Skip to contentSkip to conversation

net/tcp: Add TCP-AO support

PATCH In review Dmitry Safonov 4 participants 36 messages Jul 21, 2023 – Jul 26, 2023

Subsystems (5)
lore
Share & tools
lore
#0 lore raw

cover net/tcp: Add TCP-AO support

Hi,

This is version 8.1 of TCP-AO support. I base it on net-next as there's commit 5e5265522a9a ("tcp: annotate data-races around tcp_rsk(req)->txhash") which makes a minor conflict.

The good news is that all pre-required patches have merged to Torvald's/master. Thanks to Herbert, crypto clone-tfm just works on master for all TCP-AO supported algorithms.
So, this is the first version of the patch set that has only net-related changes (well, selftests as well, but they'll be upstreamed separately).

In this version, I've finally spent time and written Documentation/ page on TCP-AO. It has Frequently Asked Questions (FAQ) on RFC 5925 - I found it very useful to answer those before writing the actual code. It provides answers to common questions that arise on a quick read of the RFC as well as how they were answered. There's also a comparison to the TCP-MD5 option, an evaluation of per-socket vs in-kernel-DB approaches and a description of uAPI provided.
I hope it will be as useful for reviewing the code as it was for writing.

The most important changes in this version are:
- CONFIG_TCP_AO implies CONFIG_IPV6 != m. I don't feel like that combination would be useful to anyone and it'd be painful to fix.
- uAPI change in TCP_AO_REPAIR (introduced in version 7): I removed {snd,rcv}_sne_seq counters. They were just copies of snd_nxt/snd_una. No reason for polluting uAPI as well as needlessly copying them.
- TCP_AO_MAX_HASH_SIZE is removed and all temporary buffers are kmalloc()'d. That also saves a couple of bytes for hmac(sha1) and cmac(aes128) traffic keys as they now are allocated with exact hash algo's digest_size.

There's an independent patch set for TCP-MD5 to verify segments on twsk:
https://lore.kernel.org/all/[email protected]/T/#u
That may be used to verify TCP-AO segments on twsk as well.

There seem to be more people that connected me off-list asking me about the status of patches and when I expect them to merge. Cc'ing more interested parties here (ping me directly if you don't want to be in copy). It would be helpful if you provide your reviews and tested-by's. As far as I'm aware, version 7 was ported to RHEL, so now there are probably more downstream kernels with TCP-AO support.

Also available as a git branch for pulling:
https://github.com/0x7f454c46/linux/tree/tcp-ao-v8.1
And another branch with selftests, that will be sent later separately:
https://github.com/0x7f454c46/linux/tree/tcp-ao-v8-with-selftests

Thanks for your time and reviews,
Dmitry

--- Changelog ---

Changes from v8:
- Rebased/retested on linux-net-next

Version 8: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v7:
- Fixed copy'n'paste typo in unsigned-md5.c selftest output
- Fix build error in tcp_v6_send_reset() (kernel test robot <[email protected]>)
- Make CONFIG_TCP_AO imply IPV6 != m
- Cleanup EXPORT_SYMBOL*() as they aren't needed with IPV6 != m
- Used scratch area instead of on-stack buffer for scatter-gather list in tcp_v{4,6}_ao_calc_key(). Fixes CONFIG_VMAP_STACK=y + CONFIG_DEBUG_SG=y
- Allocated digest_size'd buffers for traffic keys in tcp_ao_key instead of maximum-sized buffers of TCP_AO_MAX_HASH_SIZE. That will save little space per key and also potentially allow algorithms with digest size > TCP_AO_MAX_HASH_SIZE.
- Removed TCP_AO_MAX_HASH_SIZE and used kmalloc(GFP_ATOMIC) instead of on-stack hash buffer.
- Don't treat fd=0 as invalid in selftests
- Make TCP-AO selftests work with CONFIG_CRYPTO_FIPS=y
- Don't tcp_ao_compute_sne() for snd_sne on twsk: it's redundant as no data can be sent on twsk
- Get rid of {snd,rcv}_sne_seq: use snd_nxt/snd_una or rcv_nxt instead
- {rcv,snd}_sne and tcp_ao_compute_sne() now are introduced in "net/tcp: Add TCP-AO SNE support" patch
- trivial copy_to_sockptr() fixup for tcp_ao_get_repair() - it could try copying bigger struct than the kernel one (embarrassing!)
- Added Documentation/networking/tcp_ao.rst that describes:
uAPI, has FAQ on RFC 5925 and has implementation details of Linux TCP-AO

Version 7: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v6:
- Some more trivial build warnings fixups (kernel test robot <[email protected]>)
- Added TCP_AO_REPAIR setsockopt(), getsockopt()
- Allowed TCP_AO_* setsockopts if (tp->repair) is on
- Added selftests for TCP_AO_REPAIR, that also check incorrect ISNs/SNEs, which result in a broken TCP-AO connection - that verifies that both Initial Sequence Numbers and Sequence Number Extension are part of MAC generation
- Using TCP_AO_REPAIR added a selftest for SEQ numbers rollover, checking that SNE was incremented, connection is alive post-rolloever and no TCP segments with a wrong signature arrived
- Wrote a selftest for RST segments: both active reset (goes through transmit_skb()) and passive reset (goes through tcp_v{4,6}_send_reset()).
- Refactored and made readable tcp_v{4,6}_send_reset(), also adding support for TCP_LISTEN/TCP_NEW_SYN_RECV
- Dropped per-CPU ahash requests allocations in favor of Herbert's clone-tfm crypto API
- Added Donald Cassidy to Cc as he's interested in getting it into RHEL.

Version 6: https://lore.kernel.org/all/[email protected]/T/#u

iperf[3] benchmarks for version 6:

                           v6.4-rc1                 TCP-AO-v6
  TCP                      43.9 Gbits/sec           43.5 Gbits/sec
  TCP-MD5                  2.20 Gbits/sec           2.25 Gbits/sec
  TCP-AO(hmac(sha1))                                2.53 Gbits/sec
  TCP-AO(hmac(sha512))                              1.67 Gbits/sec
  TCP-AO(hmac(sha384))                              1.77 Gbits/sec
  TCP-AO(hmac(sha224))                              1.29 Gbits/sec
  TCP-AO(hmac(sha3-512))                             481 Mbits/sec
  TCP-AO(hmac(md5))                                 2.07 Gbits/sec
  TCP-AO(hmac(rmd160))                              1.01 Gbits/sec
  TCP-AO(cmac(aes128))                              2.11 Gbits/sec

Changes from v5:
- removed check for TCP_AO_KEYF_IFINDEX in delete command:
VRF might have been destroyed, there still needs to be a way to delete keys that were bound to that l3intf (should tcp_v{4,6}_parse_md5_keys() avoid the same check as well?)
- corrected copy'n'paste typo in tcp_ao_info_cmd() (assign ao_info->rnext_key)
- simplified a bit tcp_ao_copy_mkts_to_user(); added more UAPI checks for getsockopt(TCP_AO_GET_KEYS)
- More UAPI selftests in setsockopt-closed: 29 => 120
- ported TCP-AO patches on Herbert's clone-tfm changes
- adjusted iperf patch for TCP-AO UAPI changes from version 5
- added measures for TCP-AO with tcp_sigpool & clone_tfm backends

Version 5: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v4:
- Renamed tcp_ao_matched_key() => tcp_ao_established_key()
- Missed static in function definitions
(kernel test robot <[email protected]>)
- Fixed CONFIG_IPV6=m build
- Unexported tcp_md5_*_sigpool() functions
- Cleaned up tcp_ao.h: undeclared tcp_ao_cache_traffic_keys(), tcp_v4_ao_calc_key_skb(); removed tcp_v4_inbound_ao_hash()
- Marked "net/tcp: Prepare tcp_md5sig_pool for TCP-AO" as a [draft] patch
- getsockopt() now returns TCP-AO per-key counters
- Another getsockopt() now returns per-ao_info stats: counters and accept_icmps flag state
- Wired up getsockopt() returning counters to selftests
- Fixed a porting mistake: TCP-AO hash in some cases was written in TCP header without accounting for MAC length of the key, rewritting skb shared info
- Fail adding a key with L3 ifindex when !TCP_AO_KEYF_IFINDEX, instead of ignoring tcpa_ifindex (stricter UAPI check)
- Added more test-cases to setsockopt-closed.c selftest
- tcp_ao_hash_skb_data() was a copy'n'paste of tcp_md5_hash_skb_data() share it now under tcp_sigpool_hash_skb_data()
- tcp_ao_mkt_overlap_v{4,6}() deleted as they just re-invented tcp_ao_do_lookup(). That fixes an issue with multiple IPv4-mapped-IPv6 keys for different peers on a listening socket.
- getsockopt() now is tested to return correct VRF number for a key
- TCP-AO and TCP-MD5 interraction in non/default VRFs: added +19 selftests made them SKIP when CONFIG_VRF=n
- unsigned-md5 selftests now checks both scenarios:
(1) adding TCP-AO key _after_ TCP-MD5 key
(2) adding TCP-MD5 key _after_ TCP-AO key
- Added a ratelimited warning if TCP-AO key.ifindex doesn't match sk->sk_bound_dev_if - that will warn a user for potential VRF issues
- tcp_v{4,6}_parse_md5_keys() now allows adding TCP-MD5 key with ifindex=0 and TCP_MD5SIG_FLAG_IFINDEX together with TCP-AO key from another VRF
- Add TCP_AO_CMDF_AO_REQUIRED, which makes a socket TCP-AO only, rejecting TCP-MD5 keys or any unsigned TCP segments
- Remove `tcpa_' prefix for UAPI structure members
- UAPI cleanup: I've separated & renamed per-socket settings (such as ao_info flags + current/rnext set) from per-key changes:
TCP_AO => TCP_AO_ADD_KEY
TCP_AO_DEL => TCP_AO_DEL_KEY
TCP_AO_GET => TCP_AO_GET_KEYS
TCP_AO_MOD => TCP_AO_INFO, the structure is now valid for both
getsockopt() and setsockopt().
- tcp_ao_current_rnext() was split up in order to fail earlier when sndid/rcvid specified can't be set, before anything was changed in ao_info
- fetch current_key before dumping TCP-AO keys in getsockopt(TCP_AO_GET_KEYS):
it may race with changing current_key by RX, which in result might produce a dump with no current_key for userspace.
- instead of TCP_AO_CMDF_* flags, used bitfileds: the flags weren't shared between all TCP_AO_{ADD,GET,DEL}_KEY{,S}, so bitfields are more descriptive here
- use READ_ONCE()/WRITE_ONCE() for current_key and rnext_key more consistently; document in comment the rules for accessing them
- selftests: check all setsockopts()/getsockopts() support extending option structs

Version 4: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v3:
- TCP_MD5 dynamic static key enable/disable patches merged separately [4]
- crypto_pool patches were nacked [5], so instead this patch set extends TCP-MD5-sigpool to be used for TCP-AO as well as for TCP-MD5
- Added missing `static' for tcp_v6_ao_calc_key()
(kernel test robot <[email protected]>)
- Removed CONFIG_TCP_AO default=y and added "If unsure, say N."
- Don't leak ao_info and don't create an unsigned TCP socket if there was a TCP-AO key during handshake, but it was removed from listening socket while the connection was being established
- Migrate to use static_key_fast_inc_not_disabled() and check return code of static_branch_inc()
- Change some return codes to EAFNOSUPPORT for error-pathes where family is neither AF_INET nor AF_INET6
- setsockopt()s on a closed/listen socket might have created stray ao_info, remove it if connect() is called with a correct TCP-MD5 key, the same for the reverse situation: remove md5sig_info straight away from the socket if it's going to be TCP-AO connection
- IPv4-mapped-IPv6 addresses + selftest in fcnal-test.sh (by Salam)
- fix using uninitialized sisn/disn from stack - it would only make non-SYN packets fail verification on a listen socket, which are not expected anyway (kernel test robot <[email protected]>)
- implicit padding in UAPI TCP-AO structures converted to explicit (spotted-by David Laight)
- Some selftests missed zero-initializers for uapi structs on stack
- Removed tcp_ao_do_lookup_rcvid() and tcp_ao_do_lookup_sndid() in favor of unified tcp_ao_matched_key()
- Disallowed setting current/rnext keys on listen sockets - that wasn't supported and didn't affect anything, cleanup for the UAPI
- VRFs support for TCP-AO

Version 3: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v2:
- Added more missing `static' declarations for local functions (kernel test robot <[email protected]>)
- Building now with CONFIG_TCP_AO=n and CONFIG_TCP_MD5SIG=n (kernel test robot <[email protected]>)
- Now setsockopt(TCP_AO) is allowed when it's TCP_LISTEN or TCP_CLOSE state OR the key added is not the first key on a socket (by Salam)
- CONFIG_TCP_AO does not depend on CONFIG_TCP_MD5SIG anymore
- Don't leak tcp_md5_needed static branch counter when TCP-MD5 key is modified/changed
- TCP-AO lookups are dynamically enabled/disabled with static key when there is ao_info in the system (and when it is destroyed)
- Wired SYN cookies up to TCP-AO (by Salam)
- Fix verification for possible re-transmitted SYN packets (by Salam)
- use sockopt_lock_sock() instead of lock_sock()
(from v6.1 rebase, commit d51bbff2aba7)
- use sockptr_t in getsockopt(TCP_AO_GET)
(from v6.1 rebase, commit 34704ef024ae)
- Fixed reallocating crypto_pool's scratch area by IPI while crypto_pool_get() was get by another CPU
- selftests on older kernels (or with CONFIG_TCP_AO=n) should exit with SKIP, not FAIL (Shuah Khan <[email protected]>)
- selftests that check interaction between TCP-AO and TCP-MD5 now SKIP when CONFIG_TCP_MD5SIG=n
- Measured the performance of different hashing algorithms for TCP-AO and compare with TCP-MD5 performance. This is done with hacky patches to iperf (see [3]). At this moment I've done it in qemu/KVM with CPU affinities set on Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz. No performance degradation was noticed before/after patches, but given the measures were done in a VM, without measuring it on a physical dut it only gives a hint of relative speed for different hash algorithms with TCP-AO. Here are results, averaging on 30 measures each:

  TCP:                    3.51Gbits/sec
  TCP-MD5:                1.12Gbits/sec
  TCP-AO(HMAC(SHA1)):     1.53Gbits/sec
  TCP-AO(CMAC(AES128)):   621Mbits/sec
  TCP-AO(HMAC(SHA512)):   1.21Gbits/sec
  TCP-AO(HMAC(SHA384)):   1.20Gbits/sec
  TCP-AO(HMAC(SHA224)):   961Mbits/sec

TCP-AO(HMAC(SHA3-512)): 157Mbits/sec
TCP-AO(HMAC(RMD160)): 659Mbits/sec
TCP-AO(HMAC(MD5): 1.12Gbits/sec
(the last one is just for fun, but may make sense as it provides the same security as TCP-MD5, but allows multiple keys and a mechanism to change them from RFC5925)

Version 2: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v1:
- Building now with CONFIG_IPV6=n (kernel test robot <[email protected]>)
- Added missing static declarations for local functions (kernel test robot <[email protected]>)
- Addressed static analyzer and review comments by Dan Carpenter (thanks, they were very useful!)
- Fix elif without defined() for !CONFIG_TCP_AO
- Recursively build selftests/net/tcp_ao (Shuah Khan), patches in:
https://lore.kernel.org/all/[email protected]/T/#u
- Don't leak crypto_pool reference when TCP-MD5 key is modified/changed
- Add TCP-AO support for nettest.c and fcnal-test.sh
(will be used for VRF testing in later versions)

Comparison between Leonard proposal and this (overview):
https://lore.kernel.org/all/[email protected]/T/#u

Version 1: https://lore.kernel.org/all/[email protected]/T/#u

This patchset implements the TCP-AO option as described in RFC5925. There is a request from industry to move away from TCP-MD5SIG and it seems the time is right to have a TCP-AO upstreamed. This TCP option is meant to replace the TCP MD5 option and address its shortcomings. Specifically, it provides more secure hashing, key rotation and support for long-lived connections (see the summary of TCP-AO advantages over TCP-MD5 in (1.3) of RFC5925). The patch series starts with six patches that are not specific to TCP-AO but implement a general crypto facility that we thought is useful to eliminate code duplication between TCP-MD5SIG and TCP-AO as well as other crypto users. These six patches are being submitted separately in a different patchset [1]. Including them here will show better the gain in code sharing. Next are 18 patches that implement the actual TCP-AO option, followed by patches implementing selftests.

The patch set was written as a collaboration of three authors (in alphabetical order): Dmitry Safonov, Francesco Ruggeri and Salam Noureddine. Additional credits should be given to Prasad Koya, who was involved in early prototyping a few years back. There is also a separate submission done by Leonard Crestez whom we thank for his efforts getting an implementation of RFC5925 submitted for review upstream [2]. This is an independent implementation that makes different design decisions.

For example, we chose a similar design to the TCP-MD5SIG implementation and used setsockopts to program per-socket keys, avoiding the extra complexity of managing a centralized key database in the kernel. A centralized database in the kernel has dubious benefits since it doesn’t eliminate per-socket setsockopts needed to specify which sockets need TCP-AO and what are the currently preferred keys. It also complicates traffic key caching and preventing deletion of in-use keys.

In this implementation, a centralized database of keys can be thought of as living in user space and user applications would have to program those keys on matching sockets. On the server side, the user application programs keys (MKTS in TCP-AO nomenclature) on the listening socket for all peers that are expected to connect. Prefix matching on the peer address is supported. When a peer issues a successful connect, all the MKTs matching the IP address of the peer are copied to the newly created socket. On the active side, when a connect() is issued all MKTs that do not match the peer are deleted from the socket since they will never match the peer. This implementation uses three setsockopt()s for adding, deleting and modifying keys on a socket. All three setsockopt()s have extensive sanity checks that prevent inconsistencies in the keys on a given socket. A getsockopt() is provided to get key information from any given socket.

Few things to note about this implementation:
- Traffic keys are cached for established connections avoiding the cost of such calculation for each packet received or sent.
- Great care has been taken to avoid deleting in-use MKTs
as required by the RFC.
- Any crypto algorithm supported by the Linux kernel can be used to calculate packet hashes.
- Fastopen works with TCP-AO but hasn’t been tested extensively.
- Tested for interop with other major networking vendors (on linux-4.19), including testing for key rotation and long lived connections.

[1]: https://lore.kernel.org/all/[email protected]/
[2]: https://lore.kernel.org/all/[email protected]/
[3]: https://github.com/0x7f454c46/iperf/tree/tcp-md5-ao
[4]: https://lore.kernel.org/all/166995421700.16716.17446147162780881407.git-patchwork-notify@kernel.org/T/#u
[5]: https://lore.kernel.org/all/[email protected]/T/#u
[6]: https://lore.kernel.org/all/[email protected]/T/#u

Cc Andy Lutomirski <[email protected]>
Cc Ard Biesheuvel <[email protected]>
Cc Bob Gilligan <[email protected]>
Cc Dan Carpenter <[email protected]>
Cc David Ahern <[email protected]>
Cc David Laight <[email protected]>
Cc "David S. Miller" <[email protected]>
Cc Dmitry Safonov <[email protected]>
Cc Donald Cassidy <[email protected]>
Cc Eric Biggers <[email protected]>
Cc Eric Dumazet <[email protected]>
Cc "Eric W. Biederman" <[email protected]>
Cc Francesco Ruggeri <[email protected]>
Cc Gaillardetz, Dominik <[email protected]>
Cc Herbert Xu <[email protected]>
Cc Hideaki YOSHIFUJI <[email protected]>
Cc Ivan Delalande <[email protected]>
Cc Jakub Kicinski <[email protected]>
Cc Leonard Crestez <[email protected]>
Cc Paolo Abeni <[email protected]>
Cc Salam Noureddine <[email protected]>
Cc Tetreault, Francois <[email protected]>

Dmitry Safonov (23):
net/tcp: Prepare tcp_md5sig_pool for TCP-AO net/tcp: Add TCP-AO config and structures net/tcp: Introduce TCP_AO setsockopt()s
net/tcp: Prevent TCP-MD5 with TCP-AO being set net/tcp: Calculate TCP-AO traffic keys
net/tcp: Add TCP-AO sign to outgoing packets net/tcp: Add tcp_parse_auth_options()
net/tcp: Add AO sign to RST packets
net/tcp: Add TCP-AO sign to twsk
net/tcp: Wire TCP-AO to request sockets
net/tcp: Sign SYN-ACK segments with TCP-AO net/tcp: Verify inbound TCP-AO signed segments net/tcp: Add TCP-AO segments counters
net/tcp: Add TCP-AO SNE support
net/tcp: Add tcp_hash_fail() ratelimited logs net/tcp: Ignore specific ICMPs for TCP-AO connections net/tcp: Add option for TCP-AO to (not) hash header net/tcp: Add TCP-AO getsockopt()s
net/tcp: Allow asynchronous delete for TCP-AO keys (MKTs) net/tcp: Add static_key for TCP-AO
net/tcp: Wire up l3index to TCP-AO
net/tcp: Add TCP_AO_REPAIR
Documentation/tcp: Add TCP-AO documentation

 Documentation/networking/index.rst  |    1 +
 Documentation/networking/tcp_ao.rst |  422 +++++
 include/linux/sockptr.h             |   23 +
 include/linux/tcp.h                 |   30 +-
 include/net/dropreason-core.h       |   30 +
 include/net/tcp.h                   |  224 ++-
 include/net/tcp_ao.h                |  349 ++++
 include/uapi/linux/snmp.h           |    5 +
 include/uapi/linux/tcp.h            |  105 ++
 net/ipv4/Kconfig                    |   17 +
 net/ipv4/Makefile                   |    2 +
 net/ipv4/proc.c                     |    5 +
 net/ipv4/syncookies.c               |    4 +
 net/ipv4/tcp.c                      |  236 +--
 net/ipv4/tcp_ao.c                   | 2367 +++++++++++++++++++++++++++
 net/ipv4/tcp_input.c                |   97 +-
 net/ipv4/tcp_ipv4.c                 |  333 +++-
 net/ipv4/tcp_minisocks.c            |   50 +-
 net/ipv4/tcp_output.c               |  232 ++-
 net/ipv4/tcp_sigpool.c              |  357 ++++
 net/ipv6/Makefile                   |    1 +
 net/ipv6/syncookies.c               |    5 +
 net/ipv6/tcp_ao.c                   |  167 ++
 net/ipv6/tcp_ipv6.c                 |  334 +++-
 24 files changed, 5026 insertions(+), 370 deletions(-)
 create mode 100644 Documentation/networking/tcp_ao.rst
 create mode 100644 include/net/tcp_ao.h
 create mode 100644 net/ipv4/tcp_ao.c
 create mode 100644 net/ipv4/tcp_sigpool.c
 create mode 100644 net/ipv6/tcp_ao.c

base-commit: 2da6a8041699691df4269e756253fa04a1d6a5fd

Signature · 1 line
2.41.0
#1 lore raw

PATCH 1/23 net/tcp: Prepare tcp_md5sig_pool for TCP-AO

TCP-AO, similarly to TCP-MD5, needs to allocate tfms on a slow-path, which is setsockopt() and use crypto ahash requests on fast paths, which are RX/TX softirqs. Also, it needs a temporary/scratch buffer for preparing the hash.

Rework tcp_md5sig_pool in order to support other hashing algorithms than MD5. It will make it possible to share pre-allocated crypto_ahash descriptors and scratch area between all TCP hash users.

Internally tcp_sigpool calls crypto_clone_ahash() API over pre-allocated crypto ahash tfm. Kudos to Herbert, who provided this new crypto API.

I was a little concerned over GFP_ATOMIC allocations of ahash and crypto_request in RX/TX (see tcp_sigpool_start()), so I benchmarked both "backends" with different algorithms, using patched version of iperf3[2]. On my laptop with i7-7600U @ 2.80GHz:

                         clone-tfm                per-CPU-requests
TCP-MD5                  2.25 Gbits/sec           2.30 Gbits/sec
TCP-AO(hmac(sha1))       2.53 Gbits/sec           2.54 Gbits/sec
TCP-AO(hmac(sha512))     1.67 Gbits/sec           1.64 Gbits/sec
TCP-AO(hmac(sha384))     1.77 Gbits/sec           1.80 Gbits/sec
TCP-AO(hmac(sha224))     1.29 Gbits/sec           1.30 Gbits/sec
TCP-AO(hmac(sha3-512))    481 Mbits/sec            480 Mbits/sec
TCP-AO(hmac(md5))        2.07 Gbits/sec           2.12 Gbits/sec
TCP-AO(hmac(rmd160))     1.01 Gbits/sec            995 Mbits/sec
TCP-AO(cmac(aes128))     [not supporetd yet]      2.11 Gbits/sec

So, it seems that my concerns don't have strong grounds and per-CPU crypto_request allocation can be dropped/removed from tcp_sigpool once ciphers get crypto_clone_ahash() support.

[1]: https://lore.kernel.org/all/[email protected]/T/#u
[2]: https://github.com/0x7f454c46/iperf/tree/tcp-md5-ao

---
 include/net/tcp.h        |  46 +++--
 net/ipv4/Kconfig         |   4 +
 net/ipv4/Makefile        |   1 +
 net/ipv4/tcp.c           | 135 +++------------
 net/ipv4/tcp_ipv4.c      |  97 ++++++-----
 net/ipv4/tcp_minisocks.c |  21 ++-
 net/ipv4/tcp_sigpool.c   | 357 +++++++++++++++++++++++++++++++++++++++
 net/ipv6/tcp_ipv6.c      |  58 +++----
 8 files changed, 510 insertions(+), 209 deletions(-)
 create mode 100644 net/ipv4/tcp_sigpool.c
include/net/tcp.h +32 -14
index d17cb8ab4c48..95651f9c9242 100644
@@ -1692,12 +1692,35 @@ union tcp_md5sum_block {
 #endif
 };
 
-/* - pool: digest algorithm, hash description and scratch buffer */
-struct tcp_md5sig_pool {
-	struct ahash_request	*md5_req;
-	void			*scratch;
+/*
+ * struct tcp_sigpool - per-CPU pool of ahash_requests
+ * @scratch: per-CPU temporary area, that can be used between
+ *	     tcp_sigpool_start() and tcp_sigpool_end() to perform
+ *	     crypto request
+ * @req: pre-allocated ahash request
+ */
+struct tcp_sigpool {
+	void *scratch;
+	struct ahash_request *req;
 };
+int tcp_sigpool_alloc_ahash(const char *alg, size_t scratch_size);
+void tcp_sigpool_get(unsigned int id);
+void tcp_sigpool_release(unsigned int id);
+int tcp_sigpool_hash_skb_data(struct tcp_sigpool *hp,
+			      const struct sk_buff *skb,
+			      unsigned int header_len);
 
+/**
+ * tcp_sigpool_start - disable bh and start using tcp_sigpool_ahash
+ * @id: tcp_sigpool that was previously allocated by tcp_sigpool_alloc_ahash()
+ * @c: returned tcp_sigpool for usage (uninitialized on failure)
+ */
+int tcp_sigpool_start(unsigned int id, struct tcp_sigpool *c);
+/**
+ * tcp_sigpool_end - enable bh and stop using tcp_sigpool
+ */
+void tcp_sigpool_end(struct tcp_sigpool *c);
+size_t tcp_sigpool_algo(unsigned int id, char *buf, size_t buf_len);
 /* - functions */
 int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
 			const struct sock *sk, const struct sk_buff *skb);
@@ -1753,17 +1776,12 @@ tcp_inbound_md5_hash(const struct sock *sk, const struct sk_buff *skb,
 #define tcp_twsk_md5_key(twsk)	NULL
 #endif
 
-bool tcp_alloc_md5sig_pool(void);
+int tcp_md5_alloc_sigpool(void);
+void tcp_md5_release_sigpool(void);
+void tcp_md5_add_sigpool(void);
+extern int tcp_md5_sigpool_id;
 
-struct tcp_md5sig_pool *tcp_get_md5sig_pool(void);
-static inline void tcp_put_md5sig_pool(void)
-{
-	local_bh_enable();
-}
-
-int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *,
-			  unsigned int header_len);
-int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
+int tcp_md5_hash_key(struct tcp_sigpool *hp,
 		     const struct tcp_md5sig_key *key);
 
 /* From tcp_fastopen.c */
net/ipv4/Kconfig +4 -0
index 2dfb12230f08..89e2ab023272 100644
@@ -741,10 +741,14 @@ config DEFAULT_TCP_CONG
 	default "bbr" if DEFAULT_BBR
 	default "cubic"
 
+config TCP_SIGPOOL
+	tristate
+
 config TCP_MD5SIG
 	bool "TCP: MD5 Signature Option support (RFC2385)"
 	select CRYPTO
 	select CRYPTO_MD5
+	select TCP_SIGPOOL
 	help
 	  RFC2385 specifies a method of giving MD5 protection to TCP sessions.
 	  Its main (only?) use is to protect BGP sessions between core routers
net/ipv4/Makefile +1 -0
index b18ba8ef93ad..cd760793cfcb 100644
@@ -62,6 +62,7 @@ obj-$(CONFIG_TCP_CONG_SCALABLE) += tcp_scalable.o
 obj-$(CONFIG_TCP_CONG_LP) += tcp_lp.o
 obj-$(CONFIG_TCP_CONG_YEAH) += tcp_yeah.o
 obj-$(CONFIG_TCP_CONG_ILLINOIS) += tcp_illinois.o
+obj-$(CONFIG_TCP_SIGPOOL) += tcp_sigpool.o
 obj-$(CONFIG_NET_SOCK_MSG) += tcp_bpf.o
 obj-$(CONFIG_BPF_SYSCALL) += udp_bpf.o
 obj-$(CONFIG_NETLABEL) += cipso_ipv4.o
net/ipv4/tcp.c +19 -116

160 lines not on this page · raw message

net/ipv4/tcp_ipv4.c +56 -41

194 lines not on this page · raw message

net/ipv4/tcp_minisocks.c +15 -6
index c8f2aa003387..5616b6a34bee 100644
@@ -261,10 +261,9 @@ static void tcp_time_wait_init(struct sock *sk, struct tcp_timewait_sock *tcptw)
 		tcptw->tw_md5_key = kmemdup(key, sizeof(*key), GFP_ATOMIC);
 		if (!tcptw->tw_md5_key)
 			return;
-		if (!tcp_alloc_md5sig_pool())
-			goto out_free;
 		if (!static_key_fast_inc_not_disabled(&tcp_md5_needed.key.key))
 			goto out_free;
+		tcp_md5_add_sigpool();
 	}
 	return;
 out_free:
@@ -349,16 +348,26 @@ void tcp_time_wait(struct sock *sk, int state, int timeo)
 }
 EXPORT_SYMBOL(tcp_time_wait);
 
+#ifdef CONFIG_TCP_MD5SIG
+static void tcp_md5_twsk_free_rcu(struct rcu_head *head)
+{
+	struct tcp_md5sig_key *key;
+
+	key = container_of(head, struct tcp_md5sig_key, rcu);
+	kfree(key);
+	static_branch_slow_dec_deferred(&tcp_md5_needed);
+	tcp_md5_release_sigpool();
+}
+#endif
+
 void tcp_twsk_destructor(struct sock *sk)
 {
 #ifdef CONFIG_TCP_MD5SIG
 	if (static_branch_unlikely(&tcp_md5_needed.key)) {
 		struct tcp_timewait_sock *twsk = tcp_twsk(sk);
 
-		if (twsk->tw_md5_key) {
-			kfree_rcu(twsk->tw_md5_key, rcu);
-			static_branch_slow_dec_deferred(&tcp_md5_needed);
-		}
+		if (twsk->tw_md5_key)
+			call_rcu(&twsk->tw_md5_key->rcu, tcp_md5_twsk_free_rcu);
 	}
 #endif
 }
net/ipv4/tcp_sigpool.c new file +357 -0

357 lines not on this page · raw message

net/ipv6/tcp_ipv6.c +26 -32

110 lines not on this page · raw message

Signature · 1 line
2.41.0
#2 lore raw

Re: [PATCH v8.1 net-next 01/23] net/tcp: Prepare tcp_md5sig_pool for TCP-AO

On Fri, Jul 21, 2023 at 05:18:52PM +0100, Dmitry Safonov wrote:

Hi Dmitry,

some minor nits from my side.

...

include/net/tcp.h union tcp_md5sum_block { · line 1720

Quoted 8 lines
include/net/tcp.h +8 -0
+/**
+ * tcp_sigpool_start - disable bh and start using tcp_sigpool_ahash
+ * @id: tcp_sigpool that was previously allocated by tcp_sigpool_alloc_ahash()
+ * @c: returned tcp_sigpool for usage (uninitialized on failure)
+ */
+int tcp_sigpool_start(unsigned int id, struct tcp_sigpool *c);
+/**
+ * tcp_sigpool_end - enable bh and stop using tcp_sigpool

nit: as this is a kernel doc, please document @c here.

Quoted 6 lines

+ */

include/net/tcp.h +2 -0
+void tcp_sigpool_end(struct tcp_sigpool *c);
+size_t tcp_sigpool_algo(unsigned int id, char *buf, size_t buf_len);
 /* - functions */
 int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
 			const struct sock *sk, const struct sk_buff *skb);

...

net/ipv4/tcp_ipv4.c int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, · line 1448

Quoted 9 lines
net/ipv4/tcp_ipv4.c +1 -2
@@ -1439,8 +1443,7 @@ int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
 			const struct sock *sk,
 			const struct sk_buff *skb)
 {
-	struct tcp_md5sig_pool *hp;
-	struct ahash_request *req;
+	struct tcp_sigpool hp;
 	const struct tcphdr *th = tcp_hdr(skb);
 	__be32 saddr, daddr;

nit: please consider using reverse xmas tree - longest line to shortest -

     for these local variable declarations.

	const struct tcphdr *th = tcp_hdr(skb);
	struct tcp_sigpool hp;
	__be32 saddr, daddr;

Likewise, elsewhere, when it can be done without excess churn.

...

#3 lore raw

Re: [PATCH v8.1 net-next 01/23] net/tcp: Prepare tcp_md5sig_pool for TCP-AO

Hi Simon,

Quoted 8 lines · Simon Horman

On Fri, Jul 21, 2023 at 05:18:52PM +0100, Dmitry Safonov wrote:

Hi Dmitry,

some minor nits from my side.

...

Quoted 10 lines

include/net/tcp.h union tcp_md5sum_block { · line 1720

Quoted 8 lines
include/net/tcp.h +8 -0
+/**
+ * tcp_sigpool_start - disable bh and start using tcp_sigpool_ahash
+ * @id: tcp_sigpool that was previously allocated by tcp_sigpool_alloc_ahash()
+ * @c: returned tcp_sigpool for usage (uninitialized on failure)
+ */
+int tcp_sigpool_start(unsigned int id, struct tcp_sigpool *c);
+/**
+ * tcp_sigpool_end - enable bh and stop using tcp_sigpool

nit: as this is a kernel doc, please document @c here.

Thanks, yeah, I also noticed that on netdev/kdoc and there are some other nits on the patchwork that Intel's build bot didn't report to my surprise. Will address them in v9.

Quoted 20 lines
Quoted 6 lines

+ */

include/net/tcp.h +2 -0
+void tcp_sigpool_end(struct tcp_sigpool *c);
+size_t tcp_sigpool_algo(unsigned int id, char *buf, size_t buf_len);
 /* - functions */
 int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
 			const struct sock *sk, const struct sk_buff *skb);

...

net/ipv4/tcp_ipv4.c int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, · line 1448

Quoted 9 lines
net/ipv4/tcp_ipv4.c +1 -2
@@ -1439,8 +1443,7 @@ int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
 			const struct sock *sk,
 			const struct sk_buff *skb)
 {
-	struct tcp_md5sig_pool *hp;
-	struct ahash_request *req;
+	struct tcp_sigpool hp;
 	const struct tcphdr *th = tcp_hdr(skb);
 	__be32 saddr, daddr;

nit: please consider using reverse xmas tree - longest line to shortest -

Quoted 7 lines
     for these local variable declarations.

	const struct tcphdr *th = tcp_hdr(skb);
	struct tcp_sigpool hp;
	__be32 saddr, daddr;

Likewise, elsewhere, when it can be done without excess churn.

Yeah, fail enough, I usually keep it Xmas-like, but sometimes they slip in unnoticed. I'll take a look over the patches.

Thanks,
Dmitry

#4 lore raw

Re: [PATCH v8.1 net-next 01/23] net/tcp: Prepare tcp_md5sig_pool for TCP-AO

Quoted 22 lines · Dmitry Safonov

Hi Simon,

Quoted 8 lines · Simon Horman

On Fri, Jul 21, 2023 at 05:18:52PM +0100, Dmitry Safonov wrote:

Hi Dmitry,

some minor nits from my side.

...

