freebsd-dev/sys/netinet/tcp_lro.c

1880 lines
48 KiB
C
Raw Normal View History

/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 2007, Myricom Inc.
* Copyright (c) 2008, Intel Corporation.
* Copyright (c) 2012 The FreeBSD Foundation
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
* Copyright (c) 2016-2021 Mellanox Technologies.
* All rights reserved.
*
* Portions of this software were developed by Bjoern Zeeb
* under sponsorship from the FreeBSD Foundation.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
2008-06-11 22:12:50 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_inet.h"
#include "opt_inet6.h"
2008-06-11 22:12:50 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
2008-06-11 22:12:50 +00:00
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/sockbuf.h>
#include <sys/sysctl.h>
2008-06-11 22:12:50 +00:00
#include <net/if.h>
#include <net/if_var.h>
2008-06-11 22:12:50 +00:00
#include <net/ethernet.h>
#include <net/bpf.h>
#include <net/vnet.h>
2008-06-11 22:12:50 +00:00
#include <netinet/in_systm.h>
#include <netinet/in.h>
#include <netinet/ip6.h>
2008-06-11 22:12:50 +00:00
#include <netinet/ip.h>
#include <netinet/ip_var.h>
#include <netinet/in_pcb.h>
#include <netinet6/in6_pcb.h>
2008-06-11 22:12:50 +00:00
#include <netinet/tcp.h>
#include <netinet/tcp_seq.h>
2008-06-11 22:12:50 +00:00
#include <netinet/tcp_lro.h>
#include <netinet/tcp_var.h>
#include <netinet/tcpip.h>
#include <netinet/tcp_hpts.h>
#include <netinet/tcp_log_buf.h>
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
#include <netinet/udp.h>
#include <netinet6/ip6_var.h>
2008-06-11 22:12:50 +00:00
#include <machine/in_cksum.h>
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
static MALLOC_DEFINE(M_LRO, "LRO", "LRO control structures");
2008-06-11 22:12:50 +00:00
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
#define TCP_LRO_TS_OPTION \
ntohl((TCPOPT_NOP << 24) | (TCPOPT_NOP << 16) | \
(TCPOPT_TIMESTAMP << 8) | TCPOLEN_TIMESTAMP)
2008-06-11 22:12:50 +00:00
static void tcp_lro_rx_done(struct lro_ctrl *lc);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
static int tcp_lro_rx_common(struct lro_ctrl *lc, struct mbuf *m,
uint32_t csum, bool use_hash);
#ifdef TCPHPTS
static bool do_bpf_strip_and_compress(struct inpcb *, struct lro_ctrl *,
struct lro_entry *, struct mbuf **, struct mbuf **, struct mbuf **, bool *, bool);
#endif
SYSCTL_NODE(_net_inet_tcp, OID_AUTO, lro, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
"TCP LRO");
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
static long tcplro_stacks_wanting_mbufq;
counter_u64_t tcp_inp_lro_direct_queue;
counter_u64_t tcp_inp_lro_wokeup_queue;
counter_u64_t tcp_inp_lro_compressed;
counter_u64_t tcp_inp_lro_locks_taken;
counter_u64_t tcp_extra_mbuf;
counter_u64_t tcp_would_have_but;
counter_u64_t tcp_comp_total;
counter_u64_t tcp_uncomp_total;
static unsigned tcp_lro_entries = TCP_LRO_ENTRIES;
SYSCTL_UINT(_net_inet_tcp_lro, OID_AUTO, entries,
CTLFLAG_RDTUN | CTLFLAG_MPSAFE, &tcp_lro_entries, 0,
"default number of LRO entries");
SYSCTL_COUNTER_U64(_net_inet_tcp_lro, OID_AUTO, fullqueue, CTLFLAG_RD,
&tcp_inp_lro_direct_queue, "Number of lro's fully queued to transport");
SYSCTL_COUNTER_U64(_net_inet_tcp_lro, OID_AUTO, wokeup, CTLFLAG_RD,
&tcp_inp_lro_wokeup_queue, "Number of lro's where we woke up transport via hpts");
SYSCTL_COUNTER_U64(_net_inet_tcp_lro, OID_AUTO, compressed, CTLFLAG_RD,
&tcp_inp_lro_compressed, "Number of lro's compressed and sent to transport");
SYSCTL_COUNTER_U64(_net_inet_tcp_lro, OID_AUTO, lockcnt, CTLFLAG_RD,
&tcp_inp_lro_locks_taken, "Number of lro's inp_wlocks taken");
SYSCTL_COUNTER_U64(_net_inet_tcp_lro, OID_AUTO, extra_mbuf, CTLFLAG_RD,
&tcp_extra_mbuf, "Number of times we had an extra compressed ack dropped into the tp");
SYSCTL_COUNTER_U64(_net_inet_tcp_lro, OID_AUTO, would_have_but, CTLFLAG_RD,
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
&tcp_would_have_but, "Number of times we would have had an extra compressed, but mget failed");
SYSCTL_COUNTER_U64(_net_inet_tcp_lro, OID_AUTO, with_m_ackcmp, CTLFLAG_RD,
&tcp_comp_total, "Number of mbufs queued with M_ACKCMP flags set");
SYSCTL_COUNTER_U64(_net_inet_tcp_lro, OID_AUTO, without_m_ackcmp, CTLFLAG_RD,
&tcp_uncomp_total, "Number of mbufs queued without M_ACKCMP");
void
tcp_lro_reg_mbufq(void)
{
atomic_fetchadd_long(&tcplro_stacks_wanting_mbufq, 1);
}
void
tcp_lro_dereg_mbufq(void)
{
atomic_fetchadd_long(&tcplro_stacks_wanting_mbufq, -1);
}
static __inline void
tcp_lro_active_insert(struct lro_ctrl *lc, struct lro_head *bucket,
struct lro_entry *le)
{
LIST_INSERT_HEAD(&lc->lro_active, le, next);
LIST_INSERT_HEAD(bucket, le, hash_next);
}
static __inline void
tcp_lro_active_remove(struct lro_entry *le)
{
LIST_REMOVE(le, next); /* active list */
LIST_REMOVE(le, hash_next); /* hash bucket */
}
2008-06-11 22:12:50 +00:00
int
tcp_lro_init(struct lro_ctrl *lc)
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
{
return (tcp_lro_init_args(lc, NULL, tcp_lro_entries, 0));
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
}
int
tcp_lro_init_args(struct lro_ctrl *lc, struct ifnet *ifp,
unsigned lro_entries, unsigned lro_mbufs)
2008-06-11 22:12:50 +00:00
{
struct lro_entry *le;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
size_t size;
unsigned i, elements;
2008-06-11 22:12:50 +00:00
lc->lro_bad_csum = 0;
lc->lro_queued = 0;
lc->lro_flushed = 0;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
lc->lro_mbuf_count = 0;
lc->lro_mbuf_max = lro_mbufs;
lc->lro_cnt = lro_entries;
lc->lro_ackcnt_lim = TCP_LRO_ACKCNT_MAX;
lc->lro_length_lim = TCP_LRO_LENGTH_MAX;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
lc->ifp = ifp;
LIST_INIT(&lc->lro_free);
LIST_INIT(&lc->lro_active);
2008-06-11 22:12:50 +00:00
/* create hash table to accelerate entry lookup */
if (lro_entries > lro_mbufs)
elements = lro_entries;
else
elements = lro_mbufs;
lc->lro_hash = phashinit_flags(elements, M_LRO, &lc->lro_hashsz,
HASH_NOWAIT);
if (lc->lro_hash == NULL) {
memset(lc, 0, sizeof(*lc));
return (ENOMEM);
}
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* compute size to allocate */
size = (lro_mbufs * sizeof(struct lro_mbuf_sort)) +
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
(lro_entries * sizeof(*le));
lc->lro_mbuf_data = (struct lro_mbuf_sort *)
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
malloc(size, M_LRO, M_NOWAIT | M_ZERO);
/* check for out of memory */
if (lc->lro_mbuf_data == NULL) {
free(lc->lro_hash, M_LRO);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
memset(lc, 0, sizeof(*lc));
return (ENOMEM);
}
/* compute offset for LRO entries */
le = (struct lro_entry *)
(lc->lro_mbuf_data + lro_mbufs);
/* setup linked list */
for (i = 0; i != lro_entries; i++)
LIST_INSERT_HEAD(&lc->lro_free, le + i, next);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
return (0);
2008-06-11 22:12:50 +00:00
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
struct vxlan_header {
uint32_t vxlh_flags;
