Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'ipsec-next-2025-11-18' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next
Steffen Klassert says:
====================
pull request (net-next): ipsec-next 2025-11-18
1) Relax a lock contention bottleneck to improve IPsec crypto
offload performance. From Jianbo Liu.
2) Deprecate pfkey, the interface will be removed in 2027.
3) Update xfrm documentation and move it to ipsec maintainance.
From Bagas Sanjaya.
* tag 'ipsec-next-2025-11-18' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next:
MAINTAINERS: Add entry for XFRM documentation
net: Move XFRM documentation into its own subdirectory
Documentation: xfrm_sync: Number the fifth section
Documentation: xfrm_sysctl: Trim trailing colon in section heading
Documentation: xfrm_sync: Trim excess section heading characters
Documentation: xfrm_sync: Properly reindent list text
Documentation: xfrm_device: Separate hardware offload sublists
Documentation: xfrm_device: Use numbered list for offloading steps
Documentation: xfrm_device: Wrap iproute2 snippets in literal code block
pfkey: Deprecate pfkey
xfrm: Skip redundant replay recheck for the hardware offload path
xfrm: Refactor xfrm_input lock to reduce contention with RSS
====================
Link: https://patch.msgid.link/20251118092610.2223552-1-steffen.klassert@secunet.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
+103
-80
+1
-4
Documentation/networking/index.rst
+1
-4
Documentation/networking/index.rst
+13
Documentation/networking/xfrm/index.rst
+13
Documentation/networking/xfrm/index.rst
+12
-8
Documentation/networking/xfrm_device.rst
Documentation/networking/xfrm/xfrm_device.rst
+12
-8
Documentation/networking/xfrm_device.rst
Documentation/networking/xfrm/xfrm_device.rst
···
20
20
Device interface allows NIC drivers to offer to the stack access to the
21
21
hardware offload.
22
22
23
-
Right now, there are two types of hardware offload that kernel supports.
23
+
Right now, there are two types of hardware offload that kernel supports:
24
+
24
25
* IPsec crypto offload:
26
+
25
27
* NIC performs encrypt/decrypt
26
28
* Kernel does everything else
29
+
27
30
* IPsec packet offload:
31
+
28
32
* NIC performs encrypt/decrypt
29
33
* NIC does encapsulation
30
34
* Kernel and NIC have SA and policy in-sync
···
38
34
Userland access to the offload is typically through a system such as
39
35
libreswan or KAME/raccoon, but the iproute2 'ip xfrm' command set can
40
36
be handy when experimenting. An example command might look something
41
-
like this for crypto offload:
37
+
like this for crypto offload::
42
38
43
39
ip x s add proto esp dst 14.0.0.70 src 14.0.0.52 spi 0x07 mode transport \
44
40
reqid 0x07 replay-window 32 \
···
46
42
sel src 14.0.0.52/24 dst 14.0.0.70/24 proto tcp \
47
43
offload dev eth4 dir in
48
44
49
-
and for packet offload
45
+
and for packet offload::
50
46
51
47
ip x s add proto esp dst 14.0.0.70 src 14.0.0.52 spi 0x07 mode transport \
52
48
reqid 0x07 replay-window 32 \
···
157
153
IPsec headers are still in the packet data; they are removed later up
158
154
the stack in xfrm_input().
159
155
160
-
find and hold the SA that was used to the Rx skb::
156
+
1. Find and hold the SA that was used to the Rx skb::
161
157
162
-
get spi, protocol, and destination IP from packet headers
158
+
/* get spi, protocol, and destination IP from packet headers */
163
159
xs = find xs from (spi, protocol, dest_IP)
164
160
xfrm_state_hold(xs);
165
161
166
-
store the state information into the skb::
162
+
2. Store the state information into the skb::
167
163
168
164
sp = secpath_set(skb);
169
165
if (!sp) return;
170
166
sp->xvec[sp->len++] = xs;
171
167
sp->olen++;
172
168
173
-
indicate the success and/or error status of the offload::
169
+
3. Indicate the success and/or error status of the offload::
174
170
175
171
xo = xfrm_offload(skb);
176
172
xo->flags = CRYPTO_DONE;
177
173
xo->status = crypto_status;
178
174
179
-
hand the packet to napi_gro_receive() as usual
175
+
4. Hand the packet to napi_gro_receive() as usual.
180
176
181
177
In ESN mode, xdo_dev_state_advance_esn() is called from
182
178
xfrm_replay_advance_esn() for RX, and xfrm_replay_overflow_offload_esn for TX.
