Pass non-zero mapID into UpdateStackDeltaPages to fix cleanups

[bobrik]

Previously it picked up zero from return values, which resulted in no deletes happening in practice, resulting in duplicate warnings and likely failure to profile for affected mappings.

[bobrik]

Logging before the bug was fixed:

$ sudo journalctl -u ebpf-profiler --since '2026-03-21 00:22:38' | grep 'c10a998c845402'
Mar 21 00:27:02 44m298 ebpf-profiler[1610714]: report executable: fileID = c10a998c8454026e44c511a78c0e5011, name = llvm-gsymutil, build id = 3b11b277b1178b0b8f7533469e3d71bb24e39492
Mar 21 00:27:02 44m298 ebpf-profiler[1610714]: time=2026-03-21T00:27:02.721Z level=INFO msg="async update: fileID = c10a998c8454026ec10a998c8454026e, outer = HashOfMaps(exe_id_to_9_stack_deltas)#42, inner = Array(exe_id_to_9_stack_deltas_inner)#984"
Mar 21 00:27:02 44m298 ebpf-profiler[1610714]: time=2026-03-21T00:27:02.721Z level=ERROR msg="loadDeltas: UpdateStackDeltaPages failed for fileID c10a998c8454026ec10a998c8454026e mapID 9, attempting rollback with ref.MapID=0 (bug: should be mapID=9): batch update: key too big for map: argument list too long"
Mar 21 00:28:33 44m298 ebpf-profiler[1610714]: time=2026-03-21T00:28:33.458Z level=ERROR msg="loadDeltas: UpdateStackDeltaPages failed for fileID c10a998c8454026ec10a998c8454026e mapID 9, attempting rollback with ref.MapID=0 (bug: should be mapID=9): batch update: key too big for map: argument list too long"
Mar 21 00:28:33 44m298 ebpf-profiler[1610714]: time=2026-03-21T00:28:33.458Z level=INFO msg="async update: fileID = c10a998c8454026ec10a998c8454026e, outer = HashOfMaps(exe_id_to_9_stack_deltas)#42, inner = Array(exe_id_to_9_stack_deltas_inner)#997"
Mar 21 00:28:33 44m298 ebpf-profiler[1610714]: time=2026-03-21T00:28:33.458Z level=ERROR msg="Outer map update failure: update: key already exists | fileID = c10a998c8454026ec10a998c8454026e, outer = HashOfMaps(exe_id_to_9_stack_deltas)#42, inner = Array(exe_id_to_9_stack_deltas_inner)#997"

I'm not exactly sure what's causing E2BIG here and whether that's also an issue.

[pid 1611513] bpf(BPF_MAP_UPDATE_BATCH, {batch={keys=0x27e65848a120, values=0x27e66090cfa0, count=2, map_fd=31, elem_flags=BPF_ANY, flags=0x1}} <unfinished ...>
[pid 1610983] <... tgkill resumed>)     = 0
[pid 1611513] <... bpf resumed> => {batch={count=0}}, 56) = -1 E2BIG (Argument list too long)

[bobrik]

I think this was surfaced by #1144, but I'm not exactly sure on that point. It's good to fix it regardless.

[fabled]

lgtm