uint32_t vxlh_vni;
};
static inline void *
tcp_lro_low_level_parser(void *ptr, struct lro_parser *parser, bool update_data, bool is_vxlan)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
const struct ether_vlan_header *eh;
void *old;
uint16_t eth_type;
if (update_data)
memset(parser, 0, sizeof(*parser));
old = ptr;
if (is_vxlan) {
const struct vxlan_header *vxh;
vxh = ptr;
ptr = (uint8_t *)ptr + sizeof(*vxh);
if (update_data) {
parser->data.vxlan_vni =
vxh->vxlh_vni & htonl(0xffffff00);
}
}
eh = ptr;
if (__predict_false(eh->evl_encap_proto == htons(ETHERTYPE_VLAN))) {
eth_type = eh->evl_proto;
if (update_data) {
/* strip priority and keep VLAN ID only */
parser->data.vlan_id = eh->evl_tag & htons(EVL_VLID_MASK);
}
/* advance to next header */
ptr = (uint8_t *)ptr + ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN;
} else {
eth_type = eh->evl_encap_proto;
/* advance to next header */
ptr = (uint8_t *)ptr + ETHER_HDR_LEN;
}
switch (eth_type) {
#ifdef INET
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
case htons(ETHERTYPE_IP):
parser->ip4 = ptr;
/* Ensure there are no IPv4 options. */
if ((parser->ip4->ip_hl << 2) != sizeof (*parser->ip4))
break;
/* .. and the packet is not fragmented. */
if (parser->ip4->ip_off & htons(IP_MF|IP_OFFMASK))
break;
ptr = (uint8_t *)ptr + (parser->ip4->ip_hl << 2);
if (update_data) {
parser->data.s_addr.v4 = parser->ip4->ip_src;
parser->data.d_addr.v4 = parser->ip4->ip_dst;
}
switch (parser->ip4->ip_p) {
case IPPROTO_UDP:
parser->udp = ptr;
if (update_data) {
parser->data.lro_type = LRO_TYPE_IPV4_UDP;
parser->data.s_port = parser->udp->uh_sport;
parser->data.d_port = parser->udp->uh_dport;
} else {
MPASS(parser->data.lro_type == LRO_TYPE_IPV4_UDP);
}
ptr = ((uint8_t *)ptr + sizeof(*parser->udp));
parser->total_hdr_len = (uint8_t *)ptr - (uint8_t *)old;
return (ptr);
case IPPROTO_TCP:
parser->tcp = ptr;
if (update_data) {
parser->data.lro_type = LRO_TYPE_IPV4_TCP;
parser->data.s_port = parser->tcp->th_sport;
parser->data.d_port = parser->tcp->th_dport;
} else {
MPASS(parser->data.lro_type == LRO_TYPE_IPV4_TCP);
}
ptr = (uint8_t *)ptr + (parser->tcp->th_off << 2);
parser->total_hdr_len = (uint8_t *)ptr - (uint8_t *)old;
return (ptr);
default:
break;
}
break;
#endif
#ifdef INET6
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
case htons(ETHERTYPE_IPV6):
parser->ip6 = ptr;
ptr = (uint8_t *)ptr + sizeof(*parser->ip6);
if (update_data) {
parser->data.s_addr.v6 = parser->ip6->ip6_src;
parser->data.d_addr.v6 = parser->ip6->ip6_dst;
}
switch (parser->ip6->ip6_nxt) {
case IPPROTO_UDP:
parser->udp = ptr;
if (update_data) {
parser->data.lro_type = LRO_TYPE_IPV6_UDP;
parser->data.s_port = parser->udp->uh_sport;
parser->data.d_port = parser->udp->uh_dport;
} else {
MPASS(parser->data.lro_type == LRO_TYPE_IPV6_UDP);
}
ptr = (uint8_t *)ptr + sizeof(*parser->udp);
parser->total_hdr_len = (uint8_t *)ptr - (uint8_t *)old;
return (ptr);
case IPPROTO_TCP:
parser->tcp = ptr;
if (update_data) {
parser->data.lro_type = LRO_TYPE_IPV6_TCP;
parser->data.s_port = parser->tcp->th_sport;
parser->data.d_port = parser->tcp->th_dport;
} else {
MPASS(parser->data.lro_type == LRO_TYPE_IPV6_TCP);
}
ptr = (uint8_t *)ptr + (parser->tcp->th_off << 2);
parser->total_hdr_len = (uint8_t *)ptr - (uint8_t *)old;
return (ptr);
default:
break;
}
break;
#endif
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
default:
break;
}
/* Invalid packet - cannot parse */
return (NULL);
}
static const int vxlan_csum = CSUM_INNER_L3_CALC | CSUM_INNER_L3_VALID |
CSUM_INNER_L4_CALC | CSUM_INNER_L4_VALID;
static inline struct lro_parser *
tcp_lro_parser(struct mbuf *m, struct lro_parser *po, struct lro_parser *pi, bool update_data)
{
void *data_ptr;
/* Try to parse outer headers first. */
data_ptr = tcp_lro_low_level_parser(m->m_data, po, update_data, false);
if (data_ptr == NULL || po->total_hdr_len > m->m_len)
return (NULL);
if (update_data) {
/* Store VLAN ID, if any. */
if (__predict_false(m->m_flags & M_VLANTAG)) {
po->data.vlan_id =
htons(m->m_pkthdr.ether_vtag) & htons(EVL_VLID_MASK);
}
}
switch (po->data.lro_type) {
case LRO_TYPE_IPV4_UDP:
case LRO_TYPE_IPV6_UDP:
/* Check for VXLAN headers. */
if ((m->m_pkthdr.csum_flags & vxlan_csum) != vxlan_csum)
break;
/* Try to parse inner headers. */
data_ptr = tcp_lro_low_level_parser(data_ptr, pi, update_data, true);
if (data_ptr == NULL || pi->total_hdr_len > m->m_len)
break;
/* Verify supported header types. */
switch (pi->data.lro_type) {
case LRO_TYPE_IPV4_TCP:
case LRO_TYPE_IPV6_TCP:
return (pi);
default:
break;
}
break;
case LRO_TYPE_IPV4_TCP:
case LRO_TYPE_IPV6_TCP:
if (update_data)
memset(pi, 0, sizeof(*pi));
return (po);
default:
break;
}
return (NULL);
}
static inline int
tcp_lro_trim_mbuf_chain(struct mbuf *m, const struct lro_parser *po)
{
int len;
switch (po->data.lro_type) {
#ifdef INET
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
case LRO_TYPE_IPV4_TCP:
len = ((uint8_t *)po->ip4 - (uint8_t *)m->m_data) +
ntohs(po->ip4->ip_len);
break;
#endif
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
#ifdef INET6
case LRO_TYPE_IPV6_TCP:
len = ((uint8_t *)po->ip6 - (uint8_t *)m->m_data) +
ntohs(po->ip6->ip6_plen) + sizeof(*po->ip6);
break;
#endif
default:
return (TCP_LRO_CANNOT);
}
/*
* If the frame is padded beyond the end of the IP packet,
* then trim the extra bytes off:
*/
if (__predict_true(m->m_pkthdr.len == len)) {
return (0);
} else if (m->m_pkthdr.len > len) {
m_adj(m, len - m->m_pkthdr.len);
return (0);
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
return (TCP_LRO_CANNOT);
}
static struct tcphdr *
tcp_lro_get_th(struct mbuf *m)
{
return ((struct tcphdr *)((uint8_t *)m->m_data + m->m_pkthdr.lro_tcp_h_off));
}
static void
lro_free_mbuf_chain(struct mbuf *m)
{
struct mbuf *save;
while (m) {
save = m->m_nextpkt;
m->m_nextpkt = NULL;
m_freem(m);
m = save;
}
}
2008-06-11 22:12:50 +00:00
void
tcp_lro_free(struct lro_ctrl *lc)
2008-06-11 22:12:50 +00:00
{
struct lro_entry *le;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
unsigned x;
/* reset LRO free list */
LIST_INIT(&lc->lro_free);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* free active mbufs, if any */
while ((le = LIST_FIRST(&lc->lro_active)) != NULL) {
tcp_lro_active_remove(le);
lro_free_mbuf_chain(le->m_head);
}
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* free hash table */
free(lc->lro_hash, M_LRO);
lc->lro_hash = NULL;
lc->lro_hashsz = 0;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* free mbuf array, if any */
for (x = 0; x != lc->lro_mbuf_count; x++)
m_freem(lc->lro_mbuf_data[x].mb);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
lc->lro_mbuf_count = 0;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* free allocated memory, if any */
free(lc->lro_mbuf_data, M_LRO);
lc->lro_mbuf_data = NULL;
}
2008-06-11 22:12:50 +00:00
static uint16_t
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_lro_rx_csum_tcphdr(const struct tcphdr *th)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
