_timer,
|
|
|
|
jiffies + transport->rto))
|
|
|
|
sctp_transport_hold(transport);
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* This transport has been assigned to an association.
|
|
|
|
* Initialize fields from the association or from the sock itself.
|
|
|
|
* Register the reference count in the association.
|
|
|
|
*/
|
|
|
|
void sctp_transport_set_owner(struct sctp_transport *transport,
|
|
|
|
struct sctp_association *asoc)
|
|
|
|
{
|
|
|
|
transport->asoc = asoc;
|
|
|
|
sctp_association_hold(asoc);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Initialize the pmtu of a transport. */
|
2011-04-27 05:51:31 +08:00
|
|
|
void sctp_transport_pmtu(struct sctp_transport *transport, struct sock *sk)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2011-04-27 05:54:17 +08:00
|
|
|
/* If we don't have a fresh route, look one up */
|
2012-07-20 03:31:33 +08:00
|
|
|
if (!transport->dst || transport->dst->obsolete) {
|
2017-02-07 05:14:13 +08:00
|
|
|
sctp_transport_dst_release(transport);
|
2011-04-27 05:54:17 +08:00
|
|
|
transport->af_specific->get_dst(transport, &transport->saddr,
|
2011-05-07 07:32:47 +08:00
|
|
|
&transport->fl, sk);
|
2011-04-27 05:54:17 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2018-04-27 03:58:59 +08:00
|
|
|
if (transport->param_flags & SPP_PMTUD_DISABLE) {
|
|
|
|
struct sctp_association *asoc = transport->asoc;
|
|
|
|
|
|
|
|
if (!transport->pathmtu && asoc && asoc->pathmtu)
|
|
|
|
transport->pathmtu = asoc->pathmtu;
|
|
|
|
if (transport->pathmtu)
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2018-04-27 03:58:57 +08:00
|
|
|
if (transport->dst)
|
|
|
|
transport->pathmtu = sctp_dst_mtu(transport->dst);
|
|
|
|
else
|
2005-12-23 03:36:46 +08:00
|
|
|
transport->pathmtu = SCTP_DEFAULT_MAXSEGMENT;
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
sctp: fix the handling of ICMP Frag Needed for too small MTUs
syzbot reported a hang involving SCTP, on which it kept flooding dmesg
with the message:
[ 246.742374] sctp: sctp_transport_update_pmtu: Reported pmtu 508 too
low, using default minimum of 512
That happened because whenever SCTP hits an ICMP Frag Needed, it tries
to adjust to the new MTU and triggers an immediate retransmission. But
it didn't consider the fact that MTUs smaller than the SCTP minimum MTU
allowed (512) would not cause the PMTU to change, and issued the
retransmission anyway (thus leading to another ICMP Frag Needed, and so
on).
As IPv4 (ip_rt_min_pmtu=556) and IPv6 (IPV6_MIN_MTU=1280) minimum MTU
are higher than that, sctp_transport_update_pmtu() is changed to
re-fetch the PMTU that got set after our request, and with that, detect
if there was an actual change or not.
The fix, thus, skips the immediate retransmission if the received ICMP
resulted in no change, in the hope that SCTP will select another path.
Note: The value being used for the minimum MTU (512,
SCTP_DEFAULT_MINSEGMENT) is not right and instead it should be (576,
SCTP_MIN_PMTU), but such change belongs to another patch.
Changes from v1:
- do not disable PMTU discovery, in the light of commit
06ad391919b2 ("[SCTP] Don't disable PMTU discovery when mtu is small")
and as suggested by Xin Long.
- changed the way to break the rtx loop by detecting if the icmp
resulted in a change or not
Changes from v2:
none
See-also: https://lkml.org/lkml/2017/12/22/811
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-01-05 21:17:18 +08:00
|
|
|
bool sctp_transport_update_pmtu(struct sctp_transport *t, u32 pmtu)
|
2007-06-08 01:47:03 +08:00
|
|
|
{
|
2017-04-04 13:39:55 +08:00
|
|
|
struct dst_entry *dst = sctp_transport_dst_check(t);
|
2018-09-20 17:27:28 +08:00
|
|
|
struct sock *sk = t->asoc->base.sk;
|
sctp: fix the handling of ICMP Frag Needed for too small MTUs
syzbot reported a hang involving SCTP, on which it kept flooding dmesg
with the message:
[ 246.742374] sctp: sctp_transport_update_pmtu: Reported pmtu 508 too
low, using default minimum of 512
That happened because whenever SCTP hits an ICMP Frag Needed, it tries
to adjust to the new MTU and triggers an immediate retransmission. But
it didn't consider the fact that MTUs smaller than the SCTP minimum MTU
allowed (512) would not cause the PMTU to change, and issued the
retransmission anyway (thus leading to another ICMP Frag Needed, and so
on).
As IPv4 (ip_rt_min_pmtu=556) and IPv6 (IPV6_MIN_MTU=1280) minimum MTU
are higher than that, sctp_transport_update_pmtu() is changed to
re-fetch the PMTU that got set after our request, and with that, detect
if there was an actual change or not.
The fix, thus, skips the immediate retransmission if the received ICMP
resulted in no change, in the hope that SCTP will select another path.
Note: The value being used for the minimum MTU (512,
SCTP_DEFAULT_MINSEGMENT) is not right and instead it should be (576,
SCTP_MIN_PMTU), but such change belongs to another patch.
Changes from v1:
- do not disable PMTU discovery, in the light of commit
06ad391919b2 ("[SCTP] Don't disable PMTU discovery when mtu is small")
and as suggested by Xin Long.
- changed the way to break the rtx loop by detecting if the icmp
resulted in a change or not
Changes from v2:
none
See-also: https://lkml.org/lkml/2017/12/22/811
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-01-05 21:17:18 +08:00
|
|
|
bool change = true;
|
2007-06-08 01:47:03 +08:00
|
|
|
|
|
|
|
if (unlikely(pmtu < SCTP_DEFAULT_MINSEGMENT)) {
|
sctp: fix the handling of ICMP Frag Needed for too small MTUs
syzbot reported a hang involving SCTP, on which it kept flooding dmesg
with the message:
[ 246.742374] sctp: sctp_transport_update_pmtu: Reported pmtu 508 too
low, using default minimum of 512
That happened because whenever SCTP hits an ICMP Frag Needed, it tries
to adjust to the new MTU and triggers an immediate retransmission. But
it didn't consider the fact that MTUs smaller than the SCTP minimum MTU
allowed (512) would not cause the PMTU to change, and issued the
retransmission anyway (thus leading to another ICMP Frag Needed, and so
on).
As IPv4 (ip_rt_min_pmtu=556) and IPv6 (IPV6_MIN_MTU=1280) minimum MTU
are higher than that, sctp_transport_update_pmtu() is changed to
re-fetch the PMTU that got set after our request, and with that, detect
if there was an actual change or not.
The fix, thus, skips the immediate retransmission if the received ICMP
resulted in no change, in the hope that SCTP will select another path.
Note: The value being used for the minimum MTU (512,
SCTP_DEFAULT_MINSEGMENT) is not right and instead it should be (576,
SCTP_MIN_PMTU), but such change belongs to another patch.
