wip: new(driver/modern_bpf): home-made bpf_loop for sendmmsg and recvmmsg.

[FedeDP]

What type of PR is this?

/kind feature

Any specific area of the project related to this PR?

/area driver-modern-bpf

Does this PR require a change in the driver versions?

What this PR does / why we need it:

For sendmmsg and recvmmsg in modern_bpf probe we could not use bpf_loop helper because it caused verifier issues on kernels prior to 5.13 (see #2027 (comment)).
Therefore we used a loop up to 16; then we noticed that 16 was too high, thus we lowered the limit to 8 (600fefb).
This means we can only read first 8 messages sent through sendmmsg and recvmmsg.

This PR's scope is to increase the limit up to 256 (in the first proposed draft, i limit it to 64).
The idea is to build an X_MACRO that let us easily chain tail-calls.
EBPF tail call limit is MAX_TAIL_CALL_CNT, ie 32 on older kernels, and 33 nowadays. For now, i capped the implementation to 8 (and each tail-call loops 8 times, this 64 total).

The good: we support up to 8x32=256messages (that is a far better situation than now)
The bad: we need to extract network args at each iteration because we can't share state between tail calls
The ugly: well, the code gets a bit convoluted but it does the trick. It's just plain old C anyway :) 

Note also that to make it a little bit less verbose, i could also create a new header with all the macros and share them between the 2 source files, since they are basically identical (naming aside). I decided to keep it as easy as possible, but it is a possibility (and it would be more future proof since we would only have a single place to update if needed).

Which issue(s) this PR fixes:

Fixes #

Special notes for your reviewer:

Putting it in wip to allow a discussion.
THIS IS NOT FOR 0.20.0.

Does this PR introduce a user-facing change?:

NONE

[FedeDP]

/milestone TBD

[poiana]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: FedeDP

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• driver/OWNERS [FedeDP]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[FedeDP]

I want to hear @Andreagit97 and @Molter73 opinions on this one :)

[github-actions]

Perf diff from master - unit tests

Warning:
Processed 36803 events and lost 1 chunks!

Check IO/CPU overload!

    11.27%     -0.69%  [.] sinsp::next
     1.53%     +0.52%  [.] next
     1.75%     -0.41%  [.] libsinsp::sinsp_suppress::process_event
     5.75%     -0.26%  [.] next_event_from_file
     0.47%     +0.26%  [.] std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >::_M_construct<char*>

Heap diff from master - unit tests

peak heap memory consumption: 0B
peak RSS (including heaptrack overhead): 0B
total memory leaked: 0B

Heap diff from master - scap file

peak heap memory consumption: 0B
peak RSS (including heaptrack overhead): 0B
total memory leaked: 0B

Benchmarks diff from master

Comparing gbench_data.json to /root/actions-runner/_work/libs/libs/build/gbench_data.json
Benchmark                                                         Time             CPU      Time Old      Time New       CPU Old       CPU New
----------------------------------------------------------------------------------------------------------------------------------------------
BM_sinsp_split_mean                                            +0.0125         +0.0125           146           148           146           148
BM_sinsp_split_median                                          +0.0168         +0.0168           146           148           146           148
BM_sinsp_split_stddev                                          +0.7774         +0.7748             1             1             1             1
BM_sinsp_split_cv                                              +0.7555         +0.7529             0             0             0             0
BM_sinsp_concatenate_paths_relative_path_mean                  -0.0333         -0.0332            61            59            61            59
BM_sinsp_concatenate_paths_relative_path_median                -0.0336         -0.0336            61            59            61            59
BM_sinsp_concatenate_paths_relative_path_stddev                +0.5974         +0.5989             1             1             1             1
BM_sinsp_concatenate_paths_relative_path_cv                    +0.6523         +0.6539             0             0             0             0
BM_sinsp_concatenate_paths_empty_path_mean                     -0.0452         -0.0452            25            24            25            24
BM_sinsp_concatenate_paths_empty_path_median                   -0.0437         -0.0437            25            24            25            24
BM_sinsp_concatenate_paths_empty_path_stddev                   -0.6335         -0.6332             0             0             0             0
BM_sinsp_concatenate_paths_empty_path_cv                       -0.6162         -0.6158             0             0             0             0
BM_sinsp_concatenate_paths_absolute_path_mean                  -0.0114         -0.0114            64            63            64            63
BM_sinsp_concatenate_paths_absolute_path_median                -0.0136         -0.0136            64            63            64            63
BM_sinsp_concatenate_paths_absolute_path_stddev                +0.8003         +0.8003             0             1             0             1
BM_sinsp_concatenate_paths_absolute_path_cv                    +0.8210         +0.8210             0             0             0             0
BM_sinsp_split_container_image_mean                            -0.0052         -0.0051           390           388           390           388
BM_sinsp_split_container_image_median                          -0.0034         -0.0034           389           388           389           388
BM_sinsp_split_container_image_stddev                          -0.1510         -0.1514             3             3             3             3
BM_sinsp_split_container_image_cv                              -0.1466         -0.1470             0             0             0             0