const uint16_t *ptr;
uint32_t csum;
uint16_t len;
csum = -th->th_sum; /* exclude checksum field */
len = th->th_off;
ptr = (const uint16_t *)th;
while (len--) {
csum += *ptr;
ptr++;
csum += *ptr;
ptr++;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
return (csum);
}
static uint16_t
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_lro_rx_csum_data(const struct lro_parser *pa, uint16_t tcp_csum)
{
uint32_t c;
uint16_t cs;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
c = tcp_csum;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
switch (pa->data.lro_type) {
#ifdef INET6
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
case LRO_TYPE_IPV6_TCP:
/* Compute full pseudo IPv6 header checksum. */
cs = in6_cksum_pseudo(pa->ip6, ntohs(pa->ip6->ip6_plen), pa->ip6->ip6_nxt, 0);
break;
#endif
#ifdef INET
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
case LRO_TYPE_IPV4_TCP:
/* Compute full pseudo IPv4 header checsum. */
cs = in_addword(ntohs(pa->ip4->ip_len) - sizeof(*pa->ip4), IPPROTO_TCP);
cs = in_pseudo(pa->ip4->ip_src.s_addr, pa->ip4->ip_dst.s_addr, htons(cs));
break;
#endif
default:
cs = 0; /* Keep compiler happy. */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
break;
2008-06-11 22:12:50 +00:00
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Complement checksum. */
cs = ~cs;
c += cs;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Remove TCP header checksum. */
cs = ~tcp_lro_rx_csum_tcphdr(pa->tcp);
c += cs;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Compute checksum remainder. */
while (c > 0xffff)
c = (c >> 16) + (c & 0xffff);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
return (c);
2008-06-11 22:12:50 +00:00
}
static void
tcp_lro_rx_done(struct lro_ctrl *lc)
{
struct lro_entry *le;
while ((le = LIST_FIRST(&lc->lro_active)) != NULL) {
tcp_lro_active_remove(le);
tcp_lro_flush(lc, le);
}
}
void
tcp_lro_flush_inactive(struct lro_ctrl *lc, const struct timeval *timeout)
{
struct lro_entry *le, *le_tmp;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
sbintime_t sbt;
if (LIST_EMPTY(&lc->lro_active))
return;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* get timeout time */
sbt = getsbinuptime() - tvtosbt(*timeout);
LIST_FOREACH_SAFE(le, &lc->lro_active, next, le_tmp) {
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (sbt >= le->alloc_time) {
tcp_lro_active_remove(le);
tcp_lro_flush(lc, le);
}
}
}
#ifdef INET
static int
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_lro_rx_ipv4(struct lro_ctrl *lc, struct mbuf *m, struct ip *ip4)
2008-06-11 22:12:50 +00:00
{
uint16_t csum;
/* Legacy IP has a header checksum that needs to be correct. */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
if (__predict_false((m->m_pkthdr.csum_flags & CSUM_IP_VALID) == 0)) {
lc->lro_bad_csum++;
return (TCP_LRO_CANNOT);
}
} else {
csum = in_cksum_hdr(ip4);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (__predict_false(csum != 0)) {
lc->lro_bad_csum++;
return (TCP_LRO_CANNOT);
}
}
return (0);
}
#endif
#ifdef TCPHPTS
static void
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_lro_log(struct tcpcb *tp, const struct lro_ctrl *lc,
const struct lro_entry *le, const struct mbuf *m,
int frm, int32_t tcp_data_len, uint32_t th_seq,
uint32_t th_ack, uint16_t th_win)
{
if (tp->t_logstate != TCP_LOG_STATE_OFF) {
union tcp_log_stackspecific log;
struct timeval tv;
uint32_t cts;
cts = tcp_get_usecs(&tv);
memset(&log, 0, sizeof(union tcp_log_stackspecific));
log.u_bbr.flex8 = frm;
log.u_bbr.flex1 = tcp_data_len;
if (m)
log.u_bbr.flex2 = m->m_pkthdr.len;
else
log.u_bbr.flex2 = 0;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
log.u_bbr.flex3 = le->m_head->m_pkthdr.lro_nsegs;
log.u_bbr.flex4 = le->m_head->m_pkthdr.lro_tcp_d_len;
if (le->m_head) {
log.u_bbr.flex5 = le->m_head->m_pkthdr.len;
log.u_bbr.delRate = le->m_head->m_flags;
log.u_bbr.rttProp = le->m_head->m_pkthdr.rcv_tstmp;
}
log.u_bbr.inflight = th_seq;
log.u_bbr.timeStamp = cts;
log.u_bbr.epoch = le->next_seq;
log.u_bbr.delivered = th_ack;
log.u_bbr.lt_epoch = le->ack_seq;
log.u_bbr.pacing_gain = th_win;
log.u_bbr.cwnd_gain = le->window;
log.u_bbr.cur_del_rate = (uintptr_t)m;
log.u_bbr.bw_inuse = (uintptr_t)le->m_head;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
log.u_bbr.flex6 = sbttous(lc->lro_last_queue_time);
log.u_bbr.flex7 = le->compressed;
log.u_bbr.pacing_gain = le->uncompressed;
if (in_epoch(net_epoch_preempt))
log.u_bbr.inhpts = 1;
else
log.u_bbr.inhpts = 0;
TCP_LOG_EVENTP(tp, NULL,
&tp->t_inpcb->inp_socket->so_rcv,
&tp->t_inpcb->inp_socket->so_snd,
TCP_LOG_LRO, 0,
0, &log, false, &tv);
}
}
#endif
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
static inline void
tcp_lro_assign_and_checksum_16(uint16_t *ptr, uint16_t value, uint16_t *psum)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
uint32_t csum;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
csum = 0xffff - *ptr + value;
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
*ptr = value;
*psum = csum;
}
static uint16_t
tcp_lro_update_checksum(const struct lro_parser *pa, const struct lro_entry *le,
uint16_t payload_len, uint16_t delta_sum)
{
uint32_t csum;
uint16_t tlen;
uint16_t temp[5] = {};
switch (pa->data.lro_type) {
case LRO_TYPE_IPV4_TCP:
/* Compute new IPv4 length. */
tlen = (pa->ip4->ip_hl << 2) + (pa->tcp->th_off << 2) + payload_len;
tcp_lro_assign_and_checksum_16(&pa->ip4->ip_len, htons(tlen), &temp[0]);
/* Subtract delta from current IPv4 checksum. */
csum = pa->ip4->ip_sum + 0xffff - temp[0];
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
tcp_lro_assign_and_checksum_16(&pa->ip4->ip_sum, csum, &temp[1]);
goto update_tcp_header;
case LRO_TYPE_IPV6_TCP:
/* Compute new IPv6 length. */
tlen = (pa->tcp->th_off << 2) + payload_len;
tcp_lro_assign_and_checksum_16(&pa->ip6->ip6_plen, htons(tlen), &temp[0]);
goto update_tcp_header;
case LRO_TYPE_IPV4_UDP:
/* Compute new IPv4 length. */
tlen = (pa->ip4->ip_hl << 2) + sizeof(*pa->udp) + payload_len;
tcp_lro_assign_and_checksum_16(&pa->ip4->ip_len, htons(tlen), &temp[0]);
/* Subtract delta from current IPv4 checksum. */
csum = pa->ip4->ip_sum + 0xffff - temp[0];
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
tcp_lro_assign_and_checksum_16(&pa->ip4->ip_sum, csum, &temp[1]);
goto update_udp_header;
case LRO_TYPE_IPV6_UDP:
/* Compute new IPv6 length. */
tlen = sizeof(*pa->udp) + payload_len;
tcp_lro_assign_and_checksum_16(&pa->ip6->ip6_plen, htons(tlen), &temp[0]);
goto update_udp_header;
default:
return (0);
}
update_tcp_header:
/* Compute current TCP header checksum. */
temp[2] = tcp_lro_rx_csum_tcphdr(pa->tcp);
/* Incorporate the latest ACK into the TCP header. */
pa->tcp->th_ack = le->ack_seq;
pa->tcp->th_win = le->window;
/* Incorporate latest timestamp into the TCP header. */
if (le->timestamp != 0) {
uint32_t *ts_ptr;
ts_ptr = (uint32_t *)(pa->tcp + 1);
ts_ptr[1] = htonl(le->tsval);
ts_ptr[2] = le->tsecr;
}
/* Compute new TCP header checksum. */
temp[3] = tcp_lro_rx_csum_tcphdr(pa->tcp);
