This patch is to add reconf_enable field in all of asoc ep and netns
to indicate if they support stream reset.
When initializing, asoc reconf_enable get the default value from ep
reconf_enable which is from netns netns reconf_enable by default.
It is also to add reconf_capable in asoc peer part to know if peer
supports reconf_enable, the value is set if ext params have reconf
chunk support when processing init chunk, just as rfc6525 section
5.1.1 demands.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is to add a primitive based on sctp primitive frame for
sending stream reconf request. It works as the other primitives,
and create a SCTP_CMD_REPLY command to send the request chunk out.
sctp_primitive_RECONF would be the api to send a reconf request
chunk.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is to add a per transport timer based on sctp timer frame
for stream reconf chunk retransmission. It would start after sending
a reconf request chunk, and stop after receiving the response chunk.
If the timer expires, besides retransmitting the reconf request chunk,
it would also do the same thing with data RTO timer. like to increase
the appropriate error counts, and perform threshold management, possibly
destroying the asoc if sctp retransmission thresholds are exceeded, just
as section 5.1.1 describes.
This patch is also to add asoc strreset_chunk, it is used to save the
reconf request chunk, so that it can be retransmitted, and to check if
the response is really for this request by comparing the information
inside with the response chunk as well.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch is to add asoc strreset_outseq and strreset_inseq for
saving the reconf request sequence, initialize them when create
assoc and process init, and also to define Incoming and Outgoing
SSN Reset Request Parameter described in rfc6525 section 4.1 and
4.2, As they can be in one same chunk as section rfc6525 3.1-3
describes, it makes them in one function.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sctp stream reconf, described in RFC 6525, needs a structure to
save per stream information in assoc, like stream state.
In the future, sctp stream scheduler also needs it to save some
stream scheduler params and queues.
This patchset is to prepare the stream array in assoc for stream
reconf. It defines sctp_stream that includes stream arrays inside
to replace ssnmap.
Note that we use different structures for IN and OUT streams, as
the members in per OUT stream will get more and more different
from per IN stream.
v1->v2:
- put these patches into a smaller group.
v2->v3:
- define sctp_stream to contain stream arrays, and create stream.c
to put stream-related functions.
- merge 3 patches into 1, as new sctp_stream has the same name
with before.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is no reason to use this cascading. It doesn't add anything.
Let's remove it and simplify.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This was entirely automated, using the script by Al:
PATT='^[[:blank:]]*#[[:blank:]]*include[[:blank:]]*<asm/uaccess.h>'
sed -i -e "s!$PATT!#include <linux/uaccess.h>!" \
$(git grep -l "$PATT"|grep -v ^include/linux/uaccess.h)
to do the replacement at the end of the merge window.
Requested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now sctp transport rhashtable uses hash(lport, dport, daddr) as the key
to hash a node to one chain. If in one host thousands of assocs connect
to one server with the same lport and different laddrs (although it's
not a normal case), all the transports would be hashed into the same
chain.
It may cause to keep returning -EBUSY when inserting a new node, as the
chain is too long and sctp inserts a transport node in a loop, which
could even lead to system hangs there.
The new rhlist interface works for this case that there are many nodes
with the same key in one chain. It puts them into a list then makes this
list be as a node of the chain.
This patch is to replace rhashtable_ interface with rhltable_ interface.
Since a chain would not be too long and it would not return -EBUSY with
this fix when inserting a node, the reinsert loop is also removed here.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prior to this patch, in rx path, before calling lock_sock, it needed to
hold assoc when got it by __sctp_lookup_association, in case other place
would free/put assoc.
But in __sctp_lookup_association, it lookup and hold transport, then got
assoc by transport->assoc, then hold assoc and put transport. It means
it didn't hold transport, yet it was returned and later on directly
assigned to chunk->transport.
Without the protection of sock lock, the transport may be freed/put by
other places, which would cause a use-after-free issue.
This patch is to fix this issue by holding transport instead of assoc.
As holding transport can make sure to access assoc is also safe, and
actually it looks up assoc by searching transport rhashtable, to hold
transport here makes more sense.
Note that the function will be renamed later on on another patch.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The prsctp polices include ttl expires policy already, we should remove
the old ttl expires codes, and just adjust the new polices' codes to be
compatible with the old one for users.
