]> git.ipfire.org Git - thirdparty/suricata.git/log
thirdparty/suricata.git
15 months agoconf: avoid quadratic complexity 10874/head
Philippe Antoine [Thu, 21 Mar 2024 08:38:25 +0000 (09:38 +0100)] 
conf: avoid quadratic complexity

Ticket: 6878

Follow up on 15649424a76d01eb332d85620ffc4956d4f3d9be

When adding many sequence nodes, either from start or scalar event

We add "sequence nodes" whose name is an integer cf sequence_node_name
and then run ConfNodeLookupChild to see if it had been already set
(from the command line cf comment in the code)
And ConfNodeLookupChild iterates the whole linked list...

1. We add node 1
2. To add node 2, we check if node 1 equals this new node
3. To add node 3, we check if nodes 1, or 2 equals this new node's name
And so on...

This commits avoids these checks ig the list is empty at the beginning

(cherry picked from commit 240e068b81275e287ffe6555d4c457a0b3916066)

15 months agoflow/inject: Select thread_id by flow flag 10861/head
Jeff Lucovsky [Mon, 15 Apr 2024 14:17:17 +0000 (10:17 -0400)] 
flow/inject: Select thread_id by flow flag

Issue: 6957

Rather than selecting the thread_id index by packets traveling to the
server, use the flow flags. If the flow has been reversed, the second
slot is represents the thread id to be used.

(cherry picked from commit c305ed149c58818779c504b6666fe9c2cb0be4ff)

15 months agoschema: Add stats.capture and in_iface properties
Arne Welzel [Tue, 20 Feb 2024 11:50:40 +0000 (12:50 +0100)] 
schema: Add stats.capture and in_iface properties

New suricata-verify test listens on loopback interface, resulting
in the capture and in_iface fields in the stats and event objects.

(cherry picked from commit f9cf87a003d273ec175590e2ffec053d2672af95)

15 months agostats: Fix non-worker stats missing
Arne Welzel [Sat, 17 Feb 2024 17:19:27 +0000 (18:19 +0100)] 
stats: Fix non-worker stats missing

Commit b8b8aa69b49ac0dd222446c28d00a50f9fd7d716 used tm_name of the
first StatsRecord of a thread block as key for the "threads" object.
However, depending on the type of thread, tm_name can be NULL and would
result in no entry being included for that thread at all. This caused
non-worker metrics to vanish from the "threads" object in the
dump-counters output.

This patch fixes this by remembering the first occurrence of a valid
tm_name within the per-thread block and adds another unittest to
cover this scenario.

(cherry picked from commit f17204191d3bb2201e6b6b1c4cf2e7a96148e8cd)

15 months agopcap: support LINKTYPE_IPV6 (229)
Victor Julien [Fri, 12 Apr 2024 09:02:13 +0000 (11:02 +0200)] 
pcap: support LINKTYPE_IPV6 (229)

This is just another variant of DLT_RAW.

Ticket: #6943.
(cherry picked from commit 76322368ed3ef89c04082939c58535c7234d7173)

15 months agodefrag: fix wrong datalink being logged
Victor Julien [Sat, 23 Mar 2024 19:17:54 +0000 (20:17 +0100)] 
defrag: fix wrong datalink being logged

Eve's packet_info.linktype should correctly indicated what the `packet`
field contains. Until now it was using DLT_RAW even if Ethernet or other
L2+ headers were present.

This commit records the datalink of the packet creating the first
fragment, which can include the L2+ header data.

Bug: #6887.
(cherry picked from commit 49c67b2bb1baa84b7105bca82afe6909be890855)

15 months agodefrag: match up v4 and v6 packet setup
Victor Julien [Sun, 3 Dec 2023 20:15:36 +0000 (21:15 +0100)] 
defrag: match up v4 and v6 packet setup

v4 was doing redundant recursion level setup.

v6 was missing PKT_REBUILT_FRAGMENT flag.

(cherry picked from commit af97316f42c6616536dc8012577827fee9a56f11)

15 months agodoc: add note about fast_pattern w base64_data 10846/head
Shivani Bhardwaj [Wed, 3 Apr 2024 07:12:11 +0000 (12:42 +0530)] 
doc: add note about fast_pattern w base64_data

Bug 6859

15 months agobase64_data: inform of use w fast_pattern
Shivani Bhardwaj [Wed, 3 Apr 2024 07:02:12 +0000 (12:32 +0530)] 
base64_data: inform of use w fast_pattern

Bug 6859

15 months agojsonbuilder: fix serialization of nan and inf 10799/head
Jason Ish [Fri, 5 Apr 2024 14:35:36 +0000 (08:35 -0600)] 
jsonbuilder: fix serialization of nan and inf

When outputting a float, check if its infinity, or not a number and
output a null instead.

Using a null was chosen as this is what serde_yaml, Firefox, Chrome,
Node, etc. do.

Ticket: #6921
(cherry picked from commit 71f59e529c3ceaba977ec5d3492bb8b3d3184cd1)

15 months agodetect/http: fix compile warning in body tests 10793/head
Victor Julien [Thu, 28 Mar 2024 09:43:46 +0000 (10:43 +0100)] 
detect/http: fix compile warning in body tests

When --enable-unittests w/o --enable-debug is used.

(cherry picked from commit e651cf922a02f5882593a23bd2ed9327a5e8d2cc)

15 months agohost-info: remove pcre2_substring_list_free use
Victor Julien [Thu, 4 Apr 2024 15:51:48 +0000 (17:51 +0200)] 
host-info: remove pcre2_substring_list_free use

Function prototype has changed in a recent release. Rather than dealing
with detecting that, fall back to our regular pattern of using
pcre2_substring_copy_bynumber().

Bug: #6918.
(cherry picked from commit b224209f45317abfbea1a1f43f17fa63ae904745)

15 months agoconfigure: .git can be a file as well
Jason Ish [Fri, 5 Apr 2024 16:33:14 +0000 (10:33 -0600)] 
configure: .git can be a file as well

In worktree scenarios, .git is a file. Assuming its a directory causes
the release date to check the ChangeLog instead of the last commit,
while not a big issue, can be confusing.