/* Compute new TCP checksum. */
csum = pa->tcp->th_sum + 0xffff - delta_sum +
0xffff - temp[0] + 0xffff - temp[3] + temp[2];
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
/* Assign new TCP checksum. */
tcp_lro_assign_and_checksum_16(&pa->tcp->th_sum, csum, &temp[4]);
/* Compute all modififications affecting next checksum. */
csum = temp[0] + temp[1] + 0xffff - temp[2] +
temp[3] + temp[4] + delta_sum;
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
/* Return delta checksum to next stage, if any. */
return (csum);
update_udp_header:
tlen = sizeof(*pa->udp) + payload_len;
/* Assign new UDP length and compute checksum delta. */
tcp_lro_assign_and_checksum_16(&pa->udp->uh_ulen, htons(tlen), &temp[2]);
/* Check if there is a UDP checksum. */
if (__predict_false(pa->udp->uh_sum != 0)) {
/* Compute new UDP checksum. */
csum = pa->udp->uh_sum + 0xffff - delta_sum +
0xffff - temp[0] + 0xffff - temp[2];
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
/* Assign new UDP checksum. */
tcp_lro_assign_and_checksum_16(&pa->udp->uh_sum, csum, &temp[3]);
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Compute all modififications affecting next checksum. */
csum = temp[0] + temp[1] + temp[2] + temp[3] + delta_sum;
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
/* Return delta checksum to next stage, if any. */
return (csum);
}
static void
tcp_flush_out_entry(struct lro_ctrl *lc, struct lro_entry *le)
{
/* Check if we need to recompute any checksums. */
if (le->m_head->m_pkthdr.lro_nsegs > 1) {
uint16_t csum;
switch (le->inner.data.lro_type) {
case LRO_TYPE_IPV4_TCP:
csum = tcp_lro_update_checksum(&le->inner, le,
le->m_head->m_pkthdr.lro_tcp_d_len,
le->m_head->m_pkthdr.lro_tcp_d_csum);
csum = tcp_lro_update_checksum(&le->outer, NULL,
le->m_head->m_pkthdr.lro_tcp_d_len +
le->inner.total_hdr_len, csum);
le->m_head->m_pkthdr.csum_flags = CSUM_DATA_VALID |
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
CSUM_PSEUDO_HDR | CSUM_IP_CHECKED | CSUM_IP_VALID;
le->m_head->m_pkthdr.csum_data = 0xffff;
break;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
case LRO_TYPE_IPV6_TCP:
csum = tcp_lro_update_checksum(&le->inner, le,
le->m_head->m_pkthdr.lro_tcp_d_len,
le->m_head->m_pkthdr.lro_tcp_d_csum);
csum = tcp_lro_update_checksum(&le->outer, NULL,
le->m_head->m_pkthdr.lro_tcp_d_len +
le->inner.total_hdr_len, csum);
le->m_head->m_pkthdr.csum_flags = CSUM_DATA_VALID |
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
CSUM_PSEUDO_HDR;
le->m_head->m_pkthdr.csum_data = 0xffff;
break;
case LRO_TYPE_NONE:
switch (le->outer.data.lro_type) {
case LRO_TYPE_IPV4_TCP:
csum = tcp_lro_update_checksum(&le->outer, le,
le->m_head->m_pkthdr.lro_tcp_d_len,
le->m_head->m_pkthdr.lro_tcp_d_csum);
le->m_head->m_pkthdr.csum_flags = CSUM_DATA_VALID |
CSUM_PSEUDO_HDR | CSUM_IP_CHECKED | CSUM_IP_VALID;
le->m_head->m_pkthdr.csum_data = 0xffff;
break;
case LRO_TYPE_IPV6_TCP:
csum = tcp_lro_update_checksum(&le->outer, le,
le->m_head->m_pkthdr.lro_tcp_d_len,
le->m_head->m_pkthdr.lro_tcp_d_csum);
le->m_head->m_pkthdr.csum_flags = CSUM_DATA_VALID |
CSUM_PSEUDO_HDR;
le->m_head->m_pkthdr.csum_data = 0xffff;
break;
default:
break;
}
break;
default:
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
break;
}
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/*
* Break any chain, this is not set to NULL on the singleton
* case m_nextpkt points to m_head. Other case set them
* m_nextpkt to NULL in push_and_replace.
*/
le->m_head->m_nextpkt = NULL;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
lc->lro_queued += le->m_head->m_pkthdr.lro_nsegs;
(*lc->ifp->if_input)(lc->ifp, le->m_head);
}
static void
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_set_entry_to_mbuf(struct lro_ctrl *lc, struct lro_entry *le,
struct mbuf *m, struct tcphdr *th)
{
uint32_t *ts_ptr;
uint16_t tcp_data_len;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
uint16_t tcp_opt_len;
ts_ptr = (uint32_t *)(th + 1);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_opt_len = (th->th_off << 2);
tcp_opt_len -= sizeof(*th);
/* Check if there is a timestamp option. */
if (tcp_opt_len == 0 ||
__predict_false(tcp_opt_len != TCPOLEN_TSTAMP_APPA ||
*ts_ptr != TCP_LRO_TS_OPTION)) {
/* We failed to find the timestamp option. */
le->timestamp = 0;
} else {
le->timestamp = 1;
le->tsval = ntohl(*(ts_ptr + 1));
le->tsecr = *(ts_ptr + 2);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
}
tcp_data_len = m->m_pkthdr.lro_tcp_d_len;
/* Pull out TCP sequence numbers and window size. */
le->next_seq = ntohl(th->th_seq) + tcp_data_len;
le->ack_seq = th->th_ack;
le->window = th->th_win;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Setup new data pointers. */
le->m_head = m;
le->m_tail = m_last(m);
}
static void
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_push_and_replace(struct lro_ctrl *lc, struct lro_entry *le, struct mbuf *m)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
struct lro_parser *pa;
/*
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
* Push up the stack of the current entry
* and replace it with "m".
*/
struct mbuf *msave;
/* Grab off the next and save it */
msave = le->m_head->m_nextpkt;
le->m_head->m_nextpkt = NULL;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Now push out the old entry */
tcp_flush_out_entry(lc, le);
/* Re-parse new header, should not fail. */
pa = tcp_lro_parser(m, &le->outer, &le->inner, false);
KASSERT(pa != NULL,
("tcp_push_and_replace: LRO parser failed on m=%p\n", m));
/*
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
* Now to replace the data properly in the entry
* we have to reset the TCP header and
* other fields.
*/
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_set_entry_to_mbuf(lc, le, m, pa->tcp);
/* Restore the next list */
m->m_nextpkt = msave;
}
static void
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_lro_mbuf_append_pkthdr(struct mbuf *m, const struct mbuf *p)
{
uint32_t csum;
if (m->m_pkthdr.lro_nsegs == 1) {
/* Compute relative checksum. */
csum = p->m_pkthdr.lro_tcp_d_csum;
} else {
/* Merge TCP data checksums. */
csum = (uint32_t)m->m_pkthdr.lro_tcp_d_csum +
(uint32_t)p->m_pkthdr.lro_tcp_d_csum;
while (csum > 0xffff)
csum = (csum >> 16) + (csum & 0xffff);
}
/* Update various counters. */
m->m_pkthdr.len += p->m_pkthdr.lro_tcp_d_len;
m->m_pkthdr.lro_tcp_d_csum = csum;
m->m_pkthdr.lro_tcp_d_len += p->m_pkthdr.lro_tcp_d_len;
m->m_pkthdr.lro_nsegs += p->m_pkthdr.lro_nsegs;
}
static void
tcp_lro_condense(struct lro_ctrl *lc, struct lro_entry *le)
{
/*
* Walk through the mbuf chain we
* have on tap and compress/condense
* as required.
*/
uint32_t *ts_ptr;
struct mbuf *m;
struct tcphdr *th;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
uint32_t tcp_data_len_total;
uint32_t tcp_data_seg_total;
uint16_t tcp_data_len;
uint16_t tcp_opt_len;
/*
* First we must check the lead (m_head)
* we must make sure that it is *not*
* something that should be sent up
* right away (sack etc).
