linux/net/sctp/sm_sideeffect.c

1808 lines
51 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-or-later
/* SCTP kernel implementation
* (C) Copyright IBM Corp. 2001, 2004
* Copyright (c) 1999 Cisco, Inc.
* Copyright (c) 1999-2001 Motorola, Inc.
*
* This file is part of the SCTP kernel implementation
*
* These functions work with the state functions in sctp_sm_statefuns.c
* to implement that state operations. These functions implement the
* steps which require modifying existing data structures.
*
* Please send any bug reports or fixes you make to the
* email address(es):
* lksctp developers <linux-sctp@vger.kernel.org>
*
* Written or modified by:
* La Monte H.P. Yarroll <piggy@acm.org>
* Karl Knutson <karl@athena.chicago.il.us>
* Jon Grimm <jgrimm@austin.ibm.com>
* Hui Huang <hui.huang@nokia.com>
* Dajiang Zhang <dajiang.zhang@nokia.com>
* Daisy Chang <daisyc@us.ibm.com>
* Sridhar Samudrala <sri@us.ibm.com>
* Ardelle Fan <ardelle.fan@intel.com>
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/skbuff.h>
#include <linux/types.h>
#include <linux/socket.h>
#include <linux/ip.h>
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/gfp.h>
#include <net/sock.h>
#include <net/sctp/sctp.h>
#include <net/sctp/sm.h>
sctp: introduce stream scheduler foundations This patch introduces the hooks necessary to do stream scheduling, as per RFC Draft ndata. It also introduces the first scheduler, which is what we do today but now factored out: first come first served (FCFS). With stream scheduling now we have to track which chunk was enqueued on which stream and be able to select another other than the in front of the main outqueue. So we introduce a list on sctp_stream_out_ext structure for this purpose. We reuse sctp_chunk->transmitted_list space for the list above, as the chunk cannot belong to the two lists at the same time. By using the union in there, we can have distinct names for these moments. sctp_sched_ops are the operations expected to be implemented by each scheduler. The dequeueing is a bit particular to this implementation but it is to match how we dequeue packets today. We first dequeue and then check if it fits the packet and if not, we requeue it at head. Thus why we don't have a peek operation but have dequeue_done instead, which is called once the chunk can be safely considered as transmitted. The check removed from sctp_outq_flush is now performed by sctp_stream_outq_migrate, which is only called during assoc setup. (sctp_sendmsg() also checks for it) The only operation that is foreseen but not yet added here is a way to signalize that a new packet is starting or that the packet is done, for round robin scheduler per packet, but is intentionally left to the patch that actually implements it. Support for I-DATA chunks, also described in this RFC, with user message interleaving is straightforward as it just requires the schedulers to probe for the feature and ignore datamsg boundaries when dequeueing. See-also: https://tools.ietf.org/html/draft-ietf-tsvwg-sctp-ndata-13 Tested-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2017-10-04 06:20:13 +08:00
#include <net/sctp/stream_sched.h>
static int sctp_cmd_interpreter(enum sctp_event_type event_type,
union sctp_subtype subtype,
enum sctp_state state,
struct sctp_endpoint *ep,
struct sctp_association *asoc,
void *event_arg,
enum sctp_disposition status,
struct sctp_cmd_seq *commands,
gfp_t gfp);
static int sctp_side_effects(enum sctp_event_type event_type,
union sctp_subtype subtype,
enum sctp_state state,
struct sctp_endpoint *ep,
struct sctp_association **asoc,
void *event_arg,
enum sctp_disposition status,
struct sctp_cmd_seq *commands,
gfp_t gfp);
/********************************************************************
* Helper functions
********************************************************************/
/* A helper function for delayed processing of INET ECN CE bit. */
static void sctp_do_ecn_ce_work(struct sctp_association *asoc,
__u32 lowest_tsn)
{
/* Save the TSN away for comparison when we receive CWR */
asoc->last_ecne_tsn = lowest_tsn;
asoc->need_ecne = 1;
}
/* Helper function for delayed processing of SCTP ECNE chunk. */
/* RFC 2960 Appendix A
*
* RFC 2481 details a specific bit for a sender to send in
* the header of its next outbound TCP segment to indicate to
* its peer that it has reduced its congestion window. This
* is termed the CWR bit. For SCTP the same indication is made
* by including the CWR chunk. This chunk contains one data
* element, i.e. the TSN number that was sent in the ECNE chunk.
* This element represents the lowest TSN number in the datagram
* that was originally marked with the CE bit.
*/
static struct sctp_chunk *sctp_do_ecn_ecne_work(struct sctp_association *asoc,
__u32 lowest_tsn,
struct sctp_chunk *chunk)
{
struct sctp_chunk *repl;
/* Our previously transmitted packet ran into some congestion
* so we should take action by reducing cwnd and ssthresh
* and then ACK our peer that we we've done so by
* sending a CWR.
*/
/* First, try to determine if we want to actually lower
* our cwnd variables. Only lower them if the ECNE looks more
* recent than the last response.
*/
if (TSN_lt(asoc->last_cwr_tsn, lowest_tsn)) {
struct sctp_transport *transport;
/* Find which transport's congestion variables
* need to be adjusted.
*/
transport = sctp_assoc_lookup_tsn(asoc, lowest_tsn);
/* Update the congestion variables. */
if (transport)
sctp_transport_lower_cwnd(transport,
SCTP_LOWER_CWND_ECNE);
asoc->last_cwr_tsn = lowest_tsn;
}
/* Always try to quiet the other end. In case of lost CWR,
* resend last_cwr_tsn.
*/
repl = sctp_make_cwr(asoc, asoc->last_cwr_tsn, chunk);
/* If we run out of memory, it will look like a lost CWR. We'll
* get back in sync eventually.
*/
return repl;
}
/* Helper function to do delayed processing of ECN CWR chunk. */
static void sctp_do_ecn_cwr_work(struct sctp_association *asoc,
__u32 lowest_tsn)
{
/* Turn off ECNE getting auto-prepended to every outgoing
* packet
*/
asoc->need_ecne = 0;
}
/* Generate SACK if necessary. We call this at the end of a packet. */
static int sctp_gen_sack(struct sctp_association *asoc, int force,
struct sctp_cmd_seq *commands)
{
struct sctp_transport *trans = asoc->peer.last_data_from;
__u32 ctsn, max_tsn_seen;
struct sctp_chunk *sack;
int error = 0;
if (force ||
(!trans && (asoc->param_flags & SPP_SACKDELAY_DISABLE)) ||
(trans && (trans->param_flags & SPP_SACKDELAY_DISABLE)))
asoc->peer.sack_needed = 1;
ctsn = sctp_tsnmap_get_ctsn(&asoc->peer.tsn_map);
max_tsn_seen = sctp_tsnmap_get_max_tsn_seen(&asoc->peer.tsn_map);
/* From 12.2 Parameters necessary per association (i.e. the TCB):
*
* Ack State : This flag indicates if the next received packet
* : is to be responded to with a SACK. ...
* : When DATA chunks are out of order, SACK's
* : are not delayed (see Section 6).
*
* [This is actually not mentioned in Section 6, but we
* implement it here anyway. --piggy]
*/
if (max_tsn_seen != ctsn)
asoc->peer.sack_needed = 1;
/* From 6.2 Acknowledgement on Reception of DATA Chunks:
*
* Section 4.2 of [RFC2581] SHOULD be followed. Specifically,
* an acknowledgement SHOULD be generated for at least every
* second packet (not every second DATA chunk) received, and
* SHOULD be generated within 200 ms of the arrival of any
* unacknowledged DATA chunk. ...
*/
if (!asoc->peer.sack_needed) {
asoc->peer.sack_cnt++;
/* Set the SACK delay timeout based on the
* SACK delay for the last transport
* data was received from, or the default
* for the association.
*/
if (trans) {
/* We will need a SACK for the next packet. */
if (asoc->peer.sack_cnt >= trans->sackfreq - 1)
asoc->peer.sack_needed = 1;
asoc->timeouts[SCTP_EVENT_TIMEOUT_SACK] =
trans->sackdelay;
} else {
/* We will need a SACK for the next packet. */
if (asoc->peer.sack_cnt >= asoc->sackfreq - 1)
asoc->peer.sack_needed = 1;
asoc->timeouts[SCTP_EVENT_TIMEOUT_SACK] =
asoc->sackdelay;
}
/* Restart the SACK timer. */
sctp_add_cmd_sf(commands, SCTP_CMD_TIMER_RESTART,
SCTP_TO(SCTP_EVENT_TIMEOUT_SACK));
} else {
__u32 old_a_rwnd = asoc->a_rwnd;
asoc->a_rwnd = asoc->rwnd;
sack = sctp_make_sack(asoc);
if (!sack) {
asoc->a_rwnd = old_a_rwnd;
goto nomem;
}
asoc->peer.sack_needed = 0;
asoc->peer.sack_cnt = 0;
sctp_add_cmd_sf(commands, SCTP_CMD_REPLY, SCTP_CHUNK(sack));
/* Stop the SACK timer. */
sctp_add_cmd_sf(commands, SCTP_CMD_TIMER_STOP,
SCTP_TO(SCTP_EVENT_TIMEOUT_SACK));
}
return error;
nomem:
error = -ENOMEM;
return error;
}
/* When the T3-RTX timer expires, it calls this function to create the
* relevant state machine event.
*/
void sctp_generate_t3_rtx_event(struct timer_list *t)
{
struct sctp_transport *transport =
from_timer(transport, t, T3_rtx_timer);
struct sctp_association *asoc = transport->asoc;
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
struct sock *sk = asoc->base.sk;
struct net *net = sock_net(sk);
int error;
/* Check whether a task is in the sock. */
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
bh_lock_sock(sk);
if (sock_owned_by_user(sk)) {
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: sock is busy\n", __func__);
/* Try again later. */
if (!mod_timer(&transport->T3_rtx_timer, jiffies + (HZ/20)))
sctp_transport_hold(transport);
goto out_unlock;
}
/* Run through the state machine. */
error = sctp_do_sm(net, SCTP_EVENT_T_TIMEOUT,
SCTP_ST_TIMEOUT(SCTP_EVENT_TIMEOUT_T3_RTX),
asoc->state,
asoc->ep, asoc,
transport, GFP_ATOMIC);
if (error)
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
sk->sk_err = -error;
out_unlock:
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
bh_unlock_sock(sk);
sctp_transport_put(transport);
}
/* This is a sa interface for producing timeout events. It works
* for timeouts which use the association as their parameter.
*/
static void sctp_generate_timeout_event(struct sctp_association *asoc,
enum sctp_event_timeout timeout_type)
{
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
struct sock *sk = asoc->base.sk;
struct net *net = sock_net(sk);
int error = 0;
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
bh_lock_sock(sk);
if (sock_owned_by_user(sk)) {
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: sock is busy: timer %d\n", __func__,
timeout_type);
/* Try again later. */
if (!mod_timer(&asoc->timers[timeout_type], jiffies + (HZ/20)))
sctp_association_hold(asoc);
goto out_unlock;
}
/* Is this association really dead and just waiting around for
* the timer to let go of the reference?