Changes from v1:
- do not disable PMTU discovery, in the light of commit
06ad391919b2 ("[SCTP] Don't disable PMTU discovery when mtu is small")
and as suggested by Xin Long.
- changed the way to break the rtx loop by detecting if the icmp
resulted in a change or not
Changes from v2:
none
See-also: https://lkml.org/lkml/2017/12/22/811
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-01-05 21:17:18 +08:00
|
|
|
pr_warn_ratelimited("%s: Reported pmtu %d too low, using default minimum of %d\n",
|
|
|
|
__func__, pmtu, SCTP_DEFAULT_MINSEGMENT);
|
|
|
|
/* Use default minimum segment instead */
|
|
|
|
pmtu = SCTP_DEFAULT_MINSEGMENT;
|
2007-06-08 01:47:03 +08:00
|
|
|
}
|
sctp: fix the handling of ICMP Frag Needed for too small MTUs
syzbot reported a hang involving SCTP, on which it kept flooding dmesg
with the message:
[ 246.742374] sctp: sctp_transport_update_pmtu: Reported pmtu 508 too
low, using default minimum of 512
That happened because whenever SCTP hits an ICMP Frag Needed, it tries
to adjust to the new MTU and triggers an immediate retransmission. But
it didn't consider the fact that MTUs smaller than the SCTP minimum MTU
allowed (512) would not cause the PMTU to change, and issued the
retransmission anyway (thus leading to another ICMP Frag Needed, and so
on).
As IPv4 (ip_rt_min_pmtu=556) and IPv6 (IPV6_MIN_MTU=1280) minimum MTU
are higher than that, sctp_transport_update_pmtu() is changed to
re-fetch the PMTU that got set after our request, and with that, detect
if there was an actual change or not.
The fix, thus, skips the immediate retransmission if the received ICMP
resulted in no change, in the hope that SCTP will select another path.
Note: The value being used for the minimum MTU (512,
SCTP_DEFAULT_MINSEGMENT) is not right and instead it should be (576,
SCTP_MIN_PMTU), but such change belongs to another patch.
Changes from v1:
- do not disable PMTU discovery, in the light of commit
06ad391919b2 ("[SCTP] Don't disable PMTU discovery when mtu is small")
and as suggested by Xin Long.
- changed the way to break the rtx loop by detecting if the icmp
resulted in a change or not
Changes from v2:
none
See-also: https://lkml.org/lkml/2017/12/22/811
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-01-05 21:17:18 +08:00
|
|
|
pmtu = SCTP_TRUNC4(pmtu);
|
2007-06-08 01:47:03 +08:00
|
|
|
|
2012-07-16 18:57:14 +08:00
|
|
|
if (dst) {
|
2018-09-20 17:27:28 +08:00
|
|
|
struct sctp_pf *pf = sctp_get_pf_specific(dst->ops->family);
|
|
|
|
union sctp_addr addr;
|
|
|
|
|
|
|
|
pf->af->from_sk(&addr, sk);
|
|
|
|
pf->to_sk_daddr(&t->ipaddr, sk);
|
2019-12-22 10:51:09 +08:00
|
|
|
dst->ops->update_pmtu(dst, sk, NULL, pmtu, true);
|
2018-09-20 17:27:28 +08:00
|
|
|
pf->to_sk_daddr(&addr, sk);
|
|
|
|
|
2012-07-16 18:57:14 +08:00
|
|
|
dst = sctp_transport_dst_check(t);
|
|
|
|
}
|
2017-04-04 13:39:55 +08:00
|
|
|
|
sctp: fix the handling of ICMP Frag Needed for too small MTUs
syzbot reported a hang involving SCTP, on which it kept flooding dmesg
with the message:
[ 246.742374] sctp: sctp_transport_update_pmtu: Reported pmtu 508 too
low, using default minimum of 512
That happened because whenever SCTP hits an ICMP Frag Needed, it tries
to adjust to the new MTU and triggers an immediate retransmission. But
it didn't consider the fact that MTUs smaller than the SCTP minimum MTU
allowed (512) would not cause the PMTU to change, and issued the
retransmission anyway (thus leading to another ICMP Frag Needed, and so
on).
As IPv4 (ip_rt_min_pmtu=556) and IPv6 (IPV6_MIN_MTU=1280) minimum MTU
are higher than that, sctp_transport_update_pmtu() is changed to
re-fetch the PMTU that got set after our request, and with that, detect
if there was an actual change or not.
The fix, thus, skips the immediate retransmission if the received ICMP
resulted in no change, in the hope that SCTP will select another path.
Note: The value being used for the minimum MTU (512,
SCTP_DEFAULT_MINSEGMENT) is not right and instead it should be (576,
SCTP_MIN_PMTU), but such change belongs to another patch.
Changes from v1:
- do not disable PMTU discovery, in the light of commit
06ad391919b2 ("[SCTP] Don't disable PMTU discovery when mtu is small")
and as suggested by Xin Long.
- changed the way to break the rtx loop by detecting if the icmp
resulted in a change or not
Changes from v2:
none
See-also: https://lkml.org/lkml/2017/12/22/811
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-01-05 21:17:18 +08:00
|
|
|
if (!dst) {
|
2018-09-20 17:27:28 +08:00
|
|
|
t->af_specific->get_dst(t, &t->saddr, &t->fl, sk);
|
sctp: fix the handling of ICMP Frag Needed for too small MTUs
syzbot reported a hang involving SCTP, on which it kept flooding dmesg
with the message:
[ 246.742374] sctp: sctp_transport_update_pmtu: Reported pmtu 508 too
low, using default minimum of 512
That happened because whenever SCTP hits an ICMP Frag Needed, it tries
to adjust to the new MTU and triggers an immediate retransmission. But
it didn't consider the fact that MTUs smaller than the SCTP minimum MTU
allowed (512) would not cause the PMTU to change, and issued the
retransmission anyway (thus leading to another ICMP Frag Needed, and so
on).
As IPv4 (ip_rt_min_pmtu=556) and IPv6 (IPV6_MIN_MTU=1280) minimum MTU
are higher than that, sctp_transport_update_pmtu() is changed to
re-fetch the PMTU that got set after our request, and with that, detect
if there was an actual change or not.
The fix, thus, skips the immediate retransmission if the received ICMP
resulted in no change, in the hope that SCTP will select another path.
Note: The value being used for the minimum MTU (512,
SCTP_DEFAULT_MINSEGMENT) is not right and instead it should be (576,
SCTP_MIN_PMTU), but such change belongs to another patch.
Changes from v1:
- do not disable PMTU discovery, in the light of commit
06ad391919b2 ("[SCTP] Don't disable PMTU discovery when mtu is small")
and as suggested by Xin Long.