*/
again:
m = le->m_head->m_nextpkt;
if (m == NULL) {
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Just one left. */
return;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
th = tcp_lro_get_th(m);
tcp_opt_len = (th->th_off << 2);
tcp_opt_len -= sizeof(*th);
ts_ptr = (uint32_t *)(th + 1);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (tcp_opt_len != 0 && __predict_false(tcp_opt_len != TCPOLEN_TSTAMP_APPA ||
*ts_ptr != TCP_LRO_TS_OPTION)) {
/*
* Its not the timestamp. We can't
* use this guy as the head.
*/
le->m_head->m_nextpkt = m->m_nextpkt;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_push_and_replace(lc, le, m);
goto again;
}
if ((th->th_flags & ~(TH_ACK | TH_PUSH)) != 0) {
/*
* Make sure that previously seen segements/ACKs are delivered
* before this segment, e.g. FIN.
*/
le->m_head->m_nextpkt = m->m_nextpkt;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_push_and_replace(lc, le, m);
goto again;
}
while((m = le->m_head->m_nextpkt) != NULL) {
/*
* condense m into le, first
* pull m out of the list.
*/
le->m_head->m_nextpkt = m->m_nextpkt;
m->m_nextpkt = NULL;
/* Setup my data */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_data_len = m->m_pkthdr.lro_tcp_d_len;
th = tcp_lro_get_th(m);
ts_ptr = (uint32_t *)(th + 1);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_opt_len = (th->th_off << 2);
tcp_opt_len -= sizeof(*th);
tcp_data_len_total = le->m_head->m_pkthdr.lro_tcp_d_len + tcp_data_len;
tcp_data_seg_total = le->m_head->m_pkthdr.lro_nsegs + m->m_pkthdr.lro_nsegs;
if (tcp_data_seg_total >= lc->lro_ackcnt_lim ||
tcp_data_len_total >= lc->lro_length_lim) {
/* Flush now if appending will result in overflow. */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_push_and_replace(lc, le, m);
goto again;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (tcp_opt_len != 0 &&
__predict_false(tcp_opt_len != TCPOLEN_TSTAMP_APPA ||
*ts_ptr != TCP_LRO_TS_OPTION)) {
/*
* Maybe a sack in the new one? We need to
* start all over after flushing the
* current le. We will go up to the beginning
* and flush it (calling the replace again possibly
* or just returning).
*/
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_push_and_replace(lc, le, m);
goto again;
}
if ((th->th_flags & ~(TH_ACK | TH_PUSH)) != 0) {
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_push_and_replace(lc, le, m);
goto again;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (tcp_opt_len != 0) {
uint32_t tsval = ntohl(*(ts_ptr + 1));
/* Make sure timestamp values are increasing. */
if (TSTMP_GT(le->tsval, tsval)) {
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_push_and_replace(lc, le, m);
goto again;
}
le->tsval = tsval;
le->tsecr = *(ts_ptr + 2);
}
/* Try to append the new segment. */
if (__predict_false(ntohl(th->th_seq) != le->next_seq ||
(tcp_data_len == 0 &&
le->ack_seq == th->th_ack &&
le->window == th->th_win))) {
/* Out of order packet or duplicate ACK. */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_push_and_replace(lc, le, m);
goto again;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (tcp_data_len != 0 ||
SEQ_GT(ntohl(th->th_ack), ntohl(le->ack_seq))) {
le->next_seq += tcp_data_len;
le->ack_seq = th->th_ack;
le->window = th->th_win;
} else if (th->th_ack == le->ack_seq) {
le->window = WIN_MAX(le->window, th->th_win);
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (tcp_data_len == 0) {
m_freem(m);
continue;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Merge TCP data checksum and length to head mbuf. */
tcp_lro_mbuf_append_pkthdr(le->m_head, m);
/*
* Adjust the mbuf so that m_data points to the first byte of
* the ULP payload. Adjust the mbuf to avoid complications and
* append new segment to existing mbuf chain.
*/
m_adj(m, m->m_pkthdr.len - tcp_data_len);
m_demote_pkthdr(m);
le->m_tail->m_next = m;
le->m_tail = m_last(m);
}
}
#ifdef TCPHPTS
static void
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_queue_pkts(struct inpcb *inp, struct tcpcb *tp, struct lro_entry *le)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
INP_WLOCK_ASSERT(inp);
if (tp->t_in_pkt == NULL) {
/* Nothing yet there */
tp->t_in_pkt = le->m_head;
tp->t_tail_pkt = le->m_last_mbuf;
} else {
/* Already some there */
tp->t_tail_pkt->m_nextpkt = le->m_head;
tp->t_tail_pkt = le->m_last_mbuf;
}
le->m_head = NULL;
le->m_last_mbuf = NULL;
}
static struct mbuf *
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_lro_get_last_if_ackcmp(struct lro_ctrl *lc, struct lro_entry *le,
struct inpcb *inp, int32_t *new_m)
{
struct tcpcb *tp;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
struct mbuf *m;
tp = intotcpcb(inp);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (__predict_false(tp == NULL))
return (NULL);
/* Look at the last mbuf if any in queue */
m = tp->t_tail_pkt;
if (m != NULL && (m->m_flags & M_ACKCMP) != 0) {
if (M_TRAILINGSPACE(m) >= sizeof(struct tcp_ackent)) {
tcp_lro_log(tp, lc, le, NULL, 23, 0, 0, 0, 0);
*new_m = 0;
counter_u64_add(tcp_extra_mbuf, 1);
return (m);
} else {
/* Mark we ran out of space */
inp->inp_flags2 |= INP_MBUF_L_ACKS;
}
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Decide mbuf size. */
if (inp->inp_flags2 & INP_MBUF_L_ACKS)
m = m_getcl(M_NOWAIT, MT_DATA, M_ACKCMP | M_PKTHDR);
else
m = m_gethdr(M_NOWAIT, MT_DATA);
if (__predict_false(m == NULL)) {
counter_u64_add(tcp_would_have_but, 1);
return (NULL);
}
counter_u64_add(tcp_comp_total, 1);
m->m_flags |= M_ACKCMP;
*new_m = 1;
return (m);
}
static struct inpcb *
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_lro_lookup(struct ifnet *ifp, struct lro_parser *pa)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
struct inpcb *inp;
NET_EPOCH_ASSERT();
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
switch (pa->data.lro_type) {
#ifdef INET6
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
case LRO_TYPE_IPV6_TCP:
inp = in6_pcblookup(&V_tcbinfo,
&pa->data.s_addr.v6,
pa->data.s_port,
&pa->data.d_addr.v6,
pa->data.d_port,
INPLOOKUP_WLOCKPCB,
ifp);
break;
#endif
#ifdef INET
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
case LRO_TYPE_IPV4_TCP:
inp = in_pcblookup(&V_tcbinfo,
pa->data.s_addr.v4,
pa->data.s_port,
pa->data.d_addr.v4,
pa->data.d_port,
INPLOOKUP_WLOCKPCB,
ifp);
break;
#endif
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
default:
inp = NULL;
break;
}
return (inp);
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
static inline bool
tcp_lro_ack_valid(struct mbuf *m, struct tcphdr *th, uint32_t **ppts, bool *other_opts)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/*
* This function returns two bits of valuable information.
* a) Is what is present capable of being ack-compressed,
* we can ack-compress if there is no options or just
* a timestamp option, and of course the th_flags must
* be correct as well.
* b) Our other options present such as SACK. This is
* used to determine if we want to wakeup or not.
*/
bool ret = true;
switch (th->th_off << 2) {
case (sizeof(*th) + TCPOLEN_TSTAMP_APPA):
*ppts = (uint32_t *)(th + 1);
/* Check if we have only one timestamp option. */
if (**ppts == TCP_LRO_TS_OPTION)
*other_opts = false;
else {
*other_opts = true;
ret = false;
}
break;
case (sizeof(*th)):
/* No options. */
*ppts = NULL;
*other_opts = false;
break;
default:
*ppts = NULL;
*other_opts = true;
ret = false;
break;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* For ACKCMP we only accept ACK, PUSH, ECE and CWR. */
if ((th->th_flags & ~(TH_ACK | TH_PUSH | TH_ECE | TH_CWR)) != 0)
ret = false;
/* If it has data on it we cannot compress it */
if (m->m_pkthdr.lro_tcp_d_len)
ret = false;
/* ACK flag must be set. */
if (!(th->th_flags & TH_ACK))
ret = false;
return (ret);
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
static int
tcp_lro_flush_tcphpts(struct lro_ctrl *lc, struct lro_entry *le)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
struct inpcb *inp;
struct tcpcb *tp;
struct mbuf **pp, *cmp, *mv_to;
bool bpf_req, should_wake;
/* Check if packet doesn't belongs to our network interface. */
if ((tcplro_stacks_wanting_mbufq == 0) ||
(le->outer.data.vlan_id != 0) ||
(le->inner.data.lro_type != LRO_TYPE_NONE))
return (TCP_LRO_CANNOT);
#ifdef INET6
/*
* Be proactive about unspecified IPv6 address in source. As
* we use all-zero to indicate unbounded/unconnected pcb,
* unspecified IPv6 address can be used to confuse us.