Quoted 10 lines

include/net/tcp.h union tcp_md5sum_block { · line 1720

Quoted 8 lines
include/net/tcp.h +8 -0
+/**
+ * tcp_sigpool_start - disable bh and start using tcp_sigpool_ahash
+ * @id: tcp_sigpool that was previously allocated by tcp_sigpool_alloc_ahash()
+ * @c: returned tcp_sigpool for usage (uninitialized on failure)
+ */
+int tcp_sigpool_start(unsigned int id, struct tcp_sigpool *c);
+/**
+ * tcp_sigpool_end - enable bh and stop using tcp_sigpool

nit: as this is a kernel doc, please document @c here.

Quoted 3 lines

Thanks, yeah, I also noticed that on netdev/kdoc and there are some other nits on the patchwork that Intel's build bot didn't report to my surprise. Will address them in v9.

Always room for improvement :)

Quoted 22 lines
Quoted 22 lines
Quoted 6 lines

+ */

include/net/tcp.h +2 -0
+void tcp_sigpool_end(struct tcp_sigpool *c);
+size_t tcp_sigpool_algo(unsigned int id, char *buf, size_t buf_len);
 /* - functions */
 int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
 			const struct sock *sk, const struct sk_buff *skb);

...

net/ipv4/tcp_ipv4.c int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key, · line 1448

Quoted 9 lines
net/ipv4/tcp_ipv4.c +1 -2
@@ -1439,8 +1443,7 @@ int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
 			const struct sock *sk,
 			const struct sk_buff *skb)
 {
-	struct tcp_md5sig_pool *hp;
-	struct ahash_request *req;
+	struct tcp_sigpool hp;
 	const struct tcphdr *th = tcp_hdr(skb);
 	__be32 saddr, daddr;

nit: please consider using reverse xmas tree - longest line to shortest -
for these local variable declarations.

Quoted 8 lines
Quoted 5 lines
	const struct tcphdr *th = tcp_hdr(skb);
	struct tcp_sigpool hp;
	__be32 saddr, daddr;

Likewise, elsewhere, when it can be done without excess churn.

Yeah, fail enough, I usually keep it Xmas-like, but sometimes they slip in unnoticed. I'll take a look over the patches.

Thanks. I also flagged this in my review of 3/23,
as I hadn't noticed your response here.

#5 lore raw

PATCH 2/23 net/tcp: Add TCP-AO config and structures

In reply to Dmitry Safonov

Introduce new kernel config option and common structures as well as helpers to be used by TCP-AO code.

---
 include/linux/tcp.h      |  9 +++-
 include/net/tcp.h        |  8 +---
 include/net/tcp_ao.h     | 91 ++++++++++++++++++++++++++++++++++++++++
 include/uapi/linux/tcp.h |  2 +
 net/ipv4/Kconfig         | 13 ++++++
 5 files changed, 115 insertions(+), 8 deletions(-)
 create mode 100644 include/net/tcp_ao.h
include/linux/tcp.h +7 -2
index d16abdb3541a..9eb15e798c32 100644
@@ -437,13 +437,18 @@ struct tcp_sock {
 	bool	syn_smc;	/* SYN includes SMC */
 #endif
 
-#ifdef CONFIG_TCP_MD5SIG
-/* TCP AF-Specific parts; only used by MD5 Signature support so far */
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
+/* TCP AF-Specific parts; only used by TCP-AO/MD5 Signature support so far */
 	const struct tcp_sock_af_ops	*af_specific;
 
+#ifdef CONFIG_TCP_MD5SIG
 /* TCP MD5 Signature Option information */
 	struct tcp_md5sig_info	__rcu *md5sig_info;
 #endif
+#ifdef CONFIG_TCP_AO
+	struct tcp_ao_info	__rcu *ao_info;
+#endif
+#endif
 
 /* TCP fastopen related information */
 	struct tcp_fastopen_request *fastopen_req;
include/net/tcp.h +2 -6
index 95651f9c9242..d1230ee888fd 100644
@@ -37,6 +37,7 @@
 #include <net/snmp.h>
 #include <net/ip.h>
 #include <net/tcp_states.h>
+#include <net/tcp_ao.h>
 #include <net/inet_ecn.h>
 #include <net/dst.h>
 #include <net/mptcp.h>
@@ -1643,12 +1644,7 @@ static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
 	tp->retransmit_skb_hint = NULL;
 }
 
-union tcp_md5_addr {
-	struct in_addr  a4;
-#if IS_ENABLED(CONFIG_IPV6)
-	struct in6_addr	a6;
-#endif
-};
+#define tcp_md5_addr tcp_ao_addr
 
 /* - key database */
 struct tcp_md5sig_key {
include/net/tcp_ao.h new file +91 -0

91 lines not on this page · raw message

include/uapi/linux/tcp.h +2 -0
index 879eeb0a084b..5655bfe28b8d 100644
@@ -348,6 +348,8 @@ struct tcp_diag_md5sig {
 	__u8	tcpm_key[TCP_MD5SIG_MAXKEYLEN];
 };
 
+#define TCP_AO_MAXKEYLEN	80
+
 /* setsockopt(fd, IPPROTO_TCP, TCP_ZEROCOPY_RECEIVE, ...) */
 
 #define TCP_RECEIVE_ZEROCOPY_FLAG_TLB_CLEAN_HINT 0x1
net/ipv4/Kconfig +13 -0
index 89e2ab023272..8e94ed7c56a0 100644
@@ -744,6 +744,19 @@ config DEFAULT_TCP_CONG
 config TCP_SIGPOOL
 	tristate
 
+config TCP_AO
+	bool "TCP: Authentication Option (RFC5925)"
+	select CRYPTO
+	select TCP_SIGPOOL
+	depends on 64BIT && IPV6 != m # seq-number extension needs WRITE_ONCE(u64)
+	help
+	  TCP-AO specifies the use of stronger Message Authentication Codes (MACs),
+	  protects against replays for long-lived TCP connections, and
+	  provides more details on the association of security with TCP
+	  connections than TCP MD5 (See RFC5925)
+
+	  If unsure, say N.
+
 config TCP_MD5SIG
 	bool "TCP: MD5 Signature Option support (RFC2385)"
 	select CRYPTO
Signature · 1 line
2.41.0
#6 lore raw

PATCH 3/23 net/tcp: Introduce TCP_AO setsockopt()s

In reply to Dmitry Safonov

Add 3 setsockopt()s:
1. TCP_AO_ADD_KEY to add a new Master Key Tuple (MKT) on a socket
2. TCP_AO_DEL_KEY to delete present MKT from a socket
3. TCP_AO_INFO to change flags, Current_key/RNext_key on a TCP-AO sk

Userspace has to introduce keys on every socket it wants to use TCP-AO option on, similarly to TCP_MD5SIG/TCP_MD5SIG_EXT.
RFC5925 prohibits definition of MKTs that would match the same peer, so do sanity checks on the data provided by userspace. Be as conservative as possible, including refusal of defining MKT on an established connection with no AO, removing the key in-use and etc.

(1) and (2) are to be used by userspace key manager to add/remove keys.
(3) main purpose is to set RNext_key, which (as prescribed by RFC5925) is the KeyID that will be requested in TCP-AO header from the peer to sign their segments with.

At this moment the life of ao_info ends in tcp_v4_destroy_sock().

---
 include/linux/sockptr.h  |  23 ++
 include/net/tcp.h        |   6 +
 include/net/tcp_ao.h     |  18 +-
 include/uapi/linux/tcp.h |  46 +++
 net/ipv4/Makefile        |   1 +
 net/ipv4/tcp.c           |  17 +
 net/ipv4/tcp_ao.c        | 816 +++++++++++++++++++++++++++++++++++++++
 net/ipv4/tcp_ipv4.c      |  10 +-
 net/ipv6/Makefile        |   1 +
 net/ipv6/tcp_ao.c        |  19 +
 net/ipv6/tcp_ipv6.c      |  39 +-
 11 files changed, 978 insertions(+), 18 deletions(-)
 create mode 100644 net/ipv4/tcp_ao.c
 create mode 100644 net/ipv6/tcp_ao.c
include/linux/sockptr.h +23 -0
index bae5e2369b4f..307961b41541 100644
@@ -55,6 +55,29 @@ static inline int copy_from_sockptr(void *dst, sockptr_t src, size_t size)
 	return copy_from_sockptr_offset(dst, src, 0, size);
 }
 
+static inline int copy_struct_from_sockptr(void *dst, size_t ksize,
+		sockptr_t src, size_t usize)
+{
+	size_t size = min(ksize, usize);
+	size_t rest = max(ksize, usize) - size;
+
+	if (!sockptr_is_kernel(src))
+		return copy_struct_from_user(dst, ksize, src.user, size);
+
+	if (usize < ksize) {
+		memset(dst + size, 0, rest);
+	} else if (usize > ksize) {
+		char *p = src.kernel;
+
+		while (rest--) {
+			if (*p++)
+				return -E2BIG;
+		}
+	}
+	memcpy(dst, src.kernel, size);
+	return 0;
+}
+
 static inline int copy_to_sockptr_offset(sockptr_t dst, size_t offset,
 		const void *src, size_t size)
 {
include/net/tcp.h +6 -0
index d1230ee888fd..98b970c078d1 100644
@@ -2122,6 +2122,12 @@ struct tcp_sock_af_ops {
 				     sockptr_t optval,
 				     int optlen);
 #endif
+#ifdef CONFIG_TCP_AO
+	int			(*ao_parse)(struct sock *sk,
+					    int optname,
+					    sockptr_t optval,
+					    int optlen);
+#endif
 };
 
 struct tcp_request_sock_ops {
include/net/tcp_ao.h +17 -1
index 3870cc29af10..1739116c6902 100644
@@ -82,10 +82,26 @@ struct tcp_ao_info {
 	 */
 	struct tcp_ao_key	*current_key;
 	struct tcp_ao_key	*rnext_key;
-	u32			flags;
+	u32			ao_required	:1,
+				__unused	:31;
 	__be32			lisn;
 	__be32			risn;
 	struct rcu_head		rcu;
 };
 
+#ifdef CONFIG_TCP_AO
+int tcp_parse_ao(struct sock *sk, int cmd, unsigned short int family,
+		 sockptr_t optval, int optlen);
+void tcp_ao_destroy_sock(struct sock *sk);
+/* ipv4 specific functions */
+int tcp_v4_parse_ao(struct sock *sk, int optname, sockptr_t optval, int optlen);
+/* ipv6 specific functions */
+int tcp_v6_parse_ao(struct sock *sk, int cmd,
+		    sockptr_t optval, int optlen);
+#else
+static inline void tcp_ao_destroy_sock(struct sock *sk)
+{
+}
+#endif
+
 #endif /* _TCP_AO_H */
include/uapi/linux/tcp.h +46 -0

58 lines not on this page · raw message

net/ipv4/Makefile +1 -0
index cd760793cfcb..e144a02a6a61 100644
@@ -69,6 +69,7 @@ obj-$(CONFIG_NETLABEL) += cipso_ipv4.o
 
 obj-$(CONFIG_XFRM) += xfrm4_policy.o xfrm4_state.o xfrm4_input.o \
 		      xfrm4_output.o xfrm4_protocol.o
+obj-$(CONFIG_TCP_AO) += tcp_ao.o
 
 ifeq ($(CONFIG_BPF_JIT),y)
 obj-$(CONFIG_BPF_SYSCALL) += bpf_tcp_ca.o
net/ipv4/tcp.c +17 -0
index ca2c74c282ad..d0b0565fa11e 100644
@@ -3608,6 +3608,23 @@ int do_tcp_setsockopt(struct sock *sk, int level, int optname,
 		__tcp_sock_set_quickack(sk, val);
 		break;
 
+#ifdef CONFIG_TCP_AO
+	case TCP_AO_ADD_KEY:
+	case TCP_AO_DEL_KEY:
+	case TCP_AO_INFO: {
+		/* If this is the first TCP-AO setsockopt() on the socket,
+		 * sk_state has to be LISTEN or CLOSE
+		 */
+		if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) ||
+		    rcu_dereference_protected(tcp_sk(sk)->ao_info,
+					      lockdep_sock_is_held(sk)))
+			err = tp->af_specific->ao_parse(sk, optname, optval,
+							optlen);
+		else
+			err = -EISCONN;
+		break;
+	}
+#endif
 #ifdef CONFIG_TCP_MD5SIG
 	case TCP_MD5SIG:
 	case TCP_MD5SIG_EXT:
net/ipv4/tcp_ao.c new file +816 -0

816 lines not on this page · raw message

net/ipv4/tcp_ipv4.c +8 -2
index b3b8a11daa71..8804eb801caf 100644
@@ -2269,11 +2269,16 @@ const struct inet_connection_sock_af_ops ipv4_specific = {
 };
 EXPORT_SYMBOL(ipv4_specific);
 
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
+#ifdef CONFIG_TCP_MD5SIG
 	.md5_lookup		= tcp_v4_md5_lookup,
 	.calc_md5_hash		= tcp_v4_md5_hash_skb,
 	.md5_parse		= tcp_v4_parse_md5_keys,
+#endif
+#ifdef CONFIG_TCP_AO
+	.ao_parse		= tcp_v4_parse_ao,
+#endif
 };
 #endif
 
@@ -2288,7 +2293,7 @@ static int tcp_v4_init_sock(struct sock *sk)
 
 	icsk->icsk_af_ops = &ipv4_specific;
 
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 	tcp_sk(sk)->af_specific = &tcp_sock_ipv4_specific;
 #endif
 
@@ -2339,6 +2344,7 @@ void tcp_v4_destroy_sock(struct sock *sk)
 		rcu_assign_pointer(tp->md5sig_info, NULL);
 	}
 #endif
+	tcp_ao_destroy_sock(sk);
 
 	/* Clean up a referenced TCP bind bucket. */
 	if (inet_csk(sk)->icsk_bind_hash)
net/ipv6/Makefile +1 -0
index 3036a45e8a1e..d283c59df4c1 100644
@@ -52,4 +52,5 @@ obj-$(subst m,y,$(CONFIG_IPV6)) += inet6_hashtables.o
 ifneq ($(CONFIG_IPV6),)
 obj-$(CONFIG_NET_UDP_TUNNEL) += ip6_udp_tunnel.o
 obj-y += mcast_snoop.o
+obj-$(CONFIG_TCP_AO) += tcp_ao.o
 endif
net/ipv6/tcp_ao.c new file +19 -0
new file mode 100644
index 000000000000..049ddbabe049
@@ -0,0 +1,19 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * INET		An implementation of the TCP Authentication Option (TCP-AO).
+ *		See RFC5925.
+ *
+ * Authors:	Dmitry Safonov <[email protected]>
+ *		Francesco Ruggeri <[email protected]>
+ *		Salam Noureddine <[email protected]>
+ */
+#include <linux/tcp.h>
+
+#include <net/tcp.h>
+#include <net/ipv6.h>
+
+int tcp_v6_parse_ao(struct sock *sk, int cmd,
+		    sockptr_t optval, int optlen)
+{
+	return tcp_parse_ao(sk, cmd, AF_INET6, optval, optlen);
+}
net/ipv6/tcp_ipv6.c +24 -15
index afa8db712dfb..96ec9c03d9eb 100644
@@ -76,16 +76,9 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb);
 
 static const struct inet_connection_sock_af_ops ipv6_mapped;
 const struct inet_connection_sock_af_ops ipv6_specific;
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific;
 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific;
-#else
-static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
-						   const struct in6_addr *addr,
-						   int l3index)
-{
-	return NULL;
-}
 #endif
 
 /* Helper returning the inet6 address from a given tcp socket.
@@ -239,7 +232,7 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 		if (sk_is_mptcp(sk))
 			mptcpv6_handle_mapped(sk, true);
 		sk->sk_backlog_rcv = tcp_v4_do_rcv;
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 		tp->af_specific = &tcp_sock_ipv6_mapped_specific;
 #endif
 
@@ -252,7 +245,7 @@ static int tcp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 			if (sk_is_mptcp(sk))
 				mptcpv6_handle_mapped(sk, false);
 			sk->sk_backlog_rcv = tcp_v6_do_rcv;
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 			tp->af_specific = &tcp_sock_ipv6_specific;
 #endif
 			goto failure;
@@ -768,7 +761,13 @@ static int tcp_v6_md5_hash_skb(char *md5_hash,
 	memset(md5_hash, 0, 16);
 	return 1;
 }
-
+#else /* CONFIG_TCP_MD5SIG */
+static struct tcp_md5sig_key *tcp_v6_md5_do_lookup(const struct sock *sk,
+						   const struct in6_addr *addr,
+						   int l3index)
+{
+	return NULL;
+}
 #endif
 
 static void tcp_v6_init_req(struct request_sock *req,
@@ -1228,7 +1227,7 @@ static struct sock *tcp_v6_syn_recv_sock(const struct sock *sk, struct sk_buff *
 		if (sk_is_mptcp(newsk))
 			mptcpv6_handle_mapped(newsk, true);
 		newsk->sk_backlog_rcv = tcp_v4_do_rcv;
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 		newtp->af_specific = &tcp_sock_ipv6_mapped_specific;
 #endif
 
@@ -1892,11 +1891,16 @@ const struct inet_connection_sock_af_ops ipv6_specific = {
 	.mtu_reduced	   = tcp_v6_mtu_reduced,
 };
 
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
+#ifdef CONFIG_TCP_MD5SIG
 	.md5_lookup	=	tcp_v6_md5_lookup,
 	.calc_md5_hash	=	tcp_v6_md5_hash_skb,
 	.md5_parse	=	tcp_v6_parse_md5_keys,
+#endif
+#ifdef CONFIG_TCP_AO
+	.ao_parse	=	tcp_v6_parse_ao,
+#endif
 };
 #endif
 
@@ -1918,11 +1922,16 @@ static const struct inet_connection_sock_af_ops ipv6_mapped = {
 	.mtu_reduced	   = tcp_v4_mtu_reduced,
 };
 
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
+#ifdef CONFIG_TCP_MD5SIG
 	.md5_lookup	=	tcp_v4_md5_lookup,
 	.calc_md5_hash	=	tcp_v4_md5_hash_skb,
 	.md5_parse	=	tcp_v6_parse_md5_keys,
+#endif
+#ifdef CONFIG_TCP_AO
+	.ao_parse	=	tcp_v6_parse_ao,
+#endif
 };
 #endif
 
@@ -1937,7 +1946,7 @@ static int tcp_v6_init_sock(struct sock *sk)
 
 	icsk->icsk_af_ops = &ipv6_specific;
 
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 	tcp_sk(sk)->af_specific = &tcp_sock_ipv6_specific;
 #endif
 
Signature · 1 line
2.41.0
#7 lore raw

Re: [PATCH v8.1 net-next 03/23] net/tcp: Introduce TCP_AO setsockopt()s

On Fri, Jul 21, 2023 at 05:18:54PM +0100, Dmitry Safonov wrote:

...

Hi Dimitry,

include/linux/sockptr.h static inline int copy_from_sockptr(void *dst, sockptr_t src, size_t size) · line 59

Quoted 10 lines
include/linux/sockptr.h +2 -0
index bae5e2369b4f..307961b41541 100644
@@ -55,6 +55,29 @@ static inline int copy_from_sockptr(void *dst, sockptr_t src, size_t size)
 	return copy_from_sockptr_offset(dst, src, 0, size);
 }
 
+static inline int copy_struct_from_sockptr(void *dst, size_t ksize,
+		sockptr_t src, size_t usize)

The indentation of the two lines above is not correct, they should be aligned to the inside of the opening '(' on the preceding line.

In order to stop things being too far to the left, which is perhaps the intent of the current indention scheme, the return type of the function can be moved to it's own line.

static inline int
copy_struct_from_sockptr(void *dst, size_t ksize, sockptr_t src, size_t usize)

...

Quoted 1 line
include/net/tcp.h +0 -0

...

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+static inline int ipv4_prefix_cmp(const struct in_addr *addr1,
+				  const struct in_addr *addr2,
+				  unsigned int prefixlen)
+{
+	__be32 mask = inet_make_mask(prefixlen);
+
+	if ((addr1->s_addr & mask) == (addr2->s_addr & mask))
+		return 0;
+	return ((addr1->s_addr & mask) > (addr2->s_addr & mask)) ? 1 : -1;
+}

Above, '>' is operating on two big endian values. But typically such maths operates on host byte order values.

Flagged by Sparse.

...

include/net/tcp.h

Quoted 3 lines
include/net/tcp.h +3 -0
+static struct tcp_ao_key *__tcp_ao_do_lookup(const struct sock *sk,
+		const union tcp_ao_addr *addr, int family, u8 prefix,
+		int sndid, int rcvid, u16 port)

Same comment about indentation as above.

static struct tcp_ao_key *
__tcp_ao_do_lookup(const struct sock *sk, const union tcp_ao_addr *addr,
int family, u8 prefix, int sndid, int rcvid, u16 port)

...

include/net/tcp.h

Quoted 3 lines
include/net/tcp.h +3 -0
+struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
+				    const union tcp_ao_addr *addr,
+				    int family, int sndid, int rcvid, u16 port)

Should tcp_ao_do_lookup be static?
It seems to only be used in this file.

...

include/net/tcp.h

Quoted 5 lines
include/net/tcp.h +5 -0
+static int tcp_ao_verify_port(struct sock *sk, u16 port)
+{
+	struct inet_sock *inet = inet_sk(sk);
+
+	if (port != 0) /* FIXME */

I guess this should be fixed :)

include/net/tcp.h

Quoted 4 lines
include/net/tcp.h +4 -0
+		return -EINVAL;
+
+	/* Check that MKT port is consistent with socket */
+	if (port != 0 && inet->inet_dport != 0 && port != inet->inet_dport)

port is host byte order, but inet->inet_dport is big endian. This does not seem correct.

Quoted 4 lines
include/net/tcp.h +4 -0
+		return -EINVAL;
+
+	return 0;
+}

...

Quoted 90 lines
include/net/tcp.h +90 -0
+static int tcp_ao_parse_crypto(struct tcp_ao_add *cmd, struct tcp_ao_key *key)
+{
+	unsigned int syn_tcp_option_space;
+	bool is_kdf_aes_128_cmac = false;
+	struct crypto_ahash *tfm;
+	struct tcp_sigpool hp;
+	void *tmp_key = NULL;
+	int err;
+
+	/* RFC5926, 3.1.1.2. KDF_AES_128_CMAC */
+	if (!strcmp("cmac(aes128)", cmd->alg_name)) {
+		strscpy(cmd->alg_name, "cmac(aes)", sizeof(cmd->alg_name));
+		is_kdf_aes_128_cmac = (cmd->keylen != 16);
+		tmp_key = kmalloc(cmd->keylen, GFP_KERNEL);
+		if (!tmp_key)
+			return -ENOMEM;
+	}
+
+	key->maclen = cmd->maclen ?: 12; /* 12 is the default in RFC5925 */
+
+	/* Check: maclen + tcp-ao header <= (MAX_TCP_OPTION_SPACE - mss
+	 *					- tstamp - wscale - sackperm),
+	 * see tcp_syn_options(), tcp_synack_options(), commit 33ad798c924b.
+	 *
+	 * In order to allow D-SACK with TCP-AO, the header size should be:
+	 * (MAX_TCP_OPTION_SPACE - TCPOLEN_TSTAMP_ALIGNED
+	 *			- TCPOLEN_SACK_BASE_ALIGNED
+	 *			- 2 * TCPOLEN_SACK_PERBLOCK) = 8 (maclen = 4),
+	 * see tcp_established_options().
+	 *
+	 * RFC5925, 2.2:
+	 * Typical MACs are 96-128 bits (12-16 bytes), but any length
+	 * that fits in the header of the segment being authenticated
+	 * is allowed.
+	 *
+	 * RFC5925, 7.6:
+	 * TCP-AO continues to consume 16 bytes in non-SYN segments,
+	 * leaving a total of 24 bytes for other options, of which
+	 * the timestamp consumes 10.  This leaves 14 bytes, of which 10
+	 * are used for a single SACK block. When two SACK blocks are used,
+	 * such as to handle D-SACK, a smaller TCP-AO MAC would be required
+	 * to make room for the additional SACK block (i.e., to leave 18
+	 * bytes for the D-SACK variant of the SACK option) [RFC2883].
+	 * Note that D-SACK is not supportable in TCP MD5 in the presence
+	 * of timestamps, because TCP MD5’s MAC length is fixed and too
+	 * large to leave sufficient option space.
+	 */
+	syn_tcp_option_space = MAX_TCP_OPTION_SPACE;
+	syn_tcp_option_space -= TCPOLEN_TSTAMP_ALIGNED;
+	syn_tcp_option_space -= TCPOLEN_WSCALE_ALIGNED;
+	syn_tcp_option_space -= TCPOLEN_SACKPERM_ALIGNED;
+	if (tcp_ao_len(key) > syn_tcp_option_space) {
+		err = -EMSGSIZE;
+		goto err_kfree;
+	}
+
+	key->keylen = cmd->keylen;
+	memcpy(key->key, cmd->key, cmd->keylen);
+
+	err = tcp_sigpool_start(key->tcp_sigpool_id, &hp);
+	if (err)
+		goto err_kfree;
+
+	tfm = crypto_ahash_reqtfm(hp.req);
+	if (is_kdf_aes_128_cmac) {
+		void *scratch = hp.scratch;
+		struct scatterlist sg;
+
+		memcpy(tmp_key, cmd->key, cmd->keylen);
+		sg_init_one(&sg, tmp_key, cmd->keylen);
+
+		/* Using zero-key of 16 bytes as described in RFC5926 */
+		memset(scratch, 0, 16);
+		err = crypto_ahash_setkey(tfm, scratch, 16);
+		if (err)
+			goto err_pool_end;
+
+		err = crypto_ahash_init(hp.req);
+		if (err)
+			goto err_pool_end;
+
+		ahash_request_set_crypt(hp.req, &sg, key->key, cmd->keylen);
+		err = crypto_ahash_update(hp.req);
+		if (err)
+			goto err_pool_end;
+
+		err |= crypto_ahash_final(hp.req);
+		if (err)
+			goto err_pool_end;
+		key->keylen = 16;

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+	}
+
+	err = crypto_ahash_setkey(tfm, key->key, key->keylen);
+	if (err)
+		goto err_pool_end;
+
+	tcp_sigpool_end(&hp);
+
+	if (tcp_ao_maclen(key) > key->digest_size)
+		return -EINVAL;

tmp_key appears to be leaked here.

include/net/tcp.h

Quoted 2 lines
include/net/tcp.h +2 -0
+
+	return 0;

And here.

This is flagged by Smatch.

Quoted 7 lines
include/net/tcp.h +7 -0
+
+err_pool_end:
+	tcp_sigpool_end(&hp);
+err_kfree:
+	kfree(tmp_key);
+	return err;
+}

...

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+static int tcp_ao_add_cmd(struct sock *sk, unsigned short int family,
+			  sockptr_t optval, int optlen)
+{
+	struct tcp_ao_info *ao_info;
+	union tcp_ao_addr *addr;
+	struct tcp_ao_key *key;
+	struct tcp_ao_add cmd;
+	int ret;
+	bool first = false;
+	u16 port;

Please use reverse xmas tree - longest line to shortest - for local variable declarations in new Networking code.

...

Quoted 45 lines
include/net/tcp.h +45 -0
+static int tcp_ao_del_cmd(struct sock *sk, unsigned short int family,
+			  sockptr_t optval, int optlen)
+{
+	struct tcp_ao_key *key, *new_current = NULL, *new_rnext = NULL;
+	struct tcp_ao_info *ao_info;
+	union tcp_ao_addr *addr;
+	struct tcp_ao_del cmd;
+	int err;
+	__u8 prefix;
+	__be16 port;
+	int addr_len;
+
+	if (optlen < sizeof(cmd))
+		return -EINVAL;
+
+	err = copy_struct_from_sockptr(&cmd, sizeof(cmd), optval, optlen);
+	if (err)
+		return err;
+
+	if (cmd.reserved != 0 || cmd.reserved2 != 0)
+		return -EINVAL;
+
+	if (cmd.set_current || cmd.set_rnext) {
+		if (!tcp_ao_can_set_current_rnext(sk))
+			return -EINVAL;
+	}
+
+	ao_info = setsockopt_ao_info(sk);
+	if (IS_ERR(ao_info))
+		return PTR_ERR(ao_info);
+	if (!ao_info)
+		return -ENOENT;
+
+	/* For sockets in TCP_CLOSED it's possible set keys that aren't
+	 * matching the future peer (address/port/VRF/etc),
+	 * tcp_ao_connect_init() will choose a correct matching MKT
+	 * if there's any.
+	 */
+	if (cmd.set_current) {
+		new_current = tcp_ao_established_key(ao_info, cmd.current_key, -1);
+		if (!new_current)
+			return -ENOENT;
+	}
+	if (cmd.set_rnext) {
+		new_rnext = tcp_ao_established_key(ao_info, -1, cmd.rnext);

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+		if (!new_rnext)
+			return -ENOENT;
+	}
+
+	if (family == AF_INET) {
+		struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.addr;
+
+		addr = (union tcp_ao_addr *)&sin->sin_addr;
+		addr_len = sizeof(struct in_addr);
+		port = ntohs(sin->sin_port);

port is big endian, but here it is assigned a host-byte order value. It looks like port should be u16 rather than __bbe16.

As flagged by Smatch.

include/net/tcp.h

Quoted 13 lines
include/net/tcp.h +13 -0
+	} else {
+		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.addr;
+		struct in6_addr *addr6 = &sin6->sin6_addr;
+
+		if (ipv6_addr_v4mapped(addr6)) {
+			addr = (union tcp_ao_addr *)&addr6->s6_addr32[3];
+			addr_len = sizeof(struct in_addr);
+			family = AF_INET;
+		} else {
+			addr = (union tcp_ao_addr *)addr6;
+			addr_len = sizeof(struct in6_addr);
+		}
+		port = ntohs(sin6->sin6_port);

Ditto.

Quoted 7 lines
include/net/tcp.h +7 -0
+	}
+	prefix = cmd.prefix;
+
+	/* We could choose random present key here for current/rnext
+	 * but that's less predictable. Let's be strict and don't
+	 * allow removing a key that's in use. RFC5925 doesn't
+	 * specify how-to coordinate key removal, but says:

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+	 * "It is presumed that an MKT affecting a particular
+	 * connection cannot be destroyed during an active connection"
+	 */
+	hlist_for_each_entry_rcu(key, &ao_info->head, node) {
+		if (cmd.sndid != key->sndid ||
+		    cmd.rcvid != key->rcvid)
+			continue;
+
+		if (family != key->family ||
+		    prefix != key->prefixlen ||

o

include/net/tcp.h

Quoted 1 line
include/net/tcp.h +1 -0
+		    port != key->port ||

There is a similar problem here too.
port is host byte order but key->port is big endian.

Quoted 11 lines
include/net/tcp.h +11 -0
+		    memcmp(addr, &key->addr, addr_len))
+			continue;
+
+		if (key == new_current || key == new_rnext)
+			continue;
+
+		return tcp_ao_delete_key(sk, ao_info, key,
+					  new_current, new_rnext);
+	}
+	return -ENOENT;
+}

...

#8 lore raw

Re: [PATCH v8.1 net-next 03/23] net/tcp: Introduce TCP_AO setsockopt()s

Hi Simon,

Quoted 21 lines · Simon Horman

On Fri, Jul 21, 2023 at 05:18:54PM +0100, Dmitry Safonov wrote:

...

Hi Dimitry,

include/linux/sockptr.h static inline int copy_from_sockptr(void *dst, sockptr_t src, size_t size) · line 59

Quoted 10 lines
include/linux/sockptr.h +2 -0
index bae5e2369b4f..307961b41541 100644
@@ -55,6 +55,29 @@ static inline int copy_from_sockptr(void *dst, sockptr_t src, size_t size)
 	return copy_from_sockptr_offset(dst, src, 0, size);
 }
 
+static inline int copy_struct_from_sockptr(void *dst, size_t ksize,
+		sockptr_t src, size_t usize)

The indentation of the two lines above is not correct, they should be aligned to the inside of the opening '(' on the preceding line.

Quoted 6 lines

In order to stop things being too far to the left, which is perhaps the intent of the current indention scheme, the return type of the function can be moved to it's own line.

static inline int
copy_struct_from_sockptr(void *dst, size_t ksize, sockptr_t src, size_t usize)

Well, that would be a bit more GNU coding-style alike. Which I don't mind, I can do that. Albeit it's a bit contrary to an example from kernel's coding-style, where it seems preferred to keep it on the same line with function name and rather not to indent argument list, see (6.1), second example with action().

Yet, I don't feel particularly strong on either of options, so I can just do as you suggest.

Quoted 21 lines

...

Quoted 1 line
include/net/tcp.h +0 -0

...

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+static inline int ipv4_prefix_cmp(const struct in_addr *addr1,
+				  const struct in_addr *addr2,
+				  unsigned int prefixlen)
+{
+	__be32 mask = inet_make_mask(prefixlen);
+
+	if ((addr1->s_addr & mask) == (addr2->s_addr & mask))
+		return 0;
+	return ((addr1->s_addr & mask) > (addr2->s_addr & mask)) ? 1 : -1;
+}

Above, '>' is operating on two big endian values. But typically such maths operates on host byte order values.

Quoted 1 line

Flagged by Sparse.

Yeah, the function just has to provide any way to compare keys. So, it's not very important, but just to silence Sparse I can convert them to host's byte order before the comparison.

Quoted 13 lines

...

include/net/tcp.h

Quoted 3 lines
include/net/tcp.h +3 -0
+static struct tcp_ao_key *__tcp_ao_do_lookup(const struct sock *sk,
+		const union tcp_ao_addr *addr, int family, u8 prefix,
+		int sndid, int rcvid, u16 port)

Same comment about indentation as above.

static struct tcp_ao_key *
__tcp_ao_do_lookup(const struct sock *sk, const union tcp_ao_addr *addr,
int family, u8 prefix, int sndid, int rcvid, u16 port)

Quoted 8 lines

...

include/net/tcp.h

Quoted 3 lines
include/net/tcp.h +3 -0
+struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
+				    const union tcp_ao_addr *addr,
+				    int family, int sndid, int rcvid, u16 port)

Should tcp_ao_do_lookup be static?
It seems to only be used in this file.

Yeah, indeed. I think, I noticed previously, but probably managed to forget. Will fix.

Quoted 10 lines

...

include/net/tcp.h

Quoted 5 lines
include/net/tcp.h +5 -0
+static int tcp_ao_verify_port(struct sock *sk, u16 port)
+{
+	struct inet_sock *inet = inet_sk(sk);
+
+	if (port != 0) /* FIXME */

I guess this should be fixed :)

Fair enough. I think, what I'll do is to remove from these initial patches TCP-port from uAPI: we've expected that it will be useful to implement port-matching, but so far none from customers requested it. So, it was left as reserved member in uAPI, not meant to be used just yet.

Separately, as I've made UAPI structures for setsockopt() extendable, see copy_struct_from_sockptr() and the extendable syscall ideas (unfortunately, not in Documentation/):
https://lpc.events/event/7/contributions/657/attachments/639/1159/extensible_syscalls.pdf
https://lwn.net/Articles/830666/

So, as those structs can be extended in future, it won't be any hard to add port-matching on the top of the patch set. RFC5925 is permissive on how IP address and TCP-port matching may be performed:

: TCP connection identifier. A TCP socket pair, i.e., a local IP : address, a remote IP address, a TCP local port, and a TCP remote port. : Values can be partially specified using ranges (e.g., 2-30), masks : (e.g., 0xF0), wildcards (e.g., "*"), or any other suitable indication.

I can see some utility of TCP-AO key port-range matching and it seems most useful/flexible, so I'll add that. Unsure if that will go in version 9 or rather later (even post-merge).

I probably have to add something on that mater to
Documentation/networking/tcp_ao.rst as well.

Quoted 7 lines

include/net/tcp.h

Quoted 4 lines
include/net/tcp.h +4 -0
+		return -EINVAL;
+
+	/* Check that MKT port is consistent with socket */
+	if (port != 0 && inet->inet_dport != 0 && port != inet->inet_dport)

port is host byte order, but inet->inet_dport is big endian. This does not seem correct.

Thanks.

Quoted 111 lines
Quoted 5 lines
include/net/tcp.h +4 -0
+		return -EINVAL;
+
+	return 0;
+}

...