- changed the way to break the rtx loop by detecting if the icmp
resulted in a change or not
Changes from v2:
none
See-also: https://lkml.org/lkml/2017/12/22/811
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-01-05 21:17:18 +08:00
|
|
|
dst = t->dst;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (dst) {
|
|
|
|
/* Re-fetch, as under layers may have a higher minimum size */
|
2018-07-03 16:30:47 +08:00
|
|
|
pmtu = sctp_dst_mtu(dst);
|
sctp: fix the handling of ICMP Frag Needed for too small MTUs
syzbot reported a hang involving SCTP, on which it kept flooding dmesg
with the message:
[ 246.742374] sctp: sctp_transport_update_pmtu: Reported pmtu 508 too
low, using default minimum of 512
That happened because whenever SCTP hits an ICMP Frag Needed, it tries
to adjust to the new MTU and triggers an immediate retransmission. But
it didn't consider the fact that MTUs smaller than the SCTP minimum MTU
allowed (512) would not cause the PMTU to change, and issued the
retransmission anyway (thus leading to another ICMP Frag Needed, and so
on).
As IPv4 (ip_rt_min_pmtu=556) and IPv6 (IPV6_MIN_MTU=1280) minimum MTU
are higher than that, sctp_transport_update_pmtu() is changed to
re-fetch the PMTU that got set after our request, and with that, detect
if there was an actual change or not.
The fix, thus, skips the immediate retransmission if the received ICMP
resulted in no change, in the hope that SCTP will select another path.
Note: The value being used for the minimum MTU (512,
SCTP_DEFAULT_MINSEGMENT) is not right and instead it should be (576,
SCTP_MIN_PMTU), but such change belongs to another patch.
Changes from v1:
- do not disable PMTU discovery, in the light of commit
06ad391919b2 ("[SCTP] Don't disable PMTU discovery when mtu is small")
and as suggested by Xin Long.
- changed the way to break the rtx loop by detecting if the icmp
resulted in a change or not
Changes from v2:
none
See-also: https://lkml.org/lkml/2017/12/22/811
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2018-01-05 21:17:18 +08:00
|
|
|
change = t->pathmtu != pmtu;
|
|
|
|
}
|
|
|
|
t->pathmtu = pmtu;
|
|
|
|
|
|
|
|
return change;
|
2007-06-08 01:47:03 +08:00
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* Caches the dst entry and source address for a transport's destination
|
|
|
|
* address.
|
|
|
|
*/
|
|
|
|
void sctp_transport_route(struct sctp_transport *transport,
|
|
|
|
union sctp_addr *saddr, struct sctp_sock *opt)
|
|
|
|
{
|
|
|
|
struct sctp_association *asoc = transport->asoc;
|
|
|
|
struct sctp_af *af = transport->af_specific;
|
|
|
|
|
2018-04-27 03:58:59 +08:00
|
|
|
sctp_transport_dst_release(transport);
|
2011-05-07 07:32:47 +08:00
|
|
|
af->get_dst(transport, saddr, &transport->fl, sctp_opt2sk(opt));
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (saddr)
|
|
|
|
memcpy(&transport->saddr, saddr, sizeof(union sctp_addr));
|
|
|
|
else
|
2011-05-07 07:32:47 +08:00
|
|
|
af->get_saddr(opt, transport, &transport->fl);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
2018-04-27 03:58:59 +08:00
|
|
|
sctp_transport_pmtu(transport, sctp_opt2sk(opt));
|
|
|
|
|
|
|
|
/* Initialize sk->sk_rcv_saddr, if the transport is the
|
|
|
|
* association's active path for getsockname().
|
|
|
|
*/
|
|
|
|
if (transport->dst && asoc &&
|
|
|
|
(!asoc->peer.primary_path || transport == asoc->peer.active_path))
|
|
|
|
opt->pf->to_sk_saddr(&transport->saddr, asoc->base.sk);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Hold a reference to a transport. */
|
2016-01-22 01:49:07 +08:00
|
|
|
int sctp_transport_hold(struct sctp_transport *transport)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2017-07-04 20:53:27 +08:00
|
|
|
return refcount_inc_not_zero(&transport->refcnt);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Release a reference to a transport and clean up
|
|
|
|
* if there are no more references.
|
|
|
|
*/
|
|
|
|
void sctp_transport_put(struct sctp_transport *transport)
|
|
|
|
{
|
2017-07-04 20:53:27 +08:00
|
|
|
if (refcount_dec_and_test(&transport->refcnt))
|
2005-04-17 06:20:36 +08:00
|
|
|
sctp_transport_destroy(transport);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Update transport's RTO based on the newly calculated RTT. */
|
|
|
|
void sctp_transport_update_rto(struct sctp_transport *tp, __u32 rtt)
|
|
|
|
{
|
net: sctp: rework debugging framework to use pr_debug and friends
We should get rid of all own SCTP debug printk macros and use the ones
that the kernel offers anyway instead. This makes the code more readable
and conform to the kernel code, and offers all the features of dynamic
debbuging that pr_debug() et al has, such as only turning on/off portions
of debug messages at runtime through debugfs. The runtime cost of having
CONFIG_DYNAMIC_DEBUG enabled, but none of the debug statements printing,
is negligible [1]. If kernel debugging is completly turned off, then these
statements will also compile into "empty" functions.
While we're at it, we also need to change the Kconfig option as it /now/
only refers to the ifdef'ed code portions in outqueue.c that enable further
debugging/tracing of SCTP transaction fields. Also, since SCTP_ASSERT code
was enabled with this Kconfig option and has now been removed, we
transform those code parts into WARNs resp. where appropriate BUG_ONs so
that those bugs can be more easily detected as probably not many people
have SCTP debugging permanently turned on.
To turn on all SCTP debugging, the following steps are needed:
# mount -t debugfs none /sys/kernel/debug
# echo -n 'module sctp +p' > /sys/kernel/debug/dynamic_debug/control
This can be done more fine-grained on a per file, per line basis and others
as described in [2].
[1] https://www.kernel.org/doc/ols/2009/ols2009-pages-39-46.pdf
[2] Documentation/dynamic-debug-howto.txt
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-29 01:49:40 +08:00
|
|
|
if (unlikely(!tp->rto_pending))
|
|
|
|
/* We should not be doing any RTO updates unless rto_pending is set. */
|
|
|
|
pr_debug("%s: rto_pending not set on transport %p!\n", __func__, tp);
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (tp->rttvar || tp->srtt) {
|
2019-12-09 13:45:18 +08:00
|
|
|
struct net *net = tp->asoc->base.net;
|
2005-04-17 06:20:36 +08:00
|
|
|
/* 6.3.1 C3) When a new RTT measurement R' is made, set
|
|
|
|
* RTTVAR <- (1 - RTO.Beta) * RTTVAR + RTO.Beta * |SRTT - R'|
|
|
|
|
* SRTT <- (1 - RTO.Alpha) * SRTT + RTO.Alpha * R'
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* Note: The above algorithm has been rewritten to
|
|
|
|
* express rto_beta and rto_alpha as inverse powers
|
|
|
|
* of two.
|
|
|
|
* For example, assuming the default value of RTO.Alpha of
|
|
|
|
* 1/8, rto_alpha would be expressed as 3.