*
* Note that packets with unspecified IPv6 destination is
* already dropped in ip6_input.
*/
if (__predict_false(le->outer.data.lro_type == LRO_TYPE_IPV6_TCP &&
IN6_IS_ADDR_UNSPECIFIED(&le->outer.data.s_addr.v6)))
return (TCP_LRO_CANNOT);
if (__predict_false(le->inner.data.lro_type == LRO_TYPE_IPV6_TCP &&
IN6_IS_ADDR_UNSPECIFIED(&le->inner.data.s_addr.v6)))
return (TCP_LRO_CANNOT);
#endif
/* Lookup inp, if any. */
inp = tcp_lro_lookup(lc->ifp,
(le->inner.data.lro_type == LRO_TYPE_NONE) ? &le->outer : &le->inner);
if (inp == NULL)
return (TCP_LRO_CANNOT);
counter_u64_add(tcp_inp_lro_locks_taken, 1);
/* Get TCP control structure. */
tp = intotcpcb(inp);
/* Check if the inp is dead, Jim. */
if (tp == NULL ||
(inp->inp_flags & (INP_DROPPED | INP_TIMEWAIT)) ||
(inp->inp_flags2 & INP_FREED)) {
INP_WUNLOCK(inp);
return (TCP_LRO_CANNOT);
}
/* Check if the transport doesn't support the needed optimizations. */
if ((inp->inp_flags2 & (INP_SUPPORTS_MBUFQ | INP_MBUF_ACKCMP)) == 0) {
INP_WUNLOCK(inp);
return (TCP_LRO_CANNOT);
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (inp->inp_flags2 & INP_MBUF_QUEUE_READY)
should_wake = false;
else
should_wake = true;
/* Check if packets should be tapped to BPF. */
bpf_req = bpf_peers_present(lc->ifp->if_bpf);
/* Strip and compress all the incoming packets. */
cmp = NULL;
for (pp = &le->m_head; *pp != NULL; ) {
mv_to = NULL;
if (do_bpf_strip_and_compress(inp, lc, le, pp,
&cmp, &mv_to, &should_wake, bpf_req ) == false) {
/* Advance to next mbuf. */
pp = &(*pp)->m_nextpkt;
} else if (mv_to != NULL) {
/* We are asked to move pp up */
pp = &mv_to->m_nextpkt;
}
}
/* Update "m_last_mbuf", if any. */
if (pp == &le->m_head)
le->m_last_mbuf = *pp;
else
le->m_last_mbuf = __containerof(pp, struct mbuf, m_nextpkt);
/* Check if any data mbufs left. */
if (le->m_head != NULL) {
counter_u64_add(tcp_inp_lro_direct_queue, 1);
tcp_lro_log(tp, lc, le, NULL, 22, 1,
inp->inp_flags2, inp->inp_in_input, 1);
tcp_queue_pkts(inp, tp, le);
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (should_wake) {
/* Wakeup */
counter_u64_add(tcp_inp_lro_wokeup_queue, 1);
if ((*tp->t_fb->tfb_do_queued_segments)(inp->inp_socket, tp, 0))
inp = NULL;
}
if (inp != NULL)
INP_WUNLOCK(inp);
return (0); /* Success. */
}
#endif
void
tcp_lro_flush(struct lro_ctrl *lc, struct lro_entry *le)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Only optimise if there are multiple packets waiting. */
#ifdef TCPHPTS
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
int error;
CURVNET_SET(lc->ifp->if_vnet);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
error = tcp_lro_flush_tcphpts(lc, le);
CURVNET_RESTORE();
if (error != 0) {
#endif
tcp_lro_condense(lc, le);
tcp_flush_out_entry(lc, le);
#ifdef TCPHPTS
}
#endif
lc->lro_flushed++;
bzero(le, sizeof(*le));
LIST_INSERT_HEAD(&lc->lro_free, le, next);
2008-06-11 22:12:50 +00:00
}
#ifdef HAVE_INLINE_FLSLL
#define tcp_lro_msb_64(x) (1ULL << (flsll(x) - 1))
#else
static inline uint64_t
tcp_lro_msb_64(uint64_t x)
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
{
x |= (x >> 1);
x |= (x >> 2);
x |= (x >> 4);
x |= (x >> 8);
x |= (x >> 16);
x |= (x >> 32);
return (x & ~(x >> 1));
}
#endif
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/*
* The tcp_lro_sort() routine is comparable to qsort(), except it has
* a worst case complexity limit of O(MIN(N,64)*N), where N is the
* number of elements to sort and 64 is the number of sequence bits
* available. The algorithm is bit-slicing the 64-bit sequence number,
* sorting one bit at a time from the most significant bit until the
* least significant one, skipping the constant bits. This is
* typically called a radix sort.
*/
static void
tcp_lro_sort(struct lro_mbuf_sort *parray, uint32_t size)
{
struct lro_mbuf_sort temp;
uint64_t ones;
uint64_t zeros;
uint32_t x;
uint32_t y;
repeat:
/* for small arrays insertion sort is faster */
if (size <= 12) {
for (x = 1; x < size; x++) {
temp = parray[x];
for (y = x; y > 0 && temp.seq < parray[y - 1].seq; y--)
parray[y] = parray[y - 1];
parray[y] = temp;
}
return;
}
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* compute sequence bits which are constant */
ones = 0;
zeros = 0;
for (x = 0; x != size; x++) {
ones |= parray[x].seq;
zeros |= ~parray[x].seq;
}
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* compute bits which are not constant into "ones" */
ones &= zeros;
if (ones == 0)
return;
/* pick the most significant bit which is not constant */
ones = tcp_lro_msb_64(ones);
/*
* Move entries having cleared sequence bits to the beginning
* of the array:
*/
for (x = y = 0; y != size; y++) {
/* skip set bits */
if (parray[y].seq & ones)
continue;
/* swap entries */
temp = parray[x];
parray[x] = parray[y];
parray[y] = temp;
x++;
}
KASSERT(x != 0 && x != size, ("Memory is corrupted\n"));
/* sort zeros */
tcp_lro_sort(parray, x);
/* sort ones */
parray += x;
size -= x;
goto repeat;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
}
void
tcp_lro_flush_all(struct lro_ctrl *lc)
{
uint64_t seq;
uint64_t nseq;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
unsigned x;
/* check if no mbufs to flush */
if (lc->lro_mbuf_count == 0)
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
goto done;
CURVNET_SET(lc->ifp->if_vnet);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* get current time */
lc->lro_last_queue_time = getsbinuptime();
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* sort all mbufs according to stream */
tcp_lro_sort(lc->lro_mbuf_data, lc->lro_mbuf_count);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* input data into LRO engine, stream by stream */
seq = 0;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
for (x = 0; x != lc->lro_mbuf_count; x++) {
struct mbuf *mb;
/* get mbuf */
mb = lc->lro_mbuf_data[x].mb;
/* get sequence number, masking away the packet index */
nseq = lc->lro_mbuf_data[x].seq & (-1ULL << 24);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* check for new stream */
if (seq != nseq) {
seq = nseq;
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* flush active streams */
tcp_lro_rx_done(lc);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
}
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* add packet to LRO engine */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (tcp_lro_rx_common(lc, mb, 0, false) != 0) {
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* input packet to network layer */
(*lc->ifp->if_input)(lc->ifp, mb);
lc->lro_queued++;
lc->lro_flushed++;
}
}
CURVNET_RESTORE();
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
done:
/* flush active streams */
tcp_lro_rx_done(lc);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
lc->lro_mbuf_count = 0;
}
#ifdef TCPHPTS
static void
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
build_ack_entry(struct tcp_ackent *ae, struct tcphdr *th, struct mbuf *m,
uint32_t *ts_ptr, uint16_t iptos)
{
/*
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
* Given a TCP ACK, summarize it down into the small TCP ACK
* entry.