Quoted 90 lines
include/net/tcp.h +90 -0
+static int tcp_ao_parse_crypto(struct tcp_ao_add *cmd, struct tcp_ao_key *key)
+{
+	unsigned int syn_tcp_option_space;
+	bool is_kdf_aes_128_cmac = false;
+	struct crypto_ahash *tfm;
+	struct tcp_sigpool hp;
+	void *tmp_key = NULL;
+	int err;
+
+	/* RFC5926, 3.1.1.2. KDF_AES_128_CMAC */
+	if (!strcmp("cmac(aes128)", cmd->alg_name)) {
+		strscpy(cmd->alg_name, "cmac(aes)", sizeof(cmd->alg_name));
+		is_kdf_aes_128_cmac = (cmd->keylen != 16);
+		tmp_key = kmalloc(cmd->keylen, GFP_KERNEL);
+		if (!tmp_key)
+			return -ENOMEM;
+	}
+
+	key->maclen = cmd->maclen ?: 12; /* 12 is the default in RFC5925 */
+
+	/* Check: maclen + tcp-ao header <= (MAX_TCP_OPTION_SPACE - mss
+	 *					- tstamp - wscale - sackperm),
+	 * see tcp_syn_options(), tcp_synack_options(), commit 33ad798c924b.
+	 *
+	 * In order to allow D-SACK with TCP-AO, the header size should be:
+	 * (MAX_TCP_OPTION_SPACE - TCPOLEN_TSTAMP_ALIGNED
+	 *			- TCPOLEN_SACK_BASE_ALIGNED
+	 *			- 2 * TCPOLEN_SACK_PERBLOCK) = 8 (maclen = 4),
+	 * see tcp_established_options().
+	 *
+	 * RFC5925, 2.2:
+	 * Typical MACs are 96-128 bits (12-16 bytes), but any length
+	 * that fits in the header of the segment being authenticated
+	 * is allowed.
+	 *
+	 * RFC5925, 7.6:
+	 * TCP-AO continues to consume 16 bytes in non-SYN segments,
+	 * leaving a total of 24 bytes for other options, of which
+	 * the timestamp consumes 10.  This leaves 14 bytes, of which 10
+	 * are used for a single SACK block. When two SACK blocks are used,
+	 * such as to handle D-SACK, a smaller TCP-AO MAC would be required
+	 * to make room for the additional SACK block (i.e., to leave 18
+	 * bytes for the D-SACK variant of the SACK option) [RFC2883].
+	 * Note that D-SACK is not supportable in TCP MD5 in the presence
+	 * of timestamps, because TCP MD5’s MAC length is fixed and too
+	 * large to leave sufficient option space.
+	 */
+	syn_tcp_option_space = MAX_TCP_OPTION_SPACE;
+	syn_tcp_option_space -= TCPOLEN_TSTAMP_ALIGNED;
+	syn_tcp_option_space -= TCPOLEN_WSCALE_ALIGNED;
+	syn_tcp_option_space -= TCPOLEN_SACKPERM_ALIGNED;
+	if (tcp_ao_len(key) > syn_tcp_option_space) {
+		err = -EMSGSIZE;
+		goto err_kfree;
+	}
+
+	key->keylen = cmd->keylen;
+	memcpy(key->key, cmd->key, cmd->keylen);
+
+	err = tcp_sigpool_start(key->tcp_sigpool_id, &hp);
+	if (err)
+		goto err_kfree;
+
+	tfm = crypto_ahash_reqtfm(hp.req);
+	if (is_kdf_aes_128_cmac) {
+		void *scratch = hp.scratch;
+		struct scatterlist sg;
+
+		memcpy(tmp_key, cmd->key, cmd->keylen);
+		sg_init_one(&sg, tmp_key, cmd->keylen);
+
+		/* Using zero-key of 16 bytes as described in RFC5926 */
+		memset(scratch, 0, 16);
+		err = crypto_ahash_setkey(tfm, scratch, 16);
+		if (err)
+			goto err_pool_end;
+
+		err = crypto_ahash_init(hp.req);
+		if (err)
+			goto err_pool_end;
+
+		ahash_request_set_crypt(hp.req, &sg, key->key, cmd->keylen);
+		err = crypto_ahash_update(hp.req);
+		if (err)
+			goto err_pool_end;
+
+		err |= crypto_ahash_final(hp.req);
+		if (err)
+			goto err_pool_end;
+		key->keylen = 16;

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+	}
+
+	err = crypto_ahash_setkey(tfm, key->key, key->keylen);
+	if (err)
+		goto err_pool_end;
+
+	tcp_sigpool_end(&hp);
+
+	if (tcp_ao_maclen(key) > key->digest_size)
+		return -EINVAL;

tmp_key appears to be leaked here.

Quoted 6 lines

include/net/tcp.h

Quoted 2 lines
include/net/tcp.h +2 -0
+
+	return 0;

And here.

This is flagged by Smatch.

Ouch. Yeah, the change from v7 that got rid of TCP_AO_MAX_HASH_SIZE and allocated traffic keys with kmalloc(), managed to add a leak as well. Thanks, will fix.

Quoted 11 lines
Quoted 8 lines
include/net/tcp.h +7 -0
+
+err_pool_end:
+	tcp_sigpool_end(&hp);
+err_kfree:
+	kfree(tmp_key);
+	return err;
+}

...

Quoted 13 lines

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+static int tcp_ao_add_cmd(struct sock *sk, unsigned short int family,
+			  sockptr_t optval, int optlen)
+{
+	struct tcp_ao_info *ao_info;
+	union tcp_ao_addr *addr;
+	struct tcp_ao_key *key;
+	struct tcp_ao_add cmd;
+	int ret;
+	bool first = false;
+	u16 port;

Please use reverse xmas tree - longest line to shortest - for local variable declarations in new Networking code.

Will do.

Quoted 101 lines
Quoted 46 lines
include/net/tcp.h +45 -0
+static int tcp_ao_del_cmd(struct sock *sk, unsigned short int family,
+			  sockptr_t optval, int optlen)
+{
+	struct tcp_ao_key *key, *new_current = NULL, *new_rnext = NULL;
+	struct tcp_ao_info *ao_info;
+	union tcp_ao_addr *addr;
+	struct tcp_ao_del cmd;
+	int err;
+	__u8 prefix;
+	__be16 port;
+	int addr_len;
+
+	if (optlen < sizeof(cmd))
+		return -EINVAL;
+
+	err = copy_struct_from_sockptr(&cmd, sizeof(cmd), optval, optlen);
+	if (err)
+		return err;
+
+	if (cmd.reserved != 0 || cmd.reserved2 != 0)
+		return -EINVAL;
+
+	if (cmd.set_current || cmd.set_rnext) {
+		if (!tcp_ao_can_set_current_rnext(sk))
+			return -EINVAL;
+	}
+
+	ao_info = setsockopt_ao_info(sk);
+	if (IS_ERR(ao_info))
+		return PTR_ERR(ao_info);
+	if (!ao_info)
+		return -ENOENT;
+
+	/* For sockets in TCP_CLOSED it's possible set keys that aren't
+	 * matching the future peer (address/port/VRF/etc),
+	 * tcp_ao_connect_init() will choose a correct matching MKT
+	 * if there's any.
+	 */
+	if (cmd.set_current) {
+		new_current = tcp_ao_established_key(ao_info, cmd.current_key, -1);
+		if (!new_current)
+			return -ENOENT;
+	}
+	if (cmd.set_rnext) {
+		new_rnext = tcp_ao_established_key(ao_info, -1, cmd.rnext);

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+		if (!new_rnext)
+			return -ENOENT;
+	}
+
+	if (family == AF_INET) {
+		struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.addr;
+
+		addr = (union tcp_ao_addr *)&sin->sin_addr;
+		addr_len = sizeof(struct in_addr);
+		port = ntohs(sin->sin_port);

port is big endian, but here it is assigned a host-byte order value. It looks like port should be u16 rather than __bbe16.

As flagged by Smatch.

include/net/tcp.h

Quoted 13 lines
include/net/tcp.h +13 -0
+	} else {
+		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.addr;
+		struct in6_addr *addr6 = &sin6->sin6_addr;
+
+		if (ipv6_addr_v4mapped(addr6)) {
+			addr = (union tcp_ao_addr *)&addr6->s6_addr32[3];
+			addr_len = sizeof(struct in_addr);
+			family = AF_INET;
+		} else {
+			addr = (union tcp_ao_addr *)addr6;
+			addr_len = sizeof(struct in6_addr);
+		}
+		port = ntohs(sin6->sin6_port);

Ditto.

Quoted 7 lines
include/net/tcp.h +7 -0
+	}
+	prefix = cmd.prefix;
+
+	/* We could choose random present key here for current/rnext
+	 * but that's less predictable. Let's be strict and don't
+	 * allow removing a key that's in use. RFC5925 doesn't
+	 * specify how-to coordinate key removal, but says:

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+	 * "It is presumed that an MKT affecting a particular
+	 * connection cannot be destroyed during an active connection"
+	 */
+	hlist_for_each_entry_rcu(key, &ao_info->head, node) {
+		if (cmd.sndid != key->sndid ||
+		    cmd.rcvid != key->rcvid)
+			continue;
+
+		if (family != key->family ||
+		    prefix != key->prefixlen ||

o

include/net/tcp.h

Quoted 1 line
include/net/tcp.h +1 -0
+		    port != key->port ||

There is a similar problem here too.
port is host byte order but key->port is big endian.

Quoted 13 lines
Quoted 11 lines
include/net/tcp.h +11 -0
+		    memcmp(addr, &key->addr, addr_len))
+			continue;
+
+		if (key == new_current || key == new_rnext)
+			continue;
+
+		return tcp_ao_delete_key(sk, ao_info, key,
+					  new_current, new_rnext);
+	}
+	return -ENOENT;
+}

...

Thanks,
Dmitry

#9 lore raw

Re: [PATCH v8.1 net-next 03/23] net/tcp: Introduce TCP_AO setsockopt()s

Hi Dimitry,

Quoted 31 lines · Dmitry Safonov

Hi Simon,

Quoted 21 lines · Simon Horman

On Fri, Jul 21, 2023 at 05:18:54PM +0100, Dmitry Safonov wrote:

...

Hi Dimitry,

include/linux/sockptr.h static inline int copy_from_sockptr(void *dst, sockptr_t src, size_t size) · line 59

Quoted 10 lines
include/linux/sockptr.h +2 -0
index bae5e2369b4f..307961b41541 100644
@@ -55,6 +55,29 @@ static inline int copy_from_sockptr(void *dst, sockptr_t src, size_t size)
 	return copy_from_sockptr_offset(dst, src, 0, size);
 }
 
+static inline int copy_struct_from_sockptr(void *dst, size_t ksize,
+		sockptr_t src, size_t usize)

The indentation of the two lines above is not correct, they should be aligned to the inside of the opening '(' on the preceding line.

Quoted 6 lines

In order to stop things being too far to the left, which is perhaps the intent of the current indention scheme, the return type of the function can be moved to it's own line.

static inline int
copy_struct_from_sockptr(void *dst, size_t ksize, sockptr_t src, size_t usize)

Quoted 8 lines

Well, that would be a bit more GNU coding-style alike. Which I don't mind, I can do that. Albeit it's a bit contrary to an example from kernel's coding-style, where it seems preferred to keep it on the same line with function name and rather not to indent argument list, see (6.1), second example with action().

Yet, I don't feel particularly strong on either of options, so I can just do as you suggest.

For some reason I thought the style I suggested is acceptable, at least in (some) Networking code.

In any case, I also don't feel strongly about this.

Quoted 20 lines
Quoted 20 lines

...

Quoted 1 line
include/net/tcp.h +0 -0

...

include/net/tcp.h

Quoted 10 lines
include/net/tcp.h +10 -0
+static inline int ipv4_prefix_cmp(const struct in_addr *addr1,
+				  const struct in_addr *addr2,
+				  unsigned int prefixlen)
+{
+	__be32 mask = inet_make_mask(prefixlen);
+
+	if ((addr1->s_addr & mask) == (addr2->s_addr & mask))
+		return 0;
+	return ((addr1->s_addr & mask) > (addr2->s_addr & mask)) ? 1 : -1;
+}

Above, '>' is operating on two big endian values. But typically such maths operates on host byte order values.

Quoted 5 lines
Quoted 1 line

Flagged by Sparse.

Yeah, the function just has to provide any way to compare keys. So, it's not very important, but just to silence Sparse I can convert them to host's byte order before the comparison.

If you just care about equality, then perhaps you can use an equality operator and avoid converting the endieness.

But, unless I am mistaken, '<' will give the wrong answer a little-endian host. And this feels, well ..., wrong.

Quoted 35 lines
Quoted 12 lines

...

include/net/tcp.h

Quoted 3 lines
include/net/tcp.h +3 -0
+static struct tcp_ao_key *__tcp_ao_do_lookup(const struct sock *sk,
+		const union tcp_ao_addr *addr, int family, u8 prefix,
+		int sndid, int rcvid, u16 port)

Same comment about indentation as above.

static struct tcp_ao_key *
__tcp_ao_do_lookup(const struct sock *sk, const union tcp_ao_addr *addr,
int family, u8 prefix, int sndid, int rcvid, u16 port)

Quoted 8 lines

...

include/net/tcp.h

Quoted 3 lines
include/net/tcp.h +3 -0
+struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
+				    const union tcp_ao_addr *addr,
+				    int family, int sndid, int rcvid, u16 port)

Should tcp_ao_do_lookup be static?
It seems to only be used in this file.

Yeah, indeed. I think, I noticed previously, but probably managed to forget. Will fix.

Quoted 10 lines

...

include/net/tcp.h

Quoted 5 lines
include/net/tcp.h +5 -0
+static int tcp_ao_verify_port(struct sock *sk, u16 port)
+{
+	struct inet_sock *inet = inet_sk(sk);
+
+	if (port != 0) /* FIXME */

I guess this should be fixed :)

Quoted 4 lines

Fair enough. I think, what I'll do is to remove from these initial patches TCP-port from uAPI: we've expected that it will be useful to implement port-matching, but so far none from customers requested it. So, it was left as reserved member in uAPI, not meant to be used just yet.

Sounds good to me.
Thanks.

Quoted 31 lines

Separately, as I've made UAPI structures for setsockopt() extendable, see copy_struct_from_sockptr() and the extendable syscall ideas (unfortunately, not in Documentation/):
https://lpc.events/event/7/contributions/657/attachments/639/1159/extensible_syscalls.pdf
https://lwn.net/Articles/830666/

So, as those structs can be extended in future, it won't be any hard to add port-matching on the top of the patch set. RFC5925 is permissive on how IP address and TCP-port matching may be performed:

: TCP connection identifier. A TCP socket pair, i.e., a local IP : address, a remote IP address, a TCP local port, and a TCP remote port. : Values can be partially specified using ranges (e.g., 2-30), masks : (e.g., 0xF0), wildcards (e.g., "*"), or any other suitable indication.

I can see some utility of TCP-AO key port-range matching and it seems most useful/flexible, so I'll add that. Unsure if that will go in version 9 or rather later (even post-merge).

I probably have to add something on that mater to
Documentation/networking/tcp_ao.rst as well.

Quoted 7 lines

include/net/tcp.h

Quoted 4 lines
include/net/tcp.h +4 -0
+		return -EINVAL;
+
+	/* Check that MKT port is consistent with socket */
+	if (port != 0 && inet->inet_dport != 0 && port != inet->inet_dport)

port is host byte order, but inet->inet_dport is big endian. This does not seem correct.

Thanks.

...

#10 lore raw

PATCH 4/23 net/tcp: Prevent TCP-MD5 with TCP-AO being set

In reply to Dmitry Safonov

Be as conservative as possible: if there is TCP-MD5 key for a given peer regardless of L3 interface - don't allow setting TCP-AO key for the same peer. According to RFC5925, TCP-AO is supposed to replace TCP-MD5 and there can't be any switch between both on any connected tuple. Later it can be relaxed, if there's a use, but in the beginning restrict any intersection.

Note: it's still should be possible to set both TCP-MD5 and TCP-AO keys on a listening socket for *different* peers.

---
 include/net/tcp.h     | 43 +++++++++++++++++++++++++++++++++++++--
 include/net/tcp_ao.h  | 15 ++++++++++++++
 net/ipv4/tcp_ao.c     | 40 ++++++++++++++++++++++++++++++++++++
 net/ipv4/tcp_ipv4.c   | 14 ++++++++++---
 net/ipv4/tcp_output.c | 47 +++++++++++++++++++++++++++++++++++++++++++
 net/ipv6/tcp_ao.c     | 17 ++++++++++++++++
 net/ipv6/tcp_ipv6.c   | 26 ++++++++++++++++++++----
 7 files changed, 193 insertions(+), 9 deletions(-)
include/net/tcp.h +41 -2
index 98b970c078d1..7e414873ef07 100644
@@ -1729,6 +1729,7 @@ int tcp_md5_key_copy(struct sock *sk, const union tcp_md5_addr *addr,
 
 int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr,
 		   int family, u8 prefixlen, int l3index, u8 flags);
+void tcp_clear_md5_list(struct sock *sk);
 struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
 					 const struct sock *addr_sk);
 
@@ -1737,14 +1738,23 @@ struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
 extern struct static_key_false_deferred tcp_md5_needed;
 struct tcp_md5sig_key *__tcp_md5_do_lookup(const struct sock *sk, int l3index,
 					   const union tcp_md5_addr *addr,
-					   int family);
+					   int family, bool any_l3index);
 static inline struct tcp_md5sig_key *
 tcp_md5_do_lookup(const struct sock *sk, int l3index,
 		  const union tcp_md5_addr *addr, int family)
 {
 	if (!static_branch_unlikely(&tcp_md5_needed.key))
 		return NULL;
-	return __tcp_md5_do_lookup(sk, l3index, addr, family);
+	return __tcp_md5_do_lookup(sk, l3index, addr, family, false);
+}
+
+static inline struct tcp_md5sig_key *
+tcp_md5_do_lookup_any_l3index(const struct sock *sk,
+			      const union tcp_md5_addr *addr, int family)
+{
+	if (!static_branch_unlikely(&tcp_md5_needed.key))
+		return NULL;
+	return __tcp_md5_do_lookup(sk, 0, addr, family, true);
 }
 
 enum skb_drop_reason
@@ -1762,6 +1772,13 @@ tcp_md5_do_lookup(const struct sock *sk, int l3index,
 	return NULL;
 }
 
+static inline struct tcp_md5sig_key *
+tcp_md5_do_lookup_any_l3index(const struct sock *sk,
+			      const union tcp_md5_addr *addr, int family)
+{
+	return NULL;
+}
+
 static inline enum skb_drop_reason
 tcp_inbound_md5_hash(const struct sock *sk, const struct sk_buff *skb,
 		     const void *saddr, const void *daddr,
@@ -2127,6 +2144,9 @@ struct tcp_sock_af_ops {
 					    int optname,
 					    sockptr_t optval,
 					    int optlen);
+	struct tcp_ao_key	*(*ao_lookup)(const struct sock *sk,
+					      struct sock  *addr_sk,
+					      int sndid, int rcvid);
 #endif
 };
 
@@ -2539,4 +2559,23 @@ static inline u64 tcp_transmit_time(const struct sock *sk)
 	return 0;
 }
 
+static inline bool tcp_ao_required(struct sock *sk, const void *saddr,
+				   int family)
+{
+#ifdef CONFIG_TCP_AO
+	struct tcp_ao_info *ao_info;
+	struct tcp_ao_key *ao_key;
+
+	ao_info = rcu_dereference_check(tcp_sk(sk)->ao_info,
+					lockdep_sock_is_held(sk));
+	if (!ao_info)
+		return false;
+
+	ao_key = tcp_ao_do_lookup(sk, saddr, family, -1, -1, 0);
+	if (ao_info->ao_required || ao_key)
+		return true;
+#endif
+	return false;
+}
+
 #endif	/* _TCP_H */
include/net/tcp_ao.h +15 -0
index 1739116c6902..1c24b586c876 100644
@@ -93,12 +93,27 @@ struct tcp_ao_info {
 int tcp_parse_ao(struct sock *sk, int cmd, unsigned short int family,
 		 sockptr_t optval, int optlen);
 void tcp_ao_destroy_sock(struct sock *sk);
+struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
+				    const union tcp_ao_addr *addr,
+				    int family, int sndid, int rcvid, u16 port);
 /* ipv4 specific functions */
 int tcp_v4_parse_ao(struct sock *sk, int optname, sockptr_t optval, int optlen);
+struct tcp_ao_key *tcp_v4_ao_lookup(const struct sock *sk, struct sock *addr_sk,
+				    int sndid, int rcvid);
 /* ipv6 specific functions */
+struct tcp_ao_key *tcp_v6_ao_lookup(const struct sock *sk,
+				    struct sock *addr_sk,
+				    int sndid, int rcvid);
 int tcp_v6_parse_ao(struct sock *sk, int cmd,
 		    sockptr_t optval, int optlen);
 #else
+static inline struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
+		const union tcp_ao_addr *addr,
+		int family, int sndid, int rcvid, u16 port)
+{
+	return NULL;
+}
+
 static inline void tcp_ao_destroy_sock(struct sock *sk)
 {
 }
net/ipv4/tcp_ao.c +40 -0
index 2b2696e24dcc..c2be72beab70 100644
@@ -171,6 +171,14 @@ void tcp_ao_destroy_sock(struct sock *sk)
 	kfree_rcu(ao, rcu);
 }
 
+struct tcp_ao_key *tcp_v4_ao_lookup(const struct sock *sk, struct sock *addr_sk,
+				    int sndid, int rcvid)
+{
+	union tcp_ao_addr *addr = (union tcp_ao_addr *)&addr_sk->sk_daddr;
+
+	return tcp_ao_do_lookup(sk, addr, AF_INET, sndid, rcvid, 0);
+}
+
 static bool tcp_ao_can_set_current_rnext(struct sock *sk)
 {
 	/* There aren't current/rnext keys on TCP_LISTEN sockets */
@@ -521,6 +529,10 @@ static int tcp_ao_add_cmd(struct sock *sk, unsigned short int family,
 			return -EINVAL;
 	}
 
+	/* Don't allow keys for peers that have a matching TCP-MD5 key */
+	if (tcp_md5_do_lookup_any_l3index(sk, addr, family))
+		return -EKEYREJECTED;
+
 	ao_info = setsockopt_ao_info(sk);
 	if (IS_ERR(ao_info))
 		return PTR_ERR(ao_info);
@@ -720,6 +732,31 @@ static int tcp_ao_del_cmd(struct sock *sk, unsigned short int family,
 	return -ENOENT;
 }
 
+/* cmd.ao_required makes a socket TCP-AO only.
+ * Don't allow any md5 keys for any l3intf on the socket together with it.
+ * Restricting it early in setsockopt() removes a check for
+ * ao_info->ao_required on inbound tcp segment fast-path.
+ */
+static inline int tcp_ao_required_verify(struct sock *sk)
+{
+#ifdef CONFIG_TCP_MD5SIG
+	const struct tcp_md5sig_info *md5sig;
+
+	if (!static_branch_unlikely(&tcp_md5_needed.key))
+		return 0;
+
+	md5sig = rcu_dereference_check(tcp_sk(sk)->md5sig_info,
+				       lockdep_sock_is_held(sk));
+	if (!md5sig)
+		return 0;
+
+	if (rcu_dereference_check(hlist_first_rcu(&md5sig->head),
+				  lockdep_sock_is_held(sk)))
+		return 1;
+#endif
+	return 0;
+}
+
 static int tcp_ao_info_cmd(struct sock *sk, unsigned short int family,
 			   sockptr_t optval, int optlen)
 {
@@ -754,6 +791,9 @@ static int tcp_ao_info_cmd(struct sock *sk, unsigned short int family,
 		first = true;
 	}
 
+	if (cmd.ao_required && tcp_ao_required_verify(sk))
+		return -EKEYREJECTED;
+
 	/* For sockets in TCP_CLOSED it's possible set keys that aren't
 	 * matching the future peer (address/port/VRF/etc),
 	 * tcp_ao_connect_init() will choose a correct matching MKT
net/ipv4/tcp_ipv4.c +11 -3
index 8804eb801caf..65e9bcac45f9 100644
@@ -1081,7 +1081,7 @@ static bool better_md5_match(struct tcp_md5sig_key *old, struct tcp_md5sig_key *
 /* Find the Key structure for an address.  */
 struct tcp_md5sig_key *__tcp_md5_do_lookup(const struct sock *sk, int l3index,
 					   const union tcp_md5_addr *addr,
-					   int family)
+					   int family, bool any_l3index)
 {
 	const struct tcp_sock *tp = tcp_sk(sk);
 	struct tcp_md5sig_key *key;
@@ -1100,7 +1100,8 @@ struct tcp_md5sig_key *__tcp_md5_do_lookup(const struct sock *sk, int l3index,
 				 lockdep_sock_is_held(sk)) {
 		if (key->family != family)
 			continue;
-		if (key->flags & TCP_MD5SIG_FLAG_IFINDEX && key->l3index != l3index)
+		if (!any_l3index && key->flags & TCP_MD5SIG_FLAG_IFINDEX &&
+		    key->l3index != l3index)
 			continue;
 		if (family == AF_INET) {
 			mask = inet_make_mask(key->prefixlen);
@@ -1312,7 +1313,7 @@ int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr, int family,
 }
 EXPORT_SYMBOL(tcp_md5_do_del);
 
-static void tcp_clear_md5_list(struct sock *sk)
+void tcp_clear_md5_list(struct sock *sk)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct tcp_md5sig_key *key;
@@ -1382,6 +1383,12 @@ static int tcp_v4_parse_md5_keys(struct sock *sk, int optname,
 	if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
 		return -EINVAL;
 
+	/* Don't allow keys for peers that have a matching TCP-AO key.
+	 * See the comment in tcp_ao_add_cmd()
+	 */
+	if (tcp_ao_required(sk, addr, AF_INET))
+		return -EKEYREJECTED;
+
 	return tcp_md5_do_add(sk, addr, AF_INET, prefixlen, l3index, flags,
 			      cmd.tcpm_key, cmd.tcpm_keylen);
 }
@@ -2277,6 +2284,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
 	.md5_parse		= tcp_v4_parse_md5_keys,
 #endif
 #ifdef CONFIG_TCP_AO
+	.ao_lookup		= tcp_v4_ao_lookup,
 	.ao_parse		= tcp_v4_parse_ao,
 #endif
 };
net/ipv4/tcp_output.c +47 -0

53 lines not on this page · raw message

net/ipv6/tcp_ao.c +17 -0
index 049ddbabe049..a56ab0a9cf55 100644
@@ -12,6 +12,23 @@
 #include <net/tcp.h>
 #include <net/ipv6.h>
 
+struct tcp_ao_key *tcp_v6_ao_do_lookup(const struct sock *sk,
+				       const struct in6_addr *addr,
+				       int sndid, int rcvid)
+{
+	return tcp_ao_do_lookup(sk, (union tcp_ao_addr *)addr, AF_INET6,
+				sndid, rcvid, 0);
+}
+
+struct tcp_ao_key *tcp_v6_ao_lookup(const struct sock *sk,
+				    struct sock *addr_sk,
+				    int sndid, int rcvid)
+{
+	struct in6_addr *addr = &addr_sk->sk_v6_daddr;
+
+	return tcp_v6_ao_do_lookup(sk, addr, sndid, rcvid);
+}
+
 int tcp_v6_parse_ao(struct sock *sk, int cmd,
 		    sockptr_t optval, int optlen)
 {
net/ipv6/tcp_ipv6.c +22 -4
index 96ec9c03d9eb..5c7c9dbfdb73 100644
@@ -599,6 +599,7 @@ static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 {
 	struct tcp_md5sig cmd;
 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&cmd.tcpm_addr;
+	union tcp_ao_addr *addr;
 	int l3index = 0;
 	u8 prefixlen;
 	u8 flags;
@@ -653,13 +654,28 @@ static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 	if (cmd.tcpm_keylen > TCP_MD5SIG_MAXKEYLEN)
 		return -EINVAL;
 
-	if (ipv6_addr_v4mapped(&sin6->sin6_addr))
-		return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3],
+	if (ipv6_addr_v4mapped(&sin6->sin6_addr)) {
+		addr = (union tcp_md5_addr *)&sin6->sin6_addr.s6_addr32[3];
+
+		/* Don't allow keys for peers that have a matching TCP-AO key.
+		 * See the comment in tcp_ao_add_cmd()
+		 */
+		if (tcp_ao_required(sk, addr, AF_INET))
+			return -EKEYREJECTED;
+		return tcp_md5_do_add(sk, addr,
 				      AF_INET, prefixlen, l3index, flags,
 				      cmd.tcpm_key, cmd.tcpm_keylen);
+	}
 
-	return tcp_md5_do_add(sk, (union tcp_md5_addr *)&sin6->sin6_addr,
-			      AF_INET6, prefixlen, l3index, flags,
+	addr = (union tcp_md5_addr *)&sin6->sin6_addr;
+
+	/* Don't allow keys for peers that have a matching TCP-AO key.
+	 * See the comment in tcp_ao_add_cmd()
+	 */
+	if (tcp_ao_required(sk, addr, AF_INET6))
+		return -EKEYREJECTED;
+
+	return tcp_md5_do_add(sk, addr, AF_INET6, prefixlen, l3index, flags,
 			      cmd.tcpm_key, cmd.tcpm_keylen);
 }
 
@@ -1899,6 +1915,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
 	.md5_parse	=	tcp_v6_parse_md5_keys,
 #endif
 #ifdef CONFIG_TCP_AO
+	.ao_lookup	=	tcp_v6_ao_lookup,
 	.ao_parse	=	tcp_v6_parse_ao,
 #endif
 };
@@ -1930,6 +1947,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
 	.md5_parse	=	tcp_v6_parse_md5_keys,
 #endif
 #ifdef CONFIG_TCP_AO
+	.ao_lookup	=	tcp_v6_ao_lookup,
 	.ao_parse	=	tcp_v6_parse_ao,
 #endif
 };
Signature · 1 line
2.41.0
#11 lore raw

PATCH 5/23 net/tcp: Calculate TCP-AO traffic keys

In reply to Dmitry Safonov

Add traffic key calculation the way it's described in RFC5926. Wire it up to tcp_finish_connect() and cache the new keys straight away on already established TCP connections.

---
 include/net/tcp.h     |   5 ++
 include/net/tcp_ao.h  |  43 +++++++++-
 net/ipv4/tcp_ao.c     | 191 ++++++++++++++++++++++++++++++++++++++++++
 net/ipv4/tcp_input.c  |   1 +
 net/ipv4/tcp_ipv4.c   |   1 +
 net/ipv4/tcp_output.c |   1 +
 net/ipv6/tcp_ao.c     |  50 +++++++++++
 net/ipv6/tcp_ipv6.c   |   1 +
 8 files changed, 292 insertions(+), 1 deletion(-)
include/net/tcp.h +5 -0
index 7e414873ef07..6a2e15a94214 100644
@@ -2147,6 +2147,11 @@ struct tcp_sock_af_ops {
 	struct tcp_ao_key	*(*ao_lookup)(const struct sock *sk,
 					      struct sock  *addr_sk,
 					      int sndid, int rcvid);
+	int			(*ao_calc_key_sk)(struct tcp_ao_key *mkt,
+						  u8 *key,
+						  const struct sock *sk,
+						  __be32 sisn, __be32 disn,
+						  bool send);
 #endif
 };
 
include/net/tcp_ao.h +42 -1

69 lines not on this page · raw message

net/ipv4/tcp_ao.c +191 -0

215 lines not on this page · raw message

net/ipv4/tcp_input.c +1 -0
index 670c3dab24f2..7780ceef3415 100644
@@ -6082,6 +6082,7 @@ void tcp_finish_connect(struct sock *sk, struct sk_buff *skb)
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct inet_connection_sock *icsk = inet_csk(sk);
 
+	tcp_ao_finish_connect(sk, skb);
 	tcp_set_state(sk, TCP_ESTABLISHED);
 	icsk->icsk_ack.lrcvtime = tcp_jiffies32;
 
net/ipv4/tcp_ipv4.c +1 -0
index 65e9bcac45f9..447e0c6f132f 100644
@@ -2286,6 +2286,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
 #ifdef CONFIG_TCP_AO
 	.ao_lookup		= tcp_v4_ao_lookup,
 	.ao_parse		= tcp_v4_parse_ao,
+	.ao_calc_key_sk		= tcp_v4_ao_calc_key_sk,
 #endif
 };
 #endif
net/ipv4/tcp_output.c +1 -0
index e98083abd7fb..937516aac99a 100644
@@ -3745,6 +3745,7 @@ static void tcp_connect_init(struct sock *sk)
 	if (tp->af_specific->md5_lookup(sk, sk))
 		tp->tcp_header_len += TCPOLEN_MD5SIG_ALIGNED;
 #endif
+	tcp_ao_connect_init(sk);
 
 	/* If user gave his TCP_MAXSEG, record it to clamp */
 	if (tp->rx_opt.user_mss)
net/ipv6/tcp_ao.c +50 -0

56 lines not on this page · raw message

net/ipv6/tcp_ipv6.c +1 -0
index 5c7c9dbfdb73..1e67aaf42460 100644
@@ -1917,6 +1917,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
 #ifdef CONFIG_TCP_AO
 	.ao_lookup	=	tcp_v6_ao_lookup,
 	.ao_parse	=	tcp_v6_parse_ao,
+	.ao_calc_key_sk	=	tcp_v6_ao_calc_key_sk,
 #endif
 };
 #endif
Signature · 1 line
2.41.0
#12 lore raw

PATCH 6/23 net/tcp: Add TCP-AO sign to outgoing packets

In reply to Dmitry Safonov

Using precalculated traffic keys, sign TCP segments as prescribed by RFC5925. Per RFC, TCP header options are included in sign calculation:
"The TCP header, by default including options, and where the TCP checksum and TCP-AO MAC fields are set to zero, all in network- byte order." (5.1.3)

tcp_ao_hash_header() has exclude_options parameter to optionally exclude TCP header from hash calculation, as described in RFC5925 (9.1), this is needed for interaction with middleboxes that may change "some TCP options". This is wired up to AO key flags and setsockopt() later.

Similarly to TCP-MD5 hash TCP segment fragments.

From this moment a user can start sending TCP-AO signed segments with one of crypto ahash algorithms from supported by Linux kernel. It can have a user-specified MAC length, to either save TCP option header space or provide higher protection using a longer signature.
The inbound segments are not yet verified, TCP-AO option is ignored and they are accepted.