This patch is to remove all the old expires codes, and if prsctp polices
are not set, it will still set msg's expires_at and check the expires in
sctp_check_abandoned.
Note that asoc->prsctp_enable is set by default, so users can't feel any
difference even if they use the old expires api in userspace.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now sctp uses chunk->resent to record if a chunk is retransmitted, for
RTT measurements with retransmitted DATA chunks. chunk->sent_count was
introduced to record how many times one chunk has been sent for prsctp
RTX policy before. We actually can know if one chunk is retransmitted
by checking chunk->sent_count is greater than 1.
This patch is to remove resent from sctp_chunk and reuse sent_count
to avoid retransmitted chunks for RTT measurements.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now sctp uses chunk->prsctp_param to save the prsctp param for all the
prsctp polices, we didn't need to introduce prsctp_param to sctp_chunk.
We can just use chunk->sinfo.sinfo_timetolive for RTX and BUF polices,
and reuse msg->expires_at for TTL policy, as the prsctp polices and old
expires policy are mutual exclusive.
This patch is to remove prsctp_param from sctp_chunk, and reuse msg's
expires_at for TTL and chunk's sinfo.sinfo_timetolive for RTX and BUF
polices.
Note that sctp can't use chunk's sinfo.sinfo_timetolive for TTL policy,
as it needs a u64 variables to save the expires_at time.
This one also fixes the "netperf-Throughput_Mbps -37.2% regression"
issue.
Fixes: a6c2f79287 ("sctp: implement prsctp TTL policy")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Now pahole sctp_chunk, it has 2 memory holes:
struct sctp_chunk {
struct list_head list;
atomic_t refcnt;
/* XXX 4 bytes hole, try to pack */
...
long unsigned int prsctp_param;
int sent_count;
/* XXX 4 bytes hole, try to pack */
This patch is to move up sent_count to fill the 1st one and eliminate
the 2nd one.
It's not just another struct compaction, it also fixes the "netperf-
Throughput_Mbps -37.2% regression" issue when overloading the CPU.
Fixes: a6c2f79287 ("sctp: implement prsctp TTL policy")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Make it similar to time_before() macros:
- easier to understand
- make use of typecheck() to avoid working on unexpected variable types
(made the issue on previous patch visible)
- for _[lg]te versions, slighly faster, as the compiler used to generate
a sequence of cmp/je/cmp/js instructions and now it's sub/test/jle
(for _lte):
Before, for sctp_outq_sack:
if (primary->cacc.changeover_active) {
1f01: 80 b9 84 02 00 00 00 cmpb $0x0,0x284(%rcx)
1f08: 74 6e je 1f78 <sctp_outq_sack+0xe8>
u8 clear_cycling = 0;
if (TSN_lte(primary->cacc.next_tsn_at_change, sack_ctsn)) {
1f0a: 8b 81 80 02 00 00 mov 0x280(%rcx),%eax
return ((s) - (t)) & TSN_SIGN_BIT;
}
static inline int TSN_lte(__u32 s, __u32 t)
{
return ((s) == (t)) || (((s) - (t)) & TSN_SIGN_BIT);
1f10: 8b 7d bc mov -0x44(%rbp),%edi
1f13: 39 c7 cmp %eax,%edi
1f15: 74 25 je 1f3c <sctp_outq_sack+0xac>
1f17: 39 f8 cmp %edi,%eax
1f19: 78 21 js 1f3c <sctp_outq_sack+0xac>
primary->cacc.changeover_active = 0;
After:
if (primary->cacc.changeover_active) {
1ee7: 80 b9 84 02 00 00 00 cmpb $0x0,0x284(%rcx)
1eee: 74 73 je 1f63 <sctp_outq_sack+0xf3>
u8 clear_cycling = 0;
if (TSN_lte(primary->cacc.next_tsn_at_change, sack_ctsn)) {
1ef0: 8b 81 80 02 00 00 mov 0x280(%rcx),%eax
1ef6: 2b 45 b4 sub -0x4c(%rbp),%eax
1ef9: 85 c0 test %eax,%eax
1efb: 7e 26 jle 1f23 <sctp_outq_sack+0xb3>
primary->cacc.changeover_active = 0;
*_lt() generated pretty much the same code.