15 months agoconfigure: export release date for documentation
Jason Ish [Mon, 1 Apr 2024 16:37:49 +0000 (10:37 -0600)] 
configure: export release date for documentation

Sphinx embeds a date in the generated man pages, and to provide
reproducible builds this date needs to be provided to Sphinx,
otherwise it will use the current date.

If building from Git, extract the date from the most recent commit. In
a release, this commit would be the commit that sets the version so is
accurate.

If .git does not exist, use the most recent data found in the
ChangeLog.

The ChangeLog is not used when building from git, as the main/master
branch may not have recent enough timestamps.

This should provide a consistent date when re-building the
distribution from the same non-git archive, or from the same git
commit.

Ticket: #6911
(cherry picked from commit b58dd5e5855864217fe898b43a3e8f9aaff47ae9)

15 months agodocs/userguide: use a consistent date for reproducible builds
Jason Ish [Mon, 1 Apr 2024 16:35:39 +0000 (10:35 -0600)] 
docs/userguide: use a consistent date for reproducible builds

By default, when Sphinx generates the man pages, the current date will
be embedded in them. This can be set to a specific date with the
"today" variable. Typically the date embedded in manpages in the
release date.

To achieve this, attempt to use the environment variable, RELEASE_DATE
to set the "today" variable, reverting back to the empty string if not
set. It is up to our build system to properly set this date.

Ticket: #6911
(cherry picked from commit 51bf1c35103261d075a283d103371a2d9a8a76f7)

15 months agoconfigure: don't check ./revision, it never exists
Jason Ish [Tue, 2 Apr 2024 15:41:08 +0000 (09:41 -0600)] 
configure: don't check ./revision, it never exists

Stop checking the ./revision file for the git revision info, its never
created.

(cherry picked from commit c00c2b116fb0ca2ecdec7a95d94454bce9c18595)

15 months agodocs/conf.py: fix python escape warning
Jason Ish [Mon, 1 Apr 2024 16:34:45 +0000 (10:34 -0600)] 
docs/conf.py: fix python escape warning

/home/jason/oisf/dev/suricata/master/doc/userguide/conf.py:74: SyntaxWarning: invalid escape sequence '\('
  "AC_INIT\(\[suricata\],\s*\[(.*)?\]\)",

(cherry picked from commit 4c16032f63c85eb6bb29b37505440cfe9072c372)

15 months agodpdk: add interrupt (power-saving) mode
Lukas Sismis [Wed, 27 Sep 2023 12:38:36 +0000 (14:38 +0200)] 
dpdk: add interrupt (power-saving) mode

When the packet load is low, Suricata can run in interrupt
mode. This more resembles the classic approach of processing
packets - CPU cores run low and only fetch packets
on interrupt.

Ticket: #6696
(cherry picked from commit 2a2898053c120fa59cb792623b10c2660d8e8cde)

16 months agoflow: Swap thread_ids 10738/head 10770/head 10771/head
Jeff Lucovsky [Mon, 11 Mar 2024 18:58:07 +0000 (14:58 -0400)] 
flow: Swap thread_ids

Issue: 6835

When swapping the flow's direction, also swap the thread_ids.

This should help with the issues identified in
https://redmine.openinfosecfoundation.org/issues/2725

(cherry picked from commit 3c5745978f85f4bf049e2892c8bda167f9e53033)

16 months agoflow/inject: Ensure initialized thread value used
Jeff Lucovsky [Mon, 11 Mar 2024 18:57:16 +0000 (14:57 -0400)] 
flow/inject: Ensure initialized thread value used

Issue: 6835

When injecting a flow, ensure that the selected thread_id has been
initialized. When a flow is picked up midstream, the initialized thread
can be the second thread element.

(cherry picked from commit 9ad73faa0a52428e47412474514b125fda6aa03d)

16 months agorust/mqtt: fix clippy 1.77 warning 10721/head 10724/head
Philippe Antoine [Thu, 21 Mar 2024 21:45:41 +0000 (22:45 +0100)] 
rust/mqtt: fix clippy 1.77 warning

error: creating a mutable reference to mutable static is discouraged
   --> src/mqtt/mqtt.rs:752:23
    |
752 |     let max_msg_len = &mut MAX_MSG_LEN;
    |                       ^^^^^^^^^^^^^^^^ mutable reference to mutable static
    |
    = note: for more information, see issue #114447 <https://github.com/rust-lang/rust/issues/114447>
    = note: this will be a hard error in the 2024 edition
    = note: this mutable reference has lifetime `'static`, but if the static gets accessed (read or written) by any other means, or any other reference is created, then any further use of this mutable reference is Undefined Behavior

16 months agorust/smb: fix clippy nightly warning
Philippe Antoine [Thu, 21 Mar 2024 15:15:53 +0000 (16:15 +0100)] 
rust/smb: fix clippy nightly warning

error: unnecessary use of `to_vec`
    --> src/smb/smb.rs:1048:62
     |
1048 |         let (name, is_dcerpc) = match self.guid2name_map.get(&guid.to_vec()) {
     |                                                              ^^^^^^^^^^^^^^ help: replace it with: `guid`
     |
     = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#unnecessary_to_owned
     = note: `#[deny(clippy::unnecessary_to_owned)]` implied by `#[deny(warnings)]`

And also other uses of to_vec() on already Vec

(cherry picked from commit f7cde8f00ef6d1f92bd86a47d5888bda06941f46)

16 months agorust: fix clippy 1.77 warning
Philippe Antoine [Thu, 21 Mar 2024 15:02:23 +0000 (16:02 +0100)] 
rust: fix clippy 1.77 warning

Ticket: 6883

error: field `0` is never read
  --> src/asn1/mod.rs:36:14
   |
36 |     BerError(Err<der_parser::error::BerError>),
   |     -------- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
   |     |
   |     field in this variant
   |

(cherry picked from commit 02f2fb88333af767ab3b171643357d607f4e86f6)

16 months agorust: add MSRV as rust-version 10677/head
Victor Julien [Wed, 20 Mar 2024 10:02:44 +0000 (11:02 +0100)] 
rust: add MSRV as rust-version

Update github-actions to use it for the MSRV check.