*/
if (asoc->base.dead)
goto out_unlock;
/* Run through the state machine. */
error = sctp_do_sm(net, SCTP_EVENT_T_TIMEOUT,
SCTP_ST_TIMEOUT(timeout_type),
asoc->state, asoc->ep, asoc,
(void *)timeout_type, GFP_ATOMIC);
if (error)
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
sk->sk_err = -error;
out_unlock:
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
bh_unlock_sock(sk);
sctp_association_put(asoc);
}
static void sctp_generate_t1_cookie_event(struct timer_list *t)
{
struct sctp_association *asoc =
from_timer(asoc, t, timers[SCTP_EVENT_TIMEOUT_T1_COOKIE]);
sctp_generate_timeout_event(asoc, SCTP_EVENT_TIMEOUT_T1_COOKIE);
}
static void sctp_generate_t1_init_event(struct timer_list *t)
{
struct sctp_association *asoc =
from_timer(asoc, t, timers[SCTP_EVENT_TIMEOUT_T1_INIT]);
sctp_generate_timeout_event(asoc, SCTP_EVENT_TIMEOUT_T1_INIT);
}
static void sctp_generate_t2_shutdown_event(struct timer_list *t)
{
struct sctp_association *asoc =
from_timer(asoc, t, timers[SCTP_EVENT_TIMEOUT_T2_SHUTDOWN]);
sctp_generate_timeout_event(asoc, SCTP_EVENT_TIMEOUT_T2_SHUTDOWN);
}
static void sctp_generate_t4_rto_event(struct timer_list *t)
{
struct sctp_association *asoc =
from_timer(asoc, t, timers[SCTP_EVENT_TIMEOUT_T4_RTO]);
sctp_generate_timeout_event(asoc, SCTP_EVENT_TIMEOUT_T4_RTO);
}
static void sctp_generate_t5_shutdown_guard_event(struct timer_list *t)
{
struct sctp_association *asoc =
from_timer(asoc, t,
timers[SCTP_EVENT_TIMEOUT_T5_SHUTDOWN_GUARD]);
sctp_generate_timeout_event(asoc,
SCTP_EVENT_TIMEOUT_T5_SHUTDOWN_GUARD);
} /* sctp_generate_t5_shutdown_guard_event() */
static void sctp_generate_autoclose_event(struct timer_list *t)
{
struct sctp_association *asoc =
from_timer(asoc, t, timers[SCTP_EVENT_TIMEOUT_AUTOCLOSE]);
sctp_generate_timeout_event(asoc, SCTP_EVENT_TIMEOUT_AUTOCLOSE);
}
/* Generate a heart beat event. If the sock is busy, reschedule. Make
* sure that the transport is still valid.
*/
void sctp_generate_heartbeat_event(struct timer_list *t)
{
struct sctp_transport *transport = from_timer(transport, t, hb_timer);
struct sctp_association *asoc = transport->asoc;
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
struct sock *sk = asoc->base.sk;
struct net *net = sock_net(sk);
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
u32 elapsed, timeout;
int error = 0;
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
bh_lock_sock(sk);
if (sock_owned_by_user(sk)) {
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: sock is busy\n", __func__);
/* Try again later. */
if (!mod_timer(&transport->hb_timer, jiffies + (HZ/20)))
sctp_transport_hold(transport);
goto out_unlock;
}
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
/* Check if we should still send the heartbeat or reschedule */
elapsed = jiffies - transport->last_time_sent;
timeout = sctp_transport_timeout(transport);
if (elapsed < timeout) {
elapsed = timeout - elapsed;
if (!mod_timer(&transport->hb_timer, jiffies + elapsed))
sctp_transport_hold(transport);
goto out_unlock;
}
error = sctp_do_sm(net, SCTP_EVENT_T_TIMEOUT,
SCTP_ST_TIMEOUT(SCTP_EVENT_TIMEOUT_HEARTBEAT),
asoc->state, asoc->ep, asoc,
transport, GFP_ATOMIC);
if (error)
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
sk->sk_err = -error;
out_unlock:
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
bh_unlock_sock(sk);
sctp_transport_put(transport);
}
sctp: Fix a race between ICMP protocol unreachable and connect() ICMP protocol unreachable handling completely disregarded the fact that the user may have locked the socket. It proceeded to destroy the association, even though the user may have held the lock and had a ref on the association. This resulted in the following: Attempt to release alive inet socket f6afcc00 ========================= [ BUG: held lock freed! ] ------------------------- somenu/2672 is freeing memory f6afcc00-f6afcfff, with a lock still held there! (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c 1 lock held by somenu/2672: #0: (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c stack backtrace: Pid: 2672, comm: somenu Not tainted 2.6.32-telco #55 Call Trace: [<c1232266>] ? printk+0xf/0x11 [<c1038553>] debug_check_no_locks_freed+0xce/0xff [<c10620b4>] kmem_cache_free+0x21/0x66 [<c1185f25>] __sk_free+0x9d/0xab [<c1185f9c>] sk_free+0x1c/0x1e [<c1216e38>] sctp_association_put+0x32/0x89 [<c1220865>] __sctp_connect+0x36d/0x3f4 [<c122098a>] ? sctp_connect+0x13/0x4c [<c102d073>] ? autoremove_wake_function+0x0/0x33 [<c12209a8>] sctp_connect+0x31/0x4c [<c11d1e80>] inet_dgram_connect+0x4b/0x55 [<c11834fa>] sys_connect+0x54/0x71 [<c103a3a2>] ? lock_release_non_nested+0x88/0x239 [<c1054026>] ? might_fault+0x42/0x7c [<c1054026>] ? might_fault+0x42/0x7c [<c11847ab>] sys_socketcall+0x6d/0x178 [<c10da994>] ? trace_hardirqs_on_thunk+0xc/0x10 [<c1002959>] syscall_call+0x7/0xb This was because the sctp_wait_for_connect() would aqcure the socket lock and then proceed to release the last reference count on the association, thus cause the fully destruction path to finish freeing the socket. The simplest solution is to start a very short timer in case the socket is owned by user. When the timer expires, we can do some verification and be able to do the release properly. Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-05-06 15:56:07 +08:00
/* Handle the timeout of the ICMP protocol unreachable timer. Trigger
* the correct state machine transition that will close the association.
*/
void sctp_generate_proto_unreach_event(struct timer_list *t)
sctp: Fix a race between ICMP protocol unreachable and connect() ICMP protocol unreachable handling completely disregarded the fact that the user may have locked the socket. It proceeded to destroy the association, even though the user may have held the lock and had a ref on the association. This resulted in the following: Attempt to release alive inet socket f6afcc00 ========================= [ BUG: held lock freed! ] ------------------------- somenu/2672 is freeing memory f6afcc00-f6afcfff, with a lock still held there! (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c 1 lock held by somenu/2672: #0: (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c stack backtrace: Pid: 2672, comm: somenu Not tainted 2.6.32-telco #55 Call Trace: [<c1232266>] ? printk+0xf/0x11 [<c1038553>] debug_check_no_locks_freed+0xce/0xff [<c10620b4>] kmem_cache_free+0x21/0x66 [<c1185f25>] __sk_free+0x9d/0xab [<c1185f9c>] sk_free+0x1c/0x1e [<c1216e38>] sctp_association_put+0x32/0x89 [<c1220865>] __sctp_connect+0x36d/0x3f4 [<c122098a>] ? sctp_connect+0x13/0x4c [<c102d073>] ? autoremove_wake_function+0x0/0x33 [<c12209a8>] sctp_connect+0x31/0x4c [<c11d1e80>] inet_dgram_connect+0x4b/0x55 [<c11834fa>] sys_connect+0x54/0x71 [<c103a3a2>] ? lock_release_non_nested+0x88/0x239 [<c1054026>] ? might_fault+0x42/0x7c [<c1054026>] ? might_fault+0x42/0x7c [<c11847ab>] sys_socketcall+0x6d/0x178 [<c10da994>] ? trace_hardirqs_on_thunk+0xc/0x10 [<c1002959>] syscall_call+0x7/0xb This was because the sctp_wait_for_connect() would aqcure the socket lock and then proceed to release the last reference count on the association, thus cause the fully destruction path to finish freeing the socket. The simplest solution is to start a very short timer in case the socket is owned by user. When the timer expires, we can do some verification and be able to do the release properly. Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-05-06 15:56:07 +08:00
{
struct sctp_transport *transport =
from_timer(transport, t, proto_unreach_timer);
sctp: Fix a race between ICMP protocol unreachable and connect() ICMP protocol unreachable handling completely disregarded the fact that the user may have locked the socket. It proceeded to destroy the association, even though the user may have held the lock and had a ref on the association. This resulted in the following: Attempt to release alive inet socket f6afcc00 ========================= [ BUG: held lock freed! ] ------------------------- somenu/2672 is freeing memory f6afcc00-f6afcfff, with a lock still held there! (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c 1 lock held by somenu/2672: #0: (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c stack backtrace: Pid: 2672, comm: somenu Not tainted 2.6.32-telco #55 Call Trace: [<c1232266>] ? printk+0xf/0x11 [<c1038553>] debug_check_no_locks_freed+0xce/0xff [<c10620b4>] kmem_cache_free+0x21/0x66 [<c1185f25>] __sk_free+0x9d/0xab [<c1185f9c>] sk_free+0x1c/0x1e [<c1216e38>] sctp_association_put+0x32/0x89 [<c1220865>] __sctp_connect+0x36d/0x3f4 [<c122098a>] ? sctp_connect+0x13/0x4c [<c102d073>] ? autoremove_wake_function+0x0/0x33 [<c12209a8>] sctp_connect+0x31/0x4c [<c11d1e80>] inet_dgram_connect+0x4b/0x55 [<c11834fa>] sys_connect+0x54/0x71 [<c103a3a2>] ? lock_release_non_nested+0x88/0x239 [<c1054026>] ? might_fault+0x42/0x7c [<c1054026>] ? might_fault+0x42/0x7c [<c11847ab>] sys_socketcall+0x6d/0x178 [<c10da994>] ? trace_hardirqs_on_thunk+0xc/0x10 [<c1002959>] syscall_call+0x7/0xb This was because the sctp_wait_for_connect() would aqcure the socket lock and then proceed to release the last reference count on the association, thus cause the fully destruction path to finish freeing the socket. The simplest solution is to start a very short timer in case the socket is owned by user. When the timer expires, we can do some verification and be able to do the release properly. Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-05-06 15:56:07 +08:00
struct sctp_association *asoc = transport->asoc;
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
struct sock *sk = asoc->base.sk;
struct net *net = sock_net(sk);
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
bh_lock_sock(sk);
if (sock_owned_by_user(sk)) {
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: sock is busy\n", __func__);
sctp: Fix a race between ICMP protocol unreachable and connect() ICMP protocol unreachable handling completely disregarded the fact that the user may have locked the socket. It proceeded to destroy the association, even though the user may have held the lock and had a ref on the association. This resulted in the following: Attempt to release alive inet socket f6afcc00 ========================= [ BUG: held lock freed! ] ------------------------- somenu/2672 is freeing memory f6afcc00-f6afcfff, with a lock still held there! (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c 1 lock held by somenu/2672: #0: (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c stack backtrace: Pid: 2672, comm: somenu Not tainted 2.6.32-telco #55 Call Trace: [<c1232266>] ? printk+0xf/0x11 [<c1038553>] debug_check_no_locks_freed+0xce/0xff [<c10620b4>] kmem_cache_free+0x21/0x66 [<c1185f25>] __sk_free+0x9d/0xab [<c1185f9c>] sk_free+0x1c/0x1e [<c1216e38>] sctp_association_put+0x32/0x89 [<c1220865>] __sctp_connect+0x36d/0x3f4 [<c122098a>] ? sctp_connect+0x13/0x4c [<c102d073>] ? autoremove_wake_function+0x0/0x33 [<c12209a8>] sctp_connect+0x31/0x4c [<c11d1e80>] inet_dgram_connect+0x4b/0x55 [<c11834fa>] sys_connect+0x54/0x71 [<c103a3a2>] ? lock_release_non_nested+0x88/0x239 [<c1054026>] ? might_fault+0x42/0x7c [<c1054026>] ? might_fault+0x42/0x7c [<c11847ab>] sys_socketcall+0x6d/0x178 [<c10da994>] ? trace_hardirqs_on_thunk+0xc/0x10 [<c1002959>] syscall_call+0x7/0xb This was because the sctp_wait_for_connect() would aqcure the socket lock and then proceed to release the last reference count on the association, thus cause the fully destruction path to finish freeing the socket. The simplest solution is to start a very short timer in case the socket is owned by user. When the timer expires, we can do some verification and be able to do the release properly. Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-05-06 15:56:07 +08:00
/* Try again later. */
if (!mod_timer(&transport->proto_unreach_timer,
jiffies + (HZ/20)))
sctp_association_hold(asoc);
goto out_unlock;
}
/* Is this structure just waiting around for us to actually
* get destroyed?