Tested with gcc (GCC) 6.1.1 20160621.
This patch also removes SSN_lte as it is not used and cleanups some
comments.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
To something more meaningful these days, specially because this is
working on packet headers or lengths and which are not tied to any CPU
arch but to the protocol itself.
So, WORD_TRUNC becomes SCTP_TRUNC4 and WORD_ROUND becomes SCTP_PAD4.
Reported-by: David Laight <David.Laight@ACULAB.COM>
Reported-by: David Miller <davem@davemloft.net>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sctp_outq_flush return value is meaningless now, this patch is
to make sctp_outq_flush return void, as well as sctp_outq_fail
and sctp_outq_uncork.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Last patch "sctp: do not return the transmit err back to sctp_sendmsg"
made sctp_primitive_SEND return err only when asoc state is unavailable.
In this case, chunks are not enqueued, they have no chance to be freed if
we don't take care of them later.
This Patch is actually to revert commit 1cd4d5c432 ("sctp: remove the
unused sctp_datamsg_free()"), commit 69b5777f2e ("sctp: hold the chunks
only after the chunk is enqueued in outq") and commit 8b570dc9f7 ("sctp:
only drop the reference on the datamsg after sending a msg"), to use
sctp_datamsg_free to free the chunks of current msg.
Fixes: 8b570dc9f7 ("sctp: only drop the reference on the datamsg after sending a msg")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This function actually operates on u32 yet its paramteres were declared
as u16, causing integer truncation upon calling.
Note in patch context that ADDIP_SERIAL_SIGN_BIT is already 32 bits.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Prior to this patch, sctp defined TCP_CLOSING as SCTP_SS_CLOSING.
TCP_CLOSING is such a special sk state in TCP that inet common codes
even exclude it.
For instance, inet_accept thinks the accept sk's state never be
TCP_CLOSING, or it will give a WARN_ON. TCP works well with that
while SCTP may trigger the call trace, as CLOSING state in SCTP
has different meaning from TCP.
This fix is to change to use TCP_CLOSE_WAIT as SCTP_SS_CLOSING,
instead of TCP_CLOSING. Some side-effects could be expected,
regardless of not being used before. inet_accept will accept it
now.
I did all the func_tests in lksctp-tools and ran sctp codnomicon
fuzzer tests against this patch, no regression or failure found.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Identifying address family operations during rx path is not something
expensive but it's ugly to the eye to have it done multiple times,
specially when we already validated it during initial rx processing.
This patch takes advantage of the now shared sctp_input_cb and make the
pointer to the operations readily available.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SCTP will try to access original IP headers on sctp_recvmsg in order to
copy the addresses used. There are also other places that do similar access
to IP or even SCTP headers. But after 90017accff ("sctp: Add GSO
support") they aren't always there because they are only present in the
header skb.
SCTP handles the queueing of incoming data by cloning the incoming skb
and limiting to only the relevant payload. This clone has its cb updated
to something different and it's then queued on socket rx queue. Thus we
need to fix this in two moments.
For rx path, not related to socket queue yet, this patch uses a
partially copied sctp_input_cb to such GSO frags. This restores the
ability to access the headers for this part of the code.
Regarding the socket rx queue, it removes iif member from sctp_event and
also add a chunk pointer on it.
With these changes we're always able to reach the headers again.
The biggest change here is that now the sctp_chunk struct and the
original skb are only freed after the application consumed the buffer.
Note however that the original payload was already like this due to the
skb cloning.
For iif, SCTP's IPv4 code doesn't use it, so no change is necessary.
IPv6 now can fetch it directly from original's IPv6 CB as the original
skb is still accessible.
In the future we probably can simplify sctp_v*_skb_iif() stuff, as
sctp_v4_skb_iif() was called but it's return value not used, and now
it's not even called, but such cleanup is out of scope for this change.
Fixes: 90017accff ("sctp: Add GSO support")
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The next patch needs 8 bytes in there. sctp_ulpevent has a hole due to
bad alignment; msg_flags is using 4 bytes while it actually uses only 2, so
we shrink it, and iif member (4 bytes) which can be easily fetched from
another place once the next patch is there, so we remove it and thus
creating space for 8 bytes.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We process input path in other files too and having access to it is
nice, so move it to a header where it's shared.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
prsctp PRIO policy is a policy to abandon lower priority chunks when
asoc doesn't have enough snd buffer, so that the current chunk with
higher priority can be queued successfully.