16 months agoversion: start development towards 7.0.5 10671/head
Shivani Bhardwaj [Wed, 20 Mar 2024 07:49:51 +0000 (13:19 +0530)] 
version: start development towards 7.0.5

16 months agorelease: 7.0.4; update changelog suricata-7.0.4
Shivani Bhardwaj [Tue, 19 Mar 2024 10:42:41 +0000 (16:12 +0530)] 
release: 7.0.4; update changelog

16 months agossh: limit length for banner logs
Philippe Antoine [Mon, 19 Feb 2024 15:57:55 +0000 (16:57 +0100)] 
ssh: limit length for banner logs

Ticket: 6770
(cherry picked from commit c4b8fb7aca482d1a1555e27072ca26896b52a480)

16 months agossh: avoid quadratic complexity from long banner
Philippe Antoine [Thu, 22 Feb 2024 09:14:36 +0000 (10:14 +0100)] 
ssh: avoid quadratic complexity from long banner

Ticket: 6799

When we find an overlong banner, we get into the state just
waiting for end of line, and we just want to skip the bytes
until then.
Returning AppLayerResult::incomplete made TCP engine retain
the bytes and grow the buffer that we parsed again and again...

(cherry picked from commit 271ed2008bb7392ca2803ab6dac8952491616151)

16 months agoci: update ubuntu22.04 builds with clang14+asan 10660/head 10663/head
Philippe Antoine [Thu, 14 Mar 2024 08:00:15 +0000 (09:00 +0100)] 
ci: update ubuntu22.04 builds with clang14+asan

using a workround about ASLR

(cherry picked from commit 632e52ca2b72f8a3462b22896cb5d25d9d724f9c)

16 months agodetect: flush when setting no_inspection
Philippe Antoine [Mon, 27 Nov 2023 16:07:21 +0000 (17:07 +0100)] 
detect: flush when setting no_inspection

Ticket: 6578

When a protocol such as SSH sets no_inspection, we still have to
flush the current streams and packets that contain clear-text
for detection.

(cherry picked from commit cc3b4b01ec285d481e7ddcc7e062d1c8ec3b542c)

16 months agodetect: update packet action on protocol change
Philippe Antoine [Tue, 3 Oct 2023 10:39:13 +0000 (12:39 +0200)] 
detect: update packet action on protocol change

Ticket: #6305

When running FlowWorkerStreamTCPUpdate, one of the dequeued packet
may set the flow action to drop, without updating the not-pseudo
packet action, as is done usually with a previous call to
FlowHandlePacketUpdate

(cherry picked from commit 4c4f7ff1a2324bff9e4e09dd8354878b2455ac34)

16 months agoyaml: remove newline from error message
Victor Julien [Mon, 20 Nov 2023 08:31:40 +0000 (09:31 +0100)] 
yaml: remove newline from error message

(cherry picked from commit f53c4ab1494acad333e73965f13cde67cf663b1c)

16 months agonapatech: emit HBA deprecation only once 10644/head 10648/head 9478/head
Jonathan Perkins [Fri, 21 Jul 2023 18:27:37 +0000 (14:27 -0400)] 
napatech: emit HBA deprecation only once

Issue: 6313
This commit removes duplicate HBA deprecation messages from being
emitted.

16 months agodecode/pppoe: skip useless loop unless debug 10627/head
Philippe Antoine [Mon, 11 Mar 2024 14:58:24 +0000 (15:58 +0100)] 
decode/pppoe: skip useless loop unless debug

Looping over tags has no effects, except debug logging

(cherry picked from commit 59d18b9a68e35b6e1129368f5cdccea6764dfc9c)

16 months agodecode/pppoe: pointer cast consistency
Philippe Antoine [Mon, 19 Feb 2024 09:53:02 +0000 (10:53 +0100)] 
decode/pppoe: pointer cast consistency

Ticket: 6787

To do pointer arithmetic, we need to use uint8_t* pointer :
Pointer arithmetic in C is automatically scaled according
to the size of the data type.

Also simplifies the loop condition

(cherry picked from commit d2db8bfdca7c2483f51c35a726d73f1912638199)

16 months agodpdk: output unknown NIC's NUMA node message only on multi-node systems
Lukas Sismis [Sat, 2 Mar 2024 14:57:43 +0000 (15:57 +0100)] 
dpdk: output unknown NIC's NUMA node message only on multi-node systems

Ticket: #6715
(cherry picked from commit 566e89a0d6aa56073705f97d972fc59911d9122e)

16 months agonetmap: Avoid HW errors when using pipes 10620/head
Jeff Lucovsky [Thu, 7 Mar 2024 20:24:36 +0000 (15:24 -0500)] 
netmap: Avoid HW errors when using pipes

Issue: 6837

When using netmap pipes (with lb, for example), avoid direct hardware
related IOCTLs that will fail (not supported with pipes).

(cherry picked from commit af529a56a9bcb37bae5243236a4505fff0df268b)

16 months agoplugins: initialize plugins earlier
Jason Ish [Tue, 27 Feb 2024 22:16:08 +0000 (16:16 -0600)] 
plugins: initialize plugins earlier

Capture plugins need to be registered before LiveDeviceFinalize,
otherwise Suricata errors before the plugin gets a chance to load.

Bug: #6811
(cherry picked from commit 25d0e0b68c333029fb8263cb9a2a287f9064b5c8)

16 months agothread modules: separate initialization from registration
Jason Ish [Tue, 27 Feb 2024 22:07:33 +0000 (16:07 -0600)] 
thread modules: separate initialization from registration

Move the zero'ing to the thread module array InitGlobal in an effort
to fix capture modules.

At some point device validation moved to a point in startup before
plugins are loaded meaning that capture plugins could not be
used. Moving plugin registration early enough caused some of their
registration to be wiped out as clearing the array was done after.

Bug: #6811
(cherry picked from commit 96e61b043c9b6757ca84ddb74543073b2ef3d7e6)

16 months agodpdk: add a user warning on the change of RX/TX descriptors 10607/head
Lukas Sismis [Sat, 9 Mar 2024 12:42:33 +0000 (13:42 +0100)] 
dpdk: add a user warning on the change of RX/TX descriptors

Ticket: #6739
(cherry picked from commit 9185ca8f3801537b359a14d5ad90ca4728dca1ad)

16 months agooutput/filestore: delay snprintf until needed
Philippe Antoine [Tue, 27 Feb 2024 20:45:03 +0000 (21:45 +0100)] 
output/filestore: delay snprintf until needed

Perf optimization so that we do not call snprintf in the
common code path.