*/
if (asoc->base.dead)
goto out_unlock;
sctp_do_sm(net, SCTP_EVENT_T_OTHER,
sctp: Fix a race between ICMP protocol unreachable and connect() ICMP protocol unreachable handling completely disregarded the fact that the user may have locked the socket. It proceeded to destroy the association, even though the user may have held the lock and had a ref on the association. This resulted in the following: Attempt to release alive inet socket f6afcc00 ========================= [ BUG: held lock freed! ] ------------------------- somenu/2672 is freeing memory f6afcc00-f6afcfff, with a lock still held there! (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c 1 lock held by somenu/2672: #0: (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c stack backtrace: Pid: 2672, comm: somenu Not tainted 2.6.32-telco #55 Call Trace: [<c1232266>] ? printk+0xf/0x11 [<c1038553>] debug_check_no_locks_freed+0xce/0xff [<c10620b4>] kmem_cache_free+0x21/0x66 [<c1185f25>] __sk_free+0x9d/0xab [<c1185f9c>] sk_free+0x1c/0x1e [<c1216e38>] sctp_association_put+0x32/0x89 [<c1220865>] __sctp_connect+0x36d/0x3f4 [<c122098a>] ? sctp_connect+0x13/0x4c [<c102d073>] ? autoremove_wake_function+0x0/0x33 [<c12209a8>] sctp_connect+0x31/0x4c [<c11d1e80>] inet_dgram_connect+0x4b/0x55 [<c11834fa>] sys_connect+0x54/0x71 [<c103a3a2>] ? lock_release_non_nested+0x88/0x239 [<c1054026>] ? might_fault+0x42/0x7c [<c1054026>] ? might_fault+0x42/0x7c [<c11847ab>] sys_socketcall+0x6d/0x178 [<c10da994>] ? trace_hardirqs_on_thunk+0xc/0x10 [<c1002959>] syscall_call+0x7/0xb This was because the sctp_wait_for_connect() would aqcure the socket lock and then proceed to release the last reference count on the association, thus cause the fully destruction path to finish freeing the socket. The simplest solution is to start a very short timer in case the socket is owned by user. When the timer expires, we can do some verification and be able to do the release properly. Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-05-06 15:56:07 +08:00
SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH),
asoc->state, asoc->ep, asoc, transport, GFP_ATOMIC);
out_unlock:
sctp: Prevent soft lockup when sctp_accept() is called during a timeout event A case can occur when sctp_accept() is called by the user during a heartbeat timeout event after the 4-way handshake. Since sctp_assoc_migrate() changes both assoc->base.sk and assoc->ep, the bh_sock_lock in sctp_generate_heartbeat_event() will be taken with the listening socket but released with the new association socket. The result is a deadlock on any future attempts to take the listening socket lock. Note that this race can occur with other SCTP timeouts that take the bh_lock_sock() in the event sctp_accept() is called. BUG: soft lockup - CPU#9 stuck for 67s! [swapper:0] ... RIP: 0010:[<ffffffff8152d48e>] [<ffffffff8152d48e>] _spin_lock+0x1e/0x30 RSP: 0018:ffff880028323b20 EFLAGS: 00000206 RAX: 0000000000000002 RBX: ffff880028323b20 RCX: 0000000000000000 RDX: 0000000000000000 RSI: ffff880028323be0 RDI: ffff8804632c4b48 RBP: ffffffff8100bb93 R08: 0000000000000000 R09: 0000000000000000 R10: ffff880610662280 R11: 0000000000000100 R12: ffff880028323aa0 R13: ffff8804383c3880 R14: ffff880028323a90 R15: ffffffff81534225 FS: 0000000000000000(0000) GS:ffff880028320000(0000) knlGS:0000000000000000 CS: 0010 DS: 0018 ES: 0018 CR0: 000000008005003b CR2: 00000000006df528 CR3: 0000000001a85000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400 Process swapper (pid: 0, threadinfo ffff880616b70000, task ffff880616b6cab0) Stack: ffff880028323c40 ffffffffa01c2582 ffff880614cfb020 0000000000000000 <d> 0100000000000000 00000014383a6c44 ffff8804383c3880 ffff880614e93c00 <d> ffff880614e93c00 0000000000000000 ffff8804632c4b00 ffff8804383c38b8 Call Trace: <IRQ> [<ffffffffa01c2582>] ? sctp_rcv+0x492/0xa10 [sctp] [<ffffffff8148c559>] ? nf_iterate+0x69/0xb0 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8148c716>] ? nf_hook_slow+0x76/0x120 [<ffffffff814974a0>] ? ip_local_deliver_finish+0x0/0x2d0 [<ffffffff8149757d>] ? ip_local_deliver_finish+0xdd/0x2d0 [<ffffffff81497808>] ? ip_local_deliver+0x98/0xa0 [<ffffffff81496ccd>] ? ip_rcv_finish+0x12d/0x440 [<ffffffff81497255>] ? ip_rcv+0x275/0x350 [<ffffffff8145cfeb>] ? __netif_receive_skb+0x4ab/0x750 ... With lockdep debugging: ===================================== [ BUG: bad unlock balance detected! ] ------------------------------------- CslRx/12087 is trying to release lock (slock-AF_INET) at: [<ffffffffa01bcae0>] sctp_generate_timeout_event+0x40/0xe0 [sctp] but there are no more locks to release! other info that might help us debug this: 2 locks held by CslRx/12087: #0: (&asoc->timers[i]){+.-...}, at: [<ffffffff8108ce1f>] run_timer_softirq+0x16f/0x3e0 #1: (slock-AF_INET){+.-...}, at: [<ffffffffa01bcac3>] sctp_generate_timeout_event+0x23/0xe0 [sctp] Ensure the socket taken is also the same one that is released by saving a copy of the socket before entering the timeout event critical section. Signed-off-by: Karl Heiss <kheiss@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2015-09-25 00:15:07 +08:00
bh_unlock_sock(sk);
sctp: Fix a race between ICMP protocol unreachable and connect() ICMP protocol unreachable handling completely disregarded the fact that the user may have locked the socket. It proceeded to destroy the association, even though the user may have held the lock and had a ref on the association. This resulted in the following: Attempt to release alive inet socket f6afcc00 ========================= [ BUG: held lock freed! ] ------------------------- somenu/2672 is freeing memory f6afcc00-f6afcfff, with a lock still held there! (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c 1 lock held by somenu/2672: #0: (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c stack backtrace: Pid: 2672, comm: somenu Not tainted 2.6.32-telco #55 Call Trace: [<c1232266>] ? printk+0xf/0x11 [<c1038553>] debug_check_no_locks_freed+0xce/0xff [<c10620b4>] kmem_cache_free+0x21/0x66 [<c1185f25>] __sk_free+0x9d/0xab [<c1185f9c>] sk_free+0x1c/0x1e [<c1216e38>] sctp_association_put+0x32/0x89 [<c1220865>] __sctp_connect+0x36d/0x3f4 [<c122098a>] ? sctp_connect+0x13/0x4c [<c102d073>] ? autoremove_wake_function+0x0/0x33 [<c12209a8>] sctp_connect+0x31/0x4c [<c11d1e80>] inet_dgram_connect+0x4b/0x55 [<c11834fa>] sys_connect+0x54/0x71 [<c103a3a2>] ? lock_release_non_nested+0x88/0x239 [<c1054026>] ? might_fault+0x42/0x7c [<c1054026>] ? might_fault+0x42/0x7c [<c11847ab>] sys_socketcall+0x6d/0x178 [<c10da994>] ? trace_hardirqs_on_thunk+0xc/0x10 [<c1002959>] syscall_call+0x7/0xb This was because the sctp_wait_for_connect() would aqcure the socket lock and then proceed to release the last reference count on the association, thus cause the fully destruction path to finish freeing the socket. The simplest solution is to start a very short timer in case the socket is owned by user. When the timer expires, we can do some verification and be able to do the release properly. Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-05-06 15:56:07 +08:00
sctp_association_put(asoc);
}
/* Handle the timeout of the RE-CONFIG timer. */
void sctp_generate_reconf_event(struct timer_list *t)
{
struct sctp_transport *transport =
from_timer(transport, t, reconf_timer);
struct sctp_association *asoc = transport->asoc;
struct sock *sk = asoc->base.sk;
struct net *net = sock_net(sk);
int error = 0;
bh_lock_sock(sk);
if (sock_owned_by_user(sk)) {
pr_debug("%s: sock is busy\n", __func__);
/* Try again later. */
if (!mod_timer(&transport->reconf_timer, jiffies + (HZ / 20)))
sctp_transport_hold(transport);
goto out_unlock;
}
error = sctp_do_sm(net, SCTP_EVENT_T_TIMEOUT,
SCTP_ST_TIMEOUT(SCTP_EVENT_TIMEOUT_RECONF),
asoc->state, asoc->ep, asoc,
transport, GFP_ATOMIC);
if (error)
sk->sk_err = -error;
out_unlock:
bh_unlock_sock(sk);
sctp_transport_put(transport);
}
sctp: Fix a race between ICMP protocol unreachable and connect() ICMP protocol unreachable handling completely disregarded the fact that the user may have locked the socket. It proceeded to destroy the association, even though the user may have held the lock and had a ref on the association. This resulted in the following: Attempt to release alive inet socket f6afcc00 ========================= [ BUG: held lock freed! ] ------------------------- somenu/2672 is freeing memory f6afcc00-f6afcfff, with a lock still held there! (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c 1 lock held by somenu/2672: #0: (sk_lock-AF_INET){+.+.+.}, at: [<c122098a>] sctp_connect+0x13/0x4c stack backtrace: Pid: 2672, comm: somenu Not tainted 2.6.32-telco #55 Call Trace: [<c1232266>] ? printk+0xf/0x11 [<c1038553>] debug_check_no_locks_freed+0xce/0xff [<c10620b4>] kmem_cache_free+0x21/0x66 [<c1185f25>] __sk_free+0x9d/0xab [<c1185f9c>] sk_free+0x1c/0x1e [<c1216e38>] sctp_association_put+0x32/0x89 [<c1220865>] __sctp_connect+0x36d/0x3f4 [<c122098a>] ? sctp_connect+0x13/0x4c [<c102d073>] ? autoremove_wake_function+0x0/0x33 [<c12209a8>] sctp_connect+0x31/0x4c [<c11d1e80>] inet_dgram_connect+0x4b/0x55 [<c11834fa>] sys_connect+0x54/0x71 [<c103a3a2>] ? lock_release_non_nested+0x88/0x239 [<c1054026>] ? might_fault+0x42/0x7c [<c1054026>] ? might_fault+0x42/0x7c [<c11847ab>] sys_socketcall+0x6d/0x178 [<c10da994>] ? trace_hardirqs_on_thunk+0xc/0x10 [<c1002959>] syscall_call+0x7/0xb This was because the sctp_wait_for_connect() would aqcure the socket lock and then proceed to release the last reference count on the association, thus cause the fully destruction path to finish freeing the socket. The simplest solution is to start a very short timer in case the socket is owned by user. When the timer expires, we can do some verification and be able to do the release properly. Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2010-05-06 15:56:07 +08:00
/* Inject a SACK Timeout event into the state machine. */
static void sctp_generate_sack_event(struct timer_list *t)
{
struct sctp_association *asoc =
from_timer(asoc, t, timers[SCTP_EVENT_TIMEOUT_SACK]);
sctp_generate_timeout_event(asoc, SCTP_EVENT_TIMEOUT_SACK);
}
sctp_timer_event_t *sctp_timer_events[SCTP_NUM_TIMEOUT_TYPES] = {
[SCTP_EVENT_TIMEOUT_NONE] = NULL,
[SCTP_EVENT_TIMEOUT_T1_COOKIE] = sctp_generate_t1_cookie_event,
[SCTP_EVENT_TIMEOUT_T1_INIT] = sctp_generate_t1_init_event,
[SCTP_EVENT_TIMEOUT_T2_SHUTDOWN] = sctp_generate_t2_shutdown_event,
[SCTP_EVENT_TIMEOUT_T3_RTX] = NULL,
[SCTP_EVENT_TIMEOUT_T4_RTO] = sctp_generate_t4_rto_event,
[SCTP_EVENT_TIMEOUT_T5_SHUTDOWN_GUARD] =
sctp_generate_t5_shutdown_guard_event,
[SCTP_EVENT_TIMEOUT_HEARTBEAT] = NULL,
[SCTP_EVENT_TIMEOUT_RECONF] = NULL,
[SCTP_EVENT_TIMEOUT_SACK] = sctp_generate_sack_event,
[SCTP_EVENT_TIMEOUT_AUTOCLOSE] = sctp_generate_autoclose_event,
};
/* RFC 2960 8.2 Path Failure Detection
*
* When its peer endpoint is multi-homed, an endpoint should keep a
* error counter for each of the destination transport addresses of the
* peer endpoint.