Documentation/networking/xfrm_proc.rst
Documentation/networking/xfrm/xfrm_proc.rst
Documentation/networking/xfrm_proc.rst
Documentation/networking/xfrm/xfrm_proc.rst
+50
-47
Documentation/networking/xfrm_sync.rst
Documentation/networking/xfrm/xfrm_sync.rst
+50
-47
Documentation/networking/xfrm_sync.rst
Documentation/networking/xfrm/xfrm_sync.rst
···
1
1
.. SPDX-License-Identifier: GPL-2.0
2
2
3
-
====
4
-
XFRM
5
-
====
3
+
=========
4
+
XFRM sync
5
+
=========
6
6
7
7
The sync patches work is based on initial patches from
8
8
Krisztian <hidden@balabit.hu> and others and additional patches
···
36
36
- the replay sequence for both inbound and outbound
37
37
38
38
1) Message Structure
39
-
----------------------
39
+
--------------------
40
40
41
41
nlmsghdr:aevent_id:optional-TLVs.
42
42
···
83
83
A program needs to subscribe to multicast group XFRMNLGRP_AEVENTS
84
84
to get notified of these events.
85
85
86
-
2) TLVS reflect the different parameters:
87
-
-----------------------------------------
86
+
2) TLVS reflect the different parameters
87
+
----------------------------------------
88
88
89
89
a) byte value (XFRMA_LTIME_VAL)
90
90
91
-
This TLV carries the running/current counter for byte lifetime since
92
-
last event.
91
+
This TLV carries the running/current counter for byte lifetime since
92
+
last event.
93
93
94
-
b)replay value (XFRMA_REPLAY_VAL)
94
+
b) replay value (XFRMA_REPLAY_VAL)
95
95
96
-
This TLV carries the running/current counter for replay sequence since
97
-
last event.
96
+
This TLV carries the running/current counter for replay sequence since
97
+
last event.
98
98
99
-
c)replay threshold (XFRMA_REPLAY_THRESH)
99
+
c) replay threshold (XFRMA_REPLAY_THRESH)
100
100
101
-
This TLV carries the threshold being used by the kernel to trigger events
102
-
when the replay sequence is exceeded.
101
+
This TLV carries the threshold being used by the kernel to trigger events
102
+
when the replay sequence is exceeded.
103
103
104
104
d) expiry timer (XFRMA_ETIMER_THRESH)
105
105
106
-
This is a timer value in milliseconds which is used as the nagle
107
-
value to rate limit the events.
106
+
This is a timer value in milliseconds which is used as the nagle
107
+
value to rate limit the events.
108
108
109
-
3) Default configurations for the parameters:
110
-
---------------------------------------------
109
+
3) Default configurations for the parameters
110
+
--------------------------------------------
111
111
112
112
By default these events should be turned off unless there is
113
113
at least one listener registered to listen to the multicast
···
121
121
the two sysctls/proc entries are:
122
122
123
123
a) /proc/sys/net/core/sysctl_xfrm_aevent_etime
124
-
used to provide default values for the XFRMA_ETIMER_THRESH in incremental
125
-
units of time of 100ms. The default is 10 (1 second)
124
+
125
+
Used to provide default values for the XFRMA_ETIMER_THRESH in incremental
126
+
units of time of 100ms. The default is 10 (1 second)
126
127
127
128
b) /proc/sys/net/core/sysctl_xfrm_aevent_rseqth
128
-
used to provide default values for XFRMA_REPLAY_THRESH parameter
129
-
in incremental packet count. The default is two packets.
129
+
130
+
Used to provide default values for XFRMA_REPLAY_THRESH parameter
131
+
in incremental packet count. The default is two packets.
130
132
131
133
4) Message types
132
134
----------------
···
136
134
a) XFRM_MSG_GETAE issued by user-->kernel.
137
135
XFRM_MSG_GETAE does not carry any TLVs.
138
136
139
-
The response is a XFRM_MSG_NEWAE which is formatted based on what
140
-
XFRM_MSG_GETAE queried for.
137
+
The response is a XFRM_MSG_NEWAE which is formatted based on what
138
+
XFRM_MSG_GETAE queried for.
141
139
142
-
The response will always have XFRMA_LTIME_VAL and XFRMA_REPLAY_VAL TLVs.