Ticket: 6796
(cherry picked from commit d255a5c7a309c067e15dde01560631f4ed427db1)

16 months agooutput/filestore: remove duplicate snprintf
Philippe Antoine [Mon, 26 Feb 2024 16:28:25 +0000 (17:28 +0100)] 
output/filestore: remove duplicate snprintf

Ticket: 6796
(cherry picked from commit 7641c07af95f4373c50812890737358fda885a4c)

16 months agooutput: log tx only when there is a fresh app update
Philippe Antoine [Thu, 22 Feb 2024 08:42:41 +0000 (09:42 +0100)] 
output: log tx only when there is a fresh app update

Ticket: 6796

Similar to commit for detection
9240ae250cc369306803740279df2ab3eca6b54a

We only have more logging to do if the app update was fresh,
ie if p->app_update_direction != 0

If we have data acknowledged in one direction,
and then many packets in the other direction,
the APP_UPDATED flow flags did not get reset because we did not
run detection yet in this direction,
but there is nothing more to do after the first packet in the
other direction.

(cherry picked from commit c41540f8391b44bf4d91d32641cc608ebfecb19c)

16 months agouserguide: fix explanation about bsize ranges 10587/head
Juliana Fajardini [Wed, 24 Jan 2024 15:00:41 +0000 (12:00 -0300)] 
userguide: fix explanation about bsize ranges

Our code handles Uint ranges as exclusive, but for bsize, our
documentation stated that they're inclusive.

Cf. from uint.rs:

    DetectUintMode::DetectUintModeRange => {
        if val > x.arg1 && val < x.arg2 {
            return true;
        }
    }

Task #6708

(cherry picked from commit 244a35d539989c00eb5cbac86e7355227cd9da03)

16 months agodoc: mention the limited number of RX/TX descriptors on Intel NICs
Lukas Sismis [Sun, 11 Feb 2024 20:14:08 +0000 (21:14 +0100)] 
doc: mention the limited number of RX/TX descriptors on Intel NICs

Ticket: 6748
(cherry picked from commit 356f9ffa130fbaaf82e7e28de98bdc24fe32f945)

17 months agodpdk: max cache size should be lower than one of the constraints 10577/head
Lukas Sismis [Sun, 11 Feb 2024 19:43:37 +0000 (20:43 +0100)] 
dpdk: max cache size should be lower than one of the constraints

Ticket: 6741
(cherry picked from commit c65ff35819845a3f42c75f79d54f9ab91c5c2ec9)

17 months agodpdk: sanitize integer overflow in the configuration
Lukas Sismis [Sun, 11 Feb 2024 12:42:20 +0000 (13:42 +0100)] 
dpdk: sanitize integer overflow in the configuration

Ticket: #6737
(cherry picked from commit cc2eb2d8b77e96586a607f661c7eed9ab41076fc)

17 months agohugepages: run hugepage check only on DPDK runmode and on Linux 10553/head
Lukas Sismis [Sat, 10 Feb 2024 19:04:55 +0000 (20:04 +0100)] 
hugepages: run hugepage check only on DPDK runmode and on Linux

Previous implementation allowed FreeBSD to enter into the hugepage
analysis. It then failed with an error message because hugepage/
NUMA node paths that are used in the codebase to retrieve info about
the system are not the same with the structure in Linux.

Additionally, the messages were logged on error level. It has been
demoted to info level because the whole hugepage analysis checkup is
only for informational purposes and does not affect Suricata operation.

The hugepage analysis and the hugepage snapshots are now limited to
only run in the DPDK runmode.

Ticket: #6760
Ticket: #6762
(cherry picked from commit 4b0704db5501c76592b2e12912b82a17f95fd842)

17 months agoerrors: add SC_ENOENT error
Lukas Sismis [Sat, 10 Feb 2024 19:04:24 +0000 (20:04 +0100)] 
errors: add SC_ENOENT error

(cherry picked from commit f88bd2dd3f37adeb44351921ae6313278964ef3d)

17 months agopfring: fix leak of configuration data and in single mode 10533/head
Jason Ish [Tue, 27 Feb 2024 22:55:59 +0000 (16:55 -0600)] 
pfring: fix leak of configuration data and in single mode

Fix leak of configuration data on exit. Also, in single mode set
thread count to one instead of the CPU count.

Bug: #4734
(cherry picked from commit 5f0853599b15adafdf47399375728d70a94c9867)

17 months agopfring: fix leak of packet on exit
Jason Ish [Tue, 27 Feb 2024 23:19:54 +0000 (17:19 -0600)] 
pfring: fix leak of packet on exit

Bug: #4734
(cherry picked from commit 358f1e35eed415d57cf24500deb58d1224c67b71)

17 months agocppcheck: Address cpcheck report of an FP
Jeff Lucovsky [Sat, 25 Nov 2023 14:22:19 +0000 (09:22 -0500)] 
cppcheck: Address cpcheck report of an FP

Issue: 6527

Address the FP raised by cppcheck -- note that although the code
corectly checks to ensure that `to_shift != &sb->reqion`, the logic was
detected as a FP. Rework the code to eliminate the FP.

(cherry picked from commit 40e3514e7a6c89a786ebf17469404524fb0d2d52)

17 months agodetect/address: avoid cppcheck false positive
Victor Julien [Tue, 20 Feb 2024 15:06:08 +0000 (16:06 +0100)] 
detect/address: avoid cppcheck false positive

src/detect-engine-address.c:1386:5: error: Memory leak: map.string [memleak]
    return true;
    ^

Seems cppcheck looses track of the pointer after the unnecessary cast to
void.