|
|
|
|
*/
|
2012-08-07 15:29:57 +08:00
|
|
|
tp->rttvar = tp->rttvar - (tp->rttvar >> net->sctp.rto_beta)
|
2015-11-10 06:58:13 +08:00
|
|
|
+ (((__u32)abs((__s64)tp->srtt - (__s64)rtt)) >> net->sctp.rto_beta);
|
2012-08-07 15:29:57 +08:00
|
|
|
tp->srtt = tp->srtt - (tp->srtt >> net->sctp.rto_alpha)
|
|
|
|
+ (rtt >> net->sctp.rto_alpha);
|
2005-04-17 06:20:36 +08:00
|
|
|
} else {
|
|
|
|
/* 6.3.1 C2) When the first RTT measurement R is made, set
|
|
|
|
* SRTT <- R, RTTVAR <- R/2.
|
|
|
|
*/
|
|
|
|
tp->srtt = rtt;
|
|
|
|
tp->rttvar = rtt >> 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* 6.3.1 G1) Whenever RTTVAR is computed, if RTTVAR = 0, then
|
|
|
|
* adjust RTTVAR <- G, where G is the CLOCK GRANULARITY.
|
|
|
|
*/
|
|
|
|
if (tp->rttvar == 0)
|
|
|
|
tp->rttvar = SCTP_CLOCK_GRANULARITY;
|
|
|
|
|
|
|
|
/* 6.3.1 C3) After the computation, update RTO <- SRTT + 4 * RTTVAR. */
|
|
|
|
tp->rto = tp->srtt + (tp->rttvar << 2);
|
|
|
|
|
|
|
|
/* 6.3.1 C6) Whenever RTO is computed, if it is less than RTO.Min
|
|
|
|
* seconds then it is rounded up to RTO.Min seconds.
|
|
|
|
*/
|
|
|
|
if (tp->rto < tp->asoc->rto_min)
|
|
|
|
tp->rto = tp->asoc->rto_min;
|
|
|
|
|
|
|
|
/* 6.3.1 C7) A maximum value may be placed on RTO provided it is
|
|
|
|
* at least RTO.max seconds.
|
|
|
|
*/
|
|
|
|
if (tp->rto > tp->asoc->rto_max)
|
|
|
|
tp->rto = tp->asoc->rto_max;
|
|
|
|
|
2012-12-01 12:49:42 +08:00
|
|
|
sctp_max_rto(tp->asoc, tp);
|
2005-04-17 06:20:36 +08:00
|
|
|
tp->rtt = rtt;
|
|
|
|
|
|
|
|
/* Reset rto_pending so that a new RTT measurement is started when a
|
|
|
|
* new data chunk is sent.
|
|
|
|
*/
|
|
|
|
tp->rto_pending = 0;
|
|
|
|
|
net: sctp: rework debugging framework to use pr_debug and friends
We should get rid of all own SCTP debug printk macros and use the ones
that the kernel offers anyway instead. This makes the code more readable
and conform to the kernel code, and offers all the features of dynamic
debbuging that pr_debug() et al has, such as only turning on/off portions
of debug messages at runtime through debugfs. The runtime cost of having
CONFIG_DYNAMIC_DEBUG enabled, but none of the debug statements printing,
is negligible [1]. If kernel debugging is completly turned off, then these
statements will also compile into "empty" functions.
While we're at it, we also need to change the Kconfig option as it /now/
only refers to the ifdef'ed code portions in outqueue.c that enable further
debugging/tracing of SCTP transaction fields. Also, since SCTP_ASSERT code
was enabled with this Kconfig option and has now been removed, we
transform those code parts into WARNs resp. where appropriate BUG_ONs so
that those bugs can be more easily detected as probably not many people
have SCTP debugging permanently turned on.
To turn on all SCTP debugging, the following steps are needed:
# mount -t debugfs none /sys/kernel/debug
# echo -n 'module sctp +p' > /sys/kernel/debug/dynamic_debug/control
This can be done more fine-grained on a per file, per line basis and others
as described in [2].
[1] https://www.kernel.org/doc/ols/2009/ols2009-pages-39-46.pdf
[2] Documentation/dynamic-debug-howto.txt
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-29 01:49:40 +08:00
|
|
|
pr_debug("%s: transport:%p, rtt:%d, srtt:%d rttvar:%d, rto:%ld\n",
|
|
|
|
__func__, tp, rtt, tp->srtt, tp->rttvar, tp->rto);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
/* This routine updates the transport's cwnd and partial_bytes_acked
|
|
|
|
* parameters based on the bytes acked in the received SACK.
|
|
|
|
*/
|
|
|
|
void sctp_transport_raise_cwnd(struct sctp_transport *transport,
|
|
|
|
__u32 sack_ctsn, __u32 bytes_acked)
|
|
|
|
{
|
2010-05-01 10:41:10 +08:00
|
|
|
struct sctp_association *asoc = transport->asoc;
|
2005-04-17 06:20:36 +08:00
|
|
|
__u32 cwnd, ssthresh, flight_size, pba, pmtu;
|
|
|
|
|
|
|
|
cwnd = transport->cwnd;
|
|
|
|
flight_size = transport->flight_size;
|
|
|
|
|
2008-06-05 03:38:43 +08:00
|
|
|
/* See if we need to exit Fast Recovery first */
|
2010-05-01 10:41:10 +08:00
|
|
|
if (asoc->fast_recovery &&
|
|
|
|
TSN_lte(asoc->fast_recovery_exit, sack_ctsn))
|
|
|
|
asoc->fast_recovery = 0;
|
2008-06-05 03:38:43 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
ssthresh = transport->ssthresh;
|
|
|
|
pba = transport->partial_bytes_acked;
|
2005-12-23 03:36:46 +08:00
|
|
|
pmtu = transport->asoc->pathmtu;
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
if (cwnd <= ssthresh) {
|
2008-06-05 03:38:43 +08:00
|
|
|
/* RFC 4960 7.2.1
|
|
|
|
* o When cwnd is less than or equal to ssthresh, an SCTP
|
|
|
|
* endpoint MUST use the slow-start algorithm to increase
|
|
|
|
* cwnd only if the current congestion window is being fully
|
|
|
|
* utilized, an incoming SACK advances the Cumulative TSN
|
|
|
|
* Ack Point, and the data sender is not in Fast Recovery.
|
|
|
|
* Only when these three conditions are met can the cwnd be
|
|
|
|
* increased; otherwise, the cwnd MUST not be increased.
|
|
|
|
* If these conditions are met, then cwnd MUST be increased
|
|
|
|
* by, at most, the lesser of 1) the total size of the
|
|
|
|
* previously outstanding DATA chunk(s) acknowledged, and
|
|
|
|
* 2) the destination's path MTU. This upper bound protects
|
|
|
|
* against the ACK-Splitting attack outlined in [SAVAGE99].
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
2010-05-01 10:41:10 +08:00
|
|
|
if (asoc->fast_recovery)
|
2008-06-05 03:38:43 +08:00
|
|
|
return;
|
|
|
|
|
2017-06-24 06:59:35 +08:00
|
|
|
/* The appropriate cwnd increase algorithm is performed
|
|
|
|
* if, and only if the congestion window is being fully
|
|
|
|
* utilized. Note that RFC4960 Errata 3.22 removed the
|
|
|
|
* other condition on ctsn moving.
|
|
|
|
*/
|
|
|
|
if (flight_size < cwnd)
|
|
|
|
return;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
if (bytes_acked > pmtu)
|
|
|
|
cwnd += pmtu;
|
|
|
|
else
|
|
|
|
cwnd += bytes_acked;
|
net: sctp: rework debugging framework to use pr_debug and friends
We should get rid of all own SCTP debug printk macros and use the ones
that the kernel offers anyway instead. This makes the code more readable
and conform to the kernel code, and offers all the features of dynamic
debbuging that pr_debug() et al has, such as only turning on/off portions
of debug messages at runtime through debugfs. The runtime cost of having
CONFIG_DYNAMIC_DEBUG enabled, but none of the debug statements printing,
is negligible [1]. If kernel debugging is completly turned off, then these
statements will also compile into "empty" functions.