---
 include/net/tcp.h     |   7 ++
 include/net/tcp_ao.h  |  13 ++++
 net/ipv4/tcp_ao.c     | 165 ++++++++++++++++++++++++++++++++++++++++++
 net/ipv4/tcp_ipv4.c   |   1 +
 net/ipv4/tcp_output.c | 122 ++++++++++++++++++++++++++++---
 net/ipv6/tcp_ao.c     |  28 +++++++
 net/ipv6/tcp_ipv6.c   |   2 +
 7 files changed, 328 insertions(+), 10 deletions(-)
include/net/tcp.h +7 -0
index 6a2e15a94214..d4344cc39893 100644
@@ -186,6 +186,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo);
 #define TCPOPT_SACK             5       /* SACK Block */
 #define TCPOPT_TIMESTAMP	8	/* Better RTT estimations/PAWS */
 #define TCPOPT_MD5SIG		19	/* MD5 Signature (RFC2385) */
+#define TCPOPT_AO		29	/* Authentication Option (RFC5925) */
 #define TCPOPT_MPTCP		30	/* Multipath TCP (RFC6824) */
 #define TCPOPT_FASTOPEN		34	/* Fast open (RFC7413) */
 #define TCPOPT_EXP		254	/* Experimental */
@@ -2147,6 +2148,12 @@ struct tcp_sock_af_ops {
 	struct tcp_ao_key	*(*ao_lookup)(const struct sock *sk,
 					      struct sock  *addr_sk,
 					      int sndid, int rcvid);
+	int			(*calc_ao_hash)(char *location,
+						struct tcp_ao_key *ao,
+						const struct sock *sk,
+						const struct sk_buff *skb,
+						const u8 *tkey,
+						int hash_offset, u32 sne);
 	int			(*ao_calc_key_sk)(struct tcp_ao_key *mkt,
 						  u8 *key,
 						  const struct sock *sk,
include/net/tcp_ao.h +13 -0
index 1aac82c1d22d..8d61f2690d52 100644
@@ -111,6 +111,10 @@ struct tcp6_ao_context {
 };
 struct tcp_sigpool;
 
+int tcp_ao_hash_skb(unsigned short int family,
+		    char *ao_hash, struct tcp_ao_key *key,
+		    const struct sock *sk, const struct sk_buff *skb,
+		    const u8 *tkey, int hash_offset, u32 sne);
 int tcp_parse_ao(struct sock *sk, int cmd, unsigned short int family,
 		 sockptr_t optval, int optlen);
 int tcp_ao_calc_traffic_key(struct tcp_ao_key *mkt, u8 *key, void *ctx,
@@ -126,13 +130,22 @@ struct tcp_ao_key *tcp_v4_ao_lookup(const struct sock *sk, struct sock *addr_sk,
 int tcp_v4_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 			  const struct sock *sk,
 			  __be32 sisn, __be32 disn, bool send);
+int tcp_v4_ao_hash_skb(char *ao_hash, struct tcp_ao_key *key,
+		       const struct sock *sk, const struct sk_buff *skb,
+		       const u8 *tkey, int hash_offset, u32 sne);
 /* ipv6 specific functions */
+int tcp_v6_ao_hash_pseudoheader(struct tcp_sigpool *hp,
+				const struct in6_addr *daddr,
+				const struct in6_addr *saddr, int nbytes);
 int tcp_v6_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 				 const struct sock *sk, __be32 sisn,
 				 __be32 disn, bool send);
 struct tcp_ao_key *tcp_v6_ao_lookup(const struct sock *sk,
 				    struct sock *addr_sk,
 				    int sndid, int rcvid);
+int tcp_v6_ao_hash_skb(char *ao_hash, struct tcp_ao_key *key,
+		       const struct sock *sk, const struct sk_buff *skb,
+		       const u8 *tkey, int hash_offset, u32 sne);
 int tcp_v6_parse_ao(struct sock *sk, int cmd,
 		    sockptr_t optval, int optlen);
 void tcp_ao_finish_connect(struct sock *sk, struct sk_buff *skb);
net/ipv4/tcp_ao.c +165 -0

171 lines not on this page · raw message

net/ipv4/tcp_ipv4.c +1 -0
index 447e0c6f132f..8222f14e4c32 100644
@@ -2285,6 +2285,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv4_specific = {
 #endif
 #ifdef CONFIG_TCP_AO
 	.ao_lookup		= tcp_v4_ao_lookup,
+	.calc_ao_hash		= tcp_v4_ao_hash_skb,
 	.ao_parse		= tcp_v4_parse_ao,
 	.ao_calc_key_sk		= tcp_v4_ao_calc_key_sk,
 #endif
net/ipv4/tcp_output.c +112 -10

228 lines not on this page · raw message

net/ipv6/tcp_ao.c +28 -0
index 0b490738c366..54c3714a05b4 100644
@@ -7,6 +7,7 @@
  *		Francesco Ruggeri <[email protected]>
  *		Salam Noureddine <[email protected]>
  */
+#include <crypto/hash.h>
 #include <linux/tcp.h>
 
 #include <net/tcp.h>
@@ -79,6 +80,33 @@ struct tcp_ao_key *tcp_v6_ao_lookup(const struct sock *sk,
 	return tcp_v6_ao_do_lookup(sk, addr, sndid, rcvid);
 }
 
+int tcp_v6_ao_hash_pseudoheader(struct tcp_sigpool *hp,
+				const struct in6_addr *daddr,
+				const struct in6_addr *saddr, int nbytes)
+{
+	struct tcp6_pseudohdr *bp;
+	struct scatterlist sg;
+
+	bp = hp->scratch;
+	/* 1. TCP pseudo-header (RFC2460) */
+	bp->saddr = *saddr;
+	bp->daddr = *daddr;
+	bp->len = cpu_to_be32(nbytes);
+	bp->protocol = cpu_to_be32(IPPROTO_TCP);
+
+	sg_init_one(&sg, bp, sizeof(*bp));
+	ahash_request_set_crypt(hp->req, &sg, NULL, sizeof(*bp));
+	return crypto_ahash_update(hp->req);
+}
+
+int tcp_v6_ao_hash_skb(char *ao_hash, struct tcp_ao_key *key,
+		       const struct sock *sk, const struct sk_buff *skb,
+		       const u8 *tkey, int hash_offset, u32 sne)
+{
+	return tcp_ao_hash_skb(AF_INET6, ao_hash, key, sk, skb, tkey,
+			hash_offset, sne);
+}
+
 int tcp_v6_parse_ao(struct sock *sk, int cmd,
 		    sockptr_t optval, int optlen)
 {
net/ipv6/tcp_ipv6.c +2 -0
index 1e67aaf42460..1dd7c8f3aff4 100644
@@ -1916,6 +1916,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv6_specific = {
 #endif
 #ifdef CONFIG_TCP_AO
 	.ao_lookup	=	tcp_v6_ao_lookup,
+	.calc_ao_hash	=	tcp_v6_ao_hash_skb,
 	.ao_parse	=	tcp_v6_parse_ao,
 	.ao_calc_key_sk	=	tcp_v6_ao_calc_key_sk,
 #endif
@@ -1949,6 +1950,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
 #endif
 #ifdef CONFIG_TCP_AO
 	.ao_lookup	=	tcp_v6_ao_lookup,
+	.calc_ao_hash	=	tcp_v4_ao_hash_skb,
 	.ao_parse	=	tcp_v6_parse_ao,
 #endif
 };
Signature · 1 line
2.41.0
#13 lore raw

Re: [PATCH v8.1 net-next 06/23] net/tcp: Add TCP-AO sign to outgoing packets

On Fri, Jul 21, 2023 at 05:18:57PM +0100, Dmitry Safonov wrote:

...

Hi Dmitry,

Quoted 1 line
net/ipv4/tcp_output.c +0 -0

...

net/ipv4/tcp_output.c static void tcp_options_write(struct tcphdr *th, struct tcp_sock *tp, · line 625

Quoted 6 lines
net/ipv4/tcp_output.c +2 -0
@@ -619,7 +621,33 @@ static void tcp_options_write(struct tcphdr *th, struct tcp_sock *tp,
 		opts->hash_location = (__u8 *)ptr;
 		ptr += 4;
 	}
+#ifdef CONFIG_TCP_AO
+	if (unlikely(OPTION_AO & options) && tp) {

Smatch warns that here we check if tp is NULL, but later on in the same function (existing) code uses tp unconditionally.

That code looks like this:

        if (unlikely(opts->num_sack_blocks)) {
                struct tcp_sack_block *sp = tp->rx_opt.dsack ?
                        tp->duplicate_sack : tp->selective_acks;

I would recommend running Smatch.
It points out a lot of interesting things.

net/ipv4/tcp_output.c static void tcp_options_write(struct tcphdr *th, struct tcp_sock *tp, · line 633

Quoted 8 lines
net/ipv4/tcp_output.c +7 -0
+		struct tcp_ao_key *rnext_key;
+		struct tcp_ao_info *ao_info;
+		u8 maclen;
 
+		if (WARN_ON_ONCE(!ao_key))
+			goto out_ao;
+		ao_info = rcu_dereference_check(tp->ao_info,
+				lockdep_sock_is_held(&tp->inet_conn.icsk_inet.sk));

Checkpatch complains about indentation here.

Rather than point out each case in the series,
could I ask you to run ./scripts/checkpatch.pl --strict over the patchset?

...

Quoted 4 lines
net/ipv4/tcp_output.c +0 -0
@@ -1363,6 +1424,34 @@ static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb,
 					       md5, sk, skb);
 	}
 #endif

net/ipv4/tcp_output.c static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, · line 1436

Quoted 10 lines
net/ipv4/tcp_output.c +10 -0
+#ifdef CONFIG_TCP_AO
+	if (ao) {
+		u8 *traffic_key;
+		void *tkey_buf = NULL;
+		u32 disn;
+
+		sk_gso_disable(sk);
+		if (unlikely(tcb->tcp_flags & TCPHDR_SYN)) {
+			if (tcb->tcp_flags & TCPHDR_ACK)
+				disn = ao->risn;

Sparse complains that there is an endian missmatch between disn and ao->risn ?

Rather than point out every problem flagged by Sparse, could I ask you to run it over the series?

...

#14 lore raw

Re: [PATCH v8.1 net-next 06/23] net/tcp: Add TCP-AO sign to outgoing packets

Hi Simon,

Quoted 42 lines · Simon Horman

On Fri, Jul 21, 2023 at 05:18:57PM +0100, Dmitry Safonov wrote:

...

Hi Dmitry,

Quoted 1 line
net/ipv4/tcp_output.c +0 -0

...

net/ipv4/tcp_output.c static void tcp_options_write(struct tcphdr *th, struct tcp_sock *tp, · line 625

Quoted 6 lines
net/ipv4/tcp_output.c +2 -0
@@ -619,7 +621,33 @@ static void tcp_options_write(struct tcphdr *th, struct tcp_sock *tp,
 		opts->hash_location = (__u8 *)ptr;
 		ptr += 4;
 	}
+#ifdef CONFIG_TCP_AO
+	if (unlikely(OPTION_AO & options) && tp) {

Smatch warns that here we check if tp is NULL, but later on in the same function (existing) code uses tp unconditionally.

That code looks like this:

        if (unlikely(opts->num_sack_blocks)) {
                struct tcp_sack_block *sp = tp->rx_opt.dsack ?
                        tp->duplicate_sack : tp->selective_acks;

I would recommend running Smatch.
It points out a lot of interesting things.

net/ipv4/tcp_output.c static void tcp_options_write(struct tcphdr *th, struct tcp_sock *tp, · line 633

Quoted 8 lines
net/ipv4/tcp_output.c +7 -0
+		struct tcp_ao_key *rnext_key;
+		struct tcp_ao_info *ao_info;
+		u8 maclen;
 
+		if (WARN_ON_ONCE(!ao_key))
+			goto out_ao;
+		ao_info = rcu_dereference_check(tp->ao_info,
+				lockdep_sock_is_held(&tp->inet_conn.icsk_inet.sk));

Checkpatch complains about indentation here.

Quoted 2 lines

Rather than point out each case in the series,
could I ask you to run ./scripts/checkpatch.pl --strict over the patchset?

Yeah, but then it won't fit 80 columns here. As both aren't hard requirements I tend to comply with 80 columns more than to indentation. In this particular case I'll check if it could be a helper function. If it won't make sense to separate it as a helper, I'll just move it to the same line than, breaking 80 columns limit.

Quoted 17 lines
Quoted 4 lines
net/ipv4/tcp_output.c +0 -0
@@ -1363,6 +1424,34 @@ static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb,
 					       md5, sk, skb);
 	}
 #endif

net/ipv4/tcp_output.c static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb, · line 1436

Quoted 10 lines
net/ipv4/tcp_output.c +10 -0
+#ifdef CONFIG_TCP_AO
+	if (ao) {
+		u8 *traffic_key;
+		void *tkey_buf = NULL;
+		u32 disn;
+
+		sk_gso_disable(sk);
+		if (unlikely(tcb->tcp_flags & TCPHDR_SYN)) {
+			if (tcb->tcp_flags & TCPHDR_ACK)
+				disn = ao->risn;

Sparse complains that there is an endian missmatch between disn and ao->risn ?

Quoted 2 lines

Rather than point out every problem flagged by Sparse, could I ask you to run it over the series?

Yeah, I noticed it on netdev patchwork, running it over the patches now.

Thanks,
Dmitry

#15 lore raw

Re: [PATCH v8.1 net-next 06/23] net/tcp: Add TCP-AO sign to outgoing packets

On Tue, Jul 25, 2023 at 08:10:21PM +0100, Dmitry Safonov wrote:

...

Quoted 12 lines
Quoted 12 lines

net/ipv4/tcp_output.c static void tcp_options_write(struct tcphdr *th, struct tcp_sock *tp, · line 633

Quoted 9 lines
net/ipv4/tcp_output.c +7 -0
+		struct tcp_ao_key *rnext_key;
+		struct tcp_ao_info *ao_info;
+		u8 maclen;
 
+		if (WARN_ON_ONCE(!ao_key))
+			goto out_ao;
+		ao_info = rcu_dereference_check(tp->ao_info,
+				lockdep_sock_is_held(&tp->inet_conn.icsk_inet.sk));

Checkpatch complains about indentation here.

Quoted 8 lines
Quoted 2 lines

Rather than point out each case in the series,
could I ask you to run ./scripts/checkpatch.pl --strict over the patchset?

Yeah, but then it won't fit 80 columns here. As both aren't hard requirements I tend to comply with 80 columns more than to indentation. In this particular case I'll check if it could be a helper function. If it won't make sense to separate it as a helper, I'll just move it to the same line than, breaking 80 columns limit.

Thanks, I missed the 80 column thing here.
I'd actually think that what you have is better than an extra long line. But it's your call.

...

#16 lore raw

PATCH 7/23 net/tcp: Add tcp_parse_auth_options()

In reply to Dmitry Safonov

Introduce a helper that:
(1) shares the common code with TCP-MD5 header options parsing
(2) looks for hash signature only once for both TCP-MD5 and TCP-AO
(3) fails with -EEXIST if any TCP sign option is present twice, see RFC5925 (2.2):
">> A single TCP segment MUST NOT have more than one TCP-AO in its options sequence. When multiple TCP-AOs appear, TCP MUST discard the segment."

---
 include/net/dropreason-core.h |  6 ++++++
 include/net/tcp.h             | 24 ++++++++++++++++++++-
 include/net/tcp_ao.h          | 17 ++++++++++++++-
 net/ipv4/tcp.c                |  3 ++-
 net/ipv4/tcp_input.c          | 39 ++++++++++++++++++++++++++---------
 net/ipv4/tcp_ipv4.c           | 15 +++++++++-----
 net/ipv6/tcp_ipv6.c           | 11 ++++++----
 7 files changed, 93 insertions(+), 22 deletions(-)
include/net/dropreason-core.h +6 -0
index f291a3b0f9e5..05255bffa2c8 100644
@@ -20,6 +20,7 @@
 	FN(IP_NOPROTO)			\
 	FN(SOCKET_RCVBUFF)		\
 	FN(PROTO_MEM)			\
+	FN(TCP_AUTH_HDR)		\
 	FN(TCP_MD5NOTFOUND)		\
 	FN(TCP_MD5UNEXPECTED)		\
 	FN(TCP_MD5FAILURE)		\
@@ -140,6 +141,11 @@ enum skb_drop_reason {
 	 * drop out of udp_memory_allocated.
 	 */
 	SKB_DROP_REASON_PROTO_MEM,
+	/**
+	 * @SKB_DROP_REASON_TCP_AUTH_HDR: TCP-MD5 or TCP-AO hashes are met
+	 * twice or set incorrectly.
+	 */
+	SKB_DROP_REASON_TCP_AUTH_HDR,
 	/**
 	 * @SKB_DROP_REASON_TCP_MD5NOTFOUND: no MD5 hash and one expected,
 	 * corresponding to LINUX_MIB_TCPMD5NOTFOUND
include/net/tcp.h +23 -1
index d4344cc39893..17991a6848ff 100644
@@ -428,7 +428,6 @@ int tcp_mmap(struct file *file, struct socket *sock,
 void tcp_parse_options(const struct net *net, const struct sk_buff *skb,
 		       struct tcp_options_received *opt_rx,
 		       int estab, struct tcp_fastopen_cookie *foc);
-const u8 *tcp_parse_md5sig_option(const struct tcphdr *th);
 
 /*
  *	BPF SKB-less helpers
@@ -2571,6 +2570,29 @@ static inline u64 tcp_transmit_time(const struct sock *sk)
 	return 0;
 }
 
+static inline int tcp_parse_auth_options(const struct tcphdr *th,
+		const u8 **md5_hash, const struct tcp_ao_hdr **aoh)
+{
+	const u8 *md5_tmp, *ao_tmp;
+	int ret;
+
+	ret = tcp_do_parse_auth_options(th, &md5_tmp, &ao_tmp);
+	if (ret)
+		return ret;
+
+	if (md5_hash)
+		*md5_hash = md5_tmp;
+
+	if (aoh) {
+		if (!ao_tmp)
+			*aoh = NULL;
+		else
+			*aoh = (struct tcp_ao_hdr *)(ao_tmp - 2);
+	}
+
+	return 0;
+}
+
 static inline bool tcp_ao_required(struct sock *sk, const void *saddr,
 				   int family)
 {
include/net/tcp_ao.h +16 -1
index 8d61f2690d52..e97dfdc11451 100644
@@ -150,7 +150,9 @@ int tcp_v6_parse_ao(struct sock *sk, int cmd,
 		    sockptr_t optval, int optlen);
 void tcp_ao_finish_connect(struct sock *sk, struct sk_buff *skb);
 void tcp_ao_connect_init(struct sock *sk);
-
+void tcp_ao_syncookie(struct sock *sk, const struct sk_buff *skb,
+		      struct tcp_request_sock *treq,
+		      unsigned short int family);
 #else /* CONFIG_TCP_AO */
 
 static inline struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
@@ -173,4 +175,17 @@ static inline void tcp_ao_connect_init(struct sock *sk)
 }
 #endif
 
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
+int tcp_do_parse_auth_options(const struct tcphdr *th,
+			      const u8 **md5_hash, const u8 **ao_hash);
+#else
+static int tcp_do_parse_auth_options(const struct tcphdr *th,
+				     const u8 **md5_hash, const u8 **ao_hash)
+{
+	*md5_hash = NULL;
+	*ao_hash = NULL;
+	return 0;
+}
+#endif
+
 #endif /* _TCP_AO_H */
net/ipv4/tcp.c +2 -1
index d0b0565fa11e..c90b3c6df620 100644
@@ -4395,7 +4395,8 @@ tcp_inbound_md5_hash(const struct sock *sk, const struct sk_buff *skb,
 	l3index = sdif ? dif : 0;
 
 	hash_expected = tcp_md5_do_lookup(sk, l3index, saddr, family);
-	hash_location = tcp_parse_md5sig_option(th);
+	if (tcp_parse_auth_options(th, &hash_location, NULL))
+		return SKB_DROP_REASON_TCP_AUTH_HDR;
 
 	/* We've parsed the options - do we have a hash? */
 	if (!hash_expected && !hash_location)
net/ipv4/tcp_input.c +29 -10
index 7780ceef3415..d9241385468a 100644
@@ -4212,39 +4212,58 @@ static bool tcp_fast_parse_options(const struct net *net,
 	return true;
 }
 
-#ifdef CONFIG_TCP_MD5SIG
+#if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
 /*
- * Parse MD5 Signature option
+ * Parse Signature options
  */
-const u8 *tcp_parse_md5sig_option(const struct tcphdr *th)
+int tcp_do_parse_auth_options(const struct tcphdr *th,
+			      const u8 **md5_hash, const u8 **ao_hash)
 {
 	int length = (th->doff << 2) - sizeof(*th);
 	const u8 *ptr = (const u8 *)(th + 1);
+	unsigned int minlen = TCPOLEN_MD5SIG;
+
+	if (IS_ENABLED(CONFIG_TCP_AO))
+		minlen = sizeof(struct tcp_ao_hdr) + 1;
+
+	*md5_hash = NULL;
+	*ao_hash = NULL;
 
 	/* If not enough data remaining, we can short cut */
-	while (length >= TCPOLEN_MD5SIG) {
+	while (length >= minlen) {
 		int opcode = *ptr++;
 		int opsize;
 
 		switch (opcode) {
 		case TCPOPT_EOL:
-			return NULL;
+			return 0;
 		case TCPOPT_NOP:
 			length--;
 			continue;
 		default:
 			opsize = *ptr++;
 			if (opsize < 2 || opsize > length)
-				return NULL;
-			if (opcode == TCPOPT_MD5SIG)
-				return opsize == TCPOLEN_MD5SIG ? ptr : NULL;
+				return -EINVAL;
+			if (opcode == TCPOPT_MD5SIG) {
+				if (opsize != TCPOLEN_MD5SIG)
+					return -EINVAL;
+				if (unlikely(*md5_hash || *ao_hash))
+					return -EEXIST;
+				*md5_hash = ptr;
+			} else if (opcode == TCPOPT_AO) {
+				if (opsize <= sizeof(struct tcp_ao_hdr))
+					return -EINVAL;
+				if (unlikely(*md5_hash || *ao_hash))
+					return -EEXIST;
+				*ao_hash = ptr;
+			}
 		}
 		ptr += opsize - 2;
 		length -= opsize;
 	}
-	return NULL;
+	return 0;
 }
-EXPORT_SYMBOL(tcp_parse_md5sig_option);
+EXPORT_SYMBOL(tcp_do_parse_auth_options);
 #endif
 
 /* Sorry, PAWS as specified is broken wrt. pure-ACKs -DaveM
net/ipv4/tcp_ipv4.c +10 -5
index 8222f14e4c32..0e9c6cfd4ab2 100644
@@ -670,7 +670,9 @@ EXPORT_SYMBOL(tcp_v4_send_check);
  *	Exception: precedence violation. We do not implement it in any case.
  */
 
-#ifdef CONFIG_TCP_MD5SIG
+#ifdef CONFIG_TCP_AO
+#define OPTION_BYTES MAX_TCP_OPTION_SPACE
+#elif defined(CONFIG_TCP_MD5SIG)
 #define OPTION_BYTES TCPOLEN_MD5SIG_ALIGNED
 #else
 #define OPTION_BYTES sizeof(__be32)
@@ -686,7 +688,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)
 	struct ip_reply_arg arg;
 #ifdef CONFIG_TCP_MD5SIG
 	struct tcp_md5sig_key *key = NULL;
-	const __u8 *hash_location = NULL;
+	const __u8 *md5_hash_location = NULL;
 	unsigned char newhash[16];
 	int genhash;
 	struct sock *sk1 = NULL;
@@ -727,8 +729,11 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)
 
 	net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
 #ifdef CONFIG_TCP_MD5SIG
+	/* Invalid TCP option size or twice included auth */
+	if (tcp_parse_auth_options(tcp_hdr(skb), &md5_hash_location, NULL))
+		return;
+
 	rcu_read_lock();
-	hash_location = tcp_parse_md5sig_option(th);
 	if (sk && sk_fullsock(sk)) {
 		const union tcp_md5_addr *addr;
 		int l3index;
@@ -739,7 +744,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)
 		l3index = tcp_v4_sdif(skb) ? inet_iif(skb) : 0;
 		addr = (union tcp_md5_addr *)&ip_hdr(skb)->saddr;
 		key = tcp_md5_do_lookup(sk, l3index, addr, AF_INET);
-	} else if (hash_location) {
+	} else if (md5_hash_location) {
 		const union tcp_md5_addr *addr;
 		int sdif = tcp_v4_sdif(skb);
 		int dif = inet_iif(skb);
@@ -771,7 +776,7 @@ static void tcp_v4_send_reset(const struct sock *sk, struct sk_buff *skb)
 
 
 		genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb);
-		if (genhash || memcmp(hash_location, newhash, 16) != 0)
+		if (genhash || memcmp(md5_hash_location, newhash, 16) != 0)
 			goto out;
 
 	}
net/ipv6/tcp_ipv6.c +7 -4
index 1dd7c8f3aff4..4c645160362e 100644
@@ -989,7 +989,7 @@ static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
 	u32 seq = 0, ack_seq = 0;
 	struct tcp_md5sig_key *key = NULL;
 #ifdef CONFIG_TCP_MD5SIG
-	const __u8 *hash_location = NULL;
+	const __u8 *md5_hash_location = NULL;
 	unsigned char newhash[16];
 	int genhash;
 	struct sock *sk1 = NULL;
@@ -1011,8 +1011,11 @@ static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
 
 	net = sk ? sock_net(sk) : dev_net(skb_dst(skb)->dev);
 #ifdef CONFIG_TCP_MD5SIG
+	/* Invalid TCP option size or twice included auth */
+	if (tcp_parse_auth_options(th, &md5_hash_location, NULL))
+		return;
+
 	rcu_read_lock();
-	hash_location = tcp_parse_md5sig_option(th);
 	if (sk && sk_fullsock(sk)) {
 		int l3index;
 
@@ -1021,7 +1024,7 @@ static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
 		 */
 		l3index = tcp_v6_sdif(skb) ? tcp_v6_iif_l3_slave(skb) : 0;
 		key = tcp_v6_md5_do_lookup(sk, &ipv6h->saddr, l3index);
-	} else if (hash_location) {
+	} else if (md5_hash_location) {
 		int dif = tcp_v6_iif_l3_slave(skb);
 		int sdif = tcp_v6_sdif(skb);
 		int l3index;
@@ -1050,7 +1053,7 @@ static void tcp_v6_send_reset(const struct sock *sk, struct sk_buff *skb)
 			goto out;
 
 		genhash = tcp_v6_md5_hash_skb(newhash, key, NULL, skb);
-		if (genhash || memcmp(hash_location, newhash, 16) != 0)
+		if (genhash || memcmp(md5_hash_location, newhash, 16) != 0)
 			goto out;
 	}
 #endif
Signature · 1 line
2.41.0
#17 lore raw

PATCH 8/23 net/tcp: Add AO sign to RST packets

In reply to Dmitry Safonov

Wire up sending resets to TCP-AO hashing.

---
 include/net/tcp_ao.h |  12 +++++
 net/ipv4/tcp_ao.c    | 105 ++++++++++++++++++++++++++++++++++++++++++-
 net/ipv4/tcp_ipv4.c  |  69 ++++++++++++++++++++++------
 net/ipv6/tcp_ipv6.c  |  66 +++++++++++++++++++++------
 4 files changed, 224 insertions(+), 28 deletions(-)
include/net/tcp_ao.h +12 -0
index e97dfdc11451..6e8567b5bbf8 100644
@@ -117,12 +117,24 @@ int tcp_ao_hash_skb(unsigned short int family,
 		    const u8 *tkey, int hash_offset, u32 sne);
 int tcp_parse_ao(struct sock *sk, int cmd, unsigned short int family,
 		 sockptr_t optval, int optlen);
+struct tcp_ao_key *tcp_ao_established_key(struct tcp_ao_info *ao,
+					  int sndid, int rcvid);
 int tcp_ao_calc_traffic_key(struct tcp_ao_key *mkt, u8 *key, void *ctx,
 			    unsigned int len, struct tcp_sigpool *hp);
 void tcp_ao_destroy_sock(struct sock *sk);
 struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
 				    const union tcp_ao_addr *addr,
 				    int family, int sndid, int rcvid, u16 port);
+int tcp_ao_hash_hdr(unsigned short family, char *ao_hash,
+		struct tcp_ao_key *key, const u8 *tkey,
+		const union tcp_ao_addr *daddr,
+		const union tcp_ao_addr *saddr,
+		const struct tcphdr *th, u32 sne);
+int tcp_ao_prepare_reset(const struct sock *sk, struct sk_buff *skb,
+			 const struct tcp_ao_hdr *aoh, int l3index, u32 seq,
+			 struct tcp_ao_key **key, char **traffic_key,
+			 bool *allocated_traffic_key, u8 *keyid, u32 *sne);
+
 /* ipv4 specific functions */
 int tcp_v4_parse_ao(struct sock *sk, int optname, sockptr_t optval, int optlen);
 struct tcp_ao_key *tcp_v4_ao_lookup(const struct sock *sk, struct sock *addr_sk,
net/ipv4/tcp_ao.c +103 -2

123 lines not on this page · raw message

net/ipv4/tcp_ipv4.c +56 -13

101 lines not on this page · raw message

net/ipv6/tcp_ipv6.c +53 -13

124 lines not on this page · raw message

Signature · 1 line
2.41.0
#18 lore raw

PATCH 9/23 net/tcp: Add TCP-AO sign to twsk

In reply to Dmitry Safonov

Add support for sockets in time-wait state.
ao_info as well as all keys are inherited on transition to time-wait socket. The lifetime of ao_info is now protected by ref counter, so that tcp_ao_destroy_sock() will destruct it only when the last user is gone.

---
 include/linux/tcp.h      |  3 ++
 include/net/tcp_ao.h     | 11 ++++++--
 net/ipv4/tcp_ao.c        | 46 +++++++++++++++++++++++++-----
 net/ipv4/tcp_ipv4.c      | 61 ++++++++++++++++++++++++++++++++++++----
 net/ipv4/tcp_minisocks.c |  4 ++-
 net/ipv4/tcp_output.c    |  2 +-
 net/ipv6/tcp_ipv6.c      | 41 ++++++++++++++++++++++++---
 7 files changed, 147 insertions(+), 21 deletions(-)
include/linux/tcp.h +3 -0
index 9eb15e798c32..bbc7db92f37c 100644
@@ -502,6 +502,9 @@ struct tcp_timewait_sock {
 #ifdef CONFIG_TCP_MD5SIG
 	struct tcp_md5sig_key	  *tw_md5_key;
 #endif
+#ifdef CONFIG_TCP_AO
+	struct tcp_ao_info	  *ao_info;
+#endif
 };
 
 static inline struct tcp_timewait_sock *tcp_twsk(const struct sock *sk)
include/net/tcp_ao.h +9 -2
index 6e8567b5bbf8..7f6849f16d7a 100644
@@ -86,6 +86,7 @@ struct tcp_ao_info {
 				__unused	:31;
 	__be32			lisn;
 	__be32			risn;
+	atomic_t		refcnt;		/* Protects twsk destruction */
 	struct rcu_head		rcu;
 };
 
@@ -121,7 +122,8 @@ struct tcp_ao_key *tcp_ao_established_key(struct tcp_ao_info *ao,
 					  int sndid, int rcvid);
 int tcp_ao_calc_traffic_key(struct tcp_ao_key *mkt, u8 *key, void *ctx,
 			    unsigned int len, struct tcp_sigpool *hp);
-void tcp_ao_destroy_sock(struct sock *sk);
+void tcp_ao_destroy_sock(struct sock *sk, bool twsk);
+void tcp_ao_time_wait(struct tcp_timewait_sock *tcptw, struct tcp_sock *tp);
 struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
 				    const union tcp_ao_addr *addr,
 				    int family, int sndid, int rcvid, u16 port);
@@ -174,7 +176,7 @@ static inline struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
 	return NULL;
 }
 
-static inline void tcp_ao_destroy_sock(struct sock *sk)
+static inline void tcp_ao_destroy_sock(struct sock *sk, bool twsk)
 {
 }
 
@@ -182,6 +184,11 @@ static inline void tcp_ao_finish_connect(struct sock *sk, struct sk_buff *skb)
 {
 }
 
+static inline void tcp_ao_time_wait(struct tcp_timewait_sock *tcptw,
+				    struct tcp_sock *tp)
+{
+}
+
 static inline void tcp_ao_connect_init(struct sock *sk)
 {
 }
net/ipv4/tcp_ao.c +39 -7

86 lines not on this page · raw message

net/ipv4/tcp_ipv4.c +55 -6

107 lines not on this page · raw message

net/ipv4/tcp_minisocks.c +3 -1
index 5616b6a34bee..3033f3187079 100644
@@ -279,7 +279,7 @@ static void tcp_time_wait_init(struct sock *sk, struct tcp_timewait_sock *tcptw)
 void tcp_time_wait(struct sock *sk, int state, int timeo)
 {
 	const struct inet_connection_sock *icsk = inet_csk(sk);
-	const struct tcp_sock *tp = tcp_sk(sk);
+	struct tcp_sock *tp = tcp_sk(sk);
 	struct net *net = sock_net(sk);
 	struct inet_timewait_sock *tw;
 
@@ -316,6 +316,7 @@ void tcp_time_wait(struct sock *sk, int state, int timeo)
 #endif
 
 		tcp_time_wait_init(sk, tcptw);
+		tcp_ao_time_wait(tcptw, tp);
 
 		/* Get the TIME_WAIT timeout firing. */
 		if (timeo < rto)
@@ -370,6 +371,7 @@ void tcp_twsk_destructor(struct sock *sk)
 			call_rcu(&twsk->tw_md5_key->rcu, tcp_md5_twsk_free_rcu);
 	}
 #endif
+	tcp_ao_destroy_sock(sk, true);
 }
 EXPORT_SYMBOL_GPL(tcp_twsk_destructor);
 
net/ipv4/tcp_output.c +1 -1
index 813d865095ac..0d74bb4460a6 100644
@@ -4059,7 +4059,7 @@ int tcp_connect(struct sock *sk)
 		 * then free up ao_info if allocated.
 		 */
 		if (needs_md5) {
-			tcp_ao_destroy_sock(sk);
+			tcp_ao_destroy_sock(sk, false);
 		} else if (needs_ao) {
 			tcp_clear_md5_list(sk);
 			kfree(rcu_replace_pointer(tp->md5sig_info, NULL,
net/ipv6/tcp_ipv6.c +37 -4

68 lines not on this page · raw message

Signature · 1 line
2.41.0
#19 lore raw

PATCH 10/23 net/tcp: Wire TCP-AO to request sockets

In reply to Dmitry Safonov

Now when the new request socket is created from the listening socket, it's recorded what MKT was used by the peer. tcp_rsk_used_ao() is a new helper for checking if TCP-AO option was used to create the request socket.
tcp_ao_copy_all_matching() will copy all keys that match the peer on the request socket, as well as preparing them for the usage (creating traffic keys).