Bug: #6527.
(cherry picked from commit 78703bbc865a1dd15a1ae1c4c996b0150096ed0e)

17 months agocppcheck/detect: Address cppcheck memory leak
Jeff Lucovsky [Sat, 25 Nov 2023 14:20:43 +0000 (09:20 -0500)] 
cppcheck/detect: Address cppcheck memory leak

Issue: 6527

Ensure that the `map->string` memory isn't leaked following an error
return from `HashListTableAdd`

(cherry picked from commit 8b2fd434fc04ae84317d44d0a450af3ead9a157f)

17 months agodetect: fix memory leak in error handling
Victor Julien [Tue, 20 Feb 2024 15:06:57 +0000 (16:06 +0100)] 
detect: fix memory leak in error handling

Pointed out by cppcheck:

src/detect-parse.c:1544:9: error: Memory leak: sig.init_data [memleak]
        SCFree(sig);
        ^

Bug: #6527.
(cherry picked from commit eb1d0c2cc2b673a8f751cd88546e7c37e33ab8e0)

17 months agorunmode/unix-socket: fix cppcheck warnings
Victor Julien [Sat, 17 Feb 2024 09:52:59 +0000 (10:52 +0100)] 
runmode/unix-socket: fix cppcheck warnings

src/runmode-unix-socket.c:547:9: warning: %d in format string (no. 1) requires 'int' but the argument type is 'unsigned int'. [invalidPrintfArgType_sint]
        snprintf(tstr, sizeof(tstr), "%d", cfile->tenant_id);
        ^
src/runmode-unix-socket.c:1040:5: warning: %d in format string (no. 1) requires 'int' but the argument type is 'unsigned int'. [invalidPrintfArgType_sint]
    snprintf(prefix, sizeof(prefix), "multi-detect.%d", tenant_id);
    ^
src/runmode-unix-socket.c:1189:5: warning: %d in format string (no. 1) requires 'int' but the argument type is 'unsigned int'. [invalidPrintfArgType_sint]
    snprintf(prefix, sizeof(prefix), "multi-detect.%d", tenant_id);
    ^

(cherry picked from commit 872f007a14cdea70345d381620c40ac4c8e95bca)

17 months agodetect/mpm: don't test for impossible condition
Victor Julien [Fri, 5 Jan 2024 07:29:16 +0000 (08:29 +0100)] 
detect/mpm: don't test for impossible condition

Functions can't get called with NULL signature or NULL
Signature::init_data.

(cherry picked from commit 124a8dea204b400f27ac2b8c15c994ffa996b1ea)

17 months agomqtt: Improve frame parsing w/mult. PDUs 10489/head 10508/head
Jeff Lucovsky [Wed, 3 Jan 2024 14:28:41 +0000 (09:28 -0500)] 
mqtt: Improve frame parsing w/mult. PDUs

This commit improves the mqtt parsing of frames to handle multiple PDUs.

Issue: 6592
(cherry picked from commit f9a20dafc6abe2c03dea98286c86430b447d8196)

17 months agotcp: do not assign TCP flags to pseudopackets 10484/head 10485/head
Lukas Sismis [Tue, 6 Feb 2024 10:38:19 +0000 (11:38 +0100)] 
tcp: do not assign TCP flags to pseudopackets

Previously pseudopackets were assigned with ACK flag which falsely turned
"SYN" flows to "SYN/ACK" flows when Suricata ran with raw content-matching
rules. The problem occured during the flow timeout or Suricata shutdown,
essentially, when the flow was being kicked out (with a pseudopacket).

When Suricata ran without raw content-matching rules (the ruleset did not
contain content matching keywords or it only contained keywords that are
app-layer content-matching) then raw stream reassembly tracking is turned off
(SignatureHasStreamContent()).
This in turn disabled a check in StreamNeedsReassembly() and the right edge
was not checked with the raw stream progress. In turn, it did not generate
a pseudopacket that would go through the detection engine. Suricata with
raw content-matching keywords would therefore on a flow with SYN packet only
return STREAM_HAS_UNPROCESSED_SEGMENTS_NEED_ONLY_DETECTION which would generate
the pseudopacket.

In Suricata versions <= 6.0.x, the flow output was correct because
only the commit 1bb6f44ff01363fa29488f1ae83b9368e33c2770 started to
differentiate the right edge calculation between the raw and application
layer streams. The older Suricata versions used only the application layer
right edge equation and therefore did not generate a pseudopacket.

Ticket: #6734
(cherry picked from commit ebf465a11bffcd2049e17648e6c46c067692b02a)

17 months agomulti-tenant: fix loader dead lock 10442/head
Victor Julien [Sun, 11 Feb 2024 08:29:38 +0000 (09:29 +0100)] 
multi-tenant: fix loader dead lock

A dead lock could occur at start up, where a loader thread would
get stuck on it's condition variable, while the main thread was
polling the loaders task results.

The vector to the dead lock is as follows:

main                         loader
DetectEngineMultiTenantSetup
-DetectLoaderSetupLoadTenant
--DetectLoaderQueueTask
---lock loader
---add task
---unlock loader
                        lock loader
                        check/exec tasks
                        unlock loader
---wake up threads
                        lock ctrl mutx
                        cond wait ctrl
                        unlock ctrl
-DetectLoadersSync
--lock loader
--check tasks
--unlock loader

Between the main thread unlocking the loader and waking up the
threads, it is possible that the loader has already moved ahead
but not yet entered its conditional wait. The main thread sends
its condition signal, but since the loader isn't yet waiting on
it the signal is ignored. Then when the loader does enter its
conditional wait, the signal is not sent again.

This patch updates the logic to send signals much more often.
It also makes sure that the signal is sent under lock, as the
API requires.

Bug: #6768.

Co-authored-by: Shivani Bhardwaj <shivani@oisf.net>
17 months agostats: Add unittest for basic stats serialization
Arne Welzel [Thu, 8 Feb 2024 17:23:38 +0000 (18:23 +0100)] 
stats: Add unittest for basic stats serialization

Main purpose is to validate that the 30 of bond0.30 isn't expanded into
a nested object during serialization.

(cherry picked from commit 08db0f302b3b2817d0fdfb1fe836fd16cc5e8775)

17 months agostats: Do not expand dots of tm_name
Arne Welzel [Mon, 5 Feb 2024 16:45:30 +0000 (17:45 +0100)] 
stats: Do not expand dots of tm_name

When an interface with dots is used, per worker stats are nested by the
dot-separated-components of the interface due to the usage of
OutputStats2Json().

Prevent this by using OutputStats2Json() on a per-thread specific object
and setting this object into the threads object using the
json_object_set_new() which won't do the dot expansion.