While we're at it, we also need to change the Kconfig option as it /now/
only refers to the ifdef'ed code portions in outqueue.c that enable further
debugging/tracing of SCTP transaction fields. Also, since SCTP_ASSERT code
was enabled with this Kconfig option and has now been removed, we
transform those code parts into WARNs resp. where appropriate BUG_ONs so
that those bugs can be more easily detected as probably not many people
have SCTP debugging permanently turned on.
To turn on all SCTP debugging, the following steps are needed:
# mount -t debugfs none /sys/kernel/debug
# echo -n 'module sctp +p' > /sys/kernel/debug/dynamic_debug/control
This can be done more fine-grained on a per file, per line basis and others
as described in [2].
[1] https://www.kernel.org/doc/ols/2009/ols2009-pages-39-46.pdf
[2] Documentation/dynamic-debug-howto.txt
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-29 01:49:40 +08:00
|
|
|
|
|
|
|
pr_debug("%s: slow start: transport:%p, bytes_acked:%d, "
|
|
|
|
"cwnd:%d, ssthresh:%d, flight_size:%d, pba:%d\n",
|
|
|
|
__func__, transport, bytes_acked, cwnd, ssthresh,
|
|
|
|
flight_size, pba);
|
2005-04-17 06:20:36 +08:00
|
|
|
} else {
|
|
|
|
/* RFC 2960 7.2.2 Whenever cwnd is greater than ssthresh,
|
2017-06-24 06:59:34 +08:00
|
|
|
* upon each SACK arrival, increase partial_bytes_acked
|
|
|
|
* by the total number of bytes of all new chunks
|
|
|
|
* acknowledged in that SACK including chunks
|
|
|
|
* acknowledged by the new Cumulative TSN Ack and by Gap
|
|
|
|
* Ack Blocks. (updated by RFC4960 Errata 3.22)
|
2005-04-17 06:20:36 +08:00
|
|
|
*
|
2017-06-24 06:59:35 +08:00
|
|
|
* When partial_bytes_acked is greater than cwnd and
|
|
|
|
* before the arrival of the SACK the sender had less
|
|
|
|
* bytes of data outstanding than cwnd (i.e., before
|
|
|
|
* arrival of the SACK, flightsize was less than cwnd),
|
|
|
|
* reset partial_bytes_acked to cwnd. (RFC 4960 Errata
|
|
|
|
* 3.26)
|
|
|
|
*
|
2017-06-24 06:59:33 +08:00
|
|
|
* When partial_bytes_acked is equal to or greater than
|
|
|
|
* cwnd and before the arrival of the SACK the sender
|
|
|
|
* had cwnd or more bytes of data outstanding (i.e.,
|
|
|
|
* before arrival of the SACK, flightsize was greater
|
|
|
|
* than or equal to cwnd), partial_bytes_acked is reset
|
|
|
|
* to (partial_bytes_acked - cwnd). Next, cwnd is
|
|
|
|
* increased by MTU. (RFC 4960 Errata 3.12)
|
2005-04-17 06:20:36 +08:00
|
|
|
*/
|
|
|
|
pba += bytes_acked;
|
2017-06-24 06:59:35 +08:00
|
|
|
if (pba > cwnd && flight_size < cwnd)
|
|
|
|
pba = cwnd;
|
|
|
|
if (pba >= cwnd && flight_size >= cwnd) {
|
2017-06-24 06:59:33 +08:00
|
|
|
pba = pba - cwnd;
|
2005-04-17 06:20:36 +08:00
|
|
|
cwnd += pmtu;
|
|
|
|
}
|
net: sctp: rework debugging framework to use pr_debug and friends
We should get rid of all own SCTP debug printk macros and use the ones
that the kernel offers anyway instead. This makes the code more readable
and conform to the kernel code, and offers all the features of dynamic
debbuging that pr_debug() et al has, such as only turning on/off portions
of debug messages at runtime through debugfs. The runtime cost of having
CONFIG_DYNAMIC_DEBUG enabled, but none of the debug statements printing,
is negligible [1]. If kernel debugging is completly turned off, then these
statements will also compile into "empty" functions.
While we're at it, we also need to change the Kconfig option as it /now/
only refers to the ifdef'ed code portions in outqueue.c that enable further
debugging/tracing of SCTP transaction fields. Also, since SCTP_ASSERT code
was enabled with this Kconfig option and has now been removed, we
transform those code parts into WARNs resp. where appropriate BUG_ONs so
that those bugs can be more easily detected as probably not many people
have SCTP debugging permanently turned on.
To turn on all SCTP debugging, the following steps are needed:
# mount -t debugfs none /sys/kernel/debug
# echo -n 'module sctp +p' > /sys/kernel/debug/dynamic_debug/control
This can be done more fine-grained on a per file, per line basis and others
as described in [2].
[1] https://www.kernel.org/doc/ols/2009/ols2009-pages-39-46.pdf
[2] Documentation/dynamic-debug-howto.txt
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-29 01:49:40 +08:00
|
|
|
|
|
|
|
pr_debug("%s: congestion avoidance: transport:%p, "
|
|
|
|
"bytes_acked:%d, cwnd:%d, ssthresh:%d, "
|
|
|
|
"flight_size:%d, pba:%d\n", __func__,
|
|
|
|
transport, bytes_acked, cwnd, ssthresh,
|
|
|
|
flight_size, pba);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
transport->cwnd = cwnd;
|
|
|
|
transport->partial_bytes_acked = pba;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* This routine is used to lower the transport's cwnd when congestion is
|
|
|
|
* detected.
|
|
|
|
*/
|
|
|
|
void sctp_transport_lower_cwnd(struct sctp_transport *transport,
|
2017-08-05 19:59:51 +08:00
|
|
|
enum sctp_lower_cwnd reason)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
2010-05-01 10:41:10 +08:00
|
|
|
struct sctp_association *asoc = transport->asoc;
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
switch (reason) {
|
|
|
|
case SCTP_LOWER_CWND_T3_RTX:
|
|
|
|
/* RFC 2960 Section 7.2.3, sctpimpguide
|
|
|
|
* When the T3-rtx timer expires on an address, SCTP should
|
|
|
|
* perform slow start by:
|
|
|
|
* ssthresh = max(cwnd/2, 4*MTU)
|
|
|
|
* cwnd = 1*MTU
|
|
|
|
* partial_bytes_acked = 0
|
|
|
|
*/
|
|
|
|
transport->ssthresh = max(transport->cwnd/2,
|
2010-05-01 10:41:10 +08:00
|
|
|
4*asoc->pathmtu);
|
|
|
|
transport->cwnd = asoc->pathmtu;
|
2009-09-05 06:20:58 +08:00
|
|
|
|
2010-05-01 10:41:10 +08:00
|
|
|
/* T3-rtx also clears fast recovery */
|
|
|
|
asoc->fast_recovery = 0;
|
2005-04-17 06:20:36 +08:00
|
|
|
break;
|
|
|
|
|
|
|
|
case SCTP_LOWER_CWND_FAST_RTX:
|
|
|
|
/* RFC 2960 7.2.4 Adjust the ssthresh and cwnd of the
|
|
|
|
* destination address(es) to which the missing DATA chunks
|
|
|
|
* were last sent, according to the formula described in
|
|
|
|
* Section 7.2.3.