*/
ae->timestamp = m->m_pkthdr.rcv_tstmp;
if (m->m_flags & M_TSTMP_LRO)
ae->flags = TSTMP_LRO;
else if (m->m_flags & M_TSTMP)
ae->flags = TSTMP_HDWR;
ae->seq = ntohl(th->th_seq);
ae->ack = ntohl(th->th_ack);
ae->flags |= th->th_flags;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (ts_ptr != NULL) {
ae->ts_value = ntohl(ts_ptr[1]);
ae->ts_echo = ntohl(ts_ptr[2]);
ae->flags |= HAS_TSTMP;
}
ae->win = ntohs(th->th_win);
ae->codepoint = iptos;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/*
* Do BPF tap for either ACK_CMP packets or MBUF QUEUE type packets
* and strip all, but the IPv4/IPv6 header.
*/
static bool
do_bpf_strip_and_compress(struct inpcb *inp, struct lro_ctrl *lc,
struct lro_entry *le, struct mbuf **pp, struct mbuf **cmp, struct mbuf **mv_to,
bool *should_wake, bool bpf_req)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
union {
void *ptr;
struct ip *ip4;
struct ip6_hdr *ip6;
} l3;
struct mbuf *m;
struct mbuf *nm;
struct tcphdr *th;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
struct tcp_ackent *ack_ent;
uint32_t *ts_ptr;
int32_t n_mbuf;
bool other_opts, can_compress;
uint16_t lro_type;
uint16_t iptos;
int tcp_hdr_offset;
int idx;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Get current mbuf. */
m = *pp;
/* Let the BPF see the packet */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
if (__predict_false(bpf_req))
ETHER_BPF_MTAP(lc->ifp, m);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_hdr_offset = m->m_pkthdr.lro_tcp_h_off;
lro_type = le->inner.data.lro_type;
switch (lro_type) {
case LRO_TYPE_NONE:
lro_type = le->outer.data.lro_type;
switch (lro_type) {
case LRO_TYPE_IPV4_TCP:
tcp_hdr_offset -= sizeof(*le->outer.ip4);
m->m_pkthdr.lro_etype = ETHERTYPE_IP;
break;
case LRO_TYPE_IPV6_TCP:
tcp_hdr_offset -= sizeof(*le->outer.ip6);
m->m_pkthdr.lro_etype = ETHERTYPE_IPV6;
break;
default:
goto compressed;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
break;
case LRO_TYPE_IPV4_TCP:
tcp_hdr_offset -= sizeof(*le->outer.ip4);
m->m_pkthdr.lro_etype = ETHERTYPE_IP;
break;
case LRO_TYPE_IPV6_TCP:
tcp_hdr_offset -= sizeof(*le->outer.ip6);
m->m_pkthdr.lro_etype = ETHERTYPE_IPV6;
break;
default:
goto compressed;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
MPASS(tcp_hdr_offset >= 0);
m_adj(m, tcp_hdr_offset);
m->m_flags |= M_LRO_EHDRSTRP;
m->m_flags &= ~M_ACKCMP;
m->m_pkthdr.lro_tcp_h_off -= tcp_hdr_offset;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
th = tcp_lro_get_th(m);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
th->th_sum = 0; /* TCP checksum is valid. */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Check if ACK can be compressed */
can_compress = tcp_lro_ack_valid(m, th, &ts_ptr, &other_opts);
/* Now lets look at the should wake states */
if ((other_opts == true) &&
((inp->inp_flags2 & INP_DONT_SACK_QUEUE) == 0)) {
/*
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
* If there are other options (SACK?) and the
* tcp endpoint has not expressly told us it does
* not care about SACKS, then we should wake up.
*/
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
*should_wake = true;
}
/* Is the ack compressable? */
if (can_compress == false)
goto done;
/* Does the TCP endpoint support ACK compression? */
if ((inp->inp_flags2 & INP_MBUF_ACKCMP) == 0)
goto done;
/* Lets get the TOS/traffic class field */
l3.ptr = mtod(m, void *);
switch (lro_type) {
case LRO_TYPE_IPV4_TCP:
iptos = l3.ip4->ip_tos;
break;
case LRO_TYPE_IPV6_TCP:
iptos = IPV6_TRAFFIC_CLASS(l3.ip6);
break;
default:
iptos = 0; /* Keep compiler happy. */
break;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Now lets get space if we don't have some already */
if (*cmp == NULL) {
new_one:
nm = tcp_lro_get_last_if_ackcmp(lc, le, inp, &n_mbuf);
if (__predict_false(nm == NULL))
goto done;
*cmp = nm;
if (n_mbuf) {
/*
* Link in the new cmp ack to our in-order place,
* first set our cmp ack's next to where we are.
*/
nm->m_nextpkt = m;
(*pp) = nm;
/*
* Set it up so mv_to is advanced to our
* compressed ack. This way the caller can
* advance pp to the right place.
*/
*mv_to = nm;
/*
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
* Advance it here locally as well.
*/
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
pp = &nm->m_nextpkt;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
} else {
/* We have one already we are working on */
nm = *cmp;
if (M_TRAILINGSPACE(nm) < sizeof(struct tcp_ackent)) {
/* We ran out of space */
inp->inp_flags2 |= INP_MBUF_L_ACKS;
goto new_one;
}
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
MPASS(M_TRAILINGSPACE(nm) >= sizeof(struct tcp_ackent));
counter_u64_add(tcp_inp_lro_compressed, 1);
le->compressed++;
/* We can add in to the one on the tail */
ack_ent = mtod(nm, struct tcp_ackent *);
idx = (nm->m_len / sizeof(struct tcp_ackent));
build_ack_entry(&ack_ent[idx], th, m, ts_ptr, iptos);
/* Bump the size of both pkt-hdr and len */
nm->m_len += sizeof(struct tcp_ackent);
nm->m_pkthdr.len += sizeof(struct tcp_ackent);
compressed:
/* Advance to next mbuf before freeing. */
*pp = m->m_nextpkt;
m->m_nextpkt = NULL;
m_freem(m);
return (true);
done:
counter_u64_add(tcp_uncomp_total, 1);
le->uncompressed++;
return (false);
}
#endif
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
static struct lro_head *
tcp_lro_rx_get_bucket(struct lro_ctrl *lc, struct mbuf *m, struct lro_parser *parser)
{
u_long hash;
if (M_HASHTYPE_ISHASH(m)) {
hash = m->m_pkthdr.flowid;
} else {
for (unsigned i = hash = 0; i != LRO_RAW_ADDRESS_MAX; i++)
hash += parser->data.raw[i];
}
return (&lc->lro_hash[hash % lc->lro_hashsz]);
}
static int
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
tcp_lro_rx_common(struct lro_ctrl *lc, struct mbuf *m, uint32_t csum, bool use_hash)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
struct lro_parser pi; /* inner address data */
struct lro_parser po; /* outer address data */
struct lro_parser *pa; /* current parser for TCP stream */
struct lro_entry *le;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
struct lro_head *bucket;
struct tcphdr *th;
int tcp_data_len;
int tcp_opt_len;
int error;
uint16_t tcp_data_sum;
#ifdef INET
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Quickly decide if packet cannot be LRO'ed */
if (__predict_false(V_ipforwarding != 0))
return (TCP_LRO_CANNOT);
#endif
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
#ifdef INET6
/* Quickly decide if packet cannot be LRO'ed */
if (__predict_false(V_ip6_forwarding != 0))
return (TCP_LRO_CANNOT);
#endif
/* We expect a contiguous header [eh, ip, tcp]. */
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
pa = tcp_lro_parser(m, &po, &pi, true);
if (__predict_false(pa == NULL))
return (TCP_LRO_NOT_SUPPORTED);
/* We don't expect any padding. */
error = tcp_lro_trim_mbuf_chain(m, pa);
if (__predict_false(error != 0))
return (error);
#ifdef INET
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
switch (pa->data.lro_type) {
case LRO_TYPE_IPV4_TCP:
error = tcp_lro_rx_ipv4(lc, m, pa->ip4);
if (__predict_false(error != 0))
return (error);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
break;
default:
break;
}
#endif
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* If no hardware or arrival stamp on the packet add timestamp */
if ((m->m_flags & (M_TSTMP_LRO | M_TSTMP)) == 0) {
m->m_pkthdr.rcv_tstmp = sbttons(lc->lro_last_queue_time);
m->m_flags |= M_TSTMP_LRO;
2008-06-11 22:12:50 +00:00
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Get pointer to TCP header. */
th = pa->tcp;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Don't process SYN packets. */
if (__predict_false(th->th_flags & TH_SYN))
return (TCP_LRO_CANNOT);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Get total TCP header length and compute payload length. */