This was tested by creating an interface with dots in the name
and checking the stats.

    ip link add name a.b.c type dummy

With Suricata 7.0.2, sniffing on the a.b.c interface results in the
following worker stats format:

    "threads": {
      "W#01-a": {
        "b": {
          "c": {
            "capture": {
              "kernel_packets": 0,

After this fix, the output looks as follows:

    "threads": {
      "W#01-a.b.c": {
        "capture": {
          "kernel_packets": 0,

Ticket: #6732
(cherry picked from commit b8b8aa69b49ac0dd222446c28d00a50f9fd7d716)

17 months agodetect-engine-iponly: improve ip list performance
Simon Dugas [Fri, 29 Dec 2023 16:58:50 +0000 (11:58 -0500)] 
detect-engine-iponly: improve ip list performance

The runtime complexity of insertion sort is approx. O(h*n)^2 where
h is the size of the HOME_NET and n is the number of ip only rules
that use the HOME_NET.

Replacing this with qsort significantly improves rule load time when
a large HOME_NET is used in combination with a moderate amount of ip
only rules.

(cherry picked from commit 17f9d7aeccd5a69ef2e3344386d1585d231e2933)

17 months agodetect/tls.certs: fix direction handling
Victor Julien [Sat, 27 Jan 2024 08:59:55 +0000 (09:59 +0100)] 
detect/tls.certs: fix direction handling

Direction flag was checked against wrong field, leading to undefined behavior.

Bug: #6778.
(cherry picked from commit 3c06457b74db5a2fd070fe1675c72f53423786de)

17 months agothreads/mutex: Ensure mutex held before signaling
Jeff Lucovsky [Sun, 4 Feb 2024 14:44:44 +0000 (09:44 -0500)] 
threads/mutex: Ensure mutex held before signaling

Ensure that the mutex protecting the condition variable is held before
signaling it. This ensures that the thread(s) awaiting the signal are
notified.

Issue: 6569
(cherry picked from commit 2a1a70b3089751b30f623871063ce155451d4cbc)

17 months agonetmap: Release lock to avoid deadlock
Jeff Lucovsky [Thu, 8 Feb 2024 14:12:47 +0000 (09:12 -0500)] 
netmap: Release lock to avoid deadlock

Issue: 6755

When NetmapOpen encounters an error opening the netmap device, it'll
retry a bit. When the retry limit is reached, it'll shutdown Suricata.

This commit ensures that the device list lock is not held when before
closing all open devices before terminating Suricata.

(cherry picked from commit 364adeeb0442592ceea4d329622ca5fe34bf31fd)

17 months agoutil/mime: fix memory leak
Alexey Simakov [Wed, 7 Feb 2024 14:51:00 +0000 (17:51 +0300)] 
util/mime: fix memory leak

Fix memory leak at util-decode-mime:MimeDecInitParser, which
root cause is not-freeing allocated memory for mimeMsg

Bug: #6745
(cherry picked from commit 231c892befe25c23d5d3b54e48077c8446940eb5)

17 months agodetect: respect directionality for filestore
Philippe Antoine [Thu, 25 Jan 2024 13:26:09 +0000 (14:26 +0100)] 
detect: respect directionality for filestore

Ticket: 6617

So that rules with keyword like `filestore:to_server,flow`
only store the files to server and not the ones to client...

Directionality only worked with the default scope, ie the
current file, and not the scope tx or scope flow.
For non-default scope, tx or flow, both directions were stored
whatever the directionality specified.

For these non-default scopes, this commit keeps a default
of both directions, but use only one direction if specified.

Need to split flag FLOWFILE_STORE per direction, so that Suricata
can retain this (optional) directional info from the filestore
keyword.

Fixes: 79499e476979 ("app-layer: move files into transactions")
(cherry picked from commit 63caa0b40a66ecf1a34bbb6d942d4a044b7728a5)

17 months agoapp-layer/template: use a max number of txs 10416/head
Philippe Antoine [Mon, 12 Feb 2024 12:42:14 +0000 (13:42 +0100)] 
app-layer/template: use a max number of txs

Ticket: 6773
(cherry picked from commit c99d93c2574cfa0fe2b94c1ab3a2cacd5a15035c)

17 months agodetect: errors on 65k filestore signatures 10357/head
Philippe Antoine [Mon, 9 Oct 2023 12:49:54 +0000 (14:49 +0200)] 
detect: errors on 65k filestore signatures

Errors when a detection engine gets 65k filestore signatures to
avoid the hard limit to have 65k filestore per signature group
head

Ticket: #6393
(cherry picked from commit db99c45d239d5ca6e805094195f7ae39d3051e44)

17 months agodetect/http: report error on alloc failure 10350/head 10355/head
Victor Julien [Fri, 9 Feb 2024 07:09:38 +0000 (08:09 +0100)] 
detect/http: report error on alloc failure

(cherry picked from commit 7e4dba7dfbd063be554b6145b747d7879eaa1f84)

17 months agodetect/http_header: fix leak on realloc failure
Philippe Antoine [Thu, 8 Feb 2024 19:23:59 +0000 (20:23 +0100)] 
detect/http_header: fix leak on realloc failure

(cherry picked from commit b48ec8a03922f36e76f2d6d942f2963afc2a3345)

17 months agodetect-http: add superfluous alloc check for cocci
Jason Ish [Thu, 8 Feb 2024 19:21:11 +0000 (13:21 -0600)] 
detect-http: add superfluous alloc check for cocci

Add not-needed SCCalloc return check to satisfy our Cocci malloc
checks as it can't see that the caller immediately checks the return
value of this simple wrapper around SCCalloc.

(cherry picked from commit f800ed0f901488d19fccf858ba0711792bf407ff)

17 months agorust: fix clippy ptr_arg warnings
Philippe Antoine [Thu, 8 Feb 2024 14:47:23 +0000 (15:47 +0100)] 
rust: fix clippy ptr_arg warnings

error: writing `&Vec` instead of `&[_]` involves a new object where a slice will do
   --> src/dns/log.rs:371:29
    |
371 | pub fn dns_print_addr(addr: &Vec<u8>) -> std::string::String {
    |                             ^^^^^^^^ help: change this to: `&[u8]`
    |
    = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#ptr_arg

(cherry picked from commit 68b0052018079adc11ea1e35ab686c30716a8aad)

17 months agoci: authors check using OISF repo
Philippe Antoine [Wed, 7 Feb 2024 20:54:28 +0000 (21:54 +0100)] 
ci: authors check using OISF repo

As flagged critical by codescan

(cherry picked from commit 7f5e98e6df279be39a47c18ac0553a1cfebb906c)

17 months agoversion: start development towards 7.0.4 10342/head
Shivani Bhardwaj [Fri, 9 Feb 2024 04:21:21 +0000 (09:51 +0530)] 
version: start development towards 7.0.4

17 months agorelease: 7.0.3; update changelog suricata-7.0.3
Shivani Bhardwaj [Thu, 8 Feb 2024 09:12:53 +0000 (14:42 +0530)] 
release: 7.0.3; update changelog

17 months agohttp2: limit number of concurrent transactions
Philippe Antoine [Thu, 9 Nov 2023 15:15:36 +0000 (16:15 +0100)] 
http2: limit number of concurrent transactions

Ticket: 6481

Instead of just setting the old transactions to a drop state so
that they get later cleaned up by Suricata, fail creating new ones.