*
* Each time the T3-rtx timer expires on any address, or when a
* HEARTBEAT sent to an idle address is not acknowledged within a RTO,
* the error counter of that destination address will be incremented.
* When the value in the error counter exceeds the protocol parameter
* 'Path.Max.Retrans' of that destination address, the endpoint should
* mark the destination transport address as inactive, and a
* notification SHOULD be sent to the upper layer.
*
*/
static void sctp_do_8_2_transport_strike(struct sctp_cmd_seq *commands,
struct sctp_association *asoc,
struct sctp_transport *transport,
int is_hb)
{
/* The check for association's overall error counter exceeding the
* threshold is done in the state function.
*/
/* We are here due to a timer expiration. If the timer was
* not a HEARTBEAT, then normal error tracking is done.
* If the timer was a heartbeat, we only increment error counts
* when we already have an outstanding HEARTBEAT that has not
* been acknowledged.
* Additionally, some tranport states inhibit error increments.
*/
if (!is_hb) {
asoc->overall_error_count++;
if (transport->state != SCTP_INACTIVE)
transport->error_count++;
} else if (transport->hb_sent) {
if (transport->state != SCTP_UNCONFIRMED)
asoc->overall_error_count++;
if (transport->state != SCTP_INACTIVE)
transport->error_count++;
}
/* If the transport error count is greater than the pf_retrans
net: sctp: Potentially-Failed state should not be reached from unconfirmed state In current implementation it is possible to reach PF state from unconfirmed. We can interpret sctp-failover-02 in a way that PF state is meant to be reached only from active state, in the end, this is when entering PF state makes sense. Here are few quotes from sctp-failover-02, but regardless of these, same understanding can be reached from whole section 5: Section 5.1, quickfailover guide: "The PF state is an intermediate state between Active and Failed states." "Each time the T3-rtx timer expires on an active or idle destination, the error counter of that destination address will be incremented. When the value in the error counter exceeds PFMR, the endpoint should mark the destination transport address as PF." There are several concrete reasons for such interpretation. For start, rfc4960 does not take into concern quickfailover algorithm. Therefore, quickfailover must comply to 4960. Point where this compliance can be argued is following behavior: When PF is entered, association overall error counter is incremented for each missed HB. This is contradictory to rfc4960, as address, while in unconfirmed state, is subjected to probing, and while it is probed, it should not increment association overall error counter. This has as a consequence that we might end up in situation in which we drop association due path failure on unconfirmed address, in case we have wrong configuration in a way: Association.Max.Retrans == Path.Max.Retrans. Another reason is that entering PF from unconfirmed will cause a loss of address confirmed event when address is once (if) confirmed. This is fine from failover guide point of view, but it is not consistent with behavior preceding failover implementation and recommendation from 4960: 5.4. Path Verification Whenever a path is confirmed, an indication MAY be given to the upper layer. Signed-off-by: Matija Glavinic Pecotic <matija.glavinic-pecotic.ext@nsn.com> Acked-by: Vlad Yasevich <vyasevich@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2014-02-20 21:13:04 +08:00
* threshold, and less than pathmaxrtx, and if the current state
* is SCTP_ACTIVE, then mark this transport as Partially Failed,
* see SCTP Quick Failover Draft, section 5.1
*/
if (asoc->base.net->sctp.pf_enable &&
transport->state == SCTP_ACTIVE &&
transport->error_count < transport->pathmaxrxt &&
transport->error_count > transport->pf_retrans) {
sctp_assoc_control_transport(asoc, transport,
SCTP_TRANSPORT_PF,
0);
/* Update the hb timer to resend a heartbeat every 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
sctp_transport_reset_hb_timer(transport);
}
if (transport->state != SCTP_INACTIVE &&
(transport->error_count > transport->pathmaxrxt)) {
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: association:%p transport addr:%pISpc failed\n",
__func__, asoc, &transport->ipaddr.sa);
sctp_assoc_control_transport(asoc, transport,
SCTP_TRANSPORT_DOWN,
SCTP_FAILED_THRESHOLD);
}
if (transport->error_count > transport->ps_retrans &&
asoc->peer.primary_path == transport &&
asoc->peer.active_path != transport)
sctp_assoc_set_primary(asoc, asoc->peer.active_path);
/* E2) For the destination address for which the timer
* expires, set RTO <- RTO * 2 ("back off the timer"). The
* maximum value discussed in rule C7 above (RTO.max) may be
* used to provide an upper bound to this doubling operation.
*
* Special Case: the first HB doesn't trigger exponential backoff.
* The first unacknowledged HB triggers it. We do this with a flag
* that indicates that we have an outstanding HB.
*/
if (!is_hb || transport->hb_sent) {
transport->rto = min((transport->rto * 2), transport->asoc->rto_max);
sctp: Add support to per-association statistics via a new SCTP_GET_ASSOC_STATS call The current SCTP stack is lacking a mechanism to have per association statistics. This is an implementation modeled after OpenSolaris' SCTP_GET_ASSOC_STATS. Userspace part will follow on lksctp if/when there is a general ACK on this. V4: - Move ipackets++ before q->immediate.func() for consistency reasons - Move sctp_max_rto() at the end of sctp_transport_update_rto() to avoid returning bogus RTO values - return asoc->rto_min when max_obs_rto value has not changed V3: - Increase ictrlchunks in sctp_assoc_bh_rcv() as well - Move ipackets++ to sctp_inq_push() - return 0 when no rto updates took place since the last call V2: - Implement partial retrieval of stat struct to cope for future expansion - Kill the rtxpackets counter as it cannot be precise anyway - Rename outseqtsns to outofseqtsns to make it clearer that these are out of sequence unexpected TSNs - Move asoc->ipackets++ under a lock to avoid potential miscounts - Fold asoc->opackets++ into the already existing asoc check - Kill unneeded (q->asoc) test when increasing rtxchunks - Do not count octrlchunks if sending failed (SCTP_XMIT_OK != 0) - Don't count SHUTDOWNs as SACKs - Move SCTP_GET_ASSOC_STATS to the private space API - Adjust the len check in sctp_getsockopt_assoc_stats() to allow for future struct growth - Move association statistics in their own struct - Update idupchunks when we send a SACK with dup TSNs - return min_rto in max_rto when RTO has not changed. Also return the transport when max_rto last changed. Signed-off: Michele Baldessari <michele@acksyn.org> Acked-by: Vlad Yasevich <vyasevich@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-12-01 12:49:42 +08:00
sctp_max_rto(asoc, transport);
}
}
/* Worker routine to handle INIT command failure. */
static void sctp_cmd_init_failed(struct sctp_cmd_seq *commands,
struct sctp_association *asoc,
unsigned int error)
{
struct sctp_ulpevent *event;
event = sctp_ulpevent_make_assoc_change(asoc, 0, SCTP_CANT_STR_ASSOC,
(__u16)error, 0, 0, NULL,
GFP_ATOMIC);
if (event)
sctp_add_cmd_sf(commands, SCTP_CMD_EVENT_ULP,
SCTP_ULPEVENT(event));
sctp_add_cmd_sf(commands, SCTP_CMD_NEW_STATE,
SCTP_STATE(SCTP_STATE_CLOSED));
/* SEND_FAILED sent later when cleaning up the association. */
asoc->outqueue.error = error;
sctp_add_cmd_sf(commands, SCTP_CMD_DELETE_TCB, SCTP_NULL());
}
/* Worker routine to handle SCTP_CMD_ASSOC_FAILED. */
static void sctp_cmd_assoc_failed(struct sctp_cmd_seq *commands,
struct sctp_association *asoc,
enum sctp_event_type event_type,
union sctp_subtype subtype,
struct sctp_chunk *chunk,
unsigned int error)
{
struct sctp_ulpevent *event;
struct sctp_chunk *abort;
/* Cancel any partial delivery in progress. */
asoc->stream.si->abort_pd(&asoc->ulpq, GFP_ATOMIC);
if (event_type == SCTP_EVENT_T_CHUNK && subtype.chunk == SCTP_CID_ABORT)
event = sctp_ulpevent_make_assoc_change(asoc, 0, SCTP_COMM_LOST,
(__u16)error, 0, 0, chunk,
GFP_ATOMIC);
else
event = sctp_ulpevent_make_assoc_change(asoc, 0, SCTP_COMM_LOST,
(__u16)error, 0, 0, NULL,
GFP_ATOMIC);
if (event)
sctp_add_cmd_sf(commands, SCTP_CMD_EVENT_ULP,
SCTP_ULPEVENT(event));
if (asoc->overall_error_count >= asoc->max_retrans) {
abort = sctp_make_violation_max_retrans(asoc, chunk);
if (abort)
sctp_add_cmd_sf(commands, SCTP_CMD_REPLY,
SCTP_CHUNK(abort));
}
sctp_add_cmd_sf(commands, SCTP_CMD_NEW_STATE,
SCTP_STATE(SCTP_STATE_CLOSED));
/* SEND_FAILED sent later when cleaning up the association. */
asoc->outqueue.error = error;
sctp_add_cmd_sf(commands, SCTP_CMD_DELETE_TCB, SCTP_NULL());
}
/* Process an init chunk (may be real INIT/INIT-ACK or an embedded INIT
* inside the cookie. In reality, this is only used for INIT-ACK processing
* since all other cases use "temporary" associations and can do all
* their work in statefuns directly.
*/
static int sctp_cmd_process_init(struct sctp_cmd_seq *commands,
struct sctp_association *asoc,
struct sctp_chunk *chunk,
struct sctp_init_chunk *peer_init,
gfp_t gfp)
{
int error;
/* We only process the init as a sideeffect in a single
* case. This is when we process the INIT-ACK. If we
* fail during INIT processing (due to malloc problems),
* just return the error and stop processing the stack.
*/
if (!sctp_process_init(asoc, chunk, sctp_source(chunk), peer_init, gfp))
error = -ENOMEM;
else
error = 0;
return error;
}
/* Helper function to break out starting up of heartbeat timers. */
static void sctp_cmd_hb_timers_start(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc)
{
struct sctp_transport *t;
/* Start a heartbeat timer for each transport on the association.
* hold a reference on the transport to make sure none of
* the needed data structures go away.
*/
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
list_for_each_entry(t, &asoc->peer.transport_addr_list, transports)
sctp_transport_reset_hb_timer(t);
}
static void sctp_cmd_hb_timers_stop(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc)
{
struct sctp_transport *t;
/* Stop all heartbeat timers. */
list_for_each_entry(t, &asoc->peer.transport_addr_list,
transports) {
if (del_timer(&t->hb_timer))
sctp_transport_put(t);
}
}
/* Helper function to stop any pending T3-RTX timers */
static void sctp_cmd_t3_rtx_timers_stop(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc)
{
struct sctp_transport *t;
list_for_each_entry(t, &asoc->peer.transport_addr_list,
transports) {
if (del_timer(&t->T3_rtx_timer))
sctp_transport_put(t);
}
}
/* Helper function to handle the reception of an HEARTBEAT ACK. */
static void sctp_cmd_transport_on(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc,
struct sctp_transport *t,
struct sctp_chunk *chunk)
{
struct sctp_sender_hb_info *hbinfo;
int was_unconfirmed = 0;
/* 8.3 Upon the receipt of the HEARTBEAT ACK, the sender of the
* HEARTBEAT should clear the error counter of the destination
* transport address to which the HEARTBEAT was sent.