Similar to TTL/RTX policy, we will set the priority of the chunk to
prsctp_param with sinfo->sinfo_timetolive in sctp_set_prsctp_policy().
So if PRIO policy is enabled, msg->expire_at won't work.
asoc->sent_cnt_removable will record how many chunks can be checked to
remove. If priority policy is enabled, when the chunk is queued into
the out_queue, we will increase sent_cnt_removable. When the chunk is
moved to abandon_queue or dequeue and free, we will decrease
sent_cnt_removable.
In sctp_sendmsg, we will check if there is enough snd buffer for current
msg and if sent_cnt_removable is not 0. Then try to abandon chunks in
sctp_prune_prsctp when sendmsg from the retransmit/transmited queue, and
free chunks from out_queue in right order until the abandon+free size >
msg_len - sctp_wfree. For the abandon size, we have to wait until it
sends FORWARD TSN, receives the sack and the chunks are really freed.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
prsctp TTL policy is a policy to abandon chunks when they expire
at the specific time in local stack. It's similar with expires_at
in struct sctp_datamsg.
This patch uses sinfo->sinfo_timetolive to set the specific time for
TTL policy. sinfo->sinfo_timetolive is also used for msg->expires_at.
So if prsctp_enable or TTL policy is not enabled, msg->expires_at
still works as before.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This patch adds SCTP_PR_ASSOC_STATUS to sctp sockopt, which is used
to dump the prsctp statistics info from the asoc. The prsctp statistics
includes abandoned_sent/unsent from the asoc. abandoned_sent is the
count of the packets we drop packets from retransmit/transmited queue,
and abandoned_unsent is the count of the packets we drop from out_queue
according to the policy.
Note: another option for prsctp statistics dump described in rfc is
SCTP_PR_STREAM_STATUS, which is used to dump the prsctp statistics
info from each stream. But by now, linux doesn't yet have per stream
statistics info, it needs rfc6525 to be implemented. As the prsctp
statistics for each stream has to be based on per stream statistics,
we will delay it until rfc6525 is done in linux.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
According to section 4.5 of rfc7496, prsctp_enable should be per asoc.
We will add prsctp_enable to both asoc and ep, and replace the places
where it used net.sctp->prsctp_enable with asoc->prsctp_enable.
ep->prsctp_enable will be initialized with net.sctp->prsctp_enable, and
asoc->prsctp_enable will be initialized with ep->prsctp_enable. We can
also modify it's value through sockopt SCTP_PR_SUPPORTED.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SCTP has this pecualiarity that its packets cannot be just segmented to
(P)MTU. Its chunks must be contained in IP segments, padding respected.
So we can't just generate a big skb, set gso_size to the fragmentation
point and deliver it to IP layer.
This patch takes a different approach. SCTP will now build a skb as it
would be if it was received using GRO. That is, there will be a cover
skb with protocol headers and children ones containing the actual
segments, already segmented to a way that respects SCTP RFCs.
With that, we can tell skb_segment() to just split based on frag_list,
trusting its sizes are already in accordance.
This way SCTP can benefit from GSO and instead of passing several
packets through the stack, it can pass a single large packet.
v2:
- Added support for receiving GSO frames, as requested by Dave Miller.
- Clear skb->cb if packet is GSO (otherwise it's not used by SCTP)
- Added heuristics similar to what we have in TCP for not generating
single GSO packets that fills cwnd.
v3:
- consider sctphdr size in skb_gso_transport_seglen()
- rebased due to 5c7cdf339a ("gso: Remove arbitrary checks for
unsupported GSO")
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Tested-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Dave Miller pointed out that fb586f2530 ("sctp: delay calls to
sk_data_ready() as much as possible") may insert latency specially if
the receiving application is running on another CPU and that it would be
better if we signalled as early as possible.
This patch thus basically inverts the logic on fb586f2530 and signals
it as early as possible, similar to what we had before.
Fixes: fb586f2530 ("sctp: delay calls to sk_data_ready() as much as possible")
Reported-by: Dave Miller <davem@davemloft.net>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
There is nothing related to BH in SNMP counters anymore,
since linux-3.0.