This is because one call to app-layer parsing can create many
transactions, and quadratic complexity could happen in one
single app-layer parsing because of find_or_create_tx

(cherry picked from commit 80abc22f6475b6a87a33166729a871203f34d578)

17 months agopgsql: parse only PDU when type is unknown
Philippe Antoine [Tue, 17 Oct 2023 20:04:57 +0000 (22:04 +0200)] 
pgsql: parse only PDU when type is unknown

A next PDU may already be in the slice to parse.
Do not skip its parsing, ie do not use rest, but take just
the length of the pdu

(cherry picked from commit 86de7cffa7e8f06fe9d600127e7dabe89c7e81dd)

17 months agopgsql: parse auth message within its bound
Philippe Antoine [Tue, 17 Oct 2023 20:01:27 +0000 (22:01 +0200)] 
pgsql: parse auth message within its bound

If the next PDU is already in the slice next, do not use it and
restrict ourselves to the length of this PDU.
Avoids overconsumption of memory by quadratic complexity, when
having many small PDUS in one big chunk being parsed

Ticket: #6411
(cherry picked from commit f52c033e566beafb4480c139eb18662a2870464f)

17 months agodetect: fixes use-after-free with http.request_header
Philippe Antoine [Tue, 7 Nov 2023 09:33:21 +0000 (10:33 +0100)] 
detect: fixes use-after-free with http.request_header

Ticket: #6441

This keyword and the response one use a multiple inspection buffer.
But the different instances point to the same memory address
that comes from HttpHeaderGetBufferSpace and is not owned
by the transaction, and is rebuilt, which is a functional
bug in itself.

As it gets crafted, it can get reallocated if one header
is over 1024 bytes, while the previous freed pointer will still get
used for the previous headers.

(cherry picked from commit bc422c17d6961f03f673f2999a949913e89fc2d0)

17 months agosmtp: avoid creating empty transaction
Philippe Antoine [Tue, 14 Nov 2023 20:51:37 +0000 (21:51 +0100)] 
smtp: avoid creating empty transaction

Ticket: 6477

So as to avoid ending up with too many empty transactions.

This happens when Suricata sees a DATA command in the current
transaction but did not have a confirmation response for it.
Then, if Suricata receives another DATA command, it will
create another new transaction, even if the previous one
is empty. And so, a malicious client can create many empty
transactions by just sending a repeated amount of DATA commands
without having a confirmation code for them.

Suricata cannot use state->current_command == SMTP_COMMAND_DATA
to prevent this attack and needs to resort to a new boolean
is_data because the malicious client may send another dummy command
after each DATA command.

This patch leaves only one call to SMTPTransactionCreate

(cherry picked from commit 61f2e4e1e5b34dfd8ae44d1c15253e2da65f6e6a)

17 months agosmtp: config limit maximum number of live transactions
Philippe Antoine [Thu, 9 Nov 2023 11:27:59 +0000 (12:27 +0100)] 
smtp: config limit maximum number of live transactions

Ticket: #6477
(cherry picked from commit 8f73a0ac5588cb5e5c501b3c7a07cb5d35b99d92)

17 months agohttp1: configurable max number of live tx per flow
Philippe Antoine [Tue, 17 Oct 2023 13:28:53 +0000 (15:28 +0200)] 
http1: configurable max number of live tx per flow

Ticket: #5921

Co-authored-by: Jason Ish <jason.ish@oisf.net>
(cherry picked from commit 4175680a8a1c0dfaa491ee63d6e36c011d498473)

17 months agohttp1: remove transactions from their list
Philippe Antoine [Fri, 7 Apr 2023 14:02:41 +0000 (16:02 +0200)] 
http1: remove transactions from their list

instead of keeping a NULL pointer in an array

Ticket: #5921
(cherry picked from commit 8f63a8f3bffbbaf8fae4985ee5f974ab326b08c0)

17 months agohttp2: handle reassembly for continuation frames
Philippe Antoine [Thu, 25 Jan 2024 15:01:14 +0000 (16:01 +0100)] 
http2: handle reassembly for continuation frames

Ticket: 5926

HTTP2 continuation frames are defined in RFC 9113.
They allow header blocks to be split over multiple HTTP2 frames.
For Suricata to process correctly these header blocks, it
must do the reassembly of the payload of these HTTP2 frames.
Otherwise, we get incomplete decoding for headers names and/or
values while decoding a single frame.

Design is to add a field to the HTTP2 state, as the RFC states that
these continuation frames form a discrete unit :
> Field blocks MUST be transmitted as a contiguous sequence of frames,
> with no interleaved frames of any other type or from any other stream.
So, we do not have to duplicate this reassembly field per stream id.

Another design choice is to wait for the reassembly to be complete
before doing any decoding, to avoid quadratic complexity on partially
decoding of the data.

(cherry picked from commit aff54f29f8c3f583ae0524a661aa90dc7a2d3f92)

18 months agomqtt: fix logic when setting event 10278/head 10308/head
Philippe Antoine [Tue, 17 Oct 2023 08:26:57 +0000 (10:26 +0200)] 
mqtt: fix logic when setting event

Especially sets transactions to complete when we get a response
without having seen the request, so that the transactions
end up getting cleaned (instead of living/leaking in the state).

Also try to set the event on the relevant transaction, instead
of creating a new transaction just for the purpose of having
the event.