|
2007-02-09 22:25:18 +08:00
|
|
|
*
|
|
|
|
* RFC 2960 7.2.3, sctpimpguide Upon detection of packet
|
2005-04-17 06:20:36 +08:00
|
|
|
* losses from SACK (see Section 7.2.4), An endpoint
|
|
|
|
* should do the following:
|
|
|
|
* ssthresh = max(cwnd/2, 4*MTU)
|
|
|
|
* cwnd = ssthresh
|
|
|
|
* partial_bytes_acked = 0
|
|
|
|
*/
|
2010-05-01 10:41:10 +08:00
|
|
|
if (asoc->fast_recovery)
|
2008-06-05 03:38:43 +08:00
|
|
|
return;
|
|
|
|
|
|
|
|
/* Mark Fast recovery */
|
2010-05-01 10:41:10 +08:00
|
|
|
asoc->fast_recovery = 1;
|
|
|
|
asoc->fast_recovery_exit = asoc->next_tsn - 1;
|
2008-06-05 03:38:43 +08:00
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
transport->ssthresh = max(transport->cwnd/2,
|
2010-05-01 10:41:10 +08:00
|
|
|
4*asoc->pathmtu);
|
2005-04-17 06:20:36 +08:00
|
|
|
transport->cwnd = transport->ssthresh;
|
|
|
|
break;
|
|
|
|
|
|
|
|
case SCTP_LOWER_CWND_ECNE:
|
|
|
|
/* RFC 2481 Section 6.1.2.
|
|
|
|
* If the sender receives an ECN-Echo ACK packet
|
|
|
|
* then the sender knows that congestion was encountered in the
|
|
|
|
* network on the path from the sender to the receiver. The
|
|
|
|
* indication of congestion should be treated just as a
|
|
|
|
* congestion loss in non-ECN Capable TCP. That is, the TCP
|
|
|
|
* source halves the congestion window "cwnd" and reduces the
|
|
|
|
* slow start threshold "ssthresh".
|
|
|
|
* A critical condition is that TCP does not react to
|
|
|
|
* congestion indications more than once every window of
|
|
|
|
* data (or more loosely more than once every round-trip time).
|
|
|
|
*/
|
2009-03-02 17:46:13 +08:00
|
|
|
if (time_after(jiffies, transport->last_time_ecne_reduced +
|
|
|
|
transport->rtt)) {
|
2005-04-17 06:20:36 +08:00
|
|
|
transport->ssthresh = max(transport->cwnd/2,
|
2010-05-01 10:41:10 +08:00
|
|
|
4*asoc->pathmtu);
|
2005-04-17 06:20:36 +08:00
|
|
|
transport->cwnd = transport->ssthresh;
|
|
|
|
transport->last_time_ecne_reduced = jiffies;
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
|
|
|
|
case SCTP_LOWER_CWND_INACTIVE:
|
|
|
|
/* RFC 2960 Section 7.2.1, sctpimpguide
|
|
|
|
* When the endpoint does not transmit data on a given
|
|
|
|
* transport address, the cwnd of the transport address
|
|
|
|
* should be adjusted to max(cwnd/2, 4*MTU) per RTO.
|
|
|
|
* NOTE: Although the draft recommends that this check needs
|
|
|
|
* to be done every RTO interval, we do it every hearbeat
|
|
|
|
* interval.
|
|
|
|
*/
|
2009-11-24 04:53:58 +08:00
|
|
|
transport->cwnd = max(transport->cwnd/2,
|
2010-05-01 10:41:10 +08:00
|
|
|
4*asoc->pathmtu);
|
2017-06-24 06:59:36 +08:00
|
|
|
/* RFC 4960 Errata 3.27.2: also adjust sshthresh */
|
|
|
|
transport->ssthresh = transport->cwnd;
|
2005-04-17 06:20:36 +08:00
|
|
|
break;
|
2007-04-21 08:09:22 +08:00
|
|
|
}
|
2005-04-17 06:20:36 +08:00
|
|
|
|
|
|
|
transport->partial_bytes_acked = 0;
|
net: sctp: rework debugging framework to use pr_debug and friends
We should get rid of all own SCTP debug printk macros and use the ones
that the kernel offers anyway instead. This makes the code more readable
and conform to the kernel code, and offers all the features of dynamic
debbuging that pr_debug() et al has, such as only turning on/off portions
of debug messages at runtime through debugfs. The runtime cost of having
CONFIG_DYNAMIC_DEBUG enabled, but none of the debug statements printing,
is negligible [1]. If kernel debugging is completly turned off, then these
statements will also compile into "empty" functions.
While we're at it, we also need to change the Kconfig option as it /now/
only refers to the ifdef'ed code portions in outqueue.c that enable further
debugging/tracing of SCTP transaction fields. Also, since SCTP_ASSERT code
was enabled with this Kconfig option and has now been removed, we
transform those code parts into WARNs resp. where appropriate BUG_ONs so
that those bugs can be more easily detected as probably not many people
have SCTP debugging permanently turned on.
To turn on all SCTP debugging, the following steps are needed:
# mount -t debugfs none /sys/kernel/debug
# echo -n 'module sctp +p' > /sys/kernel/debug/dynamic_debug/control
This can be done more fine-grained on a per file, per line basis and others
as described in [2].
[1] https://www.kernel.org/doc/ols/2009/ols2009-pages-39-46.pdf
[2] Documentation/dynamic-debug-howto.txt
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2013-06-29 01:49:40 +08:00
|
|
|
|
|
|
|
pr_debug("%s: transport:%p, reason:%d, cwnd:%d, ssthresh:%d\n",
|
|
|
|
__func__, transport, reason, transport->cwnd,
|
|
|
|
transport->ssthresh);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
|
|
|
|
2009-11-24 04:54:00 +08:00
|
|
|
/* Apply Max.Burst limit to the congestion window:
|
|
|
|
* sctpimpguide-05 2.14.2
|
|
|
|
* D) When the time comes for the sender to
|
|
|
|
* transmit new DATA chunks, the protocol parameter Max.Burst MUST
|
|
|
|
* first be applied to limit how many new DATA chunks may be sent.