143
-
* if XFRM_AE_RTHR flag is set, then XFRMA_REPLAY_THRESH is also retrieved
144
-
* if XFRM_AE_ETHR flag is set, then XFRMA_ETIMER_THRESH is also retrieved
140
+
The response will always have XFRMA_LTIME_VAL and XFRMA_REPLAY_VAL TLVs.
141
+
142
+
* if XFRM_AE_RTHR flag is set, then XFRMA_REPLAY_THRESH is also retrieved
143
+
* if XFRM_AE_ETHR flag is set, then XFRMA_ETIMER_THRESH is also retrieved
145
144
146
145
b) XFRM_MSG_NEWAE is issued by either user space to configure
147
146
or kernel to announce events or respond to a XFRM_MSG_GETAE.
148
147
149
-
i) user --> kernel to configure a specific SA.
148
+
i) user --> kernel to configure a specific SA.
150
149
151
-
any of the values or threshold parameters can be updated by passing the
152
-
appropriate TLV.
150
+
any of the values or threshold parameters can be updated by passing the
151
+
appropriate TLV.
153
152
154
-
A response is issued back to the sender in user space to indicate success
155
-
or failure.
153
+
A response is issued back to the sender in user space to indicate success
154
+
or failure.
156
155
157
-
In the case of success, additionally an event with
158
-
XFRM_MSG_NEWAE is also issued to any listeners as described in iii).
156
+
In the case of success, additionally an event with
157
+
XFRM_MSG_NEWAE is also issued to any listeners as described in iii).
159
158
160
-
ii) kernel->user direction as a response to XFRM_MSG_GETAE
159
+
ii) kernel->user direction as a response to XFRM_MSG_GETAE
161
160
162
-
The response will always have XFRMA_LTIME_VAL and XFRMA_REPLAY_VAL TLVs.
161
+
The response will always have XFRMA_LTIME_VAL and XFRMA_REPLAY_VAL TLVs.
163
162
164
-
The threshold TLVs will be included if explicitly requested in
165
-
the XFRM_MSG_GETAE message.
163
+
The threshold TLVs will be included if explicitly requested in
164
+
the XFRM_MSG_GETAE message.
166
165
167
-
iii) kernel->user to report as event if someone sets any values or
168
-
thresholds for an SA using XFRM_MSG_NEWAE (as described in #i above).
169
-
In such a case XFRM_AE_CU flag is set to inform the user that
170
-
the change happened as a result of an update.
171
-
The message will always have XFRMA_LTIME_VAL and XFRMA_REPLAY_VAL TLVs.
166
+
iii) kernel->user to report as event if someone sets any values or
167
+
thresholds for an SA using XFRM_MSG_NEWAE (as described in #i above).
168
+
In such a case XFRM_AE_CU flag is set to inform the user that
169
+
the change happened as a result of an update.
170
+
The message will always have XFRMA_LTIME_VAL and XFRMA_REPLAY_VAL TLVs.
172
171
173
-
iv) kernel->user to report event when replay threshold or a timeout
174
-
is exceeded.
172
+
iv) kernel->user to report event when replay threshold or a timeout
173
+
is exceeded.
175
174
176
175
In such a case either XFRM_AE_CR (replay exceeded) or XFRM_AE_CE (timeout
177
176
happened) is set to inform the user what happened.
178
177
Note the two flags are mutually exclusive.
179
178
The message will always have XFRMA_LTIME_VAL and XFRMA_REPLAY_VAL TLVs.