Ticket: #6299
(cherry picked from commit 89936b6530690c6d03869b2ad8b82f9f84776f94)

18 months agodetect: merge sorted lists instead of qsort
Philippe Antoine [Mon, 18 Sep 2023 11:27:47 +0000 (13:27 +0200)] 
detect: merge sorted lists instead of qsort

Ticket: #6299

Simply because it is faster (just linear).

This is for merging match_array into tx_candidates

(cherry picked from commit 5bb8800588e7b4a09e1770f049cd88be71e2d30b)

18 months agodetect: avoids case of useless detection on txs
Philippe Antoine [Thu, 7 Sep 2023 12:33:04 +0000 (14:33 +0200)] 
detect: avoids case of useless detection on txs

When a TCP flow packet has not led to app-layer updates,
it is useless to run DetectRunTx, as there cannot be new
matches.

This happens for instance, when one side sends in a row multiple
packets which are not acked (and thus not parsed in IDS mode).

Doing so requires to move up the call to
AppLayerParserSetTransactionInspectId
so that it is run the same times DetectRunTx is run, and not in the
case where the transaction was not updated.

Ticket: 6299
(cherry picked from commit 9240ae250cc369306803740279df2ab3eca6b54a)

18 months agodoc: note what version "requires" was added in 10253/head
Jason Ish [Wed, 24 Jan 2024 15:55:37 +0000 (09:55 -0600)] 
doc: note what version "requires" was added in

(cherry picked from commit 8bf8131c31088f501f4d151f8748a817f89f7c1f)

18 months agodetect/requires: reset sigerror flags for each rule
Jason Ish [Wed, 24 Jan 2024 15:02:19 +0000 (09:02 -0600)] 
detect/requires: reset sigerror flags for each rule

"sigerror_ok" and "sigerror_requires" were not being reset after each
rule which could lead to a rule load error being incorrectly tracked
as skipped rather than failed.

Also initialize "skippedsigs" to 0 along with "goodsigs" and
"badsigs", while not directly related to this issue, could also throw
off some stats.

Ticket: #6710
(cherry picked from commit de3cbe4c90fb7fb6d54b6876ab515f791145271a)

18 months agorequirements: use libhtp 0.5.x
Jason Ish [Wed, 24 Jan 2024 20:41:47 +0000 (14:41 -0600)] 
requirements: use libhtp 0.5.x

Move to libhtp to the 0.5.x branch instead of 0.5.45.

(cherry picked from commit c3b3c11e30e1afdbde5c196a0692ee5e2c379cb1)

18 months agodoc: remove references to prehistoric versions
Lukas Sismis [Mon, 11 Dec 2023 00:47:55 +0000 (01:47 +0100)] 
doc: remove references to prehistoric versions

Remove references that are mentioning Suricata 3 or less
As a note - only one Suricata 4 reference found:
(suricata-yaml.rst:"In 4.1.x")
Fast pattern selection criteria can be internally found by inspecting
SupportFastPatternForSigMatchList and SigTableSetup functions.

Ticket: #6699
(cherry picked from commit 6e4cc79b398d91e243bd1574cbbcbbf7cd021c5d)

18 months agodpdk: rework hugepage hints to use per-numa information
Lukas Sismis [Mon, 30 Oct 2023 11:33:57 +0000 (12:33 +0100)] 
dpdk: rework hugepage hints to use per-numa information

Previous integration of hugepage analysis only fetched data
from /proc/meminfo. However this proved to be often
deceiving mainly for providing only global information and
not taking into account different hugepage sizes (e.g. 1GB
hugepages) and different NUMA nodes.

Ticket: #6697
(cherry picked from commit ca6f7c2d002cdad8eb3630860b5023560626aee0)

18 months agoflow-bypass: Set bypass thread to running state
Vincent Li [Fri, 4 Aug 2023 17:49:35 +0000 (17:49 +0000)] 
flow-bypass: Set bypass thread to running state

When running Suricata in XDP bypass mode (bypass: yes),

Suricata started up with error:
Error: threads: thread "FB" failed to start in time: flags 0003

"FB" thread does not transition from THV_INIT_DONE to THV_RUNNING.

Set "FB" thread THV_RUNNING state in BypassedFlowManager().

Bug: #6254

Signed-off-by: Vincent Li <vincent.mc.li@gmail.com>
(cherry picked from commit f80d26db0b62d4a547290a1cf536754f79927a3b)

18 months agoipfw: close(2) instead shutdown(2) of the divert(4) socket
Gleb Smirnoff [Thu, 11 Jan 2024 21:35:21 +0000 (13:35 -0800)] 
ipfw: close(2) instead shutdown(2) of the divert(4) socket

The shutdown(2) syscall would always return ENOTCONN for FreeBSD 11,
FreeBSD 12, FreeBSD 13 and FreeBSD 14.  It could do some action on the
socket in the kernel in FreeBSD 10 and before, did not test.

(cherry picked from commit b239e88c9303d83a3074ed69c4989bd136bbf8b4)

18 months agodetect: case-insensitive comparison for requires
Philippe Antoine [Thu, 4 Jan 2024 10:42:55 +0000 (11:42 +0100)] 
detect: case-insensitive comparison for requires

Ticket: 6656
(cherry picked from commit d3218385e942a4a01f3938f0cd7050f5dc7c9a6f)

18 months agosource/erf-dag: compiler warnings
Stephen Donnelly [Wed, 10 Jan 2024 03:39:34 +0000 (16:39 +1300)] 
source/erf-dag: compiler warnings

Bug: #6667.

Fix compiler warnings for function pointer parameters missing const with --enable-dag

(cherry picked from commit c28cc93e23e829dac34ca0170df754ab1063c482)

18 months agoendace: Fix source-dag timestamps
Stephen Donnelly [Mon, 11 Dec 2023 02:32:06 +0000 (15:32 +1300)] 
endace: Fix source-dag timestamps

Bug: #6618.

Fix Endace ERF to SCTime_t timestamp conversion

Fix typo preventing compilation with --enable-dag

(cherry picked from commit 879db3dbc3e93912c784375c85d88404a9371f31)

18 months agostats: add rules skipped 10211/head
Jason Ish [Tue, 19 Dec 2023 18:13:23 +0000 (12:13 -0600)] 
stats: add rules skipped

Rule skipped is a count of the number of rules that are skipped due to
missing requirements.

Feature: #6637
(cherry picked from commit b453eea1502cc5455f383c4fe82c6011fee93495)