feat: AddRingBuf() using ring_buffer__add

[mighty1231]

libbpf provides ring_buffer__add to handle multiple ringbuf event callbacks

• single RingBuffer object now handles multiple slots
• implements AddRingBuf() for ring_buffer__add wrapper
• updates ringbuffers selftest to show how to handle it

Reference

• https://github.com/torvalds/linux/blob/eb6a9339efeb6f3d2b5c86fdf2382cdc293eca2c/tools/testing/selftests/bpf/progs/test_ringbuf_multi.c
• https://github.com/torvalds/linux/blob/eb6a9339efeb6f3d2b5c86fdf2382cdc293eca2c/tools/testing/selftests/bpf/prog_tests/ringbuf_multi.c#41

[CLAassistant]

All committers have signed the CLA.

[geyslan]

@mighty1231 tks for this PR. There's a formatting issue https://github.com/aquasecurity/libbpfgo/pull/430/checks. As it is solved I'll stash it for later review. Tks again.

[mighty1231]

@geyslan I ran make fmt-fix. Now make fmt-check returns no error

EDIT: with new amended commit, make lint-check also returns no error

[mighty1231]

Failed test map-getmapsbyname seems not related to my PR, because it seems to have no ringbuffer-related code, and it runs before the ringbuffers test.

[geyslan]

@mighty1231, I suspect that self-test/iter sometimes fails due to ping usage and internal network malfunction. Well, it seems like a flaky test.

[geyslan]

@mighty1231 I tested it and it works nice. That's awesome.

I would just ask you to make the selftest more robust, since you have touched the logic of the slot (index) inside RingBuffer that's used by rwArray which .

Would you mind to improve the the respective selftest to submit a greater volume of data through a greater number of slots (ringbuffers) not reusing the same data holder? I mean, try using more than one program to submit the data.

[geyslan]

Please use the same pattern of cgo_init_ring_buf regarding saving and restoring errno.

[mighty1231]

I changed it.

[mighty1231]

@mighty1231 I tested it and it works nice. That's awesome.

I would just ask you to make the selftest more robust, since you have touched the logic of the slot (index) inside RingBuffer that's used by rwArray which .

Would you mind to improve the the respective selftest to submit a greater volume of data through a greater number of slots (ringbuffers) not reusing the same data holder? I mean, try using more than one program to submit the data.

I am confused. Which of the following did you mean?

1. More tests to check thread safety of Module *AddRingBuf(), RingBuffer *Stop(), and RingBuffer *Close(), which modifies RingBuffer.slots
2. More tests with two or more struct *RingBuffer in golang
3. More tests wtih two or more NewModuleFromFile("main.bpf.o") call
4. More tests wtih two or more probes - other than kprobe__sys_mmap from selftest/ringbuffers/main.bpf.c
5. Make similar to ringbuf_multi test from linux repository - ringbuf_multi.c and test_ringbuf_multi.c

1 may be resolved with notifying the limitation for users. If it doesn't seem right, I would appreciate any feedback you may have.
2 is the motivation of this PR. I have got completely lost to use more than one RingBuffer, so I decided to use ring_buffer__add.
3, 4 - it's difficult to guess what can be improved for it
5 seems interesting. if you mean for it, I will take it

[geyslan]

5. Make similar to ringbuf_multi test from linux repository

It would be awesome to have a huge amount of tests like ringbuf_multi. Anyway, I think it would take longer.

I meant:

• increasing numberOfEvent*Received* to a considerable amount.
• split the recvLoop for select into the amount of ringbuf slots, raising one goroutine for each.

[mighty1231]

I don't know how to split the for-select loop, since I don't know golang much. And I didn't understand what is improved with new test.

[geyslan]

@NDStrahilevitz do you mind reviewing this?

[geyslan]

@mighty1231 I'm merging this. Thanks again for your contrib.