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 75.09%. Comparing base (8362ae9) to head (8e15871).
Report is 5 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #2233   +/-   ##
=======================================
  Coverage   75.09%   75.09%           
=======================================
  Files         276      276           
  Lines       34391    34391           
  Branches     5927     5927           
=======================================
  Hits        25826    25826           
  Misses       8565     8565           

Flag	Coverage Δ
libsinsp	75.09% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[github-actions]

X64 kernel testing matrix

KERNEL	CMAKE-CONFIGURE	KMOD BUILD	KMOD SCAP-OPEN	BPF-PROBE BUILD	BPF-PROBE SCAP-OPEN	MODERN-BPF SCAP-OPEN
amazonlinux2-4.19	🟢	🟢	🟢	🟢	🟢	🟡
amazonlinux2-5.10	🟢	🟢	🟢	🟢	🟢	🟢
amazonlinux2-5.15	🟢	🟢	🟢	🟢	🟢	🟢
amazonlinux2-5.4	🟢	🟢	🟢	🟢	🟢	🟡
amazonlinux2022-5.15	🟢	🟢	🟢	🟢	🟢	🟢
amazonlinux2023-6.1	🟢	🟢	🟢	🟢	🟢	🟢
archlinux-6.0	🟢	🟢	🟢	🟢	🟢	🟢
archlinux-6.7	🟢	🟢	🟢	🟢	🟢	🟢
centos-3.10	🟢	🟢	🟢	🟡	🟡	🟡
centos-4.18	🟢	🟢	🟢	🟢	🟢	🟢
centos-5.14	🟢	🟢	🟢	🟢	🟢	🟢
fedora-5.17	🟢	🟢	🟢	🟢	🟢	🟢
fedora-5.8	🟢	🟢	🟢	🟢	🟢	🟢
fedora-6.2	🟢	🟢	🟢	🟢	🟢	🟢
oraclelinux-3.10	🟢	🟢	🟢	🟡	🟡	🟡
oraclelinux-4.14	🟢	🟢	🟢	🟢	🟢	🟡
oraclelinux-5.15	🟢	🟢	🟢	🟢	🟢	🟢
oraclelinux-5.4	🟢	🟢	🟢	🟢	🟢	🟡
ubuntu-4.15	🟢	🟢	🟢	🟢	🟢	🟡
ubuntu-5.8	🟢	🟢	🟢	🟢	🟢	🟡
ubuntu-6.5	🟢	🟢	🟢	🟢	🟢	🟢

ARM64 kernel testing matrix

KERNEL	CMAKE-CONFIGURE	KMOD BUILD	KMOD SCAP-OPEN	BPF-PROBE BUILD	BPF-PROBE SCAP-OPEN	MODERN-BPF SCAP-OPEN
amazonlinux2-5.4	🟢	🟢	🟢	🟢	🟢	🟡
amazonlinux2022-5.15	🟢	🟢	🟢	🟢	🟢	🟢
fedora-6.2	🟢	🟢	🟢	🟢	🟢	🟢
oraclelinux-4.14	🟢	🟢	🟢	🟡	🟡	🟡
oraclelinux-5.15	🟢	🟢	🟢	🟢	🟢	🟢
ubuntu-6.5	🟢	🟢	🟢	🟢	🟢	🟢

[FedeDP]

CI Build / run-e2e-tests-amd64 (bundled_deps) (pull_request) Failing after 21m

Need to understand what broke e2e-tests.

[Andreagit97]

The approach seems great! thank you! if we want to avoid macro X we can just use a single ebpf program in tail call and save the state in a PER-CPU map/array so that each tailed called prog can see the iteration number. In the end, it should change almost nothing, it would be probably just easier to debug

[FedeDP]

If we create a per-cpu map, we could also avoid to share recvmmsg_data_t structure on each tail called program, avoiding unneeded extract__network_args calls.
I will look into that; for a first draft i wanted to avoid any additional map and see how it went.