Rename helpers to use __ prefix instead of _BH prefix,
for contexts where preemption is disabled.
This more closely matches convention used to update
percpu variables.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Rename SCTP_INC_STATS_BH() to __SCTP_INC_STATS()
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the old days (before linux-3.0), SNMP counters were duplicated,
one for user context, and one for BH context.
After commit 8f0ea0fe3a ("snmp: reduce percpu needs by 50%")
we have a single copy, and what really matters is preemption being
enabled or disabled, since we use this_cpu_inc() or __this_cpu_inc()
respectively.
We therefore kill SNMP_INC_STATS_USER(), SNMP_ADD_STATS_USER(),
NET_INC_STATS_USER(), NET_ADD_STATS_USER(), SCTP_INC_STATS_USER(),
SNMP_INC_STATS64_USER(), SNMP_ADD_STATS64_USER(), TCP_ADD_STATS_USER(),
UDP_INC_STATS_USER(), UDP6_INC_STATS_USER(), and XFRM_INC_STATS_USER()
Following patches will rename __BH helpers to make clear their
usage is not tied to BH being disabled.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Conflicts were two cases of simple overlapping changes,
nothing serious.
In the UDP case, we need to add a hlist_add_tail_rcu()
to linux/rculist.h, because we've moved UDP socket handling
away from using nulls lists.
Signed-off-by: David S. Miller <davem@davemloft.net>
For some main variables in sctp.ko, we couldn't export it to other modules,
so we have to define some api to access them.
It will include sctp transport and endpoint's traversal.
There are some transport traversal functions for sctp_diag, we can also
use it for sctp_proc. cause they have the similar situation to traversal
transport.
v2->v3:
- rhashtable_walk_init need the parameter gfp, because of recent upstrem
update
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
sctp_diag will dump some important details of sctp's assoc or ep, we use
sctp_info to describe them, sctp_get_sctp_info to get them, and export
it to sctp_diag.ko.
v2->v3:
- we will not use list_for_each_safe in sctp_get_sctp_info, cause
all the callers of it will use lock_sock.
- fix the holes in struct sctp_info with __reserved* field.
because sctp_diag is a new feature, and sctp_info is just for now,
it may be changed in the future.
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SCTP already serializes access to rcvbuf through its sock lock:
sctp_recvmsg takes it right in the start and release at the end, while
rx path will also take the lock before doing any socket processing. On
sctp_rcv() it will check if there is an user using the socket and, if
there is, it will queue incoming packets to the backlog. The backlog
processing will do the same. Even timers will do such check and
re-schedule if an user is using the socket.
Simplifying this will allow us to remove sctp_skb_list_tail and get ride
of some expensive lockings. The lists that it is used on are also
mangled with functions like __skb_queue_tail and __skb_unlink in the
same context, like on sctp_ulpq_tail_event() and sctp_clear_pd().
sctp_close() will also purge those while using only the sock lock.
Therefore the lockings performed by sctp_skb_list_tail() are not
necessary. This patch removes this function and replaces its calls with
just skb_queue_splice_tail_init() instead.
The biggest gain is at sctp_ulpq_tail_event(), because the events always
contain a list, even if it's queueing a single skb and this was
triggering expensive calls to spin_lock_irqsave/_irqrestore for every
data chunk received.
As SCTP will deliver each data chunk on a corresponding recvmsg, the
more effective the change will be.
Before this patch, with chunks with 30 bytes:
netperf -t SCTP_STREAM -H 192.168.1.2 -cC -l 60 -- -m 30 -S 400000
400000 -s 400000 400000
on a 10Gbit link with 1500 MTU:
SCTP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.1.1 () port 0 AF_INET
Recv Send Send Utilization Service Demand
Socket Socket Message Elapsed Send Recv Send Recv
Size Size Size Time Throughput local remote local remote
bytes bytes bytes secs. 10^6bits/s % S % S us/KB us/KB
425984 425984 30 60.00 137.45 7.34 7.36 52.504 52.608
With it:
SCTP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.1.1 () port 0 AF_INET
Recv Send Send Utilization Service Demand
Socket Socket Message Elapsed Send Recv Send Recv
Size Size Size Time Throughput local remote local remote
bytes bytes bytes secs. 10^6bits/s % S % S us/KB us/KB
425984 425984 30 60.00 179.10 7.97 6.70 43.740 36.788
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Currently processing of multiple chunks in a single SCTP packet leads to
multiple calls to sk_data_ready, causing multiple wake up signals which
are costy and doesn't make it wake up any faster.
