Allow loading multiple specs per tracer

[rafaelroquetto]

This effectively allows a tracer to be composed of multiple independent translation units - in other words, instead of doing:

// mytracer.c

#include "file1.c"
#include "file2.c"

A tracer can now go:generate individual .c files independently, each yielding its own .o file, treated as independent translation units. Map sharing continue to work as usual due to OBI_PIN_INTERNAL.

But why?

Glad you asked! Having different translation units allow us to have multiple/different definitions for a single function. For example, sock msg programs operate on a struct sk_msg context, whilst sock skb programs operate on struct __skb_buf contexts. These have similar semantics, e.g.:

• you can pull data
• they both provide direct packet access (data and data_end pointers)
• likewise, they both have the concept of length/size

Being able to have multiple implementations allow writing common code. For instance:

//common_defs.h
// defines a common """interface"""
static __always_inline u32 ctx_len(void *ctx);
static __always_inline void ctx_pull_data(void *ctx, u32 len);
static __always_inline void *ctx_data(void *ctx);
static __always_inline void *ctx_data_end(void *ctx);

// sockmsg_defs.h
// sk_msg implementations

static __always_inline u32 ctx_len(void *ctx) {
    return ((struct sk_msg_md *)ctx)->size;
}

static __always_inline void ctx_pull_data(void *ctx, u32 len) {
    bpf_msg_pull_data(ctx, 0, len, 0);
}

static __always_inline void *ctx_data(void *ctx) {
    return ((struct sk_msg_md *)ctx)->data;
}

static __always_inline void *ctx_data_end(void *ctx) {
    return ((struct sk_msg_md *)ctx)->data_end;
}

// skb_defs.h
// skb implementations
static __always_inline u32 ctx_len(void *ctx) {
    return ((struct __sk_buff *)ctx)->len;
}

static __always_inline void ctx_pull_data(void *ctx, u32 len) {
    bpf_skb_pull_data(ctx, len);
}

static __always_inline void *ctx_data(void *ctx) {
    return ctx_skb_data((struct __sk_buff *)ctx);
}

static __always_inline void *ctx_data_end(void *ctx) {
    return ctx_skb_data_end((struct __sk_buff *)ctx);
}

// common logic code (context agnostic)
static __always_inline void init_http_request(void *ctx, struct socket_data *sk_data) {
    const u32 len = ctx_len(ctx);

    /* ... */
    const u32 nbytes = len > sizeof(info->buf) ? sizeof(info->buf) : len;

    ctx_pull_data(ctx, nbytes);

    const unsigned char *ptr = ctx_data(ctx);
    const unsigned char *e = ctx_data_end(ctx);

    /* ... */
}

The use case is exactly the one illustrated above: allowing programs featuring direct packet access to be generalised (WIP/different PR).

This PR changes tpinjector/sock_iter.c to be a separate translation unit as an example.

Checklist

• If this enhances / fixes / changes a core feature, I have updated the features documentation

[codecov]

Codecov Report

❌ Patch coverage is 43.50649% with 87 lines in your changes missing coverage. Please review.
✅ Project coverage is 43.77%. Comparing base (55f2804) to head (f506195).
⚠️ Report is 7 commits behind head on main.

Files with missing lines	Patch %	Lines
pkg/ebpf/tracer_linux.go	40.38%	26 Missing and 5 partials ⚠️
pkg/internal/ebpf/gpuevent/gpuevent.go	0.00%	13 Missing ⚠️
pkg/internal/ebpf/logenricher/logenricher.go	0.00%	13 Missing ⚠️
pkg/internal/ebpf/tpinjector/tpinjector.go	65.78%	11 Missing and 2 partials ⚠️
pkg/internal/ebpf/gotracer/gotracer.go	0.00%	12 Missing ⚠️
pkg/internal/ebpf/logger/logger.go	75.00%	2 Missing and 1 partial ⚠️
pkg/internal/ebpf/watcher/watcher.go	81.81%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1373      +/-   ##
==========================================
+ Coverage   43.75%   43.77%   +0.01%     
==========================================
  Files         310      310              
  Lines       33772    33816      +44     
==========================================
+ Hits        14777    14803      +26     
- Misses      18044    18058      +14     
- Partials      951      955       +4     

Flag	Coverage Δ
integration-test	21.60% <32.06%> (+0.01%)	⬆️
integration-test-arm	0.00% <0.00%> (ø)
integration-test-vm-x86_64-5.15.152	0.00% <0.00%> (ø)
integration-test-vm-x86_64-6.10.6	0.00% <0.00%> (ø)
k8s-integration-test	2.31% <0.00%> (+<0.01%)	⬆️
oats-test	0.00% <0.00%> (ø)
unittests	44.63% <19.08%> (-0.01%)	⬇️

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.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[mariomac]

🪄

[mmat11]

lgtm, nice!

[grcevski]

LGTM! I left some minor comments.