tcp_opt_len = (th->th_off << 2);
tcp_data_len = m->m_pkthdr.len - ((uint8_t *)th -
(uint8_t *)m->m_data) - tcp_opt_len;
tcp_opt_len -= sizeof(*th);
/* Don't process invalid TCP headers. */
if (__predict_false(tcp_opt_len < 0 || tcp_data_len < 0))
return (TCP_LRO_CANNOT);
/* Compute TCP data only checksum. */
if (tcp_data_len == 0)
tcp_data_sum = 0; /* no data, no checksum */
else if (__predict_false(csum != 0))
tcp_data_sum = tcp_lro_rx_csum_data(pa, ~csum);
else
tcp_data_sum = tcp_lro_rx_csum_data(pa, ~th->th_sum);
/* Save TCP info in mbuf. */
m->m_nextpkt = NULL;
m->m_pkthdr.rcvif = lc->ifp;
m->m_pkthdr.lro_tcp_d_csum = tcp_data_sum;
m->m_pkthdr.lro_tcp_d_len = tcp_data_len;
m->m_pkthdr.lro_tcp_h_off = ((uint8_t *)th - (uint8_t *)m->m_data);
m->m_pkthdr.lro_nsegs = 1;
/* Get hash bucket. */
if (!use_hash) {
bucket = &lc->lro_hash[0];
} else {
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
bucket = tcp_lro_rx_get_bucket(lc, m, pa);
}
/* Try to find a matching previous segment. */
LIST_FOREACH(le, bucket, hash_next) {
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Compare addresses and ports. */
if (lro_address_compare(&po.data, &le->outer.data) == false ||
lro_address_compare(&pi.data, &le->inner.data) == false)
continue;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Check if no data and old ACK. */
if (tcp_data_len == 0 &&
SEQ_LT(ntohl(th->th_ack), ntohl(le->ack_seq))) {
m_freem(m);
return (0);
2008-06-11 22:12:50 +00:00
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Mark "m" in the last spot. */
le->m_last_mbuf->m_nextpkt = m;
/* Now set the tail to "m". */
le->m_last_mbuf = m;
return (0);
2008-06-11 22:12:50 +00:00
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Try to find an empty slot. */
if (LIST_EMPTY(&lc->lro_free))
return (TCP_LRO_NO_ENTRIES);
/* Start a new segment chain. */
le = LIST_FIRST(&lc->lro_free);
LIST_REMOVE(le, next);
tcp_lro_active_insert(lc, bucket, le);
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* Make sure the headers are set. */
le->inner = pi;
le->outer = po;
/* Store time this entry was allocated. */
le->alloc_time = lc->lro_last_queue_time;
tcp_set_entry_to_mbuf(lc, le, m, th);
/* Now set the tail to "m". */
le->m_last_mbuf = m;
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
return (0);
2008-06-11 22:12:50 +00:00
}
int
tcp_lro_rx(struct lro_ctrl *lc, struct mbuf *m, uint32_t csum)
{
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
int error;
/* get current time */
lc->lro_last_queue_time = getsbinuptime();
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
CURVNET_SET(lc->ifp->if_vnet);
error = tcp_lro_rx_common(lc, m, csum, true);
CURVNET_RESTORE();
return (error);
}
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
void
tcp_lro_queue_mbuf(struct lro_ctrl *lc, struct mbuf *mb)
{
/* sanity checks */
if (__predict_false(lc->ifp == NULL || lc->lro_mbuf_data == NULL ||
lc->lro_mbuf_max == 0)) {
/* packet drop */
m_freem(mb);
return;
}
/* check if packet is not LRO capable */
if (__predict_false(mb->m_pkthdr.csum_flags == 0 ||
(lc->ifp->if_capenable & IFCAP_LRO) == 0)) {
/* input packet to network layer */
(*lc->ifp->if_input) (lc->ifp, mb);
return;
}
Add TCP LRO support for VLAN and VxLAN. This change makes the TCP LRO code more generic and flexible with regards to supporting multiple different TCP encapsulation protocols and in general lays the ground for broader TCP LRO support. The main job of the TCP LRO code is to merge TCP packets for the same flow, to reduce the number of calls to upper layers. This reduces CPU and increases performance, due to being able to send larger TSO offloaded data chunks at a time. Basically the TCP LRO makes it possible to avoid per-packet interaction by the host CPU. Because the current TCP LRO code was tightly bound and optimized for TCP/IP over ethernet only, several larger changes were needed. Also a minor bug was fixed in the flushing mechanism for inactive entries, where the expire time, "le->mtime" was not always properly set. To avoid having to re-run time consuming regression tests for every change, it was chosen to squash the following list of changes into a single commit: - Refactor parsing of all address information into the "lro_parser" structure. This easily allows to reuse parsing code for inner headers. - Speedup header data comparison. Don't compare field by field, but instead use an unsigned long array, where the fields get packed. - Refactor the IPv4/TCP/UDP checksum computations, so that they may be computed recursivly, only applying deltas as the result of updating payload data. - Make smaller inline functions doing one operation at a time instead of big functions having repeated code. - Refactor the TCP ACK compression code to only execute once per TCP LRO flush. This gives a minor performance improvement and keeps the code simple. - Use sbintime() for all time-keeping. This change also fixes flushing of inactive entries. - Try to shrink the size of the LRO entry, because it is frequently zeroed. - Removed unused TCP LRO macros. - Cleanup unused TCP LRO statistics counters while at it. - Try to use __predict_true() and predict_false() to optimise CPU branch predictions. Bump the __FreeBSD_version due to changing the "lro_ctrl" structure. Tested by: Netflix Reviewed by: rrs (transport) Differential Revision: https://reviews.freebsd.org/D29564 MFC after: 2 week Sponsored by: Mellanox Technologies // NVIDIA Networking
2021-03-31 10:36:36 +00:00
/* create sequence number */
lc->lro_mbuf_data[lc->lro_mbuf_count].seq =
(((uint64_t)M_HASHTYPE_GET(mb)) << 56) |
(((uint64_t)mb->m_pkthdr.flowid) << 24) |
((uint64_t)lc->lro_mbuf_count);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
/* enter mbuf */
lc->lro_mbuf_data[lc->lro_mbuf_count].mb = mb;
/* flush if array is full */
if (__predict_false(++lc->lro_mbuf_count == lc->lro_mbuf_max))
tcp_lro_flush_all(lc);
Add optimizing LRO wrapper: - Add optimizing LRO wrapper which pre-sorts all incoming packets according to the hash type and flowid. This prevents exhaustion of the LRO entries due to too many connections at the same time. Testing using a larger number of higher bandwidth TCP connections showed that the incoming ACK packet aggregation rate increased from ~1.3:1 to almost 3:1. Another test showed that for a number of TCP connections greater than 16 per hardware receive ring, where 8 TCP connections was the LRO active entry limit, there was a significant improvement in throughput due to being able to fully aggregate more than 8 TCP stream. For very few very high bandwidth TCP streams, the optimizing LRO wrapper will add CPU usage instead of reducing CPU usage. This is expected. Network drivers which want to use the optimizing LRO wrapper needs to call "tcp_lro_queue_mbuf()" instead of "tcp_lro_rx()" and "tcp_lro_flush_all()" instead of "tcp_lro_flush()". Further the LRO control structure must be initialized using "tcp_lro_init_args()" passing a non-zero number into the "lro_mbufs" argument. - Make LRO statistics 64-bit. Previously 32-bit integers were used for statistics which can be prone to wrap-around. Fix this while at it and update all SYSCTL's which expose LRO statistics. - Ensure all data is freed when destroying a LRO control structures, especially leftover LRO entries. - Reduce number of memory allocations needed when setting up a LRO control structure by precomputing the total amount of memory needed. - Add own memory allocation counter for LRO. - Bump the FreeBSD version to force recompilation of all KLDs due to change of the LRO control structure size. Sponsored by: Mellanox Technologies Reviewed by: gallatin, sbruno, rrs, gnn, transport Tested by: Netflix Differential Revision: https://reviews.freebsd.org/D4914
2016-01-19 15:33:28 +00:00
}
/* end */