---
 include/linux/tcp.h      |  18 ++++
 include/net/tcp.h        |   7 ++
 include/net/tcp_ao.h     |  22 ++++
 net/ipv4/syncookies.c    |   2 +
 net/ipv4/tcp_ao.c        | 219 ++++++++++++++++++++++++++++++++++++++-
 net/ipv4/tcp_input.c     |  15 +++
 net/ipv4/tcp_ipv4.c      |  65 ++++++++++--
 net/ipv4/tcp_minisocks.c |  10 ++
 net/ipv4/tcp_output.c    |  42 +++++---
 net/ipv6/syncookies.c    |   2 +
 net/ipv6/tcp_ao.c        |  21 ++++
 net/ipv6/tcp_ipv6.c      |  74 +++++++++++--
 12 files changed, 466 insertions(+), 31 deletions(-)
include/linux/tcp.h +18 -0
index bbc7db92f37c..f276e7222760 100644
@@ -165,6 +165,11 @@ struct tcp_request_sock {
 						  * after data-in-SYN.
 						  */
 	u8				syn_tos;
+#ifdef CONFIG_TCP_AO
+	u8				ao_keyid;
+	u8				ao_rcv_next;
+	u8				maclen;
+#endif
 };
 
 static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
@@ -172,6 +177,19 @@ static inline struct tcp_request_sock *tcp_rsk(const struct request_sock *req)
 	return (struct tcp_request_sock *)req;
 }
 
+static inline bool tcp_rsk_used_ao(const struct request_sock *req)
+{
+	/* The real length of MAC is saved in the request socket,
+	 * signing anything with zero-length makes no sense, so here is
+	 * a little hack..
+	 */
+#ifndef CONFIG_TCP_AO
+	return false;
+#else
+	return tcp_rsk(req)->maclen != 0;
+#endif
+}
+
 #define TCP_RMEM_TO_WIN_SCALE 8
 
 struct tcp_sock {
include/net/tcp.h +7 -0
index 17991a6848ff..b8be3aefb9c4 100644
@@ -2171,6 +2171,13 @@ struct tcp_request_sock_ops {
 					  const struct sock *sk,
 					  const struct sk_buff *skb);
 #endif
+#ifdef CONFIG_TCP_AO
+	struct tcp_ao_key	*(*ao_lookup)(const struct sock *sk,
+					       struct request_sock *req,
+					       int sndid, int rcvid);
+	int			(*ao_calc_key)(struct tcp_ao_key *mkt, u8 *key,
+						struct request_sock *sk);
+#endif
 #ifdef CONFIG_SYN_COOKIES
 	__u32 (*cookie_init_seq)(const struct sk_buff *skb,
 				 __u16 *mss);
include/net/tcp_ao.h +22 -0
index 7f6849f16d7a..b63c4f2c36ea 100644
@@ -120,6 +120,9 @@ int tcp_parse_ao(struct sock *sk, int cmd, unsigned short int family,
 		 sockptr_t optval, int optlen);
 struct tcp_ao_key *tcp_ao_established_key(struct tcp_ao_info *ao,
 					  int sndid, int rcvid);
+int tcp_ao_copy_all_matching(const struct sock *sk, struct sock *newsk,
+			     struct request_sock *req, struct sk_buff *skb,
+			     int family);
 int tcp_ao_calc_traffic_key(struct tcp_ao_key *mkt, u8 *key, void *ctx,
 			    unsigned int len, struct tcp_sigpool *hp);
 void tcp_ao_destroy_sock(struct sock *sk, bool twsk);
@@ -144,6 +147,11 @@ struct tcp_ao_key *tcp_v4_ao_lookup(const struct sock *sk, struct sock *addr_sk,
 int tcp_v4_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 			  const struct sock *sk,
 			  __be32 sisn, __be32 disn, bool send);
+int tcp_v4_ao_calc_key_rsk(struct tcp_ao_key *mkt, u8 *key,
+			   struct request_sock *req);
+struct tcp_ao_key *tcp_v4_ao_lookup_rsk(const struct sock *sk,
+					struct request_sock *req,
+					int sndid, int rcvid);
 int tcp_v4_ao_hash_skb(char *ao_hash, struct tcp_ao_key *key,
 		       const struct sock *sk, const struct sk_buff *skb,
 		       const u8 *tkey, int hash_offset, u32 sne);
@@ -154,9 +162,17 @@ int tcp_v6_ao_hash_pseudoheader(struct tcp_sigpool *hp,
 int tcp_v6_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 				 const struct sock *sk, __be32 sisn,
 				 __be32 disn, bool send);
+int tcp_v6_ao_calc_key_rsk(struct tcp_ao_key *mkt, u8 *key,
+			   struct request_sock *req);
+struct tcp_ao_key *tcp_v6_ao_do_lookup(const struct sock *sk,
+				       const struct in6_addr *addr,
+				       int sndid, int rcvid);
 struct tcp_ao_key *tcp_v6_ao_lookup(const struct sock *sk,
 				    struct sock *addr_sk,
 				    int sndid, int rcvid);
+struct tcp_ao_key *tcp_v6_ao_lookup_rsk(const struct sock *sk,
+					struct request_sock *req,
+					int sndid, int rcvid);
 int tcp_v6_ao_hash_skb(char *ao_hash, struct tcp_ao_key *key,
 		       const struct sock *sk, const struct sk_buff *skb,
 		       const u8 *tkey, int hash_offset, u32 sne);
@@ -169,6 +185,12 @@ void tcp_ao_syncookie(struct sock *sk, const struct sk_buff *skb,
 		      unsigned short int family);
 #else /* CONFIG_TCP_AO */
 
+static inline void tcp_ao_syncookie(struct sock *sk, const struct sk_buff *skb,
+		      struct tcp_request_sock *treq,
+		      unsigned short int family)
+{
+}
+
 static inline struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
 		const union tcp_ao_addr *addr,
 		int family, int sndid, int rcvid, u16 port)
net/ipv4/syncookies.c +2 -0
index dc478a0574cb..23fca22bc992 100644
@@ -394,6 +394,8 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)
 	treq->snt_synack	= 0;
 	treq->tfo_listener	= false;
 
+	tcp_ao_syncookie(sk, skb, treq, AF_INET);
+
 	if (IS_ENABLED(CONFIG_SMC))
 		ireq->smc_ok = 0;
 
net/ipv4/tcp_ao.c +218 -1

255 lines not on this page · raw message

net/ipv4/tcp_input.c +15 -0
index d9241385468a..2a7ed329f7cb 100644
@@ -6973,6 +6973,10 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops,
 	struct flowi fl;
 	u8 syncookies;
 
+#ifdef CONFIG_TCP_AO
+	const struct tcp_ao_hdr *aoh;
+#endif
+
 	syncookies = READ_ONCE(net->ipv4.sysctl_tcp_syncookies);
 
 	/* TW buckets are converted to open requests without
@@ -7058,6 +7062,17 @@ int tcp_conn_request(struct request_sock_ops *rsk_ops,
 			inet_rsk(req)->ecn_ok = 0;
 	}
 
+#ifdef CONFIG_TCP_AO
+	if (tcp_parse_auth_options(tcp_hdr(skb), NULL, &aoh))
+		goto drop_and_release; /* Invalid TCP options */
+	if (aoh) {
+		tcp_rsk(req)->maclen = aoh->length - sizeof(struct tcp_ao_hdr);
+		tcp_rsk(req)->ao_rcv_next = aoh->keyid;
+		tcp_rsk(req)->ao_keyid = aoh->rnext_keyid;
+	} else {
+		tcp_rsk(req)->maclen = 0;
+	}
+#endif
 	tcp_rsk(req)->snt_isn = isn;
 	tcp_rsk(req)->txhash = net_tx_rndhash();
 	tcp_rsk(req)->syn_tos = TCP_SKB_CB(skb)->ip_dsfield;
net/ipv4/tcp_ipv4.c +59 -6

110 lines not on this page · raw message

net/ipv4/tcp_minisocks.c +10 -0
index 3033f3187079..11b76fa601a7 100644
@@ -506,6 +506,9 @@ struct sock *tcp_create_openreq_child(const struct sock *sk,
 	const struct tcp_sock *oldtp;
 	struct tcp_sock *newtp;
 	u32 seq;
+#ifdef CONFIG_TCP_AO
+	struct tcp_ao_key *ao_key;
+#endif
 
 	if (!newsk)
 		return NULL;
@@ -584,6 +587,13 @@ struct sock *tcp_create_openreq_child(const struct sock *sk,
 	if (treq->af_specific->req_md5_lookup(sk, req_to_sk(req)))
 		newtp->tcp_header_len += TCPOLEN_MD5SIG_ALIGNED;
 #endif
+#ifdef CONFIG_TCP_AO
+	newtp->ao_info = NULL;
+	ao_key = treq->af_specific->ao_lookup(sk, req,
+				tcp_rsk(req)->ao_keyid, -1);
+	if (ao_key)
+		newtp->tcp_header_len += tcp_ao_len(ao_key);
+ #endif
 	if (skb->len >= TCP_MSS_DEFAULT + newtp->tcp_header_len)
 		newicsk->icsk_ack.last_seg_size = skb->len - newtp->tcp_header_len;
 	newtp->rx_opt.mss_clamp = req->mss;
net/ipv4/tcp_output.c +28 -14

71 lines not on this page · raw message

net/ipv6/syncookies.c +2 -0
index 5014aa663452..ad7a8caa7b2a 100644
@@ -214,6 +214,8 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
 	treq->snt_isn = cookie;
 	treq->ts_off = 0;
 	treq->txhash = net_tx_rndhash();
+	tcp_ao_syncookie(sk, skb, treq, AF_INET6);
+
 	if (IS_ENABLED(CONFIG_SMC))
 		ireq->smc_ok = 0;
 
net/ipv6/tcp_ao.c +21 -0
index 54c3714a05b4..65fa52f8c2e3 100644
@@ -63,6 +63,18 @@ int tcp_v6_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 					  htons(sk->sk_num), disn, sisn);
 }
 
+int tcp_v6_ao_calc_key_rsk(struct tcp_ao_key *mkt, u8 *key,
+			   struct request_sock *req)
+{
+	struct inet_request_sock *ireq = inet_rsk(req);
+
+	return tcp_v6_ao_calc_key(mkt, key,
+			&ireq->ir_v6_loc_addr, &ireq->ir_v6_rmt_addr,
+			htons(ireq->ir_num), ireq->ir_rmt_port,
+			htonl(tcp_rsk(req)->snt_isn),
+			htonl(tcp_rsk(req)->rcv_isn));
+}
+
 struct tcp_ao_key *tcp_v6_ao_do_lookup(const struct sock *sk,
 				       const struct in6_addr *addr,
 				       int sndid, int rcvid)
@@ -80,6 +92,15 @@ struct tcp_ao_key *tcp_v6_ao_lookup(const struct sock *sk,
 	return tcp_v6_ao_do_lookup(sk, addr, sndid, rcvid);
 }
 
+struct tcp_ao_key *tcp_v6_ao_lookup_rsk(const struct sock *sk,
+					struct request_sock *req,
+					int sndid, int rcvid)
+{
+	struct in6_addr *addr = &inet_rsk(req)->ir_v6_rmt_addr;
+
+	return tcp_v6_ao_do_lookup(sk, addr, sndid, rcvid);
+}
+
 int tcp_v6_ao_hash_pseudoheader(struct tcp_sigpool *hp,
 				const struct in6_addr *daddr,
 				const struct in6_addr *saddr, int nbytes)
net/ipv6/tcp_ipv6.c +64 -10

107 lines not on this page · raw message

Signature · 1 line
2.41.0
#20 lore raw

PATCH 11/23 net/tcp: Sign SYN-ACK segments with TCP-AO

In reply to Dmitry Safonov

Similarly to RST segments, wire SYN-ACKs to TCP-AO. tcp_rsk_used_ao() is handy here to check if the request socket used AO and needs a signature on the outgoing segments.

---
 include/net/tcp.h     |  4 ++++
 include/net/tcp_ao.h  |  6 ++++++
 net/ipv4/tcp_ao.c     | 22 ++++++++++++++++++++
 net/ipv4/tcp_ipv4.c   |  1 +
 net/ipv4/tcp_output.c | 48 +++++++++++++++++++++++++++++++++++++------
 net/ipv6/tcp_ao.c     | 22 ++++++++++++++++++++
 net/ipv6/tcp_ipv6.c   |  1 +
 7 files changed, 98 insertions(+), 6 deletions(-)
include/net/tcp.h +4 -0
index b8be3aefb9c4..ad300e5b6eb8 100644
@@ -2177,6 +2177,10 @@ struct tcp_request_sock_ops {
 					       int sndid, int rcvid);
 	int			(*ao_calc_key)(struct tcp_ao_key *mkt, u8 *key,
 						struct request_sock *sk);
+	int		(*ao_synack_hash)(char *ao_hash, struct tcp_ao_key *mkt,
+					  struct request_sock *req,
+					  const struct sk_buff *skb,
+					  int hash_offset, u32 sne);
 #endif
 #ifdef CONFIG_SYN_COOKIES
 	__u32 (*cookie_init_seq)(const struct sk_buff *skb,
include/net/tcp_ao.h +6 -0
index b63c4f2c36ea..4979c993973f 100644
@@ -144,6 +144,9 @@ int tcp_ao_prepare_reset(const struct sock *sk, struct sk_buff *skb,
 int tcp_v4_parse_ao(struct sock *sk, int optname, sockptr_t optval, int optlen);
 struct tcp_ao_key *tcp_v4_ao_lookup(const struct sock *sk, struct sock *addr_sk,
 				    int sndid, int rcvid);
+int tcp_v4_ao_synack_hash(char *ao_hash, struct tcp_ao_key *mkt,
+			  struct request_sock *req, const struct sk_buff *skb,
+			  int hash_offset, u32 sne);
 int tcp_v4_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 			  const struct sock *sk,
 			  __be32 sisn, __be32 disn, bool send);
@@ -178,6 +181,9 @@ int tcp_v6_ao_hash_skb(char *ao_hash, struct tcp_ao_key *key,
 		       const u8 *tkey, int hash_offset, u32 sne);
 int tcp_v6_parse_ao(struct sock *sk, int cmd,
 		    sockptr_t optval, int optlen);
+int tcp_v6_ao_synack_hash(char *ao_hash, struct tcp_ao_key *ao_key,
+			  struct request_sock *req, const struct sk_buff *skb,
+			  int hash_offset, u32 sne);
 void tcp_ao_finish_connect(struct sock *sk, struct sk_buff *skb);
 void tcp_ao_connect_init(struct sock *sk);
 void tcp_ao_syncookie(struct sock *sk, const struct sk_buff *skb,
net/ipv4/tcp_ao.c +22 -0
index be9349239139..0233cf363de9 100644
@@ -546,6 +546,28 @@ int tcp_v4_ao_hash_skb(char *ao_hash, struct tcp_ao_key *key,
 			       tkey, hash_offset, sne);
 }
 
+int tcp_v4_ao_synack_hash(char *ao_hash, struct tcp_ao_key *ao_key,
+			  struct request_sock *req, const struct sk_buff *skb,
+			  int hash_offset, u32 sne)
+{
+	void *hash_buf = NULL;
+	int err;
+
+	hash_buf = kmalloc(tcp_ao_digest_size(ao_key), GFP_ATOMIC);
+	if (!hash_buf)
+		return -ENOMEM;
+
+	err = tcp_v4_ao_calc_key_rsk(ao_key, hash_buf, req);
+	if (err)
+		goto out;
+
+	err = tcp_ao_hash_skb(AF_INET, ao_hash, ao_key, req_to_sk(req), skb,
+			      hash_buf, hash_offset, sne);
+out:
+	kfree(hash_buf);
+	return err;
+}
+
 struct tcp_ao_key *tcp_v4_ao_lookup_rsk(const struct sock *sk,
 					struct request_sock *req,
 					int sndid, int rcvid)
net/ipv4/tcp_ipv4.c +1 -0
index 12b952d0dcf4..884d9f854276 100644
@@ -1678,6 +1678,7 @@ const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops = {
 #ifdef CONFIG_TCP_AO
 	.ao_lookup	=	tcp_v4_ao_lookup_rsk,
 	.ao_calc_key	=	tcp_v4_ao_calc_key_rsk,
+	.ao_synack_hash	=	tcp_v4_ao_synack_hash,
 #endif
 #ifdef CONFIG_SYN_COOKIES
 	.cookie_init_seq =	cookie_v4_init_sequence,
net/ipv4/tcp_output.c +42 -6

78 lines not on this page · raw message

net/ipv6/tcp_ao.c +22 -0
index 65fa52f8c2e3..ec320781bc5c 100644
@@ -133,3 +133,25 @@ int tcp_v6_parse_ao(struct sock *sk, int cmd,
 {
 	return tcp_parse_ao(sk, cmd, AF_INET6, optval, optlen);
 }
+
+int tcp_v6_ao_synack_hash(char *ao_hash, struct tcp_ao_key *ao_key,
+			  struct request_sock *req, const struct sk_buff *skb,
+			  int hash_offset, u32 sne)
+{
+	void *hash_buf = NULL;
+	int err;
+
+	hash_buf = kmalloc(tcp_ao_digest_size(ao_key), GFP_ATOMIC);
+	if (!hash_buf)
+		return -ENOMEM;
+
+	err = tcp_v6_ao_calc_key_rsk(ao_key, hash_buf, req);
+	if (err)
+		goto out;
+
+	err = tcp_ao_hash_skb(AF_INET6, ao_hash, ao_key, req_to_sk(req), skb,
+			      hash_buf, hash_offset, sne);
+out:
+	kfree(hash_buf);
+	return err;
+}
net/ipv6/tcp_ipv6.c +1 -0
index 6905484077db..000e4e5309d3 100644
@@ -845,6 +845,7 @@ const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops = {
 #ifdef CONFIG_TCP_AO
 	.ao_lookup	=	tcp_v6_ao_lookup_rsk,
 	.ao_calc_key	=	tcp_v6_ao_calc_key_rsk,
+	.ao_synack_hash =	tcp_v6_ao_synack_hash,
 #endif
 #ifdef CONFIG_SYN_COOKIES
 	.cookie_init_seq =	cookie_v6_init_sequence,
Signature · 1 line
2.41.0
#21 lore raw

PATCH 12/23 net/tcp: Verify inbound TCP-AO signed segments

In reply to Dmitry Safonov

Now there is a common function to verify signature on TCP segments:
tcp_inbound_hash(). It has checks for all possible cross-interactions with MD5 signs as well as with unsigned segments.

The rules from RFC5925 are:
(1) Any TCP segment can have at max only one signature.
(2) TCP connections can't switch between using TCP-MD5 and TCP-AO.
(3) TCP-AO connections can't stop using AO, as well as unsigned connections can't suddenly start using AO.

---
 include/net/dropreason-core.h |  17 ++++
 include/net/tcp.h             |  53 ++++++++++-
 include/net/tcp_ao.h          |  17 ++++
 net/ipv4/tcp.c                |  39 ++------
 net/ipv4/tcp_ao.c             | 167 ++++++++++++++++++++++++++++++++++
 net/ipv4/tcp_ipv4.c           |  10 +-
 net/ipv6/tcp_ao.c             |  12 +++
 net/ipv6/tcp_ipv6.c           |  11 ++-
 8 files changed, 283 insertions(+), 43 deletions(-)
include/net/dropreason-core.h +17 -0
index 05255bffa2c8..1de089ab73fd 100644
@@ -24,6 +24,10 @@
 	FN(TCP_MD5NOTFOUND)		\
 	FN(TCP_MD5UNEXPECTED)		\
 	FN(TCP_MD5FAILURE)		\
+	FN(TCP_AONOTFOUND)		\
+	FN(TCP_AOUNEXPECTED)		\
+	FN(TCP_AOKEYNOTFOUND)		\
+	FN(TCP_AOFAILURE)		\
 	FN(SOCKET_BACKLOG)		\
 	FN(TCP_FLAGS)			\
 	FN(TCP_ZEROWINDOW)		\
@@ -161,6 +165,19 @@ enum skb_drop_reason {
 	 * to LINUX_MIB_TCPMD5FAILURE
 	 */
 	SKB_DROP_REASON_TCP_MD5FAILURE,
+	/**
+	 * @SKB_DROP_REASON_TCP_AONOTFOUND: no TCP-AO hash and one was expected
+	 */
+	SKB_DROP_REASON_TCP_AONOTFOUND,
+	/**
+	 * @SKB_DROP_REASON_TCP_AOUNEXPECTED: TCP-AO hash is present and it
+	 * was not expected.
+	 */
+	SKB_DROP_REASON_TCP_AOUNEXPECTED,
+	/** @SKB_DROP_REASON_TCP_AOKEYNOTFOUND: TCP-AO key is unknown */
+	SKB_DROP_REASON_TCP_AOKEYNOTFOUND,
+	/** @SKB_DROP_REASON_TCP_AOFAILURE: TCP-AO hash is wrong */
+	SKB_DROP_REASON_TCP_AOFAILURE,
 	/**
 	 * @SKB_DROP_REASON_SOCKET_BACKLOG: failed to add skb to socket backlog (
 	 * see LINUX_MIB_TCPBACKLOGDROP)
include/net/tcp.h +51 -2

69 lines not on this page · raw message

include/net/tcp_ao.h +17 -0
index 4979c993973f..1b7ca7bf93a3 100644
@@ -112,6 +112,9 @@ struct tcp6_ao_context {
 };
 struct tcp_sigpool;
 
+#define TCP_AO_ESTABLISHED (TCPF_ESTABLISHED|TCPF_FIN_WAIT1|TCPF_FIN_WAIT2|\
+		TCPF_CLOSE|TCPF_CLOSE_WAIT|TCPF_LAST_ACK|TCPF_CLOSING)
+
 int tcp_ao_hash_skb(unsigned short int family,
 		    char *ao_hash, struct tcp_ao_key *key,
 		    const struct sock *sk, const struct sk_buff *skb,
@@ -127,6 +130,10 @@ int tcp_ao_calc_traffic_key(struct tcp_ao_key *mkt, u8 *key, void *ctx,
 			    unsigned int len, struct tcp_sigpool *hp);
 void tcp_ao_destroy_sock(struct sock *sk, bool twsk);
 void tcp_ao_time_wait(struct tcp_timewait_sock *tcptw, struct tcp_sock *tp);
+enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
+			const struct sk_buff *skb, unsigned short int family,
+			const struct request_sock *req,
+			const struct tcp_ao_hdr *aoh);
 struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
 				    const union tcp_ao_addr *addr,
 				    int family, int sndid, int rcvid, u16 port);
@@ -162,6 +169,9 @@ int tcp_v4_ao_hash_skb(char *ao_hash, struct tcp_ao_key *key,
 int tcp_v6_ao_hash_pseudoheader(struct tcp_sigpool *hp,
 				const struct in6_addr *daddr,
 				const struct in6_addr *saddr, int nbytes);
+int tcp_v6_ao_calc_key_skb(struct tcp_ao_key *mkt, u8 *key,
+			   const struct sk_buff *skb, __be32 sisn,
+			    __be32 disn);
 int tcp_v6_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 				 const struct sock *sk, __be32 sisn,
 				 __be32 disn, bool send);
@@ -197,6 +207,13 @@ static inline void tcp_ao_syncookie(struct sock *sk, const struct sk_buff *skb,
 {
 }
 
+static inline enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
+		const struct sk_buff *skb, unsigned short int family,
+		const struct request_sock *req, const struct tcp_ao_hdr *aoh)
+{
+	return SKB_NOT_DROPPED_YET;
+}
+
 static inline struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
 		const union tcp_ao_addr *addr,
 		int family, int sndid, int rcvid, u16 port)
net/ipv4/tcp.c +8 -31
index c90b3c6df620..d7e18c528c7b 100644
@@ -4372,42 +4372,23 @@ EXPORT_SYMBOL(tcp_md5_hash_key);
 enum skb_drop_reason
 tcp_inbound_md5_hash(const struct sock *sk, const struct sk_buff *skb,
 		     const void *saddr, const void *daddr,
-		     int family, int dif, int sdif)
+		     int family, int l3index, const __u8 *hash_location)
 {
-	/*
-	 * This gets called for each TCP segment that arrives
-	 * so we want to be efficient.
+	/* This gets called for each TCP segment that has TCP-MD5 option.
 	 * We have 3 drop cases:
 	 * o No MD5 hash and one expected.
 	 * o MD5 hash and we're not expecting one.
 	 * o MD5 hash and its wrong.
 	 */
-	const __u8 *hash_location = NULL;
-	struct tcp_md5sig_key *hash_expected;
 	const struct tcphdr *th = tcp_hdr(skb);
 	const struct tcp_sock *tp = tcp_sk(sk);
-	int genhash, l3index;
+	struct tcp_md5sig_key *key;
+	int genhash;
 	u8 newhash[16];
 
-	/* sdif set, means packet ingressed via a device
-	 * in an L3 domain and dif is set to the l3mdev
-	 */
-	l3index = sdif ? dif : 0;
+	key = tcp_md5_do_lookup(sk, l3index, saddr, family);
 
-	hash_expected = tcp_md5_do_lookup(sk, l3index, saddr, family);
-	if (tcp_parse_auth_options(th, &hash_location, NULL))
-		return SKB_DROP_REASON_TCP_AUTH_HDR;
-
-	/* We've parsed the options - do we have a hash? */
-	if (!hash_expected && !hash_location)
-		return SKB_NOT_DROPPED_YET;
-
-	if (hash_expected && !hash_location) {
-		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
-		return SKB_DROP_REASON_TCP_MD5NOTFOUND;
-	}
-
-	if (!hash_expected && hash_location) {
+	if (!key && hash_location) {
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
 		return SKB_DROP_REASON_TCP_MD5UNEXPECTED;
 	}
@@ -4417,14 +4398,10 @@ tcp_inbound_md5_hash(const struct sock *sk, const struct sk_buff *skb,
 	 * IPv4-mapped case.
 	 */
 	if (family == AF_INET)
-		genhash = tcp_v4_md5_hash_skb(newhash,
-					      hash_expected,
-					      NULL, skb);
+		genhash = tcp_v4_md5_hash_skb(newhash, key, NULL, skb);
 	else
-		genhash = tp->af_specific->calc_md5_hash(newhash,
-							 hash_expected,
+		genhash = tp->af_specific->calc_md5_hash(newhash, key,
 							 NULL, skb);
-
 	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);
 		if (family == AF_INET) {
net/ipv4/tcp_ao.c +167 -0

179 lines not on this page · raw message

net/ipv4/tcp_ipv4.c +5 -5
index 884d9f854276..f09c0389ae11 100644
@@ -2200,9 +2200,9 @@ int tcp_v4_rcv(struct sk_buff *skb)
 		if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
 			drop_reason = SKB_DROP_REASON_XFRM_POLICY;
 		else
-			drop_reason = tcp_inbound_md5_hash(sk, skb,
-						   &iph->saddr, &iph->daddr,
-						   AF_INET, dif, sdif);
+			drop_reason = tcp_inbound_hash(sk, req, skb,
+						       &iph->saddr, &iph->daddr,
+						       AF_INET, dif, sdif);
 		if (unlikely(drop_reason)) {
 			sk_drops_add(sk, skb);
 			reqsk_put(req);
@@ -2279,8 +2279,8 @@ int tcp_v4_rcv(struct sk_buff *skb)
 		goto discard_and_relse;
 	}
 
-	drop_reason = tcp_inbound_md5_hash(sk, skb, &iph->saddr,
-					   &iph->daddr, AF_INET, dif, sdif);
+	drop_reason = tcp_inbound_hash(sk, NULL, skb, &iph->saddr, &iph->daddr,
+				       AF_INET, dif, sdif);
 	if (drop_reason)
 		goto discard_and_relse;
 
net/ipv6/tcp_ao.c +12 -0
index ec320781bc5c..984a75f96e20 100644
@@ -49,6 +49,18 @@ static int tcp_v6_ao_calc_key(struct tcp_ao_key *mkt, u8 *key,
 	return err;
 }
 
+int tcp_v6_ao_calc_key_skb(struct tcp_ao_key *mkt, u8 *key,
+			   const struct sk_buff *skb,
+			   __be32 sisn, __be32 disn)
+{
+	const struct ipv6hdr *iph = ipv6_hdr(skb);
+	const struct tcphdr *th = tcp_hdr(skb);
+
+	return tcp_v6_ao_calc_key(mkt, key, &iph->saddr,
+				  &iph->daddr, th->source,
+				  th->dest, sisn, disn);
+}
+
 int tcp_v6_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 			  const struct sock *sk, __be32 sisn,
 			  __be32 disn, bool send)
net/ipv6/tcp_ipv6.c +6 -5
index 000e4e5309d3..40266a3a99a6 100644
@@ -1779,9 +1779,9 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		struct sock *nsk;
 
 		sk = req->rsk_listener;
-		drop_reason = tcp_inbound_md5_hash(sk, skb,
-						   &hdr->saddr, &hdr->daddr,
-						   AF_INET6, dif, sdif);
+		drop_reason = tcp_inbound_hash(sk, req, skb,
+					       &hdr->saddr, &hdr->daddr,
+					       AF_INET6, dif, sdif);
 		if (drop_reason) {
 			sk_drops_add(sk, skb);
 			reqsk_put(req);
@@ -1854,8 +1854,8 @@ INDIRECT_CALLABLE_SCOPE int tcp_v6_rcv(struct sk_buff *skb)
 		goto discard_and_relse;
 	}
 
-	drop_reason = tcp_inbound_md5_hash(sk, skb, &hdr->saddr, &hdr->daddr,
-					   AF_INET6, dif, sdif);
+	drop_reason = tcp_inbound_hash(sk, NULL, skb, &hdr->saddr, &hdr->daddr,
+				       AF_INET6, dif, sdif);
 	if (drop_reason)
 		goto discard_and_relse;
 
@@ -2083,6 +2083,7 @@ static const struct tcp_sock_af_ops tcp_sock_ipv6_mapped_specific = {
 	.ao_lookup	=	tcp_v6_ao_lookup,
 	.calc_ao_hash	=	tcp_v4_ao_hash_skb,
 	.ao_parse	=	tcp_v6_parse_ao,
+	.ao_calc_key_sk	=	tcp_v4_ao_calc_key_sk,
 #endif
 };
 #endif
Signature · 1 line
2.41.0
#22 lore raw

PATCH 13/23 net/tcp: Add TCP-AO segments counters

In reply to Dmitry Safonov

Introduce segment counters that are useful for troubleshooting/debugging as well as for writing tests.
Now there are global snmp counters as well as per-socket and per-key.

---
 include/net/dropreason-core.h | 15 +++++++++++----
 include/net/tcp.h             | 15 +++++++++++----
 include/net/tcp_ao.h          | 10 ++++++++++
 include/uapi/linux/snmp.h     |  4 ++++
 include/uapi/linux/tcp.h      |  8 +++++++-
 net/ipv4/proc.c               |  4 ++++
 net/ipv4/tcp_ao.c             | 30 +++++++++++++++++++++++++++---
 net/ipv4/tcp_ipv4.c           |  2 +-
 net/ipv6/tcp_ipv6.c           |  4 ++--
 9 files changed, 77 insertions(+), 15 deletions(-)
include/net/dropreason-core.h +11 -4
index 1de089ab73fd..f93ab108df4e 100644
@@ -166,17 +166,24 @@ enum skb_drop_reason {
 	 */
 	SKB_DROP_REASON_TCP_MD5FAILURE,
 	/**
-	 * @SKB_DROP_REASON_TCP_AONOTFOUND: no TCP-AO hash and one was expected
+	 * @SKB_DROP_REASON_TCP_AONOTFOUND: no TCP-AO hash and one was expected,
+	 * corresponding to LINUX_MIB_TCPAOREQUIRED
 	 */
 	SKB_DROP_REASON_TCP_AONOTFOUND,
 	/**
 	 * @SKB_DROP_REASON_TCP_AOUNEXPECTED: TCP-AO hash is present and it
-	 * was not expected.
+	 * was not expected, corresponding to LINUX_MIB_TCPAOKEYNOTFOUND
 	 */
 	SKB_DROP_REASON_TCP_AOUNEXPECTED,
-	/** @SKB_DROP_REASON_TCP_AOKEYNOTFOUND: TCP-AO key is unknown */
+	/**
+	 * @SKB_DROP_REASON_TCP_AOKEYNOTFOUND: TCP-AO key is unknown,
+	 * corresponding to LINUX_MIB_TCPAOKEYNOTFOUND
+	 */
 	SKB_DROP_REASON_TCP_AOKEYNOTFOUND,
-	/** @SKB_DROP_REASON_TCP_AOFAILURE: TCP-AO hash is wrong */
+	/**
+	 * @SKB_DROP_REASON_TCP_AOFAILURE: TCP-AO hash is wrong,
+	 * corresponding to LINUX_MIB_TCPAOBAD
+	 */
 	SKB_DROP_REASON_TCP_AOFAILURE,
 	/**
 	 * @SKB_DROP_REASON_SOCKET_BACKLOG: failed to add skb to socket backlog (
include/net/tcp.h +11 -4
index 6289ee642931..a58f95237ea3 100644
@@ -2605,7 +2605,7 @@ static inline int tcp_parse_auth_options(const struct tcphdr *th,
 }
 
 static inline bool tcp_ao_required(struct sock *sk, const void *saddr,
-				   int family)
+				   int family, bool stat_inc)
 {
 #ifdef CONFIG_TCP_AO
 	struct tcp_ao_info *ao_info;
@@ -2617,8 +2617,13 @@ static inline bool tcp_ao_required(struct sock *sk, const void *saddr,
 		return false;
 
 	ao_key = tcp_ao_do_lookup(sk, saddr, family, -1, -1, 0);
-	if (ao_info->ao_required || ao_key)
+	if (ao_info->ao_required || ao_key) {
+		if (stat_inc) {
+			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOREQUIRED);
+			atomic64_inc(&ao_info->counters.ao_required);
+		}
 		return true;
+	}
 #endif
 	return false;
 }
@@ -2640,8 +2645,10 @@ tcp_inbound_hash(struct sock *sk, const struct request_sock *req,
 		return SKB_DROP_REASON_TCP_AUTH_HDR;
 
 	if (req) {
-		if (tcp_rsk_used_ao(req) != !!aoh)
+		if (tcp_rsk_used_ao(req) != !!aoh) {
+			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOBAD);
 			return SKB_DROP_REASON_TCP_AOFAILURE;
+		}
 	}
 
 	/* sdif set, means packet ingressed via a device
@@ -2656,7 +2663,7 @@ tcp_inbound_hash(struct sock *sk, const struct request_sock *req,
 		 * the last key is impossible to remove, so there's
 		 * always at least one current_key.
 		 */
-		if (tcp_ao_required(sk, saddr, family))
+		if (tcp_ao_required(sk, saddr, family, true))
 			return SKB_DROP_REASON_TCP_AONOTFOUND;
 		if (unlikely(tcp_md5_do_lookup(sk, l3index, saddr, family))) {
 			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
include/net/tcp_ao.h +10 -0
index 1b7ca7bf93a3..0a02098bddeb 100644
@@ -19,6 +19,13 @@ struct tcp_ao_hdr {
 	u8	rnext_keyid;
 };
 
+struct tcp_ao_counters {
+	atomic64_t	pkt_good;
+	atomic64_t	pkt_bad;
+	atomic64_t	key_not_found;
+	atomic64_t	ao_required;
+};
+
 struct tcp_ao_key {
 	struct hlist_node	node;
 	union tcp_ao_addr	addr;
@@ -34,6 +41,8 @@ struct tcp_ao_key {
 	u8			rcvid;
 	u8			maclen;
 	struct rcu_head		rcu;
+	atomic64_t		pkt_good;
+	atomic64_t		pkt_bad;
 	u8			traffic_keys[];
 };
 
@@ -82,6 +91,7 @@ struct tcp_ao_info {
 	 */
 	struct tcp_ao_key	*current_key;
 	struct tcp_ao_key	*rnext_key;
+	struct tcp_ao_counters	counters;
 	u32			ao_required	:1,
 				__unused	:31;
 	__be32			lisn;
include/uapi/linux/snmp.h +4 -0
index 26f33a4c253d..06ddf4cd295c 100644
@@ -296,6 +296,10 @@ enum
 	LINUX_MIB_TCPMIGRATEREQSUCCESS,		/* TCPMigrateReqSuccess */
 	LINUX_MIB_TCPMIGRATEREQFAILURE,		/* TCPMigrateReqFailure */
 	LINUX_MIB_TCPPLBREHASH,			/* TCPPLBRehash */
+	LINUX_MIB_TCPAOREQUIRED,		/* TCPAORequired */
+	LINUX_MIB_TCPAOBAD,			/* TCPAOBad */
+	LINUX_MIB_TCPAOKEYNOTFOUND,		/* TCPAOKeyNotFound */
+	LINUX_MIB_TCPAOGOOD,			/* TCPAOGood */
 	__LINUX_MIB_MAX
 };
 
include/uapi/linux/tcp.h +7 -1
index 250e0ce2cc38..3fe0612ec59a 100644
@@ -391,9 +391,15 @@ struct tcp_ao_info_opt { /* setsockopt(TCP_AO_INFO) */
 	__u32   set_current	:1,	/* corresponding ::current_key */
 		set_rnext	:1,	/* corresponding ::rnext */
 		ao_required	:1,	/* don't accept non-AO connects */
-		reserved	:29;	/* must be 0 */
+		set_counters	:1,	/* set/clear ::pkt_* counters */
+		reserved	:28;	/* must be 0 */
+	__u16	reserved2;		/* padding, must be 0 */
 	__u8	current_key;		/* KeyID to set as Current_key */
 	__u8	rnext;			/* KeyID to set as Rnext_key */
+	__u64	pkt_good;		/* verified segments */
+	__u64	pkt_bad;		/* failed verification */
+	__u64	pkt_key_not_found;	/* could not find a key to verify */
+	__u64	pkt_ao_required;	/* segments missing TCP-AO sign */
 } __attribute__((aligned(8)));
 
 /* setsockopt(fd, IPPROTO_TCP, TCP_ZEROCOPY_RECEIVE, ...) */
net/ipv4/proc.c +4 -0
index eaf1d3113b62..3f643cd29cfe 100644
@@ -298,6 +298,10 @@ static const struct snmp_mib snmp4_net_list[] = {
 	SNMP_MIB_ITEM("TCPMigrateReqSuccess", LINUX_MIB_TCPMIGRATEREQSUCCESS),
 	SNMP_MIB_ITEM("TCPMigrateReqFailure", LINUX_MIB_TCPMIGRATEREQFAILURE),
 	SNMP_MIB_ITEM("TCPPLBRehash", LINUX_MIB_TCPPLBREHASH),
+	SNMP_MIB_ITEM("TCPAORequired", LINUX_MIB_TCPAOREQUIRED),
+	SNMP_MIB_ITEM("TCPAOBad", LINUX_MIB_TCPAOBAD),
+	SNMP_MIB_ITEM("TCPAOKeyNotFound", LINUX_MIB_TCPAOKEYNOTFOUND),
+	SNMP_MIB_ITEM("TCPAOGood", LINUX_MIB_TCPAOGOOD),
 	SNMP_MIB_SENTINEL
 };
 
net/ipv4/tcp_ao.c +27 -3
index 073be1680818..5ab66928a952 100644
@@ -184,6 +184,8 @@ static struct tcp_ao_key *tcp_ao_copy_key(struct sock *sk,
 	*new_key = *key;
 	INIT_HLIST_NODE(&new_key->node);
 	tcp_sigpool_get(new_key->tcp_sigpool_id);
+	atomic64_set(&new_key->pkt_good, 0);
+	atomic64_set(&new_key->pkt_bad, 0);
 
 	return new_key;
 }
@@ -745,8 +747,12 @@ tcp_ao_verify_hash(const struct sock *sk, const struct sk_buff *skb,
 	const struct tcphdr *th = tcp_hdr(skb);
 	void *hash_buf = NULL;
 
-	if (maclen != tcp_ao_maclen(key))
+	if (maclen != tcp_ao_maclen(key)) {
+		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOBAD);
+		atomic64_inc(&info->counters.pkt_bad);
+		atomic64_inc(&key->pkt_bad);
 		return SKB_DROP_REASON_TCP_AOFAILURE;
+	}
 
 	hash_buf = kmalloc(tcp_ao_digest_size(key), GFP_ATOMIC);
 	if (!hash_buf)
@@ -756,9 +762,15 @@ tcp_ao_verify_hash(const struct sock *sk, const struct sk_buff *skb,
 	tcp_ao_hash_skb(family, hash_buf, key, sk, skb, traffic_key,
 			(phash - (u8 *)th), sne);
 	if (memcmp(phash, hash_buf, maclen)) {
+		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOBAD);
+		atomic64_inc(&info->counters.pkt_bad);
+		atomic64_inc(&key->pkt_bad);
 		kfree(hash_buf);
 		return SKB_DROP_REASON_TCP_AOFAILURE;
 	}
+	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOGOOD);
+	atomic64_inc(&info->counters.pkt_good);
+	atomic64_inc(&key->pkt_good);
 	kfree(hash_buf);
 	return SKB_NOT_DROPPED_YET;
 }
@@ -778,8 +790,10 @@ tcp_inbound_ao_hash(struct sock *sk, const struct sk_buff *skb,
 	u32 sne = 0;
 
 	info = rcu_dereference(tcp_sk(sk)->ao_info);
-	if (!info)
+	if (!info) {
+		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOKEYNOTFOUND);
 		return SKB_DROP_REASON_TCP_AOUNEXPECTED;
+	}
 
 	if (unlikely(th->syn)) {
 		sisn = th->seq;
@@ -875,6 +889,8 @@ tcp_inbound_ao_hash(struct sock *sk, const struct sk_buff *skb,
 	return ret;
 
 key_not_found:
+	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOKEYNOTFOUND);
+	atomic64_inc(&info->counters.key_not_found);
 	return SKB_DROP_REASON_TCP_AOKEYNOTFOUND;
 }
 
@@ -1460,6 +1476,8 @@ static int tcp_ao_add_cmd(struct sock *sk, unsigned short int family,
 	key->keyflags	= cmd.keyflags;
 	key->sndid	= cmd.sndid;
 	key->rcvid	= cmd.rcvid;
+	atomic64_set(&key->pkt_good, 0);
+	atomic64_set(&key->pkt_bad, 0);
 
 	ret = tcp_ao_parse_crypto(&cmd, key);
 	if (ret < 0)
@@ -1673,7 +1691,7 @@ static int tcp_ao_info_cmd(struct sock *sk, unsigned short int family,
 			return -EINVAL;
 	}
 
-	if (cmd.reserved != 0)
+	if (cmd.reserved != 0 || cmd.reserved2 != 0)
 		return -EINVAL;
 
 	ao_info = setsockopt_ao_info(sk);
@@ -1708,6 +1726,12 @@ static int tcp_ao_info_cmd(struct sock *sk, unsigned short int family,
 			goto out;
 		}
 	}
+	if (cmd.set_counters) {
+		atomic64_set(&ao_info->counters.pkt_good, cmd.pkt_good);
+		atomic64_set(&ao_info->counters.pkt_bad, cmd.pkt_bad);
+		atomic64_set(&ao_info->counters.key_not_found, cmd.pkt_key_not_found);
+		atomic64_set(&ao_info->counters.ao_required, cmd.pkt_ao_required);
+	}
 
 	ao_info->ao_required = cmd.ao_required;
 	if (new_current)
net/ipv4/tcp_ipv4.c +1 -1
index f09c0389ae11..d8d2499b5e26 100644
@@ -1528,7 +1528,7 @@ static int tcp_v4_parse_md5_keys(struct sock *sk, int optname,
 	/* Don't allow keys for peers that have a matching TCP-AO key.
 	 * See the comment in tcp_ao_add_cmd()
 	 */
-	if (tcp_ao_required(sk, addr, AF_INET))
+	if (tcp_ao_required(sk, addr, AF_INET, false))
 		return -EKEYREJECTED;
 
 	return tcp_md5_do_add(sk, addr, AF_INET, prefixlen, l3index, flags,
net/ipv6/tcp_ipv6.c +2 -2
index 40266a3a99a6..c825d7c80f5b 100644
@@ -660,7 +660,7 @@ static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 		/* Don't allow keys for peers that have a matching TCP-AO key.
 		 * See the comment in tcp_ao_add_cmd()
 		 */
-		if (tcp_ao_required(sk, addr, AF_INET))
+		if (tcp_ao_required(sk, addr, AF_INET, false))
 			return -EKEYREJECTED;
 		return tcp_md5_do_add(sk, addr,
 				      AF_INET, prefixlen, l3index, flags,
@@ -672,7 +672,7 @@ static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 	/* Don't allow keys for peers that have a matching TCP-AO key.
 	 * See the comment in tcp_ao_add_cmd()
 	 */
-	if (tcp_ao_required(sk, addr, AF_INET6))
+	if (tcp_ao_required(sk, addr, AF_INET6, false))
 		return -EKEYREJECTED;
 
 	return tcp_md5_do_add(sk, addr, AF_INET6, prefixlen, l3index, flags,
Signature · 1 line
2.41.0
#23 lore raw

PATCH 14/23 net/tcp: Add TCP-AO SNE support

In reply to Dmitry Safonov

Add Sequence Number Extension (SNE) for TCP-AO.
This is needed to protect long-living TCP-AO connections from replaying attacks after sequence number roll-over, see RFC5925 (6.2).

