forked from lei.chen/myxce
Rongsong Shen
530a02bda3
Converting COMMIT_MODE_WAIT away from a blocking wait removed the only mechanism that ever detected a commit timing out (the caller's own pthread_cond_timedwait deadline). Adds a dedicated per-commit-engine sweep thread that arms an equivalent timeout for async-registered continuations, resolving FAILED/-ETIMEDOUT within the same DEFAULT_SYNC_TIMEOUT_MS window the old blocking path used, with no thread parked to detect it. Chose a private sweep thread over the shared delaymgr: wiring delaymgr would touch cluster.c/ha.c (out of boundary) and its 200ms poll granularity vs the 500ms default timeout would add jitter, breaking the "identical failure timing" no-regression bar. Round 1 review rejected a lost-wakeup race: the sweep loop computed its next-deadline snapshot under lock, unlocked to dispatch, then slept on the stale snapshot -- a registration landing mid-dispatch had its broadcast lost (no thread parked on the condvar), so the sweep thread could sleep on an absent/late deadline indefinitely. Fixed by holding the lock continuously across scan-then-wait decisions and always re-scanning immediately before ever waiting (recheck-then-wait idiom). Round 2 review independently reproduced the round-1 bug shape against the fix to confirm it actually closes the race, not just narrows it.
815 lines
37 KiB
C
815 lines
37 KiB
C
/*
|
|
* test_commit_resolve.c — raft-apply-decoupling tasks 1.1 + 1.2.
|
|
*
|
|
* Keystone conversion, registration half (task 1.1): the commit engine gains
|
|
* commit_submit_async(), a non-blocking sibling of
|
|
* commit_submit(..., COMMIT_MODE_WAIT, ...) that registers a continuation
|
|
* (callback + context) against an appended LSN instead of parking the
|
|
* calling thread. The task-1.1 cases below cover ONLY the registration
|
|
* mechanism — a registered continuation is never invoked in those tests, by
|
|
* design (resolution did not exist yet).
|
|
*
|
|
* Asserts (task 1.1):
|
|
* (1) commit_submit_async returns immediately (0), before the LSN
|
|
* commits, and does NOT invoke the continuation inline.
|
|
* (2) Bad-argument rejection: -EINVAL for NULL engine, LSN 0, NULL fn.
|
|
* (3) The SAME admission-cap discipline commit_submit's WAIT mode uses
|
|
* (wait_admit_locked / wait_cap / inflight_waits) governs
|
|
* registration: cap exhaustion fast-rejects with -XCE_ERR_NOT_READY,
|
|
* never blocks.
|
|
* (4) That admission budget is SHARED with COMMIT_MODE_WAIT — a parked
|
|
* blocking waiter and a pending async registration draw from the
|
|
* same inflight_waits pool.
|
|
* (5) Distinct concurrent registrations occupy distinct per-LSN slots.
|
|
*
|
|
* Requirements (task 1.1): 1.3, 1.4, 9.1, 9.2
|
|
*
|
|
* Resolution half (task 1.2): commit_resolve() delivers a registered
|
|