With this patch it will note that the wake up is pending and will do it
before leaving the state machine interpreter, latest place possible to
do it realiably and cleanly.
Note that sk_data_ready events are not dependent on asocs, unlike waking
up writers.
v2: series re-checked
v3: use local vars to cleanup the code, suggested by Jakub Sitnicki
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
It wastes space and gets worse as we add new flags, so convert bit-wide
flags to a bitfield.
Currently it already saves 4 bytes in sctp_sock, which are left as holes
in it for now. The whole struct needs packing, which should be done in
another patch.
Note that do_auto_asconf cannot be merged, as explained in the comment
before it.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
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>
Use list_* helpers in sctp_list_dequeue, more readable.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the user supply a different fragmentation point or if there is a
network header that cause it to not be aligned, force it to be aligned.
Fragmentation point at a value that is not aligned is not optimal. It
causes extra padding to be used and has just no pros.
v2:
- Make use of the new WORD_TRUNC macro
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
SCTP is a protocol that is aligned to a word (4 bytes). Thus using bare
MTU can sometimes return values that are not aligned, like for loopback,
which is 65536 but ipv4_mtu() limits that to 65535. This mis-alignment
will cause the last non-aligned bytes to never be used and can cause
issues with congestion control.
So it's better to just consider a lower MTU and keep congestion control
calcs saner as they are based on PMTU.
Same applies to icmp frag needed messages, which is also fixed by this
patch.
One other effect of this is the inability to send MTU-sized packet
without queueing or fragmentation and without hitting Nagle. As the
check performed at sctp_packet_can_append_data():
if (chunk->skb->len + q->out_qlen >= transport->pathmtu - packet->overhead)
/* Enough data queued to fill a packet */
return SCTP_XMIT_OK;
with the above example of MTU, if there are no other messages queued,
one cannot send a packet that just fits one packet (65532 bytes) and
without causing DATA chunk fragmentation or a delay.
v2:
- Added WORD_TRUNC macro
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull networking updates from David Miller:
"Highlights:
1) Support more Realtek wireless chips, from Jes Sorenson.
2) New BPF types for per-cpu hash and arrap maps, from Alexei
Starovoitov.
3) Make several TCP sysctls per-namespace, from Nikolay Borisov.
4) Allow the use of SO_REUSEPORT in order to do per-thread processing
of incoming TCP/UDP connections. The muxing can be done using a
BPF program which hashes the incoming packet. From Craig Gallek.
5) Add a multiplexer for TCP streams, to provide a messaged based
interface. BPF programs can be used to determine the message
boundaries. From Tom Herbert.
6) Add 802.1AE MACSEC support, from Sabrina Dubroca.
7) Avoid factorial complexity when taking down an inetdev interface
with lots of configured addresses. We were doing things like
traversing the entire address less for each address removed, and
flushing the entire netfilter conntrack table for every address as
well.
8) Add and use SKB bulk free infrastructure, from Jesper Brouer.
9) Allow offloading u32 classifiers to hardware, and implement for
ixgbe, from John Fastabend.
10) Allow configuring IRQ coalescing parameters on a per-queue basis,
from Kan Liang.
11) Extend ethtool so that larger link mode masks can be supported.
From David Decotigny.
12) Introduce devlink, which can be used to configure port link types
(ethernet vs Infiniband, etc.), port splitting, and switch device
level attributes as a whole. From Jiri Pirko.
13) Hardware offload support for flower classifiers, from Amir Vadai.