180
179
181
-
Exceptions to threshold settings
182
-
--------------------------------
180
+
5) Exceptions to threshold settings
181
+
-----------------------------------
183
182
184
183
If you have an SA that is getting hit by traffic in bursts such that
185
184
there is a period where the timer threshold expires with no packets
+2
-2
Documentation/networking/xfrm_sysctl.rst
Documentation/networking/xfrm/xfrm_sysctl.rst
+2
-2
Documentation/networking/xfrm_sysctl.rst
Documentation/networking/xfrm/xfrm_sysctl.rst
···
4
4
XFRM Syscall
5
5
============
6
6
7
-
/proc/sys/net/core/xfrm_* Variables:
8
-
====================================
7
+
/proc/sys/net/core/xfrm_* Variables
8
+
===================================
9
9
10
10
xfrm_acq_expires - INTEGER
11
11
default 30 - hard timeout in seconds for acquire requests
+1
MAINTAINERS
+1
MAINTAINERS
···
18068
18068
S: Maintained
18069
18069
T: git git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec.git
18070
18070
T: git git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-next.git
18071
+
F: Documentation/networking/xfrm/
18071
18072
F: include/net/xfrm.h
18072
18073
F: include/uapi/linux/xfrm.h
18073
18074
F: net/ipv4/ah4.c
+2
net/key/af_key.c
+2
net/key/af_key.c
+7
-4
net/xfrm/Kconfig
+7
-4
net/xfrm/Kconfig
···
110
110
select CRYPTO_DEFLATE
111
111
112
112
config NET_KEY
113
-
tristate "PF_KEY sockets"
113
+
tristate "PF_KEY sockets (deprecated)"
114
114
select XFRM_ALGO
115
115
help
116
116
PF_KEYv2 socket family, compatible to KAME ones.
117
-
They are required if you are going to use IPsec tools ported
118
-
from KAME.
119
117
120
-
Say Y unless you know what you are doing.
118
+
The PF_KEYv2 socket interface is deprecated and
119
+
scheduled for removal. All maintained IKE daemons
120
+
no longer need PF_KEY sockets. Please use the netlink
121
+
interface (XFRM_USER) to configure IPsec.
122
+
123
+
If unsure, say N.
121
124
122
125
config NET_KEY_MIGRATE
123
126
bool "PF_KEY MIGRATE"
+15
-15
net/xfrm/xfrm_input.c
+15
-15
net/xfrm/xfrm_input.c
···
505
505
async = 1;
506
506
dev_put(skb->dev);
507
507
seq = XFRM_SKB_CB(skb)->seq.input.low;
508
+
spin_lock(&x->lock);
508
509
goto resume;
509
510
}
510
511
/* GRO call */
···
542
541
XFRM_INC_STATS(net, LINUX_MIB_XFRMINHDRERROR);
543
542
goto drop;
544
543
}
544
+
545
+
nexthdr = x->type_offload->input_tail(x, skb);
545
546
}
546
547
547
-
goto lock;
548
+
goto process;
548
549
}
549
550
550
551
family = XFRM_SPI_SKB_CB(skb)->family;
···
614
611
goto drop;
615
612
}
616
613
617
-
lock:
614
+
process:
615
+
seq_hi = htonl(xfrm_replay_seqhi(x, seq));
616
+
617
+
XFRM_SKB_CB(skb)->seq.input.low = seq;
618
+
XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
619
+
618
620
spin_lock(&x->lock);
619
621
620
622
if (unlikely(x->km.state != XFRM_STATE_VALID)) {
···
646
638
goto drop_unlock;
647
639
}
648
640
649
-
spin_unlock(&x->lock);
650
-
651
641
if (xfrm_tunnel_check(skb, x, family)) {
652
642
XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATEMODEERROR);
653
-
goto drop;
643
+
goto drop_unlock;
654
644
}
655
645
656
-
seq_hi = htonl(xfrm_replay_seqhi(x, seq));
657
-
658
-
XFRM_SKB_CB(skb)->seq.input.low = seq;
659
-
XFRM_SKB_CB(skb)->seq.input.hi = seq_hi;
660
-
661
-
if (crypto_done) {
662
-
nexthdr = x->type_offload->input_tail(x, skb);
663
-
} else {
646
+
if (!crypto_done) {
647
+
spin_unlock(&x->lock);
664
648
dev_hold(skb->dev);
665
649
666
650
nexthdr = x->type->input(x, skb);
···
660
660
return 0;
661
661
662
662
dev_put(skb->dev);
663
+
spin_lock(&x->lock);
663
664
}
664
665
resume:
665
-
spin_lock(&x->lock);
666
666
if (nexthdr < 0) {
667
667
if (nexthdr == -EBADMSG) {
668
668
xfrm_audit_state_icvfail(x, skb,
···
676
676
/* only the first xfrm gets the encap type */
677
677
encap_type = 0;
678
678
679
-
if (xfrm_replay_recheck(x, skb, seq)) {
679
+
if (!crypto_done && xfrm_replay_recheck(x, skb, seq)) {
680
680
XFRM_INC_STATS(net, LINUX_MIB_XFRMINSTATESEQERROR);
681
681
goto drop_unlock;
682
682
}