Files
zeekay d14b104790 feat(zap): v2 generic API, bufpool, transport allocator + UMA/RDMA cgo
Three orthogonal additions, one wire format.

v2 package — generic ad-hoc schema dispatch (~/work/lux/zap/v2):
- README + IMPOSSIBILITY + BENCH_RESULTS document the v1↔v2 split.
- builder/field/iter/list/pool/registry/schema/view + codegen package
  emit v1-equivalent fast paths from declarative .zap sources.
- _compile_fail_test/ + examples/ cover negative and positive cases.
- v2_test.go exercises the generic dispatch path.
- doc.go in v1 marks v1 deprecated for *new* schema authoring; existing
  v1 buffers stay parseable by both v1 and v2.

bufpool — pooled read slabs (bufpool.go + bufpool_test.go):
- Message gains optional *bufRef; Release() returns the slab to its
  pool when sourced from the pooled read path, GC-managed otherwise.
- ParseHeader returns (data, rootOffset) without allocating *Message,
  used by zapv2 to avoid two allocations per parse.
- bench_read_test.go + node_quic_stream_test.go cover both paths.

Transport allocator + UMA/RDMA cgo (transport/*):
- Transport.Caps()/Buffer interface: capability + slab-typed buffer
  contract. Bytes() vs DevicePtr() decoupled so UMA and GPUDirect
  can share the same call surface.
- DPDK PMD + GPUDirect RDMA + CUDA UMA real cgo backends gated behind
  build tags; stub paths remain default on darwin / non-Linux.
- transport_test.go covers Caps reporting + Buffer lifecycle on the
  default transport.

LLM.md + go.mod follow the additions.
2026-06-04 17:06:59 -07:00

75 lines
2.5 KiB
C

// Copyright (C) 2025-2026, Lux Industries Inc. All rights reserved.
// See the file LICENSE for licensing terms.
//
//go:build cgo && linux && dpdk
//
// DPDK PMD-backed transport, host-side glue. Linked into the luxfi/zap
// binary only when the cgo + linux + dpdk build tags are set; see
// dpdk_linux_real.go for the Go cgo bridge.
//
// We deliberately keep this file pure C and avoid heavy DPDK headers
// outside the dpdk build tag. The build tag pulls in <rte_eal.h> via
// the per-tag flag-only file in dpdk_pmd_eal.c.
#include "dpdk_pmd.h"
#include <dirent.h>
#include <errno.h>
#include <stdio.h>
#include <string.h>
#include <sys/stat.h>
// hugepages_configured returns 1 when /sys/kernel/mm/hugepages reports
// at least one configured size with nr_hugepages > 0. We don't insist
// on a specific size — DPDK's mempool driver picks at init.
static int hugepages_configured(void) {
DIR* d = opendir("/sys/kernel/mm/hugepages");
if (!d) return 0;
int ok = 0;
struct dirent* e;
while ((e = readdir(d)) != NULL) {
if (strncmp(e->d_name, "hugepages-", 10) != 0) continue;
char path[512];
snprintf(path, sizeof(path),
"/sys/kernel/mm/hugepages/%s/nr_hugepages", e->d_name);
FILE* f = fopen(path, "r");
if (!f) continue;
unsigned long n = 0;
if (fscanf(f, "%lu", &n) == 1 && n > 0) {
ok = 1;
}
fclose(f);
if (ok) break;
}
closedir(d);
return ok;
}
// luxzap_dpdk_init / shutdown are defined in dpdk_pmd_eal.c which is
// compiled only when the dpdk build tag is set. To keep this file
// always-compilable (so static analysis sees a definition) we provide
// weak fall-back stubs that return EOPNOTSUPP. The link order ensures
// the real symbols win when present.
int __attribute__((weak)) luxzap_dpdk_init(int argc, char** argv) {
(void)argc;
(void)argv;
return EOPNOTSUPP;
}
int __attribute__((weak)) luxzap_dpdk_shutdown(void) { return EOPNOTSUPP; }
int luxzap_dpdk_probe(void) {
if (!hugepages_configured()) return 1;
// We cannot tell at runtime whether librte_eal is present without
// either dlopen'ing it or attempting rte_eal_init. The build-tag
// separation in dpdk_pmd_eal.c handles that: if the dpdk tag is
// set, that file's strong symbols replace the weak stubs here, and
// a one-shot probe-init returns 0 or the rte_errno.
int rc = luxzap_dpdk_init(0, NULL);
if (rc == EOPNOTSUPP) return 2;
if (rc != 0) return 3;
// Successful probe-init — leave EAL up for the live transport.
return 0;
}