|
|
|
|
* The limit is applied by adjusting cwnd as follows:
|
|
|
|
* if ((flightsize+ Max.Burst * MTU) < cwnd)
|
|
|
|
* cwnd = flightsize + Max.Burst * MTU
|
|
|
|
*/
|
|
|
|
|
|
|
|
void sctp_transport_burst_limited(struct sctp_transport *t)
|
|
|
|
{
|
|
|
|
struct sctp_association *asoc = t->asoc;
|
|
|
|
u32 old_cwnd = t->cwnd;
|
|
|
|
u32 max_burst_bytes;
|
|
|
|
|
2013-12-04 17:32:39 +08:00
|
|
|
if (t->burst_limited || asoc->max_burst == 0)
|
2009-11-24 04:54:00 +08:00
|
|
|
return;
|
|
|
|
|
|
|
|
max_burst_bytes = t->flight_size + (asoc->max_burst * asoc->pathmtu);
|
|
|
|
if (max_burst_bytes < old_cwnd) {
|
|
|
|
t->cwnd = max_burst_bytes;
|
|
|
|
t->burst_limited = old_cwnd;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Restore the old cwnd congestion window, after the burst had it's
|
|
|
|
* desired effect.
|
|
|
|
*/
|
|
|
|
void sctp_transport_burst_reset(struct sctp_transport *t)
|
|
|
|
{
|
|
|
|
if (t->burst_limited) {
|
|
|
|
t->cwnd = t->burst_limited;
|
|
|
|
t->burst_limited = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-04-17 06:20:36 +08:00
|
|
|
/* What is the next timeout value for this transport? */
|
net: sctp: improve timer slack calculation for transport HBs
RFC4960, section 8.3 says:
On an idle destination address that is allowed to heartbeat,
it is recommended that a HEARTBEAT chunk is sent once per RTO
of that destination address plus the protocol parameter
'HB.interval', with jittering of +/- 50% of the RTO value,
and exponential backoff of the RTO if the previous HEARTBEAT
is unanswered.
Currently, we calculate jitter via sctp_jitter() function first,
and then add its result to the current RTO for the new timeout:
TMO = RTO + (RAND() % RTO) - (RTO / 2)
`------------------------^-=> sctp_jitter()
Instead, we can just simplify all this by directly calculating:
TMO = (RTO / 2) + (RAND() % RTO)
With the help of prandom_u32_max(), we don't need to open code
our own global PRNG, but can instead just make use of the per
CPU implementation of prandom with better quality numbers. Also,
we can now spare us the conditional for divide by zero check
since no div or mod operation needs to be used. Note that
prandom_u32_max() won't emit the same result as a mod operation,
but we really don't care here as we only want to have a random
number scaled into RTO interval.
Note, exponential RTO backoff is handeled elsewhere, namely in
sctp_do_8_2_transport_strike().
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-06-30 19:52:08 +08:00
|
|
|
unsigned long sctp_transport_timeout(struct sctp_transport *trans)
|
2005-04-17 06:20:36 +08:00
|
|
|
{
|
net: sctp: improve timer slack calculation for transport HBs
RFC4960, section 8.3 says:
On an idle destination address that is allowed to heartbeat,
it is recommended that a HEARTBEAT chunk is sent once per RTO
of that destination address plus the protocol parameter
'HB.interval', with jittering of +/- 50% of the RTO value,
and exponential backoff of the RTO if the previous HEARTBEAT
is unanswered.
Currently, we calculate jitter via sctp_jitter() function first,
and then add its result to the current RTO for the new timeout:
TMO = RTO + (RAND() % RTO) - (RTO / 2)
`------------------------^-=> sctp_jitter()
Instead, we can just simplify all this by directly calculating:
TMO = (RTO / 2) + (RAND() % RTO)
With the help of prandom_u32_max(), we don't need to open code
our own global PRNG, but can instead just make use of the per
CPU implementation of prandom with better quality numbers. Also,
we can now spare us the conditional for divide by zero check
since no div or mod operation needs to be used. Note that
prandom_u32_max() won't emit the same result as a mod operation,
but we really don't care here as we only want to have a random
number scaled into RTO interval.
Note, exponential RTO backoff is handeled elsewhere, namely in
sctp_do_8_2_transport_strike().
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-06-30 19:52:08 +08:00
|
|
|
/* RTO + timer slack +/- 50% of RTO */
|
sctp: avoid refreshing heartbeat timer too often
Currently on high rate SCTP streams the heartbeat timer refresh can
consume quite a lot of resources as timer updates are costly and it
contains a random factor, which a) is also costly and b) invalidates
mod_timer() optimization for not editing a timer to the same value.
It may even cause the timer to be slightly advanced, for no good reason.
As suggested by David Laight this patch now removes this timer update
from hot path by leaving the timer on and re-evaluating upon its
expiration if the heartbeat is still needed or not, similarly to what is
done for TCP. If it's not needed anymore the timer is re-scheduled to
the new timeout, considering the time already elapsed.
For this, we now record the last tx timestamp per transport, updated in
the same spots as hb timer was restarted on tx. Also split up
sctp_transport_reset_timers into sctp_transport_reset_t3_rtx and
sctp_transport_reset_hb_timer, so we can re-arm T3 without re-arming the
heartbeat one.
On loopback with MTU of 65535 and data chunks with 1636, so that we
have a considerable amount of chunks without stressing system calls,
netperf -t SCTP_STREAM -l 30, perf looked like this before:
Samples: 103K of event 'cpu-clock', Event count (approx.): 25833000000
Overhead Command Shared Object Symbol
+ 6,15% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,43% netperf [kernel.vmlinux] [k] _raw_write_unlock_irqrestore
- _raw_write_unlock_irqrestore
- 96,54% _raw_spin_unlock_irqrestore
- 36,14% mod_timer
+ 97,24% sctp_transport_reset_timers
+ 2,76% sctp_do_sm
+ 33,65% __wake_up_sync_key
+ 28,77% sctp_ulpq_tail_event
+ 1,40% del_timer
- 1,84% mod_timer
+ 99,03% sctp_transport_reset_timers
+ 0,97% sctp_do_sm
+ 1,50% sctp_ulpq_tail_event
And after this patch, now with netperf -l 60:
Samples: 230K of event 'cpu-clock', Event count (approx.): 57707250000
Overhead Command Shared Object Symbol
+ 5,65% netperf [kernel.vmlinux] [k] memcpy_erms
+ 5,59% netperf [kernel.vmlinux] [k] copy_user_enhanced_fast_string
- 5,05% netperf [kernel.vmlinux] [k] _raw_spin_unlock_irqrestore
- _raw_spin_unlock_irqrestore
+ 49,89% __wake_up_sync_key
+ 45,68% sctp_ulpq_tail_event
- 2,85% mod_timer
+ 76,51% sctp_transport_reset_t3_rtx
+ 23,49% sctp_do_sm
+ 1,55% del_timer
+ 2,50% netperf [sctp] [k] sctp_datamsg_from_user
+ 2,26% netperf [sctp] [k] sctp_sendmsg
Throughput-wise, from 6800mbps without the patch to 7050mbps with it,
~3.7%.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2016-04-07 02:15:19 +08:00
|
|
|
unsigned long timeout = trans->rto >> 1;
|
net: sctp: improve timer slack calculation for transport HBs
RFC4960, section 8.3 says:
On an idle destination address that is allowed to heartbeat,
it is recommended that a HEARTBEAT chunk is sent once per RTO
of that destination address plus the protocol parameter
'HB.interval', with jittering of +/- 50% of the RTO value,
and exponential backoff of the RTO if the previous HEARTBEAT
is unanswered.