*/
t->error_count = 0;
sctp: Enforce retransmission limit during shutdown When initiating a graceful shutdown while having data chunks on the retransmission queue with a peer which is in zero window mode the shutdown is never completed because the retransmission error count is reset periodically by the following two rules: - Do not timeout association while doing zero window probe. - Reset overall error count when a heartbeat request has been acknowledged. The graceful shutdown will wait for all outstanding TSN to be acknowledged before sending the SHUTDOWN request. This never happens due to the peer's zero window not acknowledging the continuously retransmitted data chunks. Although the error counter is incremented for each failed retransmission, the receiving of the SACK announcing the zero window clears the error count again immediately. Also heartbeat requests continue to be sent periodically. The peer acknowledges these requests causing the error counter to be reset as well. This patch changes behaviour to only reset the overall error counter for the above rules while not in shutdown. After reaching the maximum number of retransmission attempts, the T5 shutdown guard timer is scheduled to give the receiver some additional time to recover. The timer is stopped as soon as the receiver acknowledges any data. The issue can be easily reproduced by establishing a sctp association over the loopback device, constantly queueing data at the sender while not reading any at the receiver. Wait for the window to reach zero, then initiate a shutdown by killing both processes simultaneously. The association will never be freed and the chunks on the retransmission queue will be retransmitted indefinitely. Signed-off-by: Thomas Graf <tgraf@infradead.org> Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2011-07-07 08:28:35 +08:00
/*
* Although RFC4960 specifies that the overall error count must
* be cleared when a HEARTBEAT ACK is received, we make an
* exception while in SHUTDOWN PENDING. If the peer keeps its
* window shut forever, we may never be able to transmit our
* outstanding data and rely on the retransmission limit be reached
* to shutdown the association.
*/
if (t->asoc->state < SCTP_STATE_SHUTDOWN_PENDING)
sctp: Enforce retransmission limit during shutdown When initiating a graceful shutdown while having data chunks on the retransmission queue with a peer which is in zero window mode the shutdown is never completed because the retransmission error count is reset periodically by the following two rules: - Do not timeout association while doing zero window probe. - Reset overall error count when a heartbeat request has been acknowledged. The graceful shutdown will wait for all outstanding TSN to be acknowledged before sending the SHUTDOWN request. This never happens due to the peer's zero window not acknowledging the continuously retransmitted data chunks. Although the error counter is incremented for each failed retransmission, the receiving of the SACK announcing the zero window clears the error count again immediately. Also heartbeat requests continue to be sent periodically. The peer acknowledges these requests causing the error counter to be reset as well. This patch changes behaviour to only reset the overall error counter for the above rules while not in shutdown. After reaching the maximum number of retransmission attempts, the T5 shutdown guard timer is scheduled to give the receiver some additional time to recover. The timer is stopped as soon as the receiver acknowledges any data. The issue can be easily reproduced by establishing a sctp association over the loopback device, constantly queueing data at the sender while not reading any at the receiver. Wait for the window to reach zero, then initiate a shutdown by killing both processes simultaneously. The association will never be freed and the chunks on the retransmission queue will be retransmitted indefinitely. Signed-off-by: Thomas Graf <tgraf@infradead.org> Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2011-07-07 08:28:35 +08:00
t->asoc->overall_error_count = 0;
/* Clear the hb_sent flag to signal that we had a good
* acknowledgement.
*/
t->hb_sent = 0;
/* Mark the destination transport address as active if it is not so
* marked.
*/
if ((t->state == SCTP_INACTIVE) || (t->state == SCTP_UNCONFIRMED)) {
was_unconfirmed = 1;
sctp_assoc_control_transport(asoc, t, SCTP_TRANSPORT_UP,
SCTP_HEARTBEAT_SUCCESS);
}
if (t->state == SCTP_PF)
sctp_assoc_control_transport(asoc, t, SCTP_TRANSPORT_UP,
SCTP_HEARTBEAT_SUCCESS);
/* HB-ACK was received for a the proper HB. Consider this
* forward progress.
*/
if (t->dst)
sctp_transport_dst_confirm(t);
/* The receiver of the HEARTBEAT ACK should also perform an
* RTT measurement for that destination transport address
* using the time value carried in the HEARTBEAT ACK chunk.
* If the transport's rto_pending variable has been cleared,
* it was most likely due to a retransmit. However, we want
* to re-enable it to properly update the rto.
*/
if (t->rto_pending == 0)
t->rto_pending = 1;
hbinfo = (struct sctp_sender_hb_info *)chunk->skb->data;
sctp_transport_update_rto(t, (jiffies - hbinfo->sent_at));
/* Update the heartbeat 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
sctp_transport_reset_hb_timer(t);
if (was_unconfirmed && asoc->peer.transport_count == 1)
sctp_transport_immediate_rtx(t);
}
/* Helper function to process the process SACK command. */
static int sctp_cmd_process_sack(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc,
struct sctp_chunk *chunk)
{
int err = 0;
if (sctp_outq_sack(&asoc->outqueue, chunk)) {
/* There are no more TSNs awaiting SACK. */
err = sctp_do_sm(asoc->base.net, SCTP_EVENT_T_OTHER,
SCTP_ST_OTHER(SCTP_EVENT_NO_PENDING_TSN),
asoc->state, asoc->ep, asoc, NULL,
GFP_ATOMIC);
}
return err;
}
/* Helper function to set the timeout value for T2-SHUTDOWN timer and to set
* the transport for a shutdown chunk.
*/
static void sctp_cmd_setup_t2(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc,
struct sctp_chunk *chunk)
{
struct sctp_transport *t;
if (chunk->transport)
t = chunk->transport;
else {
t = sctp_assoc_choose_alter_transport(asoc,
asoc->shutdown_last_sent_to);
chunk->transport = t;
}
asoc->shutdown_last_sent_to = t;
asoc->timeouts[SCTP_EVENT_TIMEOUT_T2_SHUTDOWN] = t->rto;
}
static void sctp_cmd_assoc_update(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc,
struct sctp_association *new)
{
struct net *net = asoc->base.net;
struct sctp_chunk *abort;
if (!sctp_assoc_update(asoc, new))
return;
abort = sctp_make_abort(asoc, NULL, sizeof(struct sctp_errhdr));
if (abort) {
sctp_init_cause(abort, SCTP_ERROR_RSRC_LOW, 0);
sctp_add_cmd_sf(cmds, SCTP_CMD_REPLY, SCTP_CHUNK(abort));
}
sctp_add_cmd_sf(cmds, SCTP_CMD_SET_SK_ERR, SCTP_ERROR(ECONNABORTED));
sctp_add_cmd_sf(cmds, SCTP_CMD_ASSOC_FAILED,
SCTP_PERR(SCTP_ERROR_RSRC_LOW));
SCTP_INC_STATS(net, SCTP_MIB_ABORTEDS);
SCTP_DEC_STATS(net, SCTP_MIB_CURRESTAB);
}
/* Helper function to change the state of an association. */
static void sctp_cmd_new_state(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc,
enum sctp_state state)
{
struct sock *sk = asoc->base.sk;
asoc->state = state;
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: asoc:%p[%s]\n", __func__, asoc, sctp_state_tbl[state]);
if (sctp_style(sk, TCP)) {
/* Change the sk->sk_state of a TCP-style socket that has
* successfully completed a connect() call.
*/
if (sctp_state(asoc, ESTABLISHED) && sctp_sstate(sk, CLOSED))
inet_sk_set_state(sk, SCTP_SS_ESTABLISHED);
/* Set the RCV_SHUTDOWN flag when a SHUTDOWN is received. */
if (sctp_state(asoc, SHUTDOWN_RECEIVED) &&
sctp_sstate(sk, ESTABLISHED)) {
inet_sk_set_state(sk, SCTP_SS_CLOSING);
sk->sk_shutdown |= RCV_SHUTDOWN;
}
}
if (sctp_state(asoc, COOKIE_WAIT)) {
/* Reset init timeouts since they may have been
* increased due to timer expirations.
*/
asoc->timeouts[SCTP_EVENT_TIMEOUT_T1_INIT] =
asoc->rto_initial;
asoc->timeouts[SCTP_EVENT_TIMEOUT_T1_COOKIE] =
asoc->rto_initial;
}
Fix memory leak in sctp_process_init syzbot found the following leak in sctp_process_init BUG: memory leak unreferenced object 0xffff88810ef68400 (size 1024): comm "syz-executor273", pid 7046, jiffies 4294945598 (age 28.770s) hex dump (first 32 bytes): 1d de 28 8d de 0b 1b e3 b5 c2 f9 68 fd 1a 97 25 ..(........h...% 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................ backtrace: [<00000000a02cebbd>] kmemleak_alloc_recursive include/linux/kmemleak.h:55 [inline] [<00000000a02cebbd>] slab_post_alloc_hook mm/slab.h:439 [inline] [<00000000a02cebbd>] slab_alloc mm/slab.c:3326 [inline] [<00000000a02cebbd>] __do_kmalloc mm/slab.c:3658 [inline] [<00000000a02cebbd>] __kmalloc_track_caller+0x15d/0x2c0 mm/slab.c:3675 [<000000009e6245e6>] kmemdup+0x27/0x60 mm/util.c:119 [<00000000dfdc5d2d>] kmemdup include/linux/string.h:432 [inline] [<00000000dfdc5d2d>] sctp_process_init+0xa7e/0xc20 net/sctp/sm_make_chunk.c:2437 [<00000000b58b62f8>] sctp_cmd_process_init net/sctp/sm_sideeffect.c:682 [inline] [<00000000b58b62f8>] sctp_cmd_interpreter net/sctp/sm_sideeffect.c:1384 [inline] [<00000000b58b62f8>] sctp_side_effects net/sctp/sm_sideeffect.c:1194 [inline] [<00000000b58b62f8>] sctp_do_sm+0xbdc/0x1d60 net/sctp/sm_sideeffect.c:1165 [<0000000044e11f96>] sctp_assoc_bh_rcv+0x13c/0x200 net/sctp/associola.c:1074 [<00000000ec43804d>] sctp_inq_push+0x7f/0xb0 net/sctp/inqueue.c:95 [<00000000726aa954>] sctp_backlog_rcv+0x5e/0x2a0 net/sctp/input.c:354 [<00000000d9e249a8>] sk_backlog_rcv include/net/sock.h:950 [inline] [<00000000d9e249a8>] __release_sock+0xab/0x110 net/core/sock.c:2418 [<00000000acae44fa>] release_sock+0x37/0xd0 net/core/sock.c:2934 [<00000000963cc9ae>] sctp_sendmsg+0x2c0/0x990 net/sctp/socket.c:2122 [<00000000a7fc7565>] inet_sendmsg+0x64/0x120 net/ipv4/af_inet.c:802 [<00000000b732cbd3>] sock_sendmsg_nosec net/socket.c:652 [inline] [<00000000b732cbd3>] sock_sendmsg+0x54/0x70 net/socket.c:671 [<00000000274c57ab>] ___sys_sendmsg+0x393/0x3c0 net/socket.c:2292 [<000000008252aedb>] __sys_sendmsg+0x80/0xf0 net/socket.c:2330 [<00000000f7bf23d1>] __do_sys_sendmsg net/socket.c:2339 [inline] [<00000000f7bf23d1>] __se_sys_sendmsg net/socket.c:2337 [inline] [<00000000f7bf23d1>] __x64_sys_sendmsg+0x23/0x30 net/socket.c:2337 [<00000000a8b4131f>] do_syscall_64+0x76/0x1a0 arch/x86/entry/common.c:3 The problem was that the peer.cookie value points to an skb allocated area on the first pass through this function, at which point it is overwritten with a heap allocated value, but in certain cases, where a COOKIE_ECHO chunk is included in the packet, a second pass through sctp_process_init is made, where the cookie value is re-allocated, leaking the first allocation. Fix is to always allocate the cookie value, and free it when we are done using it. Signed-off-by: Neil Horman <nhorman@tuxdriver.com> Reported-by: syzbot+f7e9153b037eac9b1df8@syzkaller.appspotmail.com CC: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> CC: "David S. Miller" <davem@davemloft.net> CC: netdev@vger.kernel.org Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2019-06-04 04:32:59 +08:00
if (sctp_state(asoc, ESTABLISHED)) {
kfree(asoc->peer.cookie);
asoc->peer.cookie = NULL;
}
if (sctp_state(asoc, ESTABLISHED) ||
sctp_state(asoc, CLOSED) ||
sctp_state(asoc, SHUTDOWN_RECEIVED)) {
/* Wake up any processes waiting in the asoc's wait queue in
* sctp_wait_for_connect() or sctp_wait_for_sndbuf().