---
 include/net/tcp_ao.h     | 20 ++++++++++++++++++++
 net/ipv4/tcp_ao.c        | 40 +++++++++++++++++++++++++++++++++-------
 net/ipv4/tcp_input.c     | 28 ++++++++++++++++++++++++++++
 net/ipv4/tcp_ipv4.c      |  1 +
 net/ipv4/tcp_minisocks.c | 15 ++++++++++++++-
 net/ipv4/tcp_output.c    |  6 +++++-
 net/ipv6/tcp_ipv6.c      |  1 +
 7 files changed, 102 insertions(+), 9 deletions(-)
include/net/tcp_ao.h +20 -0
index 0a02098bddeb..fc5674763730 100644
@@ -96,6 +96,25 @@ struct tcp_ao_info {
 				__unused	:31;
 	__be32			lisn;
 	__be32			risn;
+	/* Sequence Number Extension (SNE) are upper 4 bytes for SEQ,
+	 * that protect TCP-AO connection from replayed old TCP segments.
+	 * See RFC5925 (6.2).
+	 * In order to get correct SNE, there's a helper tcp_ao_compute_sne().
+	 * It needs SEQ basis to understand whereabouts are lower SEQ numbers.
+	 * According to that basis vector, it can provide incremented SNE
+	 * when SEQ rolls over or provide decremented SNE when there's
+	 * a retransmitted segment from before-rolling over.
+	 * - for request sockets such basis is rcv_isn/snt_isn, which seems
+	 *   good enough as it's unexpected to receive 4 Gbytes on reqsk.
+	 * - for full sockets the basis is rcv_nxt/snd_una. snd_una is
+	 *   taken instead of snd_nxt as currently it's easier to track
+	 *   in tcp_snd_una_update(), rather than updating SNE in all
+	 *   WRITE_ONCE(tp->snd_nxt, ...)
+	 * - for time-wait sockets the basis is tw_rcv_nxt/tw_snd_nxt.
+	 *   tw_snd_nxt is not expected to change, while tw_rcv_nxt may.
+	 */
+	u32			snd_sne;
+	u32			rcv_sne;
 	atomic_t		refcnt;		/* Protects twsk destruction */
 	struct rcu_head		rcu;
 };
@@ -144,6 +163,7 @@ enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
 			const struct sk_buff *skb, unsigned short int family,
 			const struct request_sock *req,
 			const struct tcp_ao_hdr *aoh);
+u32 tcp_ao_compute_sne(u32 next_sne, u32 next_seq, u32 seq);
 struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
 				    const union tcp_ao_addr *addr,
 				    int family, int sndid, int rcvid, u16 port);
net/ipv4/tcp_ao.c +33 -7
index 5ab66928a952..5ebbfadfbfd0 100644
@@ -403,6 +403,21 @@ static int tcp_ao_hash_pseudoheader(unsigned short int family,
 	return -EAFNOSUPPORT;
 }
 
+u32 tcp_ao_compute_sne(u32 next_sne, u32 next_seq, u32 seq)
+{
+	u32 sne = next_sne;
+
+	if (before(seq, next_seq)) {
+		if (seq > next_seq)
+			sne--;
+	} else {
+		if (seq < next_seq)
+			sne++;
+	}
+
+	return sne;
+}
+
 /* tcp_ao_hash_sne(struct tcp_sigpool *hp)
  * @hp	- used for hashing
  * @sne - sne value
@@ -641,7 +656,8 @@ int tcp_ao_prepare_reset(const struct sock *sk, struct sk_buff *skb,
 
 			sisn = htonl(tcp_rsk(req)->rcv_isn);
 			disn = htonl(tcp_rsk(req)->snt_isn);
-			*sne = 0;
+			*sne = tcp_ao_compute_sne(0, tcp_rsk(req)->snt_isn,
+						  ntohl(seq));
 		} else {
 			sisn = th->seq;
 			disn = 0;
@@ -674,11 +690,15 @@ int tcp_ao_prepare_reset(const struct sock *sk, struct sk_buff *skb,
 		*keyid = (*key)->rcvid;
 	} else {
 		struct tcp_ao_key *rnext_key;
+		u32 snd_basis;
 
-		if (sk->sk_state == TCP_TIME_WAIT)
+		if (sk->sk_state == TCP_TIME_WAIT) {
 			ao_info = rcu_dereference(tcp_twsk(sk)->ao_info);
-		else
+			snd_basis = tcp_twsk(sk)->tw_snd_nxt;
+		} else {
 			ao_info = rcu_dereference(tcp_sk(sk)->ao_info);
+			snd_basis = tcp_sk(sk)->snd_una;
+		}
 		if (!ao_info)
 			return -ENOENT;
 
@@ -688,7 +708,8 @@ int tcp_ao_prepare_reset(const struct sock *sk, struct sk_buff *skb,
 		*traffic_key = snd_other_key(*key);
 		rnext_key = READ_ONCE(ao_info->rnext_key);
 		*keyid = rnext_key->rcvid;
-		*sne = 0;
+		*sne = tcp_ao_compute_sne(READ_ONCE(ao_info->snd_sne),
+					  snd_basis, ntohl(seq));
 	}
 	return 0;
 }
@@ -821,7 +842,8 @@ tcp_inbound_ao_hash(struct sock *sk, const struct sk_buff *skb,
 		if (unlikely(th->syn && !th->ack))
 			goto verify_hash;
 
-		sne = 0;
+		sne = tcp_ao_compute_sne(info->rcv_sne, tcp_sk(sk)->rcv_nxt,
+					 ntohl(th->seq));
 		/* Established socket, traffic key are cached */
 		traffic_key = rcv_other_key(key);
 		err = tcp_ao_verify_hash(sk, skb, family, info, aoh, key,
@@ -856,14 +878,16 @@ tcp_inbound_ao_hash(struct sock *sk, const struct sk_buff *skb,
 	if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV)) {
 		/* Make the initial syn the likely case here */
 		if (unlikely(req)) {
-			sne = 0;
+			sne = tcp_ao_compute_sne(0, tcp_rsk(req)->rcv_isn,
+						 ntohl(th->seq));
 			sisn = htonl(tcp_rsk(req)->rcv_isn);
 			disn = htonl(tcp_rsk(req)->snt_isn);
 		} else if (unlikely(th->ack && !th->syn)) {
 			/* Possible syncookie packet */
 			sisn = htonl(ntohl(th->seq) - 1);
 			disn = htonl(ntohl(th->ack_seq) - 1);
-			sne = 0;
+			sne = tcp_ao_compute_sne(0, ntohl(sisn),
+						 ntohl(th->seq));
 		} else if (unlikely(!th->syn)) {
 			/* no way to figure out initial sisn/disn - drop */
 			return SKB_DROP_REASON_TCP_FLAGS;
@@ -963,6 +987,7 @@ void tcp_ao_connect_init(struct sock *sk)
 		tp->tcp_header_len += tcp_ao_len(key);
 
 		ao_info->lisn = htonl(tp->write_seq);
+		ao_info->snd_sne = 0;
 	} else {
 		/* TODO: probably, it should fail to connect() here */
 		rcu_assign_pointer(tp->ao_info, NULL);
@@ -981,6 +1006,7 @@ void tcp_ao_finish_connect(struct sock *sk, struct sk_buff *skb)
 		return;
 
 	ao->risn = tcp_hdr(skb)->seq;
+	ao->rcv_sne = 0;
 
 	hlist_for_each_entry_rcu(key, &ao->head, node)
 		tcp_ao_cache_traffic_keys(sk, ao, key);
net/ipv4/tcp_input.c +28 -0
index 2a7ed329f7cb..4b37e82ab612 100644
@@ -3532,9 +3532,18 @@ static inline bool tcp_may_update_window(const struct tcp_sock *tp,
 static void tcp_snd_una_update(struct tcp_sock *tp, u32 ack)
 {
 	u32 delta = ack - tp->snd_una;
+#ifdef CONFIG_TCP_AO
+	struct tcp_ao_info *ao;
+#endif
 
 	sock_owned_by_me((struct sock *)tp);
 	tp->bytes_acked += delta;
+#ifdef CONFIG_TCP_AO
+	ao = rcu_dereference_protected(tp->ao_info,
+				       lockdep_sock_is_held((struct sock *)tp));
+	if (ao && ack < tp->snd_una)
+		ao->snd_sne++;
+#endif
 	tp->snd_una = ack;
 }
 
@@ -3542,9 +3551,18 @@ static void tcp_snd_una_update(struct tcp_sock *tp, u32 ack)
 static void tcp_rcv_nxt_update(struct tcp_sock *tp, u32 seq)
 {
 	u32 delta = seq - tp->rcv_nxt;
+#ifdef CONFIG_TCP_AO
+	struct tcp_ao_info *ao;
+#endif
 
 	sock_owned_by_me((struct sock *)tp);
 	tp->bytes_received += delta;
+#ifdef CONFIG_TCP_AO
+	ao = rcu_dereference_protected(tp->ao_info,
+				       lockdep_sock_is_held((struct sock *)tp));
+	if (ao && seq < tp->rcv_nxt)
+		ao->rcv_sne++;
+#endif
 	WRITE_ONCE(tp->rcv_nxt, seq);
 }
 
@@ -6387,6 +6405,16 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 		 * simultaneous connect with crossed SYNs.
 		 * Particularly, it can be connect to self.
 		 */
+#ifdef CONFIG_TCP_AO
+		struct tcp_ao_info *ao;
+
+		ao = rcu_dereference_protected(tp->ao_info,
+					       lockdep_sock_is_held(sk));
+		if (ao) {
+			ao->risn = th->seq;
+			ao->rcv_sne = 0;
+		}
+#endif
 		tcp_set_state(sk, TCP_SYN_RECV);
 
 		if (tp->rx_opt.saw_tstamp) {
net/ipv4/tcp_ipv4.c +1 -0
index d8d2499b5e26..37497dec5eb9 100644
@@ -1039,6 +1039,7 @@ static void tcp_v4_timewait_ack(struct sock *sk, struct sk_buff *skb)
 		struct tcp_ao_key *rnext_key;
 
 		traffic_key = snd_other_key(ao_key);
+		ao_sne = READ_ONCE(ao_info->snd_sne);
 		rnext_key = READ_ONCE(ao_info->rnext_key);
 		rcv_next = rnext_key->rcvid;
 	}
net/ipv4/tcp_minisocks.c +14 -1
index 11b76fa601a7..3b367365dc7d 100644
@@ -51,6 +51,18 @@ tcp_timewait_check_oow_rate_limit(struct inet_timewait_sock *tw,
 	return TCP_TW_SUCCESS;
 }
 
+static void twsk_rcv_nxt_update(struct tcp_timewait_sock *tcptw, u32 seq)
+{
+#ifdef CONFIG_TCP_AO
+	struct tcp_ao_info *ao;
+
+	ao = rcu_dereference(tcptw->ao_info);
+	if (unlikely(ao && seq < tcptw->tw_rcv_nxt))
+		WRITE_ONCE(ao->rcv_sne, ao->rcv_sne + 1);
+#endif
+	tcptw->tw_rcv_nxt = seq;
+}
+
 /*
  * * Main purpose of TIME-WAIT state is to close connection gracefully,
  *   when one of ends sits in LAST-ACK or CLOSING retransmitting FIN
@@ -136,7 +148,8 @@ tcp_timewait_state_process(struct inet_timewait_sock *tw, struct sk_buff *skb,
 
 		/* FIN arrived, enter true time-wait state. */
 		tw->tw_substate	  = TCP_TIME_WAIT;
-		tcptw->tw_rcv_nxt = TCP_SKB_CB(skb)->end_seq;
+		twsk_rcv_nxt_update(tcptw, TCP_SKB_CB(skb)->end_seq);
+
 		if (tmp_opt.saw_tstamp) {
 			tcptw->tw_ts_recent_stamp = ktime_get_seconds();
 			tcptw->tw_ts_recent	  = tmp_opt.rcv_tsval;
net/ipv4/tcp_output.c +5 -1
index ed1164f287b7..d87a3823fec5 100644
@@ -1443,6 +1443,7 @@ static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb,
 		u8 *traffic_key;
 		void *tkey_buf = NULL;
 		u32 disn;
+		u32 sne;
 
 		sk_gso_disable(sk);
 		if (unlikely(tcb->tcp_flags & TCPHDR_SYN)) {
@@ -1460,9 +1461,12 @@ static int __tcp_transmit_skb(struct sock *sk, struct sk_buff *skb,
 		} else {
 			traffic_key = snd_other_key(ao_key);
 		}
+		sne = tcp_ao_compute_sne(READ_ONCE(ao->snd_sne),
+					 READ_ONCE(tp->snd_una),
+					 ntohl(th->seq));
 		tp->af_specific->calc_ao_hash(opts.hash_location, ao_key, sk, skb,
 					      traffic_key,
-					      opts.hash_location - (u8 *)th, 0);
+					      opts.hash_location - (u8 *)th, sne);
 		kfree(tkey_buf);
 	}
 #endif
net/ipv6/tcp_ipv6.c +1 -0
index c825d7c80f5b..873ce0d72f34 100644
@@ -1180,6 +1180,7 @@ static void tcp_v6_timewait_ack(struct sock *sk, struct sk_buff *skb)
 		/* rcv_next switches to our rcv_next */
 		rnext_key = READ_ONCE(ao_info->rnext_key);
 		rcv_next = rnext_key->rcvid;
+		ao_sne = READ_ONCE(ao_info->snd_sne);
 	}
 #endif
 
Signature · 1 line
2.41.0
#24 lore raw

PATCH 15/23 net/tcp: Add tcp_hash_fail() ratelimited logs

In reply to Dmitry Safonov

Add a helper for logging connection-detailed messages for failed TCP hash verification (both MD5 and AO).

---
 include/net/tcp.h    | 14 ++++++++++++--
 include/net/tcp_ao.h | 29 +++++++++++++++++++++++++++++
 net/ipv4/tcp.c       | 23 +++++++++++++----------
 net/ipv4/tcp_ao.c    |  7 +++++++
 4 files changed, 61 insertions(+), 12 deletions(-)
include/net/tcp.h +12 -2
index a58f95237ea3..50636ee7ee2f 100644
@@ -2641,12 +2641,18 @@ tcp_inbound_hash(struct sock *sk, const struct request_sock *req,
 	int l3index;
 
 	/* Invalid option or two times meet any of auth options */
-	if (tcp_parse_auth_options(th, &md5_location, &aoh))
+	if (tcp_parse_auth_options(th, &md5_location, &aoh)) {
+		tcp_hash_fail("TCP segment has incorrect auth options set",
+				family, skb, "");
 		return SKB_DROP_REASON_TCP_AUTH_HDR;
+	}
 
 	if (req) {
 		if (tcp_rsk_used_ao(req) != !!aoh) {
 			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOBAD);
+			tcp_hash_fail("TCP connection can't start/end using TCP-AO",
+					family, skb, "%s",
+					!aoh ? "missing AO" : "AO signed");
 			return SKB_DROP_REASON_TCP_AOFAILURE;
 		}
 	}
@@ -2663,10 +2669,14 @@ tcp_inbound_hash(struct sock *sk, const struct request_sock *req,
 		 * the last key is impossible to remove, so there's
 		 * always at least one current_key.
 		 */
-		if (tcp_ao_required(sk, saddr, family, true))
+		if (tcp_ao_required(sk, saddr, family, true)) {
+			tcp_hash_fail("AO hash is required, but not found",
+					family, skb, "L3 index %d", l3index);
 			return SKB_DROP_REASON_TCP_AONOTFOUND;
+		}
 		if (unlikely(tcp_md5_do_lookup(sk, l3index, saddr, family))) {
 			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
+			tcp_hash_fail("MD5 Hash not found", family, skb, "");
 			return SKB_DROP_REASON_TCP_MD5NOTFOUND;
 		}
 		return SKB_NOT_DROPPED_YET;
include/net/tcp_ao.h +29 -0
index fc5674763730..801dca405df7 100644
@@ -119,6 +119,35 @@ struct tcp_ao_info {
 	struct rcu_head		rcu;
 };
 
+#define tcp_hash_fail(msg, family, skb, fmt, ...)			\
+do {									\
+	const struct tcphdr *th = tcp_hdr(skb);				\
+	char hdr_flags[5] = {};						\
+	char *f = hdr_flags;						\
+									\
+	if (th->fin)							\
+		*f++ = 'F';						\
+	if (th->syn)							\
+		*f++ = 'S';						\
+	if (th->rst)							\
+		*f++ = 'R';						\
+	if (th->ack)							\
+		*f++ = 'A';						\
+	if (f != hdr_flags)						\
+		*f = ' ';						\
+	if (family == AF_INET) {					\
+		net_info_ratelimited("%s for (%pI4, %d)->(%pI4, %d) %s" fmt "\n", \
+				msg, &ip_hdr(skb)->saddr, ntohs(th->source), \
+				&ip_hdr(skb)->daddr, ntohs(th->dest),	\
+				hdr_flags, ##__VA_ARGS__);		\
+	} else {							\
+		net_info_ratelimited("%s for [%pI6c]:%u->[%pI6c]:%u %s" fmt "\n", \
+				msg, &ipv6_hdr(skb)->saddr, ntohs(th->source), \
+				&ipv6_hdr(skb)->daddr, ntohs(th->dest),	\
+				hdr_flags, ##__VA_ARGS__);		\
+	}								\
+} while (0)
+
 #ifdef CONFIG_TCP_AO
 /* TCP-AO structures and functions */
 
net/ipv4/tcp.c +13 -10
index d7e18c528c7b..37d3eac0561b 100644
@@ -4380,7 +4380,6 @@ tcp_inbound_md5_hash(const struct sock *sk, const struct sk_buff *skb,
 	 * o MD5 hash and we're not expecting one.
 	 * o MD5 hash and its wrong.
 	 */
-	const struct tcphdr *th = tcp_hdr(skb);
 	const struct tcp_sock *tp = tcp_sk(sk);
 	struct tcp_md5sig_key *key;
 	int genhash;
@@ -4390,6 +4389,7 @@ tcp_inbound_md5_hash(const struct sock *sk, const struct sk_buff *skb,
 
 	if (!key && hash_location) {
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5UNEXPECTED);
+		tcp_hash_fail("Unexpected MD5 Hash found", family, skb, "");
 		return SKB_DROP_REASON_TCP_MD5UNEXPECTED;
 	}
 
@@ -4405,16 +4405,19 @@ tcp_inbound_md5_hash(const struct sock *sk, const struct sk_buff *skb,
 	if (genhash || memcmp(hash_location, newhash, 16) != 0) {
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5FAILURE);
 		if (family == AF_INET) {
-			net_info_ratelimited("MD5 Hash failed for (%pI4, %d)->(%pI4, %d)%s L3 index %d\n",
-					saddr, ntohs(th->source),
-					daddr, ntohs(th->dest),
-					genhash ? " tcp_v4_calc_md5_hash failed"
-					: "", l3index);
+			tcp_hash_fail("MD5 Hash failed", AF_INET, skb, "%s L3 index %d",
+				      genhash ? "tcp_v4_calc_md5_hash failed"
+				      : "", l3index);
 		} else {
-			net_info_ratelimited("MD5 Hash %s for [%pI6c]:%u->[%pI6c]:%u L3 index %d\n",
-					genhash ? "failed" : "mismatch",
-					saddr, ntohs(th->source),
-					daddr, ntohs(th->dest), l3index);
+			if (genhash) {
+				tcp_hash_fail("MD5 Hash failed",
+					      AF_INET6, skb, "L3 index %d",
+					      l3index);
+			} else {
+				tcp_hash_fail("MD5 Hash mismatch",
+					      AF_INET6, skb, "L3 index %d",
+					      l3index);
+			}
 		}
 		return SKB_DROP_REASON_TCP_MD5FAILURE;
 	}
net/ipv4/tcp_ao.c +7 -0
index 5ebbfadfbfd0..9ee7fc9f6a76 100644
@@ -772,6 +772,8 @@ tcp_ao_verify_hash(const struct sock *sk, const struct sk_buff *skb,
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOBAD);
 		atomic64_inc(&info->counters.pkt_bad);
 		atomic64_inc(&key->pkt_bad);
+		tcp_hash_fail("AO hash wrong length", family, skb,
+			      "%u != %d", maclen, tcp_ao_maclen(key));
 		return SKB_DROP_REASON_TCP_AOFAILURE;
 	}
 
@@ -786,6 +788,7 @@ tcp_ao_verify_hash(const struct sock *sk, const struct sk_buff *skb,
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOBAD);
 		atomic64_inc(&info->counters.pkt_bad);
 		atomic64_inc(&key->pkt_bad);
+		tcp_hash_fail("AO hash mismatch", family, skb, "");
 		kfree(hash_buf);
 		return SKB_DROP_REASON_TCP_AOFAILURE;
 	}
@@ -813,6 +816,8 @@ tcp_inbound_ao_hash(struct sock *sk, const struct sk_buff *skb,
 	info = rcu_dereference(tcp_sk(sk)->ao_info);
 	if (!info) {
 		NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOKEYNOTFOUND);
+		tcp_hash_fail("AO key not found", family, skb,
+			      "keyid: %u", aoh->keyid);
 		return SKB_DROP_REASON_TCP_AOUNEXPECTED;
 	}
 
@@ -915,6 +920,8 @@ tcp_inbound_ao_hash(struct sock *sk, const struct sk_buff *skb,
 key_not_found:
 	NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOKEYNOTFOUND);
 	atomic64_inc(&info->counters.key_not_found);
+	tcp_hash_fail("Requested by the peer AO key id not found",
+		      family, skb, "");
 	return SKB_DROP_REASON_TCP_AOKEYNOTFOUND;
 }
 
Signature · 1 line
2.41.0
#25 lore raw

PATCH 16/23 net/tcp: Ignore specific ICMPs for TCP-AO connections

In reply to Dmitry Safonov

Similarly to IPsec, RFC5925 prescribes:
">> A TCP-AO implementation MUST default to ignore incoming ICMPv4 messages of Type 3 (destination unreachable), Codes 2-4 (protocol unreachable, port unreachable, and fragmentation needed -- ’hard errors’), and ICMPv6 Type 1 (destination unreachable), Code 1 (administratively prohibited) and Code 4 (port unreachable) intended for connections in synchronized states (ESTABLISHED, FIN-WAIT-1, FIN- WAIT-2, CLOSE-WAIT, CLOSING, LAST-ACK, TIME-WAIT) that match MKTs."

A selftest (later in patch series) verifies that this attack is not possible in this TCP-AO implementation.

---
 include/net/tcp_ao.h      | 10 ++++++-
 include/uapi/linux/snmp.h |  1 +
 include/uapi/linux/tcp.h  |  4 ++-
 net/ipv4/proc.c           |  1 +
 net/ipv4/tcp_ao.c         | 61 +++++++++++++++++++++++++++++++++++++++
 net/ipv4/tcp_ipv4.c       |  5 ++++
 net/ipv6/tcp_ipv6.c       |  4 +++
 7 files changed, 84 insertions(+), 2 deletions(-)
include/net/tcp_ao.h +9 -1
index 801dca405df7..cb16af2a9f39 100644
@@ -24,6 +24,7 @@ struct tcp_ao_counters {
 	atomic64_t	pkt_bad;
 	atomic64_t	key_not_found;
 	atomic64_t	ao_required;
+	atomic64_t	dropped_icmp;
 };
 
 struct tcp_ao_key {
@@ -93,7 +94,8 @@ struct tcp_ao_info {
 	struct tcp_ao_key	*rnext_key;
 	struct tcp_ao_counters	counters;
 	u32			ao_required	:1,
-				__unused	:31;
+				accept_icmps	:1,
+				__unused	:30;
 	__be32			lisn;
 	__be32			risn;
 	/* Sequence Number Extension (SNE) are upper 4 bytes for SEQ,
@@ -188,6 +190,7 @@ int tcp_ao_calc_traffic_key(struct tcp_ao_key *mkt, u8 *key, void *ctx,
 			    unsigned int len, struct tcp_sigpool *hp);
 void tcp_ao_destroy_sock(struct sock *sk, bool twsk);
 void tcp_ao_time_wait(struct tcp_timewait_sock *tcptw, struct tcp_sock *tp);
+bool tcp_ao_ignore_icmp(struct sock *sk, int type, int code);
 enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
 			const struct sk_buff *skb, unsigned short int family,
 			const struct request_sock *req,
@@ -266,6 +269,11 @@ static inline void tcp_ao_syncookie(struct sock *sk, const struct sk_buff *skb,
 {
 }
 
+static inline bool tcp_ao_ignore_icmp(struct sock *sk, int type, int code)
+{
+	return false;
+}
+
 static inline enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
 		const struct sk_buff *skb, unsigned short int family,
 		const struct request_sock *req, const struct tcp_ao_hdr *aoh)
include/uapi/linux/snmp.h +1 -0
index 06ddf4cd295c..47a6b47da66f 100644
@@ -300,6 +300,7 @@ enum
 	LINUX_MIB_TCPAOBAD,			/* TCPAOBad */
 	LINUX_MIB_TCPAOKEYNOTFOUND,		/* TCPAOKeyNotFound */
 	LINUX_MIB_TCPAOGOOD,			/* TCPAOGood */
+	LINUX_MIB_TCPAODROPPEDICMPS,		/* TCPAODroppedIcmps */
 	__LINUX_MIB_MAX
 };
 
include/uapi/linux/tcp.h +3 -1
index 3fe0612ec59a..ca7ed18ce67b 100644
@@ -392,7 +392,8 @@ struct tcp_ao_info_opt { /* setsockopt(TCP_AO_INFO) */
 		set_rnext	:1,	/* corresponding ::rnext */
 		ao_required	:1,	/* don't accept non-AO connects */
 		set_counters	:1,	/* set/clear ::pkt_* counters */
-		reserved	:28;	/* must be 0 */
+		accept_icmps	:1,	/* accept incoming ICMPs */
+		reserved	:27;	/* must be 0 */
 	__u16	reserved2;		/* padding, must be 0 */
 	__u8	current_key;		/* KeyID to set as Current_key */
 	__u8	rnext;			/* KeyID to set as Rnext_key */
@@ -400,6 +401,7 @@ struct tcp_ao_info_opt { /* setsockopt(TCP_AO_INFO) */
 	__u64	pkt_bad;		/* failed verification */
 	__u64	pkt_key_not_found;	/* could not find a key to verify */
 	__u64	pkt_ao_required;	/* segments missing TCP-AO sign */
+	__u64	pkt_dropped_icmp;	/* ICMPs that were ignored */
 } __attribute__((aligned(8)));
 
 /* setsockopt(fd, IPPROTO_TCP, TCP_ZEROCOPY_RECEIVE, ...) */
net/ipv4/proc.c +1 -0
index 3f643cd29cfe..5d3c9c96773e 100644
@@ -302,6 +302,7 @@ static const struct snmp_mib snmp4_net_list[] = {
 	SNMP_MIB_ITEM("TCPAOBad", LINUX_MIB_TCPAOBAD),
 	SNMP_MIB_ITEM("TCPAOKeyNotFound", LINUX_MIB_TCPAOKEYNOTFOUND),
 	SNMP_MIB_ITEM("TCPAOGood", LINUX_MIB_TCPAOGOOD),
+	SNMP_MIB_ITEM("TCPAODroppedIcmps", LINUX_MIB_TCPAODROPPEDICMPS),
 	SNMP_MIB_SENTINEL
 };
 
net/ipv4/tcp_ao.c +61 -0

88 lines not on this page · raw message

net/ipv4/tcp_ipv4.c +5 -0
index 37497dec5eb9..054897a13575 100644
@@ -494,6 +494,8 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
 		return -ENOENT;
 	}
 	if (sk->sk_state == TCP_TIME_WAIT) {
+		/* To increase the counter of ignored icmps for TCP-AO */
+		tcp_ao_ignore_icmp(sk, type, code);
 		inet_twsk_put(inet_twsk(sk));
 		return 0;
 	}
@@ -508,6 +510,9 @@ int tcp_v4_err(struct sk_buff *skb, u32 info)
 	}
 
 	bh_lock_sock(sk);
+	if (tcp_ao_ignore_icmp(sk, type, code))
+		goto out;
+
 	/* If too many ICMPs get dropped on busy
 	 * servers this needs to be solved differently.
 	 * We do take care of PMTU discovery (RFC1191) special case :
net/ipv6/tcp_ipv6.c +4 -0
index 873ce0d72f34..9a427c0d48ab 100644
@@ -395,6 +395,8 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 	}
 
 	if (sk->sk_state == TCP_TIME_WAIT) {
+		/* To increase the counter of ignored icmps for TCP-AO */
+		tcp_ao_ignore_icmp(sk, type, code);
 		inet_twsk_put(inet_twsk(sk));
 		return 0;
 	}
@@ -406,6 +408,8 @@ static int tcp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
 	}
 
 	bh_lock_sock(sk);
+	if (tcp_ao_ignore_icmp(sk, type, code))
+		goto out;
 	if (sock_owned_by_user(sk) && type != ICMPV6_PKT_TOOBIG)
 		__NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
 
Signature · 1 line
2.41.0
#26 lore raw

PATCH 17/23 net/tcp: Add option for TCP-AO to (not) hash header

In reply to Dmitry Safonov

Provide setsockopt() key flag that makes TCP-AO exclude hashing TCP header for peers that match the key. This is needed for interraction with middleboxes that may change TCP options, see RFC5925 (9.2).

---
 include/uapi/linux/tcp.h | 5 +++++
 net/ipv4/tcp_ao.c        | 8 +++++---
 2 files changed, 10 insertions(+), 3 deletions(-)
include/uapi/linux/tcp.h +5 -0
index ca7ed18ce67b..3275ade3293a 100644
@@ -354,6 +354,11 @@ struct tcp_diag_md5sig {
 #define TCP_AO_MAXKEYLEN	80
 
 #define TCP_AO_KEYF_IFINDEX	(1 << 0)	/* L3 ifindex for VRF */
+#define TCP_AO_KEYF_EXCLUDE_OPT	(1 << 1)	/* "Indicates whether TCP
+						 *  options other than TCP-AO
+						 *  are included in the MAC
+						 *  calculation"
+						 */
 
 struct tcp_ao_add { /* setsockopt(TCP_AO_ADD_KEY) */
 	struct __kernel_sockaddr_storage addr;	/* peer's address for the key */
net/ipv4/tcp_ao.c +5 -3
index 4bde58ede63a..4fabd3ec5c71 100644
@@ -567,7 +567,8 @@ int tcp_ao_hash_hdr(unsigned short int family, char *ao_hash,
 		WARN_ON_ONCE(1);
 		goto clear_hash;
 	}
-	if (tcp_ao_hash_header(&hp, th, false,
+	if (tcp_ao_hash_header(&hp, th,
+			       !!(key->keyflags & TCP_AO_KEYF_EXCLUDE_OPT),
 			       ao_hash, hash_offset, tcp_ao_maclen(key)))
 		goto clear_hash;
 	ahash_request_set_crypt(hp.req, NULL, hash_buf, 0);
@@ -615,7 +616,8 @@ int tcp_ao_hash_skb(unsigned short int family,
 		goto clear_hash;
 	if (tcp_ao_hash_pseudoheader(family, sk, skb, &hp, skb->len))
 		goto clear_hash;
-	if (tcp_ao_hash_header(&hp, th, false,
+	if (tcp_ao_hash_header(&hp, th,
+			       !!(key->keyflags & TCP_AO_KEYF_EXCLUDE_OPT),
 			       ao_hash, hash_offset, tcp_ao_maclen(key)))
 		goto clear_hash;
 	if (tcp_sigpool_hash_skb_data(&hp, skb, th->doff << 2))
@@ -1430,7 +1432,7 @@ static struct tcp_ao_info *setsockopt_ao_info(struct sock *sk)
 	return ERR_PTR(-ESOCKTNOSUPPORT);
 }
 
-#define TCP_AO_KEYF_ALL		(0)
+#define TCP_AO_KEYF_ALL		(TCP_AO_KEYF_EXCLUDE_OPT)
 
 static struct tcp_ao_key *tcp_ao_key_alloc(struct sock *sk,
 					   struct tcp_ao_add *cmd)
Signature · 1 line
2.41.0
#27 lore raw

PATCH 18/23 net/tcp: Add TCP-AO getsockopt()s

In reply to Dmitry Safonov

Introduce getsockopt(TCP_AO_GET_KEYS) that lets a user get TCP-AO keys and their properties from a socket. The user can provide a filter to match the specific key to be dumped or ::get_all = 1 may be used to dump all keys in one syscall.

Add another getsockopt(TCP_AO_INFO) for providing per-socket/per-ao_info
stats: packet counters, Current_key/RNext_key and flags like ::ao_required and ::accept_icmps.