14) Add "Local Checksum Offload". Basically, for a tunneled packet
the checksum of the outer header is 'constant' (because with the
checksum field filled into the inner protocol header, the payload
of the outer frame checksums to 'zero'), and we can take advantage
of that in various ways. From Edward Cree"
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1548 commits)
bonding: fix bond_get_stats()
net: bcmgenet: fix dma api length mismatch
net/mlx4_core: Fix backward compatibility on VFs
phy: mdio-thunder: Fix some Kconfig typos
lan78xx: add ndo_get_stats64
lan78xx: handle statistics counter rollover
RDS: TCP: Remove unused constant
RDS: TCP: Add sysctl tunables for sndbuf/rcvbuf on rds-tcp socket
net: smc911x: convert pxa dma to dmaengine
team: remove duplicate set of flag IFF_MULTICAST
bonding: remove duplicate set of flag IFF_MULTICAST
net: fix a comment typo
ethernet: micrel: fix some error codes
ip_tunnels, bpf: define IP_TUNNEL_OPTS_MAX and use it
bpf, dst: add and use dst_tclassid helper
bpf: make skb->tc_classid also readable
net: mvneta: bm: clarify dependencies
cls_bpf: reset class and reuse major in da
ldmvsw: Checkpatch sunvnet.c and sunvnet_common.c
ldmvsw: Add ldmvsw.c driver code
...
Pull crypto update from Herbert Xu:
"Here is the crypto update for 4.6:
API:
- Convert remaining crypto_hash users to shash or ahash, also convert
blkcipher/ablkcipher users to skcipher.
- Remove crypto_hash interface.
- Remove crypto_pcomp interface.
- Add crypto engine for async cipher drivers.
- Add akcipher documentation.
- Add skcipher documentation.
Algorithms:
- Rename crypto/crc32 to avoid name clash with lib/crc32.
- Fix bug in keywrap where we zero the wrong pointer.
Drivers:
- Support T5/M5, T7/M7 SPARC CPUs in n2 hwrng driver.
- Add PIC32 hwrng driver.
- Support BCM6368 in bcm63xx hwrng driver.
- Pack structs for 32-bit compat users in qat.
- Use crypto engine in omap-aes.
- Add support for sama5d2x SoCs in atmel-sha.
- Make atmel-sha available again.
- Make sahara hashing available again.
- Make ccp hashing available again.
- Make sha1-mb available again.
- Add support for multiple devices in ccp.
- Improve DMA performance in caam.
- Add hashing support to rockchip"
* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (116 commits)
crypto: qat - remove redundant arbiter configuration
crypto: ux500 - fix checks of error code returned by devm_ioremap_resource()
crypto: atmel - fix checks of error code returned by devm_ioremap_resource()
crypto: qat - Change the definition of icp_qat_uof_regtype
hwrng: exynos - use __maybe_unused to hide pm functions
crypto: ccp - Add abstraction for device-specific calls
crypto: ccp - CCP versioning support
crypto: ccp - Support for multiple CCPs
crypto: ccp - Remove check for x86 family and model
crypto: ccp - memset request context to zero during import
lib/mpi: use "static inline" instead of "extern inline"
lib/mpi: avoid assembler warning
hwrng: bcm63xx - fix non device tree compatibility
crypto: testmgr - allow rfc3686 aes-ctr variants in fips mode.
crypto: qat - The AE id should be less than the maximal AE number
lib/mpi: Endianness fix
crypto: rockchip - add hash support for crypto engine in rk3288
crypto: xts - fix compile errors
crypto: doc - add skcipher API documentation
crypto: doc - update AEAD AD handling
...
Currently sctp_sendmsg() triggers some calls that will allocate memory
with GFP_ATOMIC even when not necessary. In the case of
sctp_packet_transmit it will allocate a linear skb that will be used to
construct the packet and this may cause sends to fail due to ENOMEM more
often than anticipated specially with big MTUs.
This patch thus allows it to inherit gfp flags from upper calls so that
it can use GFP_KERNEL if it was triggered by a sctp_sendmsg call or
similar. All others, like retransmits or flushes started from BH, are
still allocated using GFP_ATOMIC.
In netperf tests this didn't result in any performance drawbacks when
memory is not too fragmented and made it trigger ENOMEM way less often.
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Dmitry reported that sctp_add_bind_addr may read more bytes than
expected in case the parameter is a IPv4 addr supplied by the user
through calls such as sctp_bindx_add(), because it always copies
sizeof(union sctp_addr) while the buffer may be just a struct
sockaddr_in, which is smaller.
This patch then fixes it by limiting the memcpy to the min between the
union size and a (new parameter) provided addr size. Where possible this
parameter still is the size of that union, except for reading from
user-provided buffers, which then it accounts for protocol type.
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Tested-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Xin Long