*/
if (waitqueue_active(&asoc->wait))
wake_up_interruptible(&asoc->wait);
/* Wake up any processes waiting in the sk's sleep queue of
* a TCP-style or UDP-style peeled-off socket in
* sctp_wait_for_accept() or sctp_wait_for_packet().
* For a UDP-style socket, the waiters are woken up by the
* notifications.
*/
if (!sctp_style(sk, UDP))
sk->sk_state_change(sk);
}
if (sctp_state(asoc, SHUTDOWN_PENDING) &&
!sctp_outq_is_empty(&asoc->outqueue))
sctp_outq_uncork(&asoc->outqueue, GFP_ATOMIC);
}
/* Helper function to delete an association. */
static void sctp_cmd_delete_tcb(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc)
{
struct sock *sk = asoc->base.sk;
/* If it is a non-temporary association belonging to a TCP-style
* listening socket that is not closed, do not free it so that accept()
* can pick it up later.
*/
if (sctp_style(sk, TCP) && sctp_sstate(sk, LISTENING) &&
(!asoc->temp) && (sk->sk_shutdown != SHUTDOWN_MASK))
return;
sctp_association_free(asoc);
}
/*
* ADDIP Section 4.1 ASCONF Chunk Procedures
* A4) Start a T-4 RTO timer, using the RTO value of the selected
* destination address (we use active path instead of primary path just
* because primary path may be inactive.
*/
static void sctp_cmd_setup_t4(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc,
struct sctp_chunk *chunk)
{
struct sctp_transport *t;
t = sctp_assoc_choose_alter_transport(asoc, chunk->transport);
asoc->timeouts[SCTP_EVENT_TIMEOUT_T4_RTO] = t->rto;
chunk->transport = t;
}
/* Process an incoming Operation Error Chunk. */
static void sctp_cmd_process_operr(struct sctp_cmd_seq *cmds,
struct sctp_association *asoc,
struct sctp_chunk *chunk)
{
struct sctp_errhdr *err_hdr;
struct sctp_ulpevent *ev;
while (chunk->chunk_end > chunk->skb->data) {
err_hdr = (struct sctp_errhdr *)(chunk->skb->data);
ev = sctp_ulpevent_make_remote_error(asoc, chunk, 0,
GFP_ATOMIC);
if (!ev)
return;
asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
switch (err_hdr->cause) {
case SCTP_ERROR_UNKNOWN_CHUNK:
{
struct sctp_chunkhdr *unk_chunk_hdr;
unk_chunk_hdr = (struct sctp_chunkhdr *)
err_hdr->variable;
switch (unk_chunk_hdr->type) {
/* ADDIP 4.1 A9) If the peer responds to an ASCONF with
* an ERROR chunk reporting that it did not recognized
* the ASCONF chunk type, the sender of the ASCONF MUST
* NOT send any further ASCONF chunks and MUST stop its
* T-4 timer.
*/
case SCTP_CID_ASCONF:
if (asoc->peer.asconf_capable == 0)
break;
asoc->peer.asconf_capable = 0;
sctp_add_cmd_sf(cmds, SCTP_CMD_TIMER_STOP,
SCTP_TO(SCTP_EVENT_TIMEOUT_T4_RTO));
break;
default:
break;
}
break;
}
default:
break;
}
}
}
/* Helper function to remove the association non-primary peer
* transports.
*/
static void sctp_cmd_del_non_primary(struct sctp_association *asoc)
{
struct sctp_transport *t;
struct list_head *temp;
struct list_head *pos;
list_for_each_safe(pos, temp, &asoc->peer.transport_addr_list) {
t = list_entry(pos, struct sctp_transport, transports);
if (!sctp_cmp_addr_exact(&t->ipaddr,
&asoc->peer.primary_addr)) {
sctp_assoc_rm_peer(asoc, t);
}
}
}
/* Helper function to set sk_err on a 1-1 style socket. */
static void sctp_cmd_set_sk_err(struct sctp_association *asoc, int error)
{
struct sock *sk = asoc->base.sk;
if (!sctp_style(sk, UDP))
sk->sk_err = error;
}
/* Helper function to generate an association change event */
static void sctp_cmd_assoc_change(struct sctp_cmd_seq *commands,
struct sctp_association *asoc,
u8 state)
{
struct sctp_ulpevent *ev;
ev = sctp_ulpevent_make_assoc_change(asoc, 0, state, 0,
asoc->c.sinit_num_ostreams,
asoc->c.sinit_max_instreams,
NULL, GFP_ATOMIC);
if (ev)
asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
}
static void sctp_cmd_peer_no_auth(struct sctp_cmd_seq *commands,
struct sctp_association *asoc)
{
struct sctp_ulpevent *ev;
ev = sctp_ulpevent_make_authkey(asoc, 0, SCTP_AUTH_NO_AUTH, GFP_ATOMIC);
if (ev)
asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
}
/* Helper function to generate an adaptation indication event */
static void sctp_cmd_adaptation_ind(struct sctp_cmd_seq *commands,
struct sctp_association *asoc)
{
struct sctp_ulpevent *ev;
ev = sctp_ulpevent_make_adaptation_indication(asoc, GFP_ATOMIC);
if (ev)
asoc->stream.si->enqueue_event(&asoc->ulpq, ev);
}
static void sctp_cmd_t1_timer_update(struct sctp_association *asoc,
enum sctp_event_timeout timer,
char *name)
{
struct sctp_transport *t;
t = asoc->init_last_sent_to;
asoc->init_err_counter++;
if (t->init_sent_count > (asoc->init_cycle + 1)) {
asoc->timeouts[timer] *= 2;
if (asoc->timeouts[timer] > asoc->max_init_timeo) {
asoc->timeouts[timer] = asoc->max_init_timeo;
}
asoc->init_cycle++;
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: T1[%s] timeout adjustment init_err_counter:%d"
" cycle:%d timeout:%ld\n", __func__, name,
asoc->init_err_counter, asoc->init_cycle,
asoc->timeouts[timer]);
}
}
/* Send the whole message, chunk by chunk, to the outqueue.
* This way the whole message is queued up and bundling if
* encouraged for small fragments.
*/
static void sctp_cmd_send_msg(struct sctp_association *asoc,
struct sctp_datamsg *msg, gfp_t gfp)
{
struct sctp_chunk *chunk;
list_for_each_entry(chunk, &msg->chunks, frag_list)
sctp_outq_tail(&asoc->outqueue, chunk, gfp);
sctp: introduce stream scheduler foundations This patch introduces the hooks necessary to do stream scheduling, as per RFC Draft ndata. It also introduces the first scheduler, which is what we do today but now factored out: first come first served (FCFS). With stream scheduling now we have to track which chunk was enqueued on which stream and be able to select another other than the in front of the main outqueue. So we introduce a list on sctp_stream_out_ext structure for this purpose. We reuse sctp_chunk->transmitted_list space for the list above, as the chunk cannot belong to the two lists at the same time. By using the union in there, we can have distinct names for these moments. sctp_sched_ops are the operations expected to be implemented by each scheduler. The dequeueing is a bit particular to this implementation but it is to match how we dequeue packets today. We first dequeue and then check if it fits the packet and if not, we requeue it at head. Thus why we don't have a peek operation but have dequeue_done instead, which is called once the chunk can be safely considered as transmitted. The check removed from sctp_outq_flush is now performed by sctp_stream_outq_migrate, which is only called during assoc setup. (sctp_sendmsg() also checks for it) The only operation that is foreseen but not yet added here is a way to signalize that a new packet is starting or that the packet is done, for round robin scheduler per packet, but is intentionally left to the patch that actually implements it. Support for I-DATA chunks, also described in this RFC, with user message interleaving is straightforward as it just requires the schedulers to probe for the feature and ignore datamsg boundaries when dequeueing. See-also: https://tools.ietf.org/html/draft-ietf-tsvwg-sctp-ndata-13 Tested-by: Xin Long <lucien.xin@gmail.com> Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2017-10-04 06:20:13 +08:00
asoc->outqueue.sched->enqueue(&asoc->outqueue, msg);
}
/* These three macros allow us to pull the debugging code out of the
* main flow of sctp_do_sm() to keep attention focused on the real
* functionality there.
*/
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
#define debug_pre_sfn() \
pr_debug("%s[pre-fn]: ep:%p, %s, %s, asoc:%p[%s], %s\n", __func__, \
ep, sctp_evttype_tbl[event_type], (*debug_fn)(subtype), \
asoc, sctp_state_tbl[state], state_fn->name)
#define debug_post_sfn() \
pr_debug("%s[post-fn]: asoc:%p, status:%s\n", __func__, asoc, \
sctp_status_tbl[status])
#define debug_post_sfx() \
pr_debug("%s[post-sfx]: error:%d, asoc:%p[%s]\n", __func__, error, \
asoc, sctp_state_tbl[(asoc && sctp_id2assoc(ep->base.sk, \
sctp_assoc2id(asoc))) ? asoc->state : SCTP_STATE_CLOSED])
/*
* This is the master state machine processing function.
*
* If you want to understand all of lksctp, this is a
* good place to start.
*/
int sctp_do_sm(struct net *net, enum sctp_event_type event_type,
union sctp_subtype subtype, enum sctp_state state,
struct sctp_endpoint *ep, struct sctp_association *asoc,
void *event_arg, gfp_t gfp)
{
typedef const char *(printfn_t)(union sctp_subtype);
static printfn_t *table[] = {
NULL, sctp_cname, sctp_tname, sctp_oname, sctp_pname,
};
printfn_t *debug_fn __attribute__ ((unused)) = table[event_type];
const struct sctp_sm_table_entry *state_fn;
struct sctp_cmd_seq commands;
enum sctp_disposition status;
int error = 0;
/* Look up the state function, run it, and then process the
* side effects. These three steps are the heart of lksctp.
*/
state_fn = sctp_sm_lookup_event(net, event_type, state, subtype);
sctp_init_cmd_seq(&commands);
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
debug_pre_sfn();
status = state_fn->fn(net, ep, asoc, subtype, event_arg, &commands);
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
debug_post_sfn();
error = sctp_side_effects(event_type, subtype, state,
ep, &asoc, event_arg, status,
&commands, gfp);
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
debug_post_sfx();
return error;
}
/*****************************************************************
* This the master state function side effect processing function.
*****************************************************************/
static int sctp_side_effects(enum sctp_event_type event_type,
union sctp_subtype subtype,
enum sctp_state state,
struct sctp_endpoint *ep,
struct sctp_association **asoc,
void *event_arg,
enum sctp_disposition status,
struct sctp_cmd_seq *commands,
gfp_t gfp)
{
int error;
/* FIXME - Most of the dispositions left today would be categorized
* as "exceptional" dispositions. For those dispositions, it
* may not be proper to run through any of the commands at all.
* For example, the command interpreter might be run only with
* disposition SCTP_DISPOSITION_CONSUME.
*/
if (0 != (error = sctp_cmd_interpreter(event_type, subtype, state,
ep, *asoc,
event_arg, status,
commands, gfp)))
goto bail;
switch (status) {
case SCTP_DISPOSITION_DISCARD:
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: ignored sctp protocol event - state:%d, "
"event_type:%d, event_id:%d\n", __func__, state,
event_type, subtype.chunk);
break;
case SCTP_DISPOSITION_NOMEM:
/* We ran out of memory, so we need to discard this
* packet.
*/
/* BUG--we should now recover some memory, probably by
* reneging...
*/
error = -ENOMEM;
break;
case SCTP_DISPOSITION_DELETE_TCB:
case SCTP_DISPOSITION_ABORT:
/* This should now be a command. */
*asoc = NULL;
break;
case SCTP_DISPOSITION_CONSUME:
/*
* We should no longer have much work to do here as the
* real work has been done as explicit commands above.