---
 include/net/tcp_ao.h     |  12 ++
 include/uapi/linux/tcp.h |  63 +++++++--
 net/ipv4/tcp.c           |  13 ++
 net/ipv4/tcp_ao.c        | 295 +++++++++++++++++++++++++++++++++++++++
 4 files changed, 369 insertions(+), 14 deletions(-)
include/net/tcp_ao.h +12 -0
index cb16af2a9f39..9209343878e1 100644
@@ -191,6 +191,8 @@ int tcp_ao_calc_traffic_key(struct tcp_ao_key *mkt, u8 *key, void *ctx,
 void tcp_ao_destroy_sock(struct sock *sk, bool twsk);
 void tcp_ao_time_wait(struct tcp_timewait_sock *tcptw, struct tcp_sock *tp);
 bool tcp_ao_ignore_icmp(struct sock *sk, int type, int code);
+int tcp_ao_get_mkts(struct sock *sk, sockptr_t optval, sockptr_t optlen);
+int tcp_ao_get_sock_info(struct sock *sk, sockptr_t optval, sockptr_t optlen);
 enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
 			const struct sk_buff *skb, unsigned short int family,
 			const struct request_sock *req,
@@ -304,6 +306,16 @@ static inline void tcp_ao_time_wait(struct tcp_timewait_sock *tcptw,
 static inline void tcp_ao_connect_init(struct sock *sk)
 {
 }
+
+static inline int tcp_ao_get_mkts(struct sock *sk, sockptr_t optval, sockptr_t optlen)
+{
+	return -ENOPROTOOPT;
+}
+
+static inline int tcp_ao_get_sock_info(struct sock *sk, sockptr_t optval, sockptr_t optlen)
+{
+	return -ENOPROTOOPT;
+}
 #endif
 
 #if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
include/uapi/linux/tcp.h +49 -14

77 lines not on this page · raw message

net/ipv4/tcp.c +13 -0
index 37d3eac0561b..82506300da6c 100644
@@ -4289,6 +4289,19 @@ int do_tcp_getsockopt(struct sock *sk, int level,
 		return err;
 	}
 #endif
+	case TCP_AO_GET_KEYS:
+	case TCP_AO_INFO: {
+		int err;
+
+		sockopt_lock_sock(sk);
+		if (optname == TCP_AO_GET_KEYS)
+			err = tcp_ao_get_mkts(sk, optval, optlen);
+		else
+			err = tcp_ao_get_sock_info(sk, optval, optlen);
+		sockopt_release_sock(sk);
+
+		return err;
+	}
 	default:
 		return -ENOPROTOOPT;
 	}
net/ipv4/tcp_ao.c +295 -0

298 lines not on this page · raw message

Signature · 1 line
2.41.0
#28 lore raw

PATCH 19/23 net/tcp: Allow asynchronous delete for TCP-AO keys (MKTs)

In reply to Dmitry Safonov

Delete becomes very, very fast - almost free, but after setsockopt() syscall returns, the key is still alive until next RCU grace period. Which is fine for listen sockets as userspace needs to be aware of setsockopt(TCP_AO) and accept() race and resolve it with verification by getsockopt() after TCP connection was accepted.

The benchmark results (on non-loaded box, worse with more RCU work pending):

Quoted 4 lines
ok 33    Worst case delete    16384 keys: min=5ms max=10ms mean=6.93904ms stddev=0.263421
ok 34        Add a new key    16384 keys: min=1ms max=4ms mean=2.17751ms stddev=0.147564
ok 35 Remove random-search    16384 keys: min=5ms max=10ms mean=6.50243ms stddev=0.254999
ok 36         Remove async    16384 keys: min=0ms max=0ms mean=0.0296107ms stddev=0.0172078
---
 include/uapi/linux/tcp.h |  3 ++-
 net/ipv4/tcp_ao.c        | 21 ++++++++++++++++++---
 2 files changed, 20 insertions(+), 4 deletions(-)
include/uapi/linux/tcp.h +2 -1
index 1109093bbb24..979ff960fddb 100644
@@ -383,7 +383,8 @@ struct tcp_ao_del { /* setsockopt(TCP_AO_DEL_KEY) */
 	__s32	ifindex;		/* L3 dev index for VRF */
 	__u32   set_current	:1,	/* corresponding ::current_key */
 		set_rnext	:1,	/* corresponding ::rnext */
-		reserved	:30;	/* must be 0 */
+		del_async	:1,	/* only valid for listen sockets */
+		reserved	:29;	/* must be 0 */
 	__u16	reserved2;		/* padding, must be 0 */
 	__u8	prefix;			/* peer's address prefix */
 	__u8	sndid;			/* SendID for outgoing segments */
net/ipv4/tcp_ao.c +18 -3
index faa6c5c0db28..567aa9400f7d 100644
@@ -1606,7 +1606,7 @@ static int tcp_ao_add_cmd(struct sock *sk, unsigned short int family,
 }
 
 static int tcp_ao_delete_key(struct sock *sk, struct tcp_ao_info *ao_info,
-			     struct tcp_ao_key *key,
+			     bool del_async, struct tcp_ao_key *key,
 			     struct tcp_ao_key *new_current,
 			     struct tcp_ao_key *new_rnext)
 {
@@ -1614,11 +1614,24 @@ static int tcp_ao_delete_key(struct sock *sk, struct tcp_ao_info *ao_info,
 
 	hlist_del_rcu(&key->node);
 
+	/* Support for async delete on listening sockets: as they don't
+	 * need current_key/rnext_key maintaining, we don't need to check
+	 * them and we can just free all resources in RCU fashion.
+	 */
+	if (del_async) {
+		atomic_sub(tcp_ao_sizeof_key(key), &sk->sk_omem_alloc);
+		call_rcu(&key->rcu, tcp_ao_key_free_rcu);
+		return 0;
+	}
+
 	/* At this moment another CPU could have looked this key up
 	 * while it was unlinked from the list. Wait for RCU grace period,
 	 * after which the key is off-list and can't be looked up again;
 	 * the rx path [just before RCU came] might have used it and set it
 	 * as current_key (very unlikely).
+	 * Free the key with next RCU grace period (in case it was
+	 * current_key before tcp_ao_current_rnext() might have
+	 * changed it in forced-delete).
 	 */
 	synchronize_rcu();
 	if (new_current)
@@ -1689,6 +1702,8 @@ static int tcp_ao_del_cmd(struct sock *sk, unsigned short int family,
 		if (!new_rnext)
 			return -ENOENT;
 	}
+	if (cmd.del_async && sk->sk_state != TCP_LISTEN)
+		return -EINVAL;
 
 	if (family == AF_INET) {
 		struct sockaddr_in *sin = (struct sockaddr_in *)&cmd.addr;
@@ -1733,8 +1748,8 @@ static int tcp_ao_del_cmd(struct sock *sk, unsigned short int family,
 		if (key == new_current || key == new_rnext)
 			continue;
 
-		return tcp_ao_delete_key(sk, ao_info, key,
-					  new_current, new_rnext);
+		return tcp_ao_delete_key(sk, ao_info, cmd.del_async, key,
+					 new_current, new_rnext);
 	}
 	return -ENOENT;
 }
Signature · 1 line
2.41.0
#29 lore raw

PATCH 20/23 net/tcp: Add static_key for TCP-AO

In reply to Dmitry Safonov

Similarly to TCP-MD5, add a static key to TCP-AO that is patched out when there are no keys on a machine and dynamically enabled with the first setsockopt(TCP_AO) adds a key on any socket. The static key is as well dynamically disabled later when the socket is destructed.

The lifetime of enabled static key here is the same as ao_info: it is enabled on allocation, passed over from full socket to twsk and destructed when ao_info is scheduled for destruction.

---
 include/net/tcp.h    |  3 +++
 include/net/tcp_ao.h |  2 ++
 net/ipv4/tcp_ao.c    | 22 +++++++++++++++++++++
 net/ipv4/tcp_input.c | 46 +++++++++++++++++++++++++++++---------------
 4 files changed, 57 insertions(+), 16 deletions(-)
include/net/tcp.h +3 -0
index 50636ee7ee2f..4df3ee6e2722 100644
@@ -2611,6 +2611,9 @@ static inline bool tcp_ao_required(struct sock *sk, const void *saddr,
 	struct tcp_ao_info *ao_info;
 	struct tcp_ao_key *ao_key;
 
+	if (!static_branch_unlikely(&tcp_ao_needed.key))
+		return false;
+
 	ao_info = rcu_dereference_check(tcp_sk(sk)->ao_info,
 					lockdep_sock_is_held(sk));
 	if (!ao_info)
include/net/tcp_ao.h +2 -0
index 9209343878e1..e6d849d7179e 100644
@@ -152,6 +152,8 @@ do {									\
 
 #ifdef CONFIG_TCP_AO
 /* TCP-AO structures and functions */
+#include <linux/jump_label.h>
+extern struct static_key_false_deferred tcp_ao_needed;
 
 struct tcp4_ao_context {
 	__be32		saddr;
net/ipv4/tcp_ao.c +22 -0
index 567aa9400f7d..54b35ee1edf1 100644
@@ -17,6 +17,8 @@
 #include <net/ipv6.h>
 #include <net/icmp.h>
 
+DEFINE_STATIC_KEY_DEFERRED_FALSE(tcp_ao_needed, HZ);
+
 int tcp_ao_calc_traffic_key(struct tcp_ao_key *mkt, u8 *key, void *ctx,
 			    unsigned int len, struct tcp_sigpool *hp)
 {
@@ -50,6 +52,9 @@ bool tcp_ao_ignore_icmp(struct sock *sk, int type, int code)
 	struct tcp_ao_info *ao;
 	bool ignore_icmp = false;
 
+	if (!static_branch_unlikely(&tcp_ao_needed.key))
+		return false;
+
 	/* RFC5925, 7.8:
 	 * >> A TCP-AO implementation MUST default to ignore incoming ICMPv4
 	 * messages of Type 3 (destination unreachable), Codes 2-4 (protocol
@@ -188,6 +193,9 @@ static struct tcp_ao_key *__tcp_ao_do_lookup(const struct sock *sk,
 	struct tcp_ao_key *key;
 	struct tcp_ao_info *ao;
 
+	if (!static_branch_unlikely(&tcp_ao_needed.key))
+		return NULL;
+
 	ao = rcu_dereference_check(tcp_sk(sk)->ao_info,
 				   lockdep_sock_is_held(sk));
 	if (!ao)
@@ -281,6 +289,7 @@ void tcp_ao_destroy_sock(struct sock *sk, bool twsk)
 	}
 
 	kfree_rcu(ao, rcu);
+	static_branch_slow_dec_deferred(&tcp_ao_needed);
 }
 
 void tcp_ao_time_wait(struct tcp_timewait_sock *tcptw, struct tcp_sock *tp)
@@ -1142,6 +1151,11 @@ int tcp_ao_copy_all_matching(const struct sock *sk, struct sock *newsk,
 		goto free_and_exit;
 	}
 
+	if (!static_key_fast_inc_not_disabled(&tcp_ao_needed.key.key)) {
+		ret = -EUSERS;
+		goto free_and_exit;
+	}
+
 	key_head = rcu_dereference(hlist_first_rcu(&new_ao->head));
 	first_key = hlist_entry_safe(key_head, struct tcp_ao_key, node);
 
@@ -1585,6 +1599,10 @@ static int tcp_ao_add_cmd(struct sock *sk, unsigned short int family,
 
 	tcp_ao_link_mkt(ao_info, key);
 	if (first) {
+		if (!static_branch_inc(&tcp_ao_needed.key)) {
+			ret = -EUSERS;
+			goto err_free_sock;
+		}
 		sk_gso_disable(sk);
 		rcu_assign_pointer(tcp_sk(sk)->ao_info, ao_info);
 	}
@@ -1850,6 +1868,10 @@ static int tcp_ao_info_cmd(struct sock *sk, unsigned short int family,
 	if (new_rnext)
 		WRITE_ONCE(ao_info->rnext_key, new_rnext);
 	if (first) {
+		if (!static_branch_inc(&tcp_ao_needed.key)) {
+			err = -EUSERS;
+			goto out;
+		}
 		sk_gso_disable(sk);
 		rcu_assign_pointer(tcp_sk(sk)->ao_info, ao_info);
 	}
net/ipv4/tcp_input.c +30 -16

71 lines not on this page · raw message

Signature · 1 line
2.41.0
#30 lore raw

PATCH 21/23 net/tcp: Wire up l3index to TCP-AO

In reply to Dmitry Safonov

Similarly how TCP_MD5SIG_FLAG_IFINDEX works for TCP-MD5, TCP_AO_KEYF_IFINDEX is an AO-key flag that binds that MKT to a specified by L3 ifinndex. Similarly, without this flag the key will work in the default VRF l3index = 0 for connections.
To prevent AO-keys from overlapping, it's restricted to add key B for a socket that has key A, which have the same sndid/rcvid and one of the following is true:
- !(A.keyflags & TCP_AO_KEYF_IFINDEX) or !(B.keyflags & TCP_AO_KEYF_IFINDEX) so that any key is non-bound to a VRF
- A.l3index == B.l3index
both want to work for the same VRF

Additionally, it's restricted to match TCP-MD5 keys for the same peer the following way:

|--------------|--------------------|----------------|---------------|
|              | MD5 key without    |     MD5 key    |    MD5 key    |
|              |     l3index        |    l3index=0   |   l3index=N   |
|--------------|--------------------|----------------|---------------|
|  TCP-AO key  |                    |                |               |
|  without     |       reject       |    reject      |   reject      |
|  l3index     |                    |                |               |
|--------------|--------------------|----------------|---------------|
|  TCP-AO key  |                    |                |               |
|  l3index=0   |       reject       |    reject      |   allow       |
|--------------|--------------------|----------------|---------------|
|  TCP-AO key  |                    |                |               |
|  l3index=N   |       reject       |    allow       |   reject      |
|--------------|--------------------|----------------|---------------|

This is done with the help of tcp_md5_do_lookup_any_l3index() to reject adding AO key without TCP_AO_KEYF_IFINDEX if there's TCP-MD5 in any VRF. This is important for case where sysctl_tcp_l3mdev_accept = 1 Similarly, for TCP-AO lookups tcp_ao_do_lookup() may be used with l3index < 0, so that __tcp_ao_key_cmp() will match TCP-AO key in any VRF.

---
 include/net/tcp.h     |  11 +--
 include/net/tcp_ao.h  |  17 +++--
 net/ipv4/syncookies.c |   6 +-
 net/ipv4/tcp_ao.c     | 164 +++++++++++++++++++++++++++++++-----------
 net/ipv4/tcp_ipv4.c   |  13 ++--
 net/ipv6/syncookies.c |   5 +-
 net/ipv6/tcp_ao.c     |  20 +++---
 net/ipv6/tcp_ipv6.c   |  15 ++--
 8 files changed, 171 insertions(+), 80 deletions(-)
include/net/tcp.h +6 -5
index 4df3ee6e2722..4aa124ddde3a 100644
@@ -2605,7 +2605,7 @@ static inline int tcp_parse_auth_options(const struct tcphdr *th,
 }
 
 static inline bool tcp_ao_required(struct sock *sk, const void *saddr,
-				   int family, bool stat_inc)
+				   int family, int l3index, bool stat_inc)
 {
 #ifdef CONFIG_TCP_AO
 	struct tcp_ao_info *ao_info;
@@ -2619,7 +2619,7 @@ static inline bool tcp_ao_required(struct sock *sk, const void *saddr,
 	if (!ao_info)
 		return false;
 
-	ao_key = tcp_ao_do_lookup(sk, saddr, family, -1, -1, 0);
+	ao_key = tcp_ao_do_lookup(sk, l3index, saddr, family, -1, -1, 0);
 	if (ao_info->ao_required || ao_key) {
 		if (stat_inc) {
 			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPAOREQUIRED);
@@ -2672,21 +2672,22 @@ tcp_inbound_hash(struct sock *sk, const struct request_sock *req,
 		 * the last key is impossible to remove, so there's
 		 * always at least one current_key.
 		 */
-		if (tcp_ao_required(sk, saddr, family, true)) {
+		if (tcp_ao_required(sk, saddr, family, l3index, true)) {
 			tcp_hash_fail("AO hash is required, but not found",
 					family, skb, "L3 index %d", l3index);
 			return SKB_DROP_REASON_TCP_AONOTFOUND;
 		}
 		if (unlikely(tcp_md5_do_lookup(sk, l3index, saddr, family))) {
 			NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPMD5NOTFOUND);
-			tcp_hash_fail("MD5 Hash not found", family, skb, "");
+			tcp_hash_fail("MD5 Hash not found",
+				      family, skb, "L3 index %d", l3index);
 			return SKB_DROP_REASON_TCP_MD5NOTFOUND;
 		}
 		return SKB_NOT_DROPPED_YET;
 	}
 
 	if (aoh)
-		return tcp_inbound_ao_hash(sk, skb, family, req, aoh);
+		return tcp_inbound_ao_hash(sk, skb, family, req, l3index, aoh);
 
 	return tcp_inbound_md5_hash(sk, skb, saddr, daddr, family,
 				    l3index, md5_location);
include/net/tcp_ao.h +8 -9
index e6d849d7179e..b56b49a98266 100644
@@ -33,6 +33,7 @@ struct tcp_ao_key {
 	u8			key[TCP_AO_MAXKEYLEN] __tcp_ao_key_align;
 	unsigned int		tcp_sigpool_id;
 	unsigned int		digest_size;
+	int			l3index;
 	u16			port;
 	u8			prefixlen;
 	u8			family;
@@ -197,10 +198,10 @@ int tcp_ao_get_mkts(struct sock *sk, sockptr_t optval, sockptr_t optlen);
 int tcp_ao_get_sock_info(struct sock *sk, sockptr_t optval, sockptr_t optlen);
 enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
 			const struct sk_buff *skb, unsigned short int family,
-			const struct request_sock *req,
+			const struct request_sock *req, int l3index,
 			const struct tcp_ao_hdr *aoh);
 u32 tcp_ao_compute_sne(u32 next_sne, u32 next_seq, u32 seq);
-struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
+struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk, int l3index,
 				    const union tcp_ao_addr *addr,
 				    int family, int sndid, int rcvid, u16 port);
 int tcp_ao_hash_hdr(unsigned short family, char *ao_hash,
@@ -243,9 +244,6 @@ int tcp_v6_ao_calc_key_sk(struct tcp_ao_key *mkt, u8 *key,
 				 __be32 disn, bool send);
 int tcp_v6_ao_calc_key_rsk(struct tcp_ao_key *mkt, u8 *key,
 			   struct request_sock *req);
-struct tcp_ao_key *tcp_v6_ao_do_lookup(const struct sock *sk,
-				       const struct in6_addr *addr,
-				       int sndid, int rcvid);
 struct tcp_ao_key *tcp_v6_ao_lookup(const struct sock *sk,
 				    struct sock *addr_sk,
 				    int sndid, int rcvid);
@@ -264,12 +262,12 @@ void tcp_ao_finish_connect(struct sock *sk, struct sk_buff *skb);
 void tcp_ao_connect_init(struct sock *sk);
 void tcp_ao_syncookie(struct sock *sk, const struct sk_buff *skb,
 		      struct tcp_request_sock *treq,
-		      unsigned short int family);
+		      unsigned short int family, int l3index);
 #else /* CONFIG_TCP_AO */
 
 static inline void tcp_ao_syncookie(struct sock *sk, const struct sk_buff *skb,
 		      struct tcp_request_sock *treq,
-		      unsigned short int family)
+		      unsigned short int family, int l3index)
 {
 }
 
@@ -280,13 +278,14 @@ static inline bool tcp_ao_ignore_icmp(struct sock *sk, int type, int code)
 
 static inline enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
 		const struct sk_buff *skb, unsigned short int family,
-		const struct request_sock *req, const struct tcp_ao_hdr *aoh)
+		const struct request_sock *req, int l3index,
+		const struct tcp_ao_hdr *aoh)
 {
 	return SKB_NOT_DROPPED_YET;
 }
 
 static inline struct tcp_ao_key *tcp_ao_do_lookup(const struct sock *sk,
-		const union tcp_ao_addr *addr,
+		int l3index, const union tcp_ao_addr *addr,
 		int family, int sndid, int rcvid, u16 port)
 {
 	return NULL;
net/ipv4/syncookies.c +4 -2
index 23fca22bc992..213accdc3600 100644
@@ -337,6 +337,7 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)
 	struct rtable *rt;
 	__u8 rcv_wscale;
 	struct flowi4 fl4;
+	int l3index;
 	u32 tsoff = 0;
 
 	if (!READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_syncookies) ||
@@ -394,13 +395,14 @@ struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb)
 	treq->snt_synack	= 0;
 	treq->tfo_listener	= false;
 
-	tcp_ao_syncookie(sk, skb, treq, AF_INET);
-
 	if (IS_ENABLED(CONFIG_SMC))
 		ireq->smc_ok = 0;
 
 	ireq->ir_iif = inet_request_bound_dev_if(sk, skb);
 
+	l3index = l3mdev_master_ifindex_by_index(sock_net(sk), ireq->ir_iif);
+	tcp_ao_syncookie(sk, skb, treq, AF_INET, l3index);
+
 	/* We throwed the options of the initial SYN away, so we hope
 	 * the ACK carries the same options again (see RFC1122 4.2.3.8)
 	 */
net/ipv4/tcp_ao.c +122 -42

420 lines not on this page · raw message

net/ipv4/tcp_ipv4.c +7 -6
index 054897a13575..c14075254bfa 100644
@@ -1076,6 +1076,7 @@ static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
 	struct tcp_ao_key *ao_key = NULL;
 	const union tcp_md5_addr *addr;
 	u8 keyid = 0;
+	int l3index;
 	u8 *traffic_key = NULL;
 #ifdef CONFIG_TCP_AO
 	const struct tcp_ao_hdr *aoh;
@@ -1087,6 +1088,7 @@ static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
 	u32 seq = (sk->sk_state == TCP_LISTEN) ? tcp_rsk(req)->snt_isn + 1 :
 					     tcp_sk(sk)->snd_nxt;
 	addr = (union tcp_md5_addr *)&ip_hdr(skb)->saddr;
+	l3index = tcp_v4_sdif(skb) ? inet_iif(skb) : 0;
 
 	if (tcp_rsk_used_ao(req)) {
 #ifdef CONFIG_TCP_AO
@@ -1097,11 +1099,11 @@ static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
 		if (!aoh)
 			return;
 
-		ao_key = tcp_ao_do_lookup(sk, addr, AF_INET,
+		ao_key = tcp_ao_do_lookup(sk, l3index, addr, AF_INET,
 					  aoh->rnext_keyid, -1, 0);
 		if (unlikely(!ao_key)) {
 			/* Send ACK with any matching MKT for the peer */
-			ao_key = tcp_ao_do_lookup(sk, addr,
+			ao_key = tcp_ao_do_lookup(sk, l3index, addr,
 						  AF_INET, -1, -1, 0);
 			/* Matching key disappeared (user removed the key?)
 			 * let the handshake timeout.
@@ -1123,9 +1125,6 @@ static void tcp_v4_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
 		tcp_v4_ao_calc_key_rsk(ao_key, traffic_key, req);
 #endif
 	} else {
-		int l3index;
-
-		l3index = tcp_v4_sdif(skb) ? inet_iif(skb) : 0;
 		md5_key = tcp_md5_do_lookup(sk, l3index, addr, AF_INET);
 	}
 	/* RFC 7323 2.3
@@ -1485,6 +1484,7 @@ static int tcp_v4_parse_md5_keys(struct sock *sk, int optname,
 	const union tcp_md5_addr *addr;
 	u8 prefixlen = 32;
 	int l3index = 0;
+	bool l3flag;
 	u8 flags;
 
 	if (optlen < sizeof(cmd))
@@ -1497,6 +1497,7 @@ static int tcp_v4_parse_md5_keys(struct sock *sk, int optname,
 		return -EINVAL;
 
 	flags = cmd.tcpm_flags & TCP_MD5SIG_FLAG_IFINDEX;
+	l3flag = cmd.tcpm_flags & TCP_MD5SIG_FLAG_IFINDEX;
 
 	if (optname == TCP_MD5SIG_EXT &&
 	    cmd.tcpm_flags & TCP_MD5SIG_FLAG_PREFIX) {
@@ -1534,7 +1535,7 @@ static int tcp_v4_parse_md5_keys(struct sock *sk, int optname,
 	/* Don't allow keys for peers that have a matching TCP-AO key.
 	 * See the comment in tcp_ao_add_cmd()
 	 */
-	if (tcp_ao_required(sk, addr, AF_INET, false))
+	if (tcp_ao_required(sk, addr, AF_INET, l3flag ? l3index : -1, false))
 		return -EKEYREJECTED;
 
 	return tcp_md5_do_add(sk, addr, AF_INET, prefixlen, l3index, flags,
net/ipv6/syncookies.c +4 -1
index ad7a8caa7b2a..500f6ed3b8cf 100644
@@ -140,6 +140,7 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
 	struct dst_entry *dst;
 	__u8 rcv_wscale;
 	u32 tsoff = 0;
+	int l3index;
 
 	if (!READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_syncookies) ||
 	    !th->ack || th->rst)
@@ -214,7 +215,9 @@ struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb)
 	treq->snt_isn = cookie;
 	treq->ts_off = 0;
 	treq->txhash = net_tx_rndhash();
-	tcp_ao_syncookie(sk, skb, treq, AF_INET6);
+
+	l3index = l3mdev_master_ifindex_by_index(sock_net(sk), ireq->ir_iif);
+	tcp_ao_syncookie(sk, skb, treq, AF_INET6, l3index);
 
 	if (IS_ENABLED(CONFIG_SMC))
 		ireq->smc_ok = 0;
net/ipv6/tcp_ao.c +9 -11
index 984a75f96e20..ef163e8928d2 100644
@@ -87,30 +87,28 @@ int tcp_v6_ao_calc_key_rsk(struct tcp_ao_key *mkt, u8 *key,
 			htonl(tcp_rsk(req)->rcv_isn));
 }
 
-struct tcp_ao_key *tcp_v6_ao_do_lookup(const struct sock *sk,
-				       const struct in6_addr *addr,
-				       int sndid, int rcvid)
-{
-	return tcp_ao_do_lookup(sk, (union tcp_ao_addr *)addr, AF_INET6,
-				sndid, rcvid, 0);
-}
-
 struct tcp_ao_key *tcp_v6_ao_lookup(const struct sock *sk,
 				    struct sock *addr_sk,
 				    int sndid, int rcvid)
 {
 	struct in6_addr *addr = &addr_sk->sk_v6_daddr;
+	int l3index = l3mdev_master_ifindex_by_index(sock_net(sk),
+						     addr_sk->sk_bound_dev_if);
 
-	return tcp_v6_ao_do_lookup(sk, addr, sndid, rcvid);
+	return tcp_ao_do_lookup(sk, l3index, (union tcp_ao_addr *)addr,
+				AF_INET6, sndid, rcvid, 0);
 }
 
 struct tcp_ao_key *tcp_v6_ao_lookup_rsk(const struct sock *sk,
 					struct request_sock *req,
 					int sndid, int rcvid)
 {
-	struct in6_addr *addr = &inet_rsk(req)->ir_v6_rmt_addr;
+	struct inet_request_sock *ireq = inet_rsk(req);
+	struct in6_addr *addr = &ireq->ir_v6_rmt_addr;
+	int l3index = l3mdev_master_ifindex_by_index(sock_net(sk), ireq->ir_iif);
 
-	return tcp_v6_ao_do_lookup(sk, addr, sndid, rcvid);
+	return tcp_ao_do_lookup(sk, l3index, (union tcp_ao_addr *)addr,
+				AF_INET6, sndid, rcvid, 0);
 }
 
 int tcp_v6_ao_hash_pseudoheader(struct tcp_sigpool *hp,
net/ipv6/tcp_ipv6.c +11 -4
index 9a427c0d48ab..dee42cba6c25 100644
@@ -606,6 +606,7 @@ static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 	union tcp_ao_addr *addr;
 	int l3index = 0;
 	u8 prefixlen;
+	bool l3flag;
 	u8 flags;
 
 	if (optlen < sizeof(cmd))
@@ -618,6 +619,7 @@ static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 		return -EINVAL;
 
 	flags = cmd.tcpm_flags & TCP_MD5SIG_FLAG_IFINDEX;
+	l3flag = cmd.tcpm_flags & TCP_MD5SIG_FLAG_IFINDEX;
 
 	if (optname == TCP_MD5SIG_EXT &&
 	    cmd.tcpm_flags & TCP_MD5SIG_FLAG_PREFIX) {
@@ -664,7 +666,8 @@ static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 		/* Don't allow keys for peers that have a matching TCP-AO key.
 		 * See the comment in tcp_ao_add_cmd()
 		 */
-		if (tcp_ao_required(sk, addr, AF_INET, false))
+		if (tcp_ao_required(sk, addr, AF_INET,
+				    l3flag ? l3index : -1, false))
 			return -EKEYREJECTED;
 		return tcp_md5_do_add(sk, addr,
 				      AF_INET, prefixlen, l3index, flags,
@@ -676,7 +679,7 @@ static int tcp_v6_parse_md5_keys(struct sock *sk, int optname,
 	/* Don't allow keys for peers that have a matching TCP-AO key.
 	 * See the comment in tcp_ao_add_cmd()
 	 */
-	if (tcp_ao_required(sk, addr, AF_INET6, false))
+	if (tcp_ao_required(sk, addr, AF_INET6, l3flag ? l3index : -1, false))
 		return -EKEYREJECTED;
 
 	return tcp_md5_do_add(sk, addr, AF_INET6, prefixlen, l3index, flags,
@@ -1223,10 +1226,14 @@ static void tcp_v6_reqsk_send_ack(const struct sock *sk, struct sk_buff *skb,
 			return;
 		if (!aoh)
 			return;
-		ao_key = tcp_v6_ao_do_lookup(sk, addr, aoh->rnext_keyid, -1);
+		ao_key = tcp_ao_do_lookup(sk, l3index,
+					  (union tcp_ao_addr *)addr, AF_INET6,
+					  aoh->rnext_keyid, -1, 0);
 		if (unlikely(!ao_key)) {
 			/* Send ACK with any matching MKT for the peer */
-			ao_key = tcp_v6_ao_do_lookup(sk, addr, -1, -1);
+			ao_key = tcp_ao_do_lookup(sk, l3index,
+						  (union tcp_ao_addr *)addr,
+						  AF_INET6, -1, -1, 0);
 			/* Matching key disappeared (user removed the key?)
 			 * let the handshake timeout.
 			 */
Signature · 1 line
2.41.0
#31 lore raw

PATCH 22/23 net/tcp: Add TCP_AO_REPAIR

In reply to Dmitry Safonov

Add TCP_AO_REPAIR setsockopt(), getsockopt(). They let a user to repair TCP-AO ISNs/SNEs. Also let the user hack around when (tp->repair) is on and add ao_info on a socket in any supported state.
As SNEs now can be read/written at any moment, use WRITE_ONCE()/READ_ONCE() to set/read them.

---
 include/net/tcp_ao.h     | 14 ++++++
 include/uapi/linux/tcp.h |  8 ++++
 net/ipv4/tcp.c           | 24 ++++++++---
 net/ipv4/tcp_ao.c        | 92 ++++++++++++++++++++++++++++++++++++++--
 4 files changed, 127 insertions(+), 11 deletions(-)
include/net/tcp_ao.h +14 -0
index b56b49a98266..0027f4b4683d 100644
@@ -196,6 +196,8 @@ void tcp_ao_time_wait(struct tcp_timewait_sock *tcptw, struct tcp_sock *tp);
 bool tcp_ao_ignore_icmp(struct sock *sk, int type, int code);
 int tcp_ao_get_mkts(struct sock *sk, sockptr_t optval, sockptr_t optlen);
 int tcp_ao_get_sock_info(struct sock *sk, sockptr_t optval, sockptr_t optlen);
+int tcp_ao_get_repair(struct sock *sk, sockptr_t optval, sockptr_t optlen);
+int tcp_ao_set_repair(struct sock *sk, sockptr_t optval, unsigned int optlen);
 enum skb_drop_reason tcp_inbound_ao_hash(struct sock *sk,
 			const struct sk_buff *skb, unsigned short int family,
 			const struct request_sock *req, int l3index,
@@ -317,6 +319,18 @@ static inline int tcp_ao_get_sock_info(struct sock *sk, sockptr_t optval, sockpt
 {
 	return -ENOPROTOOPT;
 }
+
+static inline int tcp_ao_get_repair(struct sock *sk,
+				    sockptr_t optval, sockptr_t optlen)
+{
+	return -ENOPROTOOPT;
+}
+
+static inline int tcp_ao_set_repair(struct sock *sk,
+				    sockptr_t optval, unsigned int optlen)
+{
+	return -ENOPROTOOPT;
+}
 #endif
 
 #if defined(CONFIG_TCP_MD5SIG) || defined(CONFIG_TCP_AO)
include/uapi/linux/tcp.h +8 -0
index 979ff960fddb..624198fa170d 100644
@@ -133,6 +133,7 @@ enum {
 #define TCP_AO_DEL_KEY		39	/* Delete MKT */
 #define TCP_AO_INFO		40	/* Set/list TCP-AO per-socket options */
 #define TCP_AO_GET_KEYS		41	/* List MKT(s) */
+#define TCP_AO_REPAIR		42	/* Get/Set SNEs and ISNs */
 
 #define TCP_REPAIR_ON		1
 #define TCP_REPAIR_OFF		0
@@ -445,6 +446,13 @@ struct tcp_ao_getsockopt { /* getsockopt(TCP_AO_GET_KEYS) */
 	__u64	pkt_bad;		/* out: segments that failed verification */
 } __attribute__((aligned(8)));
 
+struct tcp_ao_repair { /* {s,g}etsockopt(TCP_AO_REPAIR) */
+	__be32			snt_isn;
+	__be32			rcv_isn;
+	__u32			snd_sne;
+	__u32			rcv_sne;
+} __attribute__((aligned(8)));
+
 /* setsockopt(fd, IPPROTO_TCP, TCP_ZEROCOPY_RECEIVE, ...) */
 
 #define TCP_RECEIVE_ZEROCOPY_FLAG_TLB_CLEAN_HINT 0x1
net/ipv4/tcp.c +17 -7
index 82506300da6c..95c9d3c9134e 100644
@@ -3608,20 +3608,28 @@ int do_tcp_setsockopt(struct sock *sk, int level, int optname,
 		__tcp_sock_set_quickack(sk, val);
 		break;
 
+	case TCP_AO_REPAIR:
+		err = tcp_ao_set_repair(sk, optval, optlen);
+		break;
 #ifdef CONFIG_TCP_AO
 	case TCP_AO_ADD_KEY:
 	case TCP_AO_DEL_KEY:
 	case TCP_AO_INFO: {
 		/* If this is the first TCP-AO setsockopt() on the socket,
-		 * sk_state has to be LISTEN or CLOSE
+		 * sk_state has to be LISTEN or CLOSE. Allow TCP_REPAIR
+		 * in any state.
 		 */
-		if (((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE)) ||
-		    rcu_dereference_protected(tcp_sk(sk)->ao_info,
+		if ((1 << sk->sk_state) & (TCPF_LISTEN | TCPF_CLOSE))
+			goto ao_parse;
+		if (rcu_dereference_protected(tcp_sk(sk)->ao_info,
 					      lockdep_sock_is_held(sk)))
-			err = tp->af_specific->ao_parse(sk, optname, optval,
-							optlen);
-		else
-			err = -EISCONN;
+			goto ao_parse;
+		if (tp->repair)
+			goto ao_parse;
+		err = -EISCONN;
+		break;
+ao_parse:
+		err = tp->af_specific->ao_parse(sk, optname, optval, optlen);
 		break;
 	}
 #endif
@@ -4289,6 +4297,8 @@ int do_tcp_getsockopt(struct sock *sk, int level,
 		return err;
 	}
 #endif
+	case TCP_AO_REPAIR:
+		return tcp_ao_get_repair(sk, optval, optlen);
 	case TCP_AO_GET_KEYS:
 	case TCP_AO_INFO: {
 		int err;
net/ipv4/tcp_ao.c +88 -4

114 lines not on this page · raw message

Signature · 1 line
2.41.0
#32 lore raw

PATCH 23/23 Documentation/tcp: Add TCP-AO documentation

In reply to Dmitry Safonov

It has Frequently Asked Questions (FAQ) on RFC 5925 - I found it very useful answering those before writing the actual code. It provides answers to common questions that arise on a quick read of the RFC, as well as how they were answered. There's also comparison to TCP-MD5 option, evaluation of per-socket vs in-kernel-DB approaches and description of uAPI provided.

Hopefully, it will be as useful for reviewing the code as it was for writing.

---
 Documentation/networking/index.rst  |   1 +
 Documentation/networking/tcp_ao.rst | 422 ++++++++++++++++++++++++++++
 2 files changed, 423 insertions(+)
 create mode 100644 Documentation/networking/tcp_ao.rst
Documentation/networking/index.rst +1 -0
index 5b75c3f7a137..69c1e53ef88b 100644
@@ -107,6 +107,7 @@ Contents:
    sysfs-tagging
    tc-actions-env-rules
    tc-queue-filters
+   tcp_ao
    tcp-thin
    team
    timestamping
Documentation/networking/tcp_ao.rst new file +422 -0

422 lines not on this page · raw message

Signature · 1 line
2.41.0
Acked-by on the v1 series
#33 lore raw
In reply to Dmitry Safonov
Quoted 348 lines · Dmitry Safonov

Hi,

This is version 8.1 of TCP-AO support. I base it on net-next as there's commit 5e5265522a9a ("tcp: annotate data-races around tcp_rsk(req)->txhash") which makes a minor conflict.

The good news is that all pre-required patches have merged to Torvald's/master. Thanks to Herbert, crypto clone-tfm just works on master for all TCP-AO supported algorithms.
So, this is the first version of the patch set that has only net-related changes (well, selftests as well, but they'll be upstreamed separately).