Currently, we calculate jitter via sctp_jitter() function first,
and then add its result to the current RTO for the new timeout:
TMO = RTO + (RAND() % RTO) - (RTO / 2)
`------------------------^-=> sctp_jitter()
Instead, we can just simplify all this by directly calculating:
TMO = (RTO / 2) + (RAND() % RTO)
With the help of prandom_u32_max(), we don't need to open code
our own global PRNG, but can instead just make use of the per
CPU implementation of prandom with better quality numbers. Also,
we can now spare us the conditional for divide by zero check
since no div or mod operation needs to be used. Note that
prandom_u32_max() won't emit the same result as a mod operation,
but we really don't care here as we only want to have a random
number scaled into RTO interval.
Note, exponential RTO backoff is handeled elsewhere, namely in
sctp_do_8_2_transport_strike().
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2014-06-30 19:52:08 +08:00
|
|
|
|
|
|
|
if (trans->state != SCTP_UNCONFIRMED &&
|
|
|
|
trans->state != SCTP_PF)
|
|
|
|
timeout += trans->hbinterval;
|
|
|
|
|
2018-06-05 12:16:58 +08:00
|
|
|
return max_t(unsigned long, timeout, HZ / 5);
|
2005-04-17 06:20:36 +08:00
|
|
|
}
|
2007-03-20 08:02:30 +08:00
|
|
|
|
|
|
|
/* Reset transport variables to their initial values */
|
|
|
|
void sctp_transport_reset(struct sctp_transport *t)
|
|
|
|
{
|
|
|
|
struct sctp_association *asoc = t->asoc;
|
|
|
|
|
|
|
|
/* RFC 2960 (bis), Section 5.2.4
|
|
|
|
* All the congestion control parameters (e.g., cwnd, ssthresh)
|
|
|
|
* related to this peer MUST be reset to their initial values
|
|
|
|
* (see Section 6.2.1)
|
|
|
|
*/
|
|
|
|
t->cwnd = min(4*asoc->pathmtu, max_t(__u32, 2*asoc->pathmtu, 4380));
|
2009-11-24 04:54:00 +08:00
|
|
|
t->burst_limited = 0;
|
2007-03-23 03:26:25 +08:00
|
|
|
t->ssthresh = asoc->peer.i.a_rwnd;
|
sctp: on T3_RTX retransmit all the in-flight chunks
When retransmitting due to T3 timeout, retransmit all the
in-flight chunks for the corresponding transport/path, including
chunks sent less then 1 rto ago.
This is the correct behaviour according to rfc4960 section 6.3.3
E3 and
"Note: Any DATA chunks that were sent to the address for which the
T3-rtx timer expired but did not fit in one MTU (rule E3 above)
should be marked for retransmission and sent as soon as cwnd
allows (normally, when a SACK arrives). ".
This fixes problems when more then one path is present and the T3
retransmission of the first chunk that timeouts stops the T3 timer
for the initial active path, leaving all the other in-flight
chunks waiting forever or until a new chunk is transmitted on the
same path and timeouts (and this will happen only if the cwnd
allows sending new chunks, but since cwnd was dropped to MTU by
the timeout => it will wait until the first heartbeat).
Example: 10 packets in flight, sent at 0.1 s intervals on the
primary path. The primary path is down and the first packet
timeouts. The first packet is retransmitted on another path, the
T3 timer for the primary path is stopped and cwnd is set to MTU.
All the other 9 in-flight packets will not be retransmitted
(unless more new packets are sent on the primary path which depend
on cwnd allowing it, and even in this case the 9 packets will be
retransmitted only after a new packet timeouts which even in the
best case would be more then RTO).
This commit reverts d0ce92910bc04e107b2f3f2048f07e94f570035d and
also removes the now unused transport->last_rto, introduced in
b6157d8e03e1e780660a328f7183bcbfa4a93a19.
p.s The problem is not only when multiple paths are there. It
can happen in a single homed environment. If the application
stops sending data, it possible to have a hung association.
Signed-off-by: Andrei Pelinescu-Onciul <andrei@iptel.org>
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2009-11-29 16:14:02 +08:00
|
|
|
t->rto = asoc->rto_initial;
|
2012-12-01 12:49:42 +08:00
|
|
|
sctp_max_rto(asoc, t);
|
2007-03-20 08:02:30 +08:00
|
|
|
t->rtt = 0;
|
|
|
|
t->srtt = 0;
|
|
|
|
t->rttvar = 0;
|
|
|
|
|
2017-02-28 06:29:06 +08:00
|
|
|
/* Reset these additional variables so that we have a clean slate. */
|
2007-03-20 08:02:30 +08:00
|
|
|
t->partial_bytes_acked = 0;
|
|
|
|
t->flight_size = 0;
|
|
|
|
t->error_count = 0;
|
|
|
|
t->rto_pending = 0;
|
2009-02-13 16:33:43 +08:00
|
|
|
t->hb_sent = 0;
|
2007-03-20 08:02:30 +08:00
|
|
|
|
|
|
|
/* Initialize the state information for SFR-CACC */
|
|
|
|
t->cacc.changeover_active = 0;
|
|
|
|
t->cacc.cycling_changeover = 0;
|
|
|
|
t->cacc.next_tsn_at_change = 0;
|
|
|
|
t->cacc.cacc_saw_newack = 0;
|
|
|
|
}
|
2011-06-17 10:03:23 +08:00
|
|
|
|
|
|
|
/* Schedule retransmission on the given transport */
|
|
|
|
void sctp_transport_immediate_rtx(struct sctp_transport *t)
|
|
|
|
{
|
|
|
|
/* Stop pending T3_rtx_timer */
|
2013-02-04 04:32:57 +08:00
|
|
|
if (del_timer(&t->T3_rtx_timer))
|
2011-06-17 10:03:23 +08:00
|
|
|
sctp_transport_put(t);
|
2013-02-04 04:32:57 +08:00
|
|
|
|
2011-06-17 10:03:23 +08:00
|
|
|
sctp_retransmit(&t->asoc->outqueue, t, SCTP_RTXR_T3_RTX);
|
|
|
|
if (!timer_pending(&t->T3_rtx_timer)) {
|
|
|
|
if (!mod_timer(&t->T3_rtx_timer, jiffies + t->rto))
|
|
|
|
sctp_transport_hold(t);
|
|
|
|
}
|
|
|
|
}
|
2017-02-07 05:14:13 +08:00
|
|
|
|
|
|
|
/* Drop dst */
|
|
|
|
void sctp_transport_dst_release(struct sctp_transport *t)
|
|
|
|
{
|
|
|
|
dst_release(t->dst);
|
|
|
|
t->dst = NULL;
|
|
|
|
t->dst_pending_confirm = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Schedule neighbour confirm */
|
|
|
|
void sctp_transport_dst_confirm(struct sctp_transport *t)
|
|
|
|
{
|
|
|
|
t->dst_pending_confirm = 1;
|
|
|
|
}
|