*/
break;
case SCTP_DISPOSITION_VIOLATION:
net_err_ratelimited("protocol violation state %d chunkid %d\n",
state, subtype.chunk);
break;
case SCTP_DISPOSITION_NOT_IMPL:
pr_warn("unimplemented feature in state %d, event_type %d, event_id %d\n",
state, event_type, subtype.chunk);
break;
case SCTP_DISPOSITION_BUG:
pr_err("bug in state %d, event_type %d, event_id %d\n",
state, event_type, subtype.chunk);
BUG();
break;
default:
pr_err("impossible disposition %d in state %d, event_type %d, event_id %d\n",
status, state, event_type, subtype.chunk);
BUG();
break;
}
bail:
return error;
}
/********************************************************************
* 2nd Level Abstractions
********************************************************************/
/* This is the side-effect interpreter. */
static int sctp_cmd_interpreter(enum sctp_event_type event_type,
union sctp_subtype subtype,
enum sctp_state state,
struct sctp_endpoint *ep,
struct sctp_association *asoc,
void *event_arg,
enum sctp_disposition status,
struct sctp_cmd_seq *commands,
gfp_t gfp)
{
struct sctp_sock *sp = sctp_sk(ep->base.sk);
struct sctp_chunk *chunk = NULL, *new_obj;
struct sctp_packet *packet;
struct sctp_sackhdr sackh;
struct timer_list *timer;
struct sctp_transport *t;
unsigned long timeout;
struct sctp_cmd *cmd;
int local_cork = 0;
int error = 0;
int force;
if (SCTP_EVENT_T_TIMEOUT != event_type)
chunk = event_arg;
/* Note: This whole file is a huge candidate for rework.
* For example, each command could either have its own handler, so
* the loop would look like:
* while (cmds)
* cmd->handle(x, y, z)
* --jgrimm
*/
while (NULL != (cmd = sctp_next_cmd(commands))) {
switch (cmd->verb) {
case SCTP_CMD_NOP:
/* Do nothing. */
break;
case SCTP_CMD_NEW_ASOC:
/* Register a new association. */
if (local_cork) {
sctp_outq_uncork(&asoc->outqueue, gfp);
local_cork = 0;
}
/* Register with the endpoint. */
asoc = cmd->obj.asoc;
BUG_ON(asoc->peer.primary_path == NULL);
sctp_endpoint_add_asoc(ep, asoc);
break;
case SCTP_CMD_UPDATE_ASSOC:
sctp_cmd_assoc_update(commands, asoc, cmd->obj.asoc);
break;
case SCTP_CMD_PURGE_OUTQUEUE:
sctp_outq_teardown(&asoc->outqueue);
break;
case SCTP_CMD_DELETE_TCB:
if (local_cork) {
sctp_outq_uncork(&asoc->outqueue, gfp);
local_cork = 0;
}
/* Delete the current association. */
sctp_cmd_delete_tcb(commands, asoc);
asoc = NULL;
break;
case SCTP_CMD_NEW_STATE:
/* Enter a new state. */
sctp_cmd_new_state(commands, asoc, cmd->obj.state);
break;
case SCTP_CMD_REPORT_TSN:
/* Record the arrival of a TSN. */
error = sctp_tsnmap_mark(&asoc->peer.tsn_map,
sctp: be more restrictive in transport selection on bundled sacks It was noticed recently that when we send data on a transport, its possible that we might bundle a sack that arrived on a different transport. While this isn't a major problem, it does go against the SHOULD requirement in section 6.4 of RFC 2960: An endpoint SHOULD transmit reply chunks (e.g., SACK, HEARTBEAT ACK, etc.) to the same destination transport address from which it received the DATA or control chunk to which it is replying. This rule should also be followed if the endpoint is bundling DATA chunks together with the reply chunk. This patch seeks to correct that. It restricts the bundling of sack operations to only those transports which have moved the ctsn of the association forward since the last sack. By doing this we guarantee that we only bundle outbound saks on a transport that has received a chunk since the last sack. This brings us into stricter compliance with the RFC. Vlad had initially suggested that we strictly allow only sack bundling on the transport that last moved the ctsn forward. While this makes sense, I was concerned that doing so prevented us from bundling in the case where we had received chunks that moved the ctsn on multiple transports. In those cases, the RFC allows us to select any of the transports having received chunks to bundle the sack on. so I've modified the approach to allow for that, by adding a state variable to each transport that tracks weather it has moved the ctsn since the last sack. This I think keeps our behavior (and performance), close enough to our current profile that I think we can do this without a sysctl knob to enable/disable it. Signed-off-by: Neil Horman <nhorman@tuxdriver.com> CC: Vlad Yaseivch <vyasevich@gmail.com> CC: David S. Miller <davem@davemloft.net> CC: linux-sctp@vger.kernel.org Reported-by: Michele Baldessari <michele@redhat.com> Reported-by: sorin serban <sserban@redhat.com> Acked-by: Vlad Yasevich <vyasevich@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-06-30 11:04:26 +08:00
cmd->obj.u32, NULL);
break;
case SCTP_CMD_REPORT_FWDTSN:
asoc->stream.si->report_ftsn(&asoc->ulpq, cmd->obj.u32);
break;
case SCTP_CMD_PROCESS_FWDTSN:
asoc->stream.si->handle_ftsn(&asoc->ulpq,
cmd->obj.chunk);
break;
case SCTP_CMD_GEN_SACK:
/* Generate a Selective ACK.
* The argument tells us whether to just count
* the packet and MAYBE generate a SACK, or
* force a SACK out.
*/
force = cmd->obj.i32;
error = sctp_gen_sack(asoc, force, commands);
break;
case SCTP_CMD_PROCESS_SACK:
/* Process an inbound SACK. */
error = sctp_cmd_process_sack(commands, asoc,
cmd->obj.chunk);
break;
case SCTP_CMD_GEN_INIT_ACK:
/* Generate an INIT ACK chunk. */
new_obj = sctp_make_init_ack(asoc, chunk, GFP_ATOMIC,
0);
if (!new_obj) {
error = -ENOMEM;
break;
}
sctp_add_cmd_sf(commands, SCTP_CMD_REPLY,
SCTP_CHUNK(new_obj));
break;
case SCTP_CMD_PEER_INIT:
/* Process a unified INIT from the peer.
* Note: Only used during INIT-ACK processing. If
* there is an error just return to the outter
* layer which will bail.
*/
error = sctp_cmd_process_init(commands, asoc, chunk,
cmd->obj.init, gfp);
break;
case SCTP_CMD_GEN_COOKIE_ECHO:
/* Generate a COOKIE ECHO chunk. */
new_obj = sctp_make_cookie_echo(asoc, chunk);
if (!new_obj) {
if (cmd->obj.chunk)
sctp_chunk_free(cmd->obj.chunk);
error = -ENOMEM;
break;
}
sctp_add_cmd_sf(commands, SCTP_CMD_REPLY,
SCTP_CHUNK(new_obj));
/* If there is an ERROR chunk to be sent along with
* the COOKIE_ECHO, send it, too.
*/
if (cmd->obj.chunk)
sctp_add_cmd_sf(commands, SCTP_CMD_REPLY,
SCTP_CHUNK(cmd->obj.chunk));
if (new_obj->transport) {
new_obj->transport->init_sent_count++;
asoc->init_last_sent_to = new_obj->transport;
}
/* FIXME - Eventually come up with a cleaner way to
* enabling COOKIE-ECHO + DATA bundling during
* multihoming stale cookie scenarios, the following
* command plays with asoc->peer.retran_path to
* avoid the problem of sending the COOKIE-ECHO and
* DATA in different paths, which could result
* in the association being ABORTed if the DATA chunk
* is processed first by the server. Checking the
* init error counter simply causes this command
* to be executed only during failed attempts of
* association establishment.
*/
if ((asoc->peer.retran_path !=
asoc->peer.primary_path) &&
(asoc->init_err_counter > 0)) {
sctp_add_cmd_sf(commands,
SCTP_CMD_FORCE_PRIM_RETRAN,
SCTP_NULL());
}
break;
case SCTP_CMD_GEN_SHUTDOWN:
/* Generate SHUTDOWN when in SHUTDOWN_SENT state.
* Reset error counts.
*/
asoc->overall_error_count = 0;
/* Generate a SHUTDOWN chunk. */
new_obj = sctp_make_shutdown(asoc, chunk);
if (!new_obj) {
error = -ENOMEM;
break;
}
sctp_add_cmd_sf(commands, SCTP_CMD_REPLY,
SCTP_CHUNK(new_obj));
break;
case SCTP_CMD_CHUNK_ULP:
/* Send a chunk to the sockets layer. */
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: sm_sideff: chunk_up:%p, ulpq:%p\n",
__func__, cmd->obj.chunk, &asoc->ulpq);
asoc->stream.si->ulpevent_data(&asoc->ulpq,
cmd->obj.chunk,
GFP_ATOMIC);
break;
case SCTP_CMD_EVENT_ULP:
/* Send a notification to the sockets layer. */
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: sm_sideff: event_up:%p, ulpq:%p\n",
__func__, cmd->obj.ulpevent, &asoc->ulpq);
asoc->stream.si->enqueue_event(&asoc->ulpq,
cmd->obj.ulpevent);
break;
case SCTP_CMD_REPLY:
/* If an caller has not already corked, do cork. */
if (!asoc->outqueue.cork) {
sctp_outq_cork(&asoc->outqueue);
local_cork = 1;
}
/* Send a chunk to our peer. */
sctp_outq_tail(&asoc->outqueue, cmd->obj.chunk, gfp);
break;
case SCTP_CMD_SEND_PKT:
/* Send a full packet to our peer. */
packet = cmd->obj.packet;
sctp_packet_transmit(packet, gfp);
sctp_ootb_pkt_free(packet);
break;
case SCTP_CMD_T1_RETRAN:
/* Mark a transport for retransmission. */
sctp_retransmit(&asoc->outqueue, cmd->obj.transport,
SCTP_RTXR_T1_RTX);
break;
case SCTP_CMD_RETRAN:
/* Mark a transport for retransmission. */
sctp_retransmit(&asoc->outqueue, cmd->obj.transport,
SCTP_RTXR_T3_RTX);
break;
case SCTP_CMD_ECN_CE:
/* Do delayed CE processing. */
sctp_do_ecn_ce_work(asoc, cmd->obj.u32);
break;
case SCTP_CMD_ECN_ECNE:
/* Do delayed ECNE processing. */
new_obj = sctp_do_ecn_ecne_work(asoc, cmd->obj.u32,
chunk);
if (new_obj)
sctp_add_cmd_sf(commands, SCTP_CMD_REPLY,
SCTP_CHUNK(new_obj));
break;
case SCTP_CMD_ECN_CWR:
/* Do delayed CWR processing. */
sctp_do_ecn_cwr_work(asoc, cmd->obj.u32);
break;
case SCTP_CMD_SETUP_T2:
sctp_cmd_setup_t2(commands, asoc, cmd->obj.chunk);
break;