In this version, I've finally spent time and written Documentation/ page on TCP-AO. It has Frequently Asked Questions (FAQ) on RFC 5925 - I found it very useful to answer those before writing the actual code. It provides answers to common questions that arise on a quick read of the RFC as well as how they were answered. There's also a comparison to the TCP-MD5 option, an evaluation of per-socket vs in-kernel-DB approaches and a description of uAPI provided.
I hope it will be as useful for reviewing the code as it was for writing.

The most important changes in this version are:
- CONFIG_TCP_AO implies CONFIG_IPV6 != m. I don't feel like that combination would be useful to anyone and it'd be painful to fix.
- uAPI change in TCP_AO_REPAIR (introduced in version 7): I removed {snd,rcv}_sne_seq counters. They were just copies of snd_nxt/snd_una. No reason for polluting uAPI as well as needlessly copying them.
- TCP_AO_MAX_HASH_SIZE is removed and all temporary buffers are kmalloc()'d. That also saves a couple of bytes for hmac(sha1) and cmac(aes128) traffic keys as they now are allocated with exact hash algo's digest_size.

There's an independent patch set for TCP-MD5 to verify segments on twsk:
https://lore.kernel.org/all/[email protected]/T/#u
That may be used to verify TCP-AO segments on twsk as well.

There seem to be more people that connected me off-list asking me about the status of patches and when I expect them to merge. Cc'ing more interested parties here (ping me directly if you don't want to be in copy). It would be helpful if you provide your reviews and tested-by's. As far as I'm aware, version 7 was ported to RHEL, so now there are probably more downstream kernels with TCP-AO support.

Also available as a git branch for pulling:
https://github.com/0x7f454c46/linux/tree/tcp-ao-v8.1
And another branch with selftests, that will be sent later separately:
https://github.com/0x7f454c46/linux/tree/tcp-ao-v8-with-selftests

Thanks for your time and reviews,
Dmitry

--- Changelog ---

Changes from v8:
- Rebased/retested on linux-net-next

Version 8: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v7:
- Fixed copy'n'paste typo in unsigned-md5.c selftest output
- Fix build error in tcp_v6_send_reset() (kernel test robot <[email protected]>)
- Make CONFIG_TCP_AO imply IPV6 != m
- Cleanup EXPORT_SYMBOL*() as they aren't needed with IPV6 != m
- Used scratch area instead of on-stack buffer for scatter-gather list in tcp_v{4,6}_ao_calc_key(). Fixes CONFIG_VMAP_STACK=y + CONFIG_DEBUG_SG=y
- Allocated digest_size'd buffers for traffic keys in tcp_ao_key instead of maximum-sized buffers of TCP_AO_MAX_HASH_SIZE. That will save little space per key and also potentially allow algorithms with digest size > TCP_AO_MAX_HASH_SIZE.
- Removed TCP_AO_MAX_HASH_SIZE and used kmalloc(GFP_ATOMIC) instead of on-stack hash buffer.
- Don't treat fd=0 as invalid in selftests
- Make TCP-AO selftests work with CONFIG_CRYPTO_FIPS=y
- Don't tcp_ao_compute_sne() for snd_sne on twsk: it's redundant as no data can be sent on twsk
- Get rid of {snd,rcv}_sne_seq: use snd_nxt/snd_una or rcv_nxt instead
- {rcv,snd}_sne and tcp_ao_compute_sne() now are introduced in "net/tcp: Add TCP-AO SNE support" patch
- trivial copy_to_sockptr() fixup for tcp_ao_get_repair() - it could try copying bigger struct than the kernel one (embarrassing!)
- Added Documentation/networking/tcp_ao.rst that describes:
uAPI, has FAQ on RFC 5925 and has implementation details of Linux TCP-AO

Version 7: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v6:
- Some more trivial build warnings fixups (kernel test robot <[email protected]>)
- Added TCP_AO_REPAIR setsockopt(), getsockopt()
- Allowed TCP_AO_* setsockopts if (tp->repair) is on
- Added selftests for TCP_AO_REPAIR, that also check incorrect ISNs/SNEs, which result in a broken TCP-AO connection - that verifies that both Initial Sequence Numbers and Sequence Number Extension are part of MAC generation
- Using TCP_AO_REPAIR added a selftest for SEQ numbers rollover, checking that SNE was incremented, connection is alive post-rolloever and no TCP segments with a wrong signature arrived
- Wrote a selftest for RST segments: both active reset (goes through transmit_skb()) and passive reset (goes through tcp_v{4,6}_send_reset()).
- Refactored and made readable tcp_v{4,6}_send_reset(), also adding support for TCP_LISTEN/TCP_NEW_SYN_RECV
- Dropped per-CPU ahash requests allocations in favor of Herbert's clone-tfm crypto API
- Added Donald Cassidy to Cc as he's interested in getting it into RHEL.

Version 6: https://lore.kernel.org/all/[email protected]/T/#u

iperf[3] benchmarks for version 6:

                           v6.4-rc1                 TCP-AO-v6
  TCP                      43.9 Gbits/sec           43.5 Gbits/sec
  TCP-MD5                  2.20 Gbits/sec           2.25 Gbits/sec
  TCP-AO(hmac(sha1))                                2.53 Gbits/sec
  TCP-AO(hmac(sha512))                              1.67 Gbits/sec
  TCP-AO(hmac(sha384))                              1.77 Gbits/sec
  TCP-AO(hmac(sha224))                              1.29 Gbits/sec
  TCP-AO(hmac(sha3-512))                             481 Mbits/sec
  TCP-AO(hmac(md5))                                 2.07 Gbits/sec
  TCP-AO(hmac(rmd160))                              1.01 Gbits/sec
  TCP-AO(cmac(aes128))                              2.11 Gbits/sec

Changes from v5:
- removed check for TCP_AO_KEYF_IFINDEX in delete command:
VRF might have been destroyed, there still needs to be a way to delete keys that were bound to that l3intf (should tcp_v{4,6}_parse_md5_keys() avoid the same check as well?)
- corrected copy'n'paste typo in tcp_ao_info_cmd() (assign ao_info->rnext_key)
- simplified a bit tcp_ao_copy_mkts_to_user(); added more UAPI checks for getsockopt(TCP_AO_GET_KEYS)
- More UAPI selftests in setsockopt-closed: 29 => 120
- ported TCP-AO patches on Herbert's clone-tfm changes
- adjusted iperf patch for TCP-AO UAPI changes from version 5
- added measures for TCP-AO with tcp_sigpool & clone_tfm backends

Version 5: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v4:
- Renamed tcp_ao_matched_key() => tcp_ao_established_key()
- Missed static in function definitions
(kernel test robot <[email protected]>)
- Fixed CONFIG_IPV6=m build
- Unexported tcp_md5_*_sigpool() functions
- Cleaned up tcp_ao.h: undeclared tcp_ao_cache_traffic_keys(), tcp_v4_ao_calc_key_skb(); removed tcp_v4_inbound_ao_hash()
- Marked "net/tcp: Prepare tcp_md5sig_pool for TCP-AO" as a [draft] patch
- getsockopt() now returns TCP-AO per-key counters
- Another getsockopt() now returns per-ao_info stats: counters and accept_icmps flag state
- Wired up getsockopt() returning counters to selftests
- Fixed a porting mistake: TCP-AO hash in some cases was written in TCP header without accounting for MAC length of the key, rewritting skb shared info
- Fail adding a key with L3 ifindex when !TCP_AO_KEYF_IFINDEX, instead of ignoring tcpa_ifindex (stricter UAPI check)
- Added more test-cases to setsockopt-closed.c selftest
- tcp_ao_hash_skb_data() was a copy'n'paste of tcp_md5_hash_skb_data() share it now under tcp_sigpool_hash_skb_data()
- tcp_ao_mkt_overlap_v{4,6}() deleted as they just re-invented tcp_ao_do_lookup(). That fixes an issue with multiple IPv4-mapped-IPv6 keys for different peers on a listening socket.
- getsockopt() now is tested to return correct VRF number for a key
- TCP-AO and TCP-MD5 interraction in non/default VRFs: added +19 selftests made them SKIP when CONFIG_VRF=n
- unsigned-md5 selftests now checks both scenarios:
(1) adding TCP-AO key _after_ TCP-MD5 key
(2) adding TCP-MD5 key _after_ TCP-AO key
- Added a ratelimited warning if TCP-AO key.ifindex doesn't match sk->sk_bound_dev_if - that will warn a user for potential VRF issues
- tcp_v{4,6}_parse_md5_keys() now allows adding TCP-MD5 key with ifindex=0 and TCP_MD5SIG_FLAG_IFINDEX together with TCP-AO key from another VRF
- Add TCP_AO_CMDF_AO_REQUIRED, which makes a socket TCP-AO only, rejecting TCP-MD5 keys or any unsigned TCP segments
- Remove `tcpa_' prefix for UAPI structure members
- UAPI cleanup: I've separated & renamed per-socket settings (such as ao_info flags + current/rnext set) from per-key changes:
TCP_AO => TCP_AO_ADD_KEY
TCP_AO_DEL => TCP_AO_DEL_KEY
TCP_AO_GET => TCP_AO_GET_KEYS
TCP_AO_MOD => TCP_AO_INFO, the structure is now valid for both
getsockopt() and setsockopt().
- tcp_ao_current_rnext() was split up in order to fail earlier when sndid/rcvid specified can't be set, before anything was changed in ao_info
- fetch current_key before dumping TCP-AO keys in getsockopt(TCP_AO_GET_KEYS):
it may race with changing current_key by RX, which in result might produce a dump with no current_key for userspace.
- instead of TCP_AO_CMDF_* flags, used bitfileds: the flags weren't shared between all TCP_AO_{ADD,GET,DEL}_KEY{,S}, so bitfields are more descriptive here
- use READ_ONCE()/WRITE_ONCE() for current_key and rnext_key more consistently; document in comment the rules for accessing them
- selftests: check all setsockopts()/getsockopts() support extending option structs

Version 4: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v3:
- TCP_MD5 dynamic static key enable/disable patches merged separately [4]
- crypto_pool patches were nacked [5], so instead this patch set extends TCP-MD5-sigpool to be used for TCP-AO as well as for TCP-MD5
- Added missing `static' for tcp_v6_ao_calc_key()
(kernel test robot <[email protected]>)
- Removed CONFIG_TCP_AO default=y and added "If unsure, say N."
- Don't leak ao_info and don't create an unsigned TCP socket if there was a TCP-AO key during handshake, but it was removed from listening socket while the connection was being established
- Migrate to use static_key_fast_inc_not_disabled() and check return code of static_branch_inc()
- Change some return codes to EAFNOSUPPORT for error-pathes where family is neither AF_INET nor AF_INET6
- setsockopt()s on a closed/listen socket might have created stray ao_info, remove it if connect() is called with a correct TCP-MD5 key, the same for the reverse situation: remove md5sig_info straight away from the socket if it's going to be TCP-AO connection
- IPv4-mapped-IPv6 addresses + selftest in fcnal-test.sh (by Salam)
- fix using uninitialized sisn/disn from stack - it would only make non-SYN packets fail verification on a listen socket, which are not expected anyway (kernel test robot <[email protected]>)
- implicit padding in UAPI TCP-AO structures converted to explicit (spotted-by David Laight)
- Some selftests missed zero-initializers for uapi structs on stack
- Removed tcp_ao_do_lookup_rcvid() and tcp_ao_do_lookup_sndid() in favor of unified tcp_ao_matched_key()
- Disallowed setting current/rnext keys on listen sockets - that wasn't supported and didn't affect anything, cleanup for the UAPI
- VRFs support for TCP-AO

Version 3: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v2:
- Added more missing `static' declarations for local functions (kernel test robot <[email protected]>)
- Building now with CONFIG_TCP_AO=n and CONFIG_TCP_MD5SIG=n (kernel test robot <[email protected]>)
- Now setsockopt(TCP_AO) is allowed when it's TCP_LISTEN or TCP_CLOSE state OR the key added is not the first key on a socket (by Salam)
- CONFIG_TCP_AO does not depend on CONFIG_TCP_MD5SIG anymore
- Don't leak tcp_md5_needed static branch counter when TCP-MD5 key is modified/changed
- TCP-AO lookups are dynamically enabled/disabled with static key when there is ao_info in the system (and when it is destroyed)
- Wired SYN cookies up to TCP-AO (by Salam)
- Fix verification for possible re-transmitted SYN packets (by Salam)
- use sockopt_lock_sock() instead of lock_sock()
(from v6.1 rebase, commit d51bbff2aba7)
- use sockptr_t in getsockopt(TCP_AO_GET)
(from v6.1 rebase, commit 34704ef024ae)
- Fixed reallocating crypto_pool's scratch area by IPI while crypto_pool_get() was get by another CPU
- selftests on older kernels (or with CONFIG_TCP_AO=n) should exit with SKIP, not FAIL (Shuah Khan <[email protected]>)
- selftests that check interaction between TCP-AO and TCP-MD5 now SKIP when CONFIG_TCP_MD5SIG=n
- Measured the performance of different hashing algorithms for TCP-AO and compare with TCP-MD5 performance. This is done with hacky patches to iperf (see [3]). At this moment I've done it in qemu/KVM with CPU affinities set on Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz. No performance degradation was noticed before/after patches, but given the measures were done in a VM, without measuring it on a physical dut it only gives a hint of relative speed for different hash algorithms with TCP-AO. Here are results, averaging on 30 measures each:

  TCP:                    3.51Gbits/sec
  TCP-MD5:                1.12Gbits/sec
  TCP-AO(HMAC(SHA1)):     1.53Gbits/sec
  TCP-AO(CMAC(AES128)):   621Mbits/sec
  TCP-AO(HMAC(SHA512)):   1.21Gbits/sec
  TCP-AO(HMAC(SHA384)):   1.20Gbits/sec
  TCP-AO(HMAC(SHA224)):   961Mbits/sec

TCP-AO(HMAC(SHA3-512)): 157Mbits/sec
TCP-AO(HMAC(RMD160)): 659Mbits/sec
TCP-AO(HMAC(MD5): 1.12Gbits/sec
(the last one is just for fun, but may make sense as it provides the same security as TCP-MD5, but allows multiple keys and a mechanism to change them from RFC5925)

Version 2: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v1:
- Building now with CONFIG_IPV6=n (kernel test robot <[email protected]>)
- Added missing static declarations for local functions (kernel test robot <[email protected]>)
- Addressed static analyzer and review comments by Dan Carpenter (thanks, they were very useful!)
- Fix elif without defined() for !CONFIG_TCP_AO
- Recursively build selftests/net/tcp_ao (Shuah Khan), patches in:
https://lore.kernel.org/all/[email protected]/T/#u
- Don't leak crypto_pool reference when TCP-MD5 key is modified/changed
- Add TCP-AO support for nettest.c and fcnal-test.sh
(will be used for VRF testing in later versions)

Comparison between Leonard proposal and this (overview):
https://lore.kernel.org/all/[email protected]/T/#u

Version 1: https://lore.kernel.org/all/[email protected]/T/#u

This patchset implements the TCP-AO option as described in RFC5925. There is a request from industry to move away from TCP-MD5SIG and it seems the time is right to have a TCP-AO upstreamed. This TCP option is meant to replace the TCP MD5 option and address its shortcomings. Specifically, it provides more secure hashing, key rotation and support for long-lived connections (see the summary of TCP-AO advantages over TCP-MD5 in (1.3) of RFC5925). The patch series starts with six patches that are not specific to TCP-AO but implement a general crypto facility that we thought is useful to eliminate code duplication between TCP-MD5SIG and TCP-AO as well as other crypto users. These six patches are being submitted separately in a different patchset [1]. Including them here will show better the gain in code sharing. Next are 18 patches that implement the actual TCP-AO option, followed by patches implementing selftests.

The patch set was written as a collaboration of three authors (in alphabetical order): Dmitry Safonov, Francesco Ruggeri and Salam Noureddine. Additional credits should be given to Prasad Koya, who was involved in early prototyping a few years back. There is also a separate submission done by Leonard Crestez whom we thank for his efforts getting an implementation of RFC5925 submitted for review upstream [2]. This is an independent implementation that makes different design decisions.

For example, we chose a similar design to the TCP-MD5SIG implementation and used setsockopts to program per-socket keys, avoiding the extra complexity of managing a centralized key database in the kernel. A centralized database in the kernel has dubious benefits since it doesn’t eliminate per-socket setsockopts needed to specify which sockets need TCP-AO and what are the currently preferred keys. It also complicates traffic key caching and preventing deletion of in-use keys.

In this implementation, a centralized database of keys can be thought of as living in user space and user applications would have to program those keys on matching sockets. On the server side, the user application programs keys (MKTS in TCP-AO nomenclature) on the listening socket for all peers that are expected to connect. Prefix matching on the peer address is supported. When a peer issues a successful connect, all the MKTs matching the IP address of the peer are copied to the newly created socket. On the active side, when a connect() is issued all MKTs that do not match the peer are deleted from the socket since they will never match the peer. This implementation uses three setsockopt()s for adding, deleting and modifying keys on a socket. All three setsockopt()s have extensive sanity checks that prevent inconsistencies in the keys on a given socket. A getsockopt() is provided to get key information from any given socket.

Few things to note about this implementation:
- Traffic keys are cached for established connections avoiding the cost of such calculation for each packet received or sent.
- Great care has been taken to avoid deleting in-use MKTs
as required by the RFC.
- Any crypto algorithm supported by the Linux kernel can be used to calculate packet hashes.
- Fastopen works with TCP-AO but hasn’t been tested extensively.
- Tested for interop with other major networking vendors (on linux-4.19), including testing for key rotation and long lived connections.

Quoted 7 lines

[1]: https://lore.kernel.org/all/[email protected]/
[2]: https://lore.kernel.org/all/[email protected]/
[3]: https://github.com/0x7f454c46/iperf/tree/tcp-md5-ao
[4]: https://lore.kernel.org/all/166995421700.16716.17446147162780881407.git-patchwork-notify@kernel.org/T/#u
[5]: https://lore.kernel.org/all/[email protected]/T/#u
[6]: https://lore.kernel.org/all/[email protected]/T/#u

For the set:

#34 lore raw
Quoted 358 lines · David Ahern
Quoted 348 lines · Dmitry Safonov

Hi,

This is version 8.1 of TCP-AO support. I base it on net-next as there's commit 5e5265522a9a ("tcp: annotate data-races around tcp_rsk(req)->txhash") which makes a minor conflict.

The good news is that all pre-required patches have merged to Torvald's/master. Thanks to Herbert, crypto clone-tfm just works on master for all TCP-AO supported algorithms.
So, this is the first version of the patch set that has only net-related changes (well, selftests as well, but they'll be upstreamed separately).

In this version, I've finally spent time and written Documentation/ page on TCP-AO. It has Frequently Asked Questions (FAQ) on RFC 5925 - I found it very useful to answer those before writing the actual code. It provides answers to common questions that arise on a quick read of the RFC as well as how they were answered. There's also a comparison to the TCP-MD5 option, an evaluation of per-socket vs in-kernel-DB approaches and a description of uAPI provided.
I hope it will be as useful for reviewing the code as it was for writing.

The most important changes in this version are:
- CONFIG_TCP_AO implies CONFIG_IPV6 != m. I don't feel like that combination would be useful to anyone and it'd be painful to fix.
- uAPI change in TCP_AO_REPAIR (introduced in version 7): I removed {snd,rcv}_sne_seq counters. They were just copies of snd_nxt/snd_una. No reason for polluting uAPI as well as needlessly copying them.
- TCP_AO_MAX_HASH_SIZE is removed and all temporary buffers are kmalloc()'d. That also saves a couple of bytes for hmac(sha1) and cmac(aes128) traffic keys as they now are allocated with exact hash algo's digest_size.

There's an independent patch set for TCP-MD5 to verify segments on twsk:
https://lore.kernel.org/all/[email protected]/T/#u
That may be used to verify TCP-AO segments on twsk as well.

There seem to be more people that connected me off-list asking me about the status of patches and when I expect them to merge. Cc'ing more interested parties here (ping me directly if you don't want to be in copy). It would be helpful if you provide your reviews and tested-by's. As far as I'm aware, version 7 was ported to RHEL, so now there are probably more downstream kernels with TCP-AO support.

Also available as a git branch for pulling:
https://github.com/0x7f454c46/linux/tree/tcp-ao-v8.1
And another branch with selftests, that will be sent later separately:
https://github.com/0x7f454c46/linux/tree/tcp-ao-v8-with-selftests

Thanks for your time and reviews,
Dmitry

--- Changelog ---

Changes from v8:
- Rebased/retested on linux-net-next

Version 8: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v7:
- Fixed copy'n'paste typo in unsigned-md5.c selftest output
- Fix build error in tcp_v6_send_reset() (kernel test robot <[email protected]>)
- Make CONFIG_TCP_AO imply IPV6 != m
- Cleanup EXPORT_SYMBOL*() as they aren't needed with IPV6 != m
- Used scratch area instead of on-stack buffer for scatter-gather list in tcp_v{4,6}_ao_calc_key(). Fixes CONFIG_VMAP_STACK=y + CONFIG_DEBUG_SG=y
- Allocated digest_size'd buffers for traffic keys in tcp_ao_key instead of maximum-sized buffers of TCP_AO_MAX_HASH_SIZE. That will save little space per key and also potentially allow algorithms with digest size > TCP_AO_MAX_HASH_SIZE.
- Removed TCP_AO_MAX_HASH_SIZE and used kmalloc(GFP_ATOMIC) instead of on-stack hash buffer.
- Don't treat fd=0 as invalid in selftests
- Make TCP-AO selftests work with CONFIG_CRYPTO_FIPS=y
- Don't tcp_ao_compute_sne() for snd_sne on twsk: it's redundant as no data can be sent on twsk
- Get rid of {snd,rcv}_sne_seq: use snd_nxt/snd_una or rcv_nxt instead
- {rcv,snd}_sne and tcp_ao_compute_sne() now are introduced in "net/tcp: Add TCP-AO SNE support" patch
- trivial copy_to_sockptr() fixup for tcp_ao_get_repair() - it could try copying bigger struct than the kernel one (embarrassing!)
- Added Documentation/networking/tcp_ao.rst that describes:
uAPI, has FAQ on RFC 5925 and has implementation details of Linux TCP-AO

Version 7: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v6:
- Some more trivial build warnings fixups (kernel test robot <[email protected]>)
- Added TCP_AO_REPAIR setsockopt(), getsockopt()
- Allowed TCP_AO_* setsockopts if (tp->repair) is on
- Added selftests for TCP_AO_REPAIR, that also check incorrect ISNs/SNEs, which result in a broken TCP-AO connection - that verifies that both Initial Sequence Numbers and Sequence Number Extension are part of MAC generation
- Using TCP_AO_REPAIR added a selftest for SEQ numbers rollover, checking that SNE was incremented, connection is alive post-rolloever and no TCP segments with a wrong signature arrived
- Wrote a selftest for RST segments: both active reset (goes through transmit_skb()) and passive reset (goes through tcp_v{4,6}_send_reset()).
- Refactored and made readable tcp_v{4,6}_send_reset(), also adding support for TCP_LISTEN/TCP_NEW_SYN_RECV
- Dropped per-CPU ahash requests allocations in favor of Herbert's clone-tfm crypto API
- Added Donald Cassidy to Cc as he's interested in getting it into RHEL.

Version 6: https://lore.kernel.org/all/[email protected]/T/#u

iperf[3] benchmarks for version 6:

                           v6.4-rc1                 TCP-AO-v6
  TCP                      43.9 Gbits/sec           43.5 Gbits/sec
  TCP-MD5                  2.20 Gbits/sec           2.25 Gbits/sec
  TCP-AO(hmac(sha1))                                2.53 Gbits/sec
  TCP-AO(hmac(sha512))                              1.67 Gbits/sec
  TCP-AO(hmac(sha384))                              1.77 Gbits/sec
  TCP-AO(hmac(sha224))                              1.29 Gbits/sec
  TCP-AO(hmac(sha3-512))                             481 Mbits/sec
  TCP-AO(hmac(md5))                                 2.07 Gbits/sec
  TCP-AO(hmac(rmd160))                              1.01 Gbits/sec
  TCP-AO(cmac(aes128))                              2.11 Gbits/sec

Changes from v5:
- removed check for TCP_AO_KEYF_IFINDEX in delete command:
VRF might have been destroyed, there still needs to be a way to delete keys that were bound to that l3intf (should tcp_v{4,6}_parse_md5_keys() avoid the same check as well?)
- corrected copy'n'paste typo in tcp_ao_info_cmd() (assign ao_info->rnext_key)
- simplified a bit tcp_ao_copy_mkts_to_user(); added more UAPI checks for getsockopt(TCP_AO_GET_KEYS)
- More UAPI selftests in setsockopt-closed: 29 => 120
- ported TCP-AO patches on Herbert's clone-tfm changes
- adjusted iperf patch for TCP-AO UAPI changes from version 5
- added measures for TCP-AO with tcp_sigpool & clone_tfm backends

Version 5: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v4:
- Renamed tcp_ao_matched_key() => tcp_ao_established_key()
- Missed static in function definitions
(kernel test robot <[email protected]>)
- Fixed CONFIG_IPV6=m build
- Unexported tcp_md5_*_sigpool() functions
- Cleaned up tcp_ao.h: undeclared tcp_ao_cache_traffic_keys(), tcp_v4_ao_calc_key_skb(); removed tcp_v4_inbound_ao_hash()
- Marked "net/tcp: Prepare tcp_md5sig_pool for TCP-AO" as a [draft] patch
- getsockopt() now returns TCP-AO per-key counters
- Another getsockopt() now returns per-ao_info stats: counters and accept_icmps flag state
- Wired up getsockopt() returning counters to selftests
- Fixed a porting mistake: TCP-AO hash in some cases was written in TCP header without accounting for MAC length of the key, rewritting skb shared info
- Fail adding a key with L3 ifindex when !TCP_AO_KEYF_IFINDEX, instead of ignoring tcpa_ifindex (stricter UAPI check)
- Added more test-cases to setsockopt-closed.c selftest
- tcp_ao_hash_skb_data() was a copy'n'paste of tcp_md5_hash_skb_data() share it now under tcp_sigpool_hash_skb_data()
- tcp_ao_mkt_overlap_v{4,6}() deleted as they just re-invented tcp_ao_do_lookup(). That fixes an issue with multiple IPv4-mapped-IPv6 keys for different peers on a listening socket.
- getsockopt() now is tested to return correct VRF number for a key
- TCP-AO and TCP-MD5 interraction in non/default VRFs: added +19 selftests made them SKIP when CONFIG_VRF=n
- unsigned-md5 selftests now checks both scenarios:
(1) adding TCP-AO key _after_ TCP-MD5 key
(2) adding TCP-MD5 key _after_ TCP-AO key
- Added a ratelimited warning if TCP-AO key.ifindex doesn't match sk->sk_bound_dev_if - that will warn a user for potential VRF issues
- tcp_v{4,6}_parse_md5_keys() now allows adding TCP-MD5 key with ifindex=0 and TCP_MD5SIG_FLAG_IFINDEX together with TCP-AO key from another VRF
- Add TCP_AO_CMDF_AO_REQUIRED, which makes a socket TCP-AO only, rejecting TCP-MD5 keys or any unsigned TCP segments
- Remove `tcpa_' prefix for UAPI structure members
- UAPI cleanup: I've separated & renamed per-socket settings (such as ao_info flags + current/rnext set) from per-key changes:
TCP_AO => TCP_AO_ADD_KEY
TCP_AO_DEL => TCP_AO_DEL_KEY
TCP_AO_GET => TCP_AO_GET_KEYS
TCP_AO_MOD => TCP_AO_INFO, the structure is now valid for both
getsockopt() and setsockopt().
- tcp_ao_current_rnext() was split up in order to fail earlier when sndid/rcvid specified can't be set, before anything was changed in ao_info
- fetch current_key before dumping TCP-AO keys in getsockopt(TCP_AO_GET_KEYS):
it may race with changing current_key by RX, which in result might produce a dump with no current_key for userspace.
- instead of TCP_AO_CMDF_* flags, used bitfileds: the flags weren't shared between all TCP_AO_{ADD,GET,DEL}_KEY{,S}, so bitfields are more descriptive here
- use READ_ONCE()/WRITE_ONCE() for current_key and rnext_key more consistently; document in comment the rules for accessing them
- selftests: check all setsockopts()/getsockopts() support extending option structs

Version 4: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v3:
- TCP_MD5 dynamic static key enable/disable patches merged separately [4]
- crypto_pool patches were nacked [5], so instead this patch set extends TCP-MD5-sigpool to be used for TCP-AO as well as for TCP-MD5
- Added missing `static' for tcp_v6_ao_calc_key()
(kernel test robot <[email protected]>)
- Removed CONFIG_TCP_AO default=y and added "If unsure, say N."
- Don't leak ao_info and don't create an unsigned TCP socket if there was a TCP-AO key during handshake, but it was removed from listening socket while the connection was being established
- Migrate to use static_key_fast_inc_not_disabled() and check return code of static_branch_inc()
- Change some return codes to EAFNOSUPPORT for error-pathes where family is neither AF_INET nor AF_INET6
- setsockopt()s on a closed/listen socket might have created stray ao_info, remove it if connect() is called with a correct TCP-MD5 key, the same for the reverse situation: remove md5sig_info straight away from the socket if it's going to be TCP-AO connection
- IPv4-mapped-IPv6 addresses + selftest in fcnal-test.sh (by Salam)
- fix using uninitialized sisn/disn from stack - it would only make non-SYN packets fail verification on a listen socket, which are not expected anyway (kernel test robot <[email protected]>)
- implicit padding in UAPI TCP-AO structures converted to explicit (spotted-by David Laight)
- Some selftests missed zero-initializers for uapi structs on stack
- Removed tcp_ao_do_lookup_rcvid() and tcp_ao_do_lookup_sndid() in favor of unified tcp_ao_matched_key()
- Disallowed setting current/rnext keys on listen sockets - that wasn't supported and didn't affect anything, cleanup for the UAPI
- VRFs support for TCP-AO

Version 3: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v2:
- Added more missing `static' declarations for local functions (kernel test robot <[email protected]>)
- Building now with CONFIG_TCP_AO=n and CONFIG_TCP_MD5SIG=n (kernel test robot <[email protected]>)
- Now setsockopt(TCP_AO) is allowed when it's TCP_LISTEN or TCP_CLOSE state OR the key added is not the first key on a socket (by Salam)
- CONFIG_TCP_AO does not depend on CONFIG_TCP_MD5SIG anymore
- Don't leak tcp_md5_needed static branch counter when TCP-MD5 key is modified/changed
- TCP-AO lookups are dynamically enabled/disabled with static key when there is ao_info in the system (and when it is destroyed)
- Wired SYN cookies up to TCP-AO (by Salam)
- Fix verification for possible re-transmitted SYN packets (by Salam)
- use sockopt_lock_sock() instead of lock_sock()
(from v6.1 rebase, commit d51bbff2aba7)
- use sockptr_t in getsockopt(TCP_AO_GET)
(from v6.1 rebase, commit 34704ef024ae)
- Fixed reallocating crypto_pool's scratch area by IPI while crypto_pool_get() was get by another CPU
- selftests on older kernels (or with CONFIG_TCP_AO=n) should exit with SKIP, not FAIL (Shuah Khan <[email protected]>)
- selftests that check interaction between TCP-AO and TCP-MD5 now SKIP when CONFIG_TCP_MD5SIG=n
- Measured the performance of different hashing algorithms for TCP-AO and compare with TCP-MD5 performance. This is done with hacky patches to iperf (see [3]). At this moment I've done it in qemu/KVM with CPU affinities set on Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz. No performance degradation was noticed before/after patches, but given the measures were done in a VM, without measuring it on a physical dut it only gives a hint of relative speed for different hash algorithms with TCP-AO. Here are results, averaging on 30 measures each:

  TCP:                    3.51Gbits/sec
  TCP-MD5:                1.12Gbits/sec
  TCP-AO(HMAC(SHA1)):     1.53Gbits/sec
  TCP-AO(CMAC(AES128)):   621Mbits/sec
  TCP-AO(HMAC(SHA512)):   1.21Gbits/sec
  TCP-AO(HMAC(SHA384)):   1.20Gbits/sec
  TCP-AO(HMAC(SHA224)):   961Mbits/sec

TCP-AO(HMAC(SHA3-512)): 157Mbits/sec
TCP-AO(HMAC(RMD160)): 659Mbits/sec
TCP-AO(HMAC(MD5): 1.12Gbits/sec
(the last one is just for fun, but may make sense as it provides the same security as TCP-MD5, but allows multiple keys and a mechanism to change them from RFC5925)

Version 2: https://lore.kernel.org/all/[email protected]/T/#u

Changes from v1:
- Building now with CONFIG_IPV6=n (kernel test robot <[email protected]>)
- Added missing static declarations for local functions (kernel test robot <[email protected]>)
- Addressed static analyzer and review comments by Dan Carpenter (thanks, they were very useful!)
- Fix elif without defined() for !CONFIG_TCP_AO
- Recursively build selftests/net/tcp_ao (Shuah Khan), patches in:
https://lore.kernel.org/all/[email protected]/T/#u
- Don't leak crypto_pool reference when TCP-MD5 key is modified/changed
- Add TCP-AO support for nettest.c and fcnal-test.sh
(will be used for VRF testing in later versions)

Comparison between Leonard proposal and this (overview):
https://lore.kernel.org/all/[email protected]/T/#u

Version 1: https://lore.kernel.org/all/[email protected]/T/#u

This patchset implements the TCP-AO option as described in RFC5925. There is a request from industry to move away from TCP-MD5SIG and it seems the time is right to have a TCP-AO upstreamed. This TCP option is meant to replace the TCP MD5 option and address its shortcomings. Specifically, it provides more secure hashing, key rotation and support for long-lived connections (see the summary of TCP-AO advantages over TCP-MD5 in (1.3) of RFC5925). The patch series starts with six patches that are not specific to TCP-AO but implement a general crypto facility that we thought is useful to eliminate code duplication between TCP-MD5SIG and TCP-AO as well as other crypto users. These six patches are being submitted separately in a different patchset [1]. Including them here will show better the gain in code sharing. Next are 18 patches that implement the actual TCP-AO option, followed by patches implementing selftests.

The patch set was written as a collaboration of three authors (in alphabetical order): Dmitry Safonov, Francesco Ruggeri and Salam Noureddine. Additional credits should be given to Prasad Koya, who was involved in early prototyping a few years back. There is also a separate submission done by Leonard Crestez whom we thank for his efforts getting an implementation of RFC5925 submitted for review upstream [2]. This is an independent implementation that makes different design decisions.

For example, we chose a similar design to the TCP-MD5SIG implementation and used setsockopts to program per-socket keys, avoiding the extra complexity of managing a centralized key database in the kernel. A centralized database in the kernel has dubious benefits since it doesn’t eliminate per-socket setsockopts needed to specify which sockets need TCP-AO and what are the currently preferred keys. It also complicates traffic key caching and preventing deletion of in-use keys.

In this implementation, a centralized database of keys can be thought of as living in user space and user applications would have to program those keys on matching sockets. On the server side, the user application programs keys (MKTS in TCP-AO nomenclature) on the listening socket for all peers that are expected to connect. Prefix matching on the peer address is supported. When a peer issues a successful connect, all the MKTs matching the IP address of the peer are copied to the newly created socket. On the active side, when a connect() is issued all MKTs that do not match the peer are deleted from the socket since they will never match the peer. This implementation uses three setsockopt()s for adding, deleting and modifying keys on a socket. All three setsockopt()s have extensive sanity checks that prevent inconsistencies in the keys on a given socket. A getsockopt() is provided to get key information from any given socket.

Few things to note about this implementation:
- Traffic keys are cached for established connections avoiding the cost of such calculation for each packet received or sent.
- Great care has been taken to avoid deleting in-use MKTs
as required by the RFC.
- Any crypto algorithm supported by the Linux kernel can be used to calculate packet hashes.
- Fastopen works with TCP-AO but hasn’t been tested extensively.
- Tested for interop with other major networking vendors (on linux-4.19), including testing for key rotation and long lived connections.

Quoted 7 lines

[1]: https://lore.kernel.org/all/[email protected]/
[2]: https://lore.kernel.org/all/[email protected]/
[3]: https://github.com/0x7f454c46/iperf/tree/tcp-md5-ao
[4]: https://lore.kernel.org/all/166995421700.16716.17446147162780881407.git-patchwork-notify@kernel.org/T/#u
[5]: https://lore.kernel.org/all/[email protected]/T/#u
[6]: https://lore.kernel.org/all/[email protected]/T/#u

Quoted 2 lines

For the set:

I will have to review this series, of course, if bots are happy.

#35 lore raw
Quoted 5 lines · Eric Dumazet
Quoted 4 lines · David Ahern
Quoted 2 lines · Dmitry Safonov

Hi,

[..]

Quoted 5 lines
Quoted 3 lines

For the set:

I will have to review this series, of course, if bots are happy.

Thanks David, Eric!
I appreciate both Acks and potential reviews.

Thanks,
Dmitry