sctp: Enforce retransmission limit during shutdown When initiating a graceful shutdown while having data chunks on the retransmission queue with a peer which is in zero window mode the shutdown is never completed because the retransmission error count is reset periodically by the following two rules: - Do not timeout association while doing zero window probe. - Reset overall error count when a heartbeat request has been acknowledged. The graceful shutdown will wait for all outstanding TSN to be acknowledged before sending the SHUTDOWN request. This never happens due to the peer's zero window not acknowledging the continuously retransmitted data chunks. Although the error counter is incremented for each failed retransmission, the receiving of the SACK announcing the zero window clears the error count again immediately. Also heartbeat requests continue to be sent periodically. The peer acknowledges these requests causing the error counter to be reset as well. This patch changes behaviour to only reset the overall error counter for the above rules while not in shutdown. After reaching the maximum number of retransmission attempts, the T5 shutdown guard timer is scheduled to give the receiver some additional time to recover. The timer is stopped as soon as the receiver acknowledges any data. The issue can be easily reproduced by establishing a sctp association over the loopback device, constantly queueing data at the sender while not reading any at the receiver. Wait for the window to reach zero, then initiate a shutdown by killing both processes simultaneously. The association will never be freed and the chunks on the retransmission queue will be retransmitted indefinitely. Signed-off-by: Thomas Graf <tgraf@infradead.org> Acked-by: Vlad Yasevich <vladislav.yasevich@hp.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2011-07-07 08:28:35 +08:00
case SCTP_CMD_TIMER_START_ONCE:
timer = &asoc->timers[cmd->obj.to];
if (timer_pending(timer))
break;
/* fall through */
case SCTP_CMD_TIMER_START:
timer = &asoc->timers[cmd->obj.to];
timeout = asoc->timeouts[cmd->obj.to];
BUG_ON(!timeout);
sctp: Don't add the shutdown timer if its already been added This BUG halt was reported a while back, but the patch somehow got missed: PID: 2879 TASK: c16adaa0 CPU: 1 COMMAND: "sctpn" #0 [f418dd28] crash_kexec at c04a7d8c #1 [f418dd7c] oops_end at c0863e02 #2 [f418dd90] do_invalid_op at c040aaca #3 [f418de28] error_code (via invalid_op) at c08631a5 EAX: f34baac0 EBX: 00000090 ECX: f418deb0 EDX: f5542950 EBP: 00000000 DS: 007b ESI: f34ba800 ES: 007b EDI: f418dea0 GS: 00e0 CS: 0060 EIP: c046fa5e ERR: ffffffff EFLAGS: 00010286 #4 [f418de5c] add_timer at c046fa5e #5 [f418de68] sctp_do_sm at f8db8c77 [sctp] #6 [f418df30] sctp_primitive_SHUTDOWN at f8dcc1b5 [sctp] #7 [f418df48] inet_shutdown at c080baf9 #8 [f418df5c] sys_shutdown at c079eedf #9 [f418df70] sys_socketcall at c079fe88 EAX: ffffffda EBX: 0000000d ECX: bfceea90 EDX: 0937af98 DS: 007b ESI: 0000000c ES: 007b EDI: b7150ae4 SS: 007b ESP: bfceea7c EBP: bfceeaa8 GS: 0033 CS: 0073 EIP: b775c424 ERR: 00000066 EFLAGS: 00000282 It appears that the side effect that starts the shutdown timer was processed multiple times, which can happen as multiple paths can trigger it. This of course leads to the BUG halt in add_timer getting called. Fix seems pretty straightforward, just check before the timer is added if its already been started. If it has mod the timer instead to min(current expiration, new expiration) Its been tested but not confirmed to fix the problem, as the issue has only occured in production environments where test kernels are enjoined from being installed. It appears to be a sane fix to me though. Also, recentely, Jere found a reproducer posted on list to confirm that this resolves the issues Signed-off-by: Neil Horman <nhorman@tuxdriver.com> CC: Vlad Yasevich <vyasevich@gmail.com> CC: "David S. Miller" <davem@davemloft.net> CC: jere.leppanen@nokia.com CC: marcelo.leitner@gmail.com CC: netdev@vger.kernel.org Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2020-05-20 04:04:05 +08:00
/*
* SCTP has a hard time with timer starts. Because we process
* timer starts as side effects, it can be hard to tell if we
* have already started a timer or not, which leads to BUG
* halts when we call add_timer. So here, instead of just starting
* a timer, if the timer is already started, and just mod
* the timer with the shorter of the two expiration times
*/
if (!timer_pending(timer))
sctp_association_hold(asoc);
timer_reduce(timer, jiffies + timeout);
break;
case SCTP_CMD_TIMER_RESTART:
timer = &asoc->timers[cmd->obj.to];
timeout = asoc->timeouts[cmd->obj.to];
if (!mod_timer(timer, jiffies + timeout))
sctp_association_hold(asoc);
break;
case SCTP_CMD_TIMER_STOP:
timer = &asoc->timers[cmd->obj.to];
if (del_timer(timer))
sctp_association_put(asoc);
break;
case SCTP_CMD_INIT_CHOOSE_TRANSPORT:
chunk = cmd->obj.chunk;
t = sctp_assoc_choose_alter_transport(asoc,
asoc->init_last_sent_to);
asoc->init_last_sent_to = t;
chunk->transport = t;
t->init_sent_count++;
/* Set the new transport as primary */
sctp_assoc_set_primary(asoc, t);
break;
case SCTP_CMD_INIT_RESTART:
/* Do the needed accounting and updates
* associated with restarting an initialization
* timer. Only multiply the timeout by two if
* all transports have been tried at the current
* timeout.
*/
sctp_cmd_t1_timer_update(asoc,
SCTP_EVENT_TIMEOUT_T1_INIT,
"INIT");
sctp_add_cmd_sf(commands, SCTP_CMD_TIMER_RESTART,
SCTP_TO(SCTP_EVENT_TIMEOUT_T1_INIT));
break;
case SCTP_CMD_COOKIEECHO_RESTART:
/* Do the needed accounting and updates
* associated with restarting an initialization
* timer. Only multiply the timeout by two if
* all transports have been tried at the current
* timeout.
*/
sctp_cmd_t1_timer_update(asoc,
SCTP_EVENT_TIMEOUT_T1_COOKIE,
"COOKIE");
/* If we've sent any data bundled with
* COOKIE-ECHO we need to resend.
*/
list_for_each_entry(t, &asoc->peer.transport_addr_list,
transports) {
sctp_retransmit_mark(&asoc->outqueue, t,
SCTP_RTXR_T1_RTX);
}
sctp_add_cmd_sf(commands,
SCTP_CMD_TIMER_RESTART,
SCTP_TO(SCTP_EVENT_TIMEOUT_T1_COOKIE));
break;
case SCTP_CMD_INIT_FAILED:
sctp_cmd_init_failed(commands, asoc, cmd->obj.u32);
break;
case SCTP_CMD_ASSOC_FAILED:
sctp_cmd_assoc_failed(commands, asoc, event_type,
subtype, chunk, cmd->obj.u32);
break;
case SCTP_CMD_INIT_COUNTER_INC:
asoc->init_err_counter++;
break;
case SCTP_CMD_INIT_COUNTER_RESET:
asoc->init_err_counter = 0;
asoc->init_cycle = 0;
list_for_each_entry(t, &asoc->peer.transport_addr_list,
transports) {
t->init_sent_count = 0;
}
break;
case SCTP_CMD_REPORT_DUP:
sctp_tsnmap_mark_dup(&asoc->peer.tsn_map,
cmd->obj.u32);
break;
case SCTP_CMD_REPORT_BAD_TAG:
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: vtag mismatch!\n", __func__);
break;
case SCTP_CMD_STRIKE:
/* Mark one strike against a transport. */
sctp_do_8_2_transport_strike(commands, asoc,
cmd->obj.transport, 0);
break;
case SCTP_CMD_TRANSPORT_IDLE:
t = cmd->obj.transport;
sctp_transport_lower_cwnd(t, SCTP_LOWER_CWND_INACTIVE);
break;
case SCTP_CMD_TRANSPORT_HB_SENT:
t = cmd->obj.transport;
sctp_do_8_2_transport_strike(commands, asoc,
t, 1);
t->hb_sent = 1;
break;
case SCTP_CMD_TRANSPORT_ON:
t = cmd->obj.transport;
sctp_cmd_transport_on(commands, asoc, t, chunk);
break;
case SCTP_CMD_HB_TIMERS_START:
sctp_cmd_hb_timers_start(commands, asoc);
break;
case SCTP_CMD_HB_TIMER_UPDATE:
t = cmd->obj.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
sctp_transport_reset_hb_timer(t);
break;
case SCTP_CMD_HB_TIMERS_STOP:
sctp_cmd_hb_timers_stop(commands, asoc);
break;
case SCTP_CMD_REPORT_ERROR:
error = cmd->obj.error;
break;
case SCTP_CMD_PROCESS_CTSN:
/* Dummy up a SACK for processing. */
sackh.cum_tsn_ack = cmd->obj.be32;
sackh.a_rwnd = htonl(asoc->peer.rwnd +
asoc->outqueue.outstanding_bytes);
sackh.num_gap_ack_blocks = 0;
sackh.num_dup_tsns = 0;
chunk->subh.sack_hdr = &sackh;
sctp_add_cmd_sf(commands, SCTP_CMD_PROCESS_SACK,
SCTP_CHUNK(chunk));
break;
case SCTP_CMD_DISCARD_PACKET:
/* We need to discard the whole packet.
* Uncork the queue since there might be
* responses pending
*/
chunk->pdiscard = 1;
if (asoc) {
sctp_outq_uncork(&asoc->outqueue, gfp);
local_cork = 0;
}
break;
case SCTP_CMD_RTO_PENDING:
t = cmd->obj.transport;
t->rto_pending = 1;
break;
case SCTP_CMD_PART_DELIVER:
asoc->stream.si->start_pd(&asoc->ulpq, GFP_ATOMIC);
break;
case SCTP_CMD_RENEGE:
asoc->stream.si->renege_events(&asoc->ulpq,
cmd->obj.chunk,
GFP_ATOMIC);
break;
case SCTP_CMD_SETUP_T4:
sctp_cmd_setup_t4(commands, asoc, cmd->obj.chunk);
break;
case SCTP_CMD_PROCESS_OPERR:
sctp_cmd_process_operr(commands, asoc, chunk);
break;
case SCTP_CMD_CLEAR_INIT_TAG:
asoc->peer.i.init_tag = 0;
break;
case SCTP_CMD_DEL_NON_PRIMARY:
sctp_cmd_del_non_primary(asoc);
break;
case SCTP_CMD_T3_RTX_TIMERS_STOP:
sctp_cmd_t3_rtx_timers_stop(commands, asoc);
break;
case SCTP_CMD_FORCE_PRIM_RETRAN:
t = asoc->peer.retran_path;
asoc->peer.retran_path = asoc->peer.primary_path;
sctp_outq_uncork(&asoc->outqueue, gfp);
local_cork = 0;
asoc->peer.retran_path = t;
break;
case SCTP_CMD_SET_SK_ERR:
sctp_cmd_set_sk_err(asoc, cmd->obj.error);
break;
case SCTP_CMD_ASSOC_CHANGE:
sctp_cmd_assoc_change(commands, asoc,
cmd->obj.u8);
break;
case SCTP_CMD_ADAPTATION_IND:
sctp_cmd_adaptation_ind(commands, asoc);
break;
case SCTP_CMD_PEER_NO_AUTH:
sctp_cmd_peer_no_auth(commands, asoc);
break;
case SCTP_CMD_ASSOC_SHKEY:
error = sctp_auth_asoc_init_active_key(asoc,
GFP_ATOMIC);
break;
case SCTP_CMD_UPDATE_INITTAG:
asoc->peer.i.init_tag = cmd->obj.u32;
break;
case SCTP_CMD_SEND_MSG:
if (!asoc->outqueue.cork) {
sctp_outq_cork(&asoc->outqueue);
local_cork = 1;
}
sctp_cmd_send_msg(asoc, cmd->obj.msg, gfp);
break;
case SCTP_CMD_PURGE_ASCONF_QUEUE:
sctp_asconf_queue_teardown(asoc);
break;
case SCTP_CMD_SET_ASOC:
if (asoc && local_cork) {
sctp_outq_uncork(&asoc->outqueue, gfp);
local_cork = 0;
}
asoc = cmd->obj.asoc;
break;
default:
pr_warn("Impossible command: %u\n",
cmd->verb);
break;
}
if (error) {
cmd = sctp_next_cmd(commands);
while (cmd) {
if (cmd->verb == SCTP_CMD_REPLY)
sctp_chunk_free(cmd->obj.chunk);
cmd = sctp_next_cmd(commands);
}
break;
}
}
/* If this is in response to a received chunk, wait until
* we are done with the packet to open the queue so that we don't
* send multiple packets in response to a single request.
*/
if (asoc && SCTP_EVENT_T_CHUNK == event_type && chunk) {
if (chunk->end_of_packet || chunk->singleton)
sctp_outq_uncork(&asoc->outqueue, gfp);
} else if (local_cork)
sctp_outq_uncork(&asoc->outqueue, gfp);
if (sp->data_ready_signalled)
sp->data_ready_signalled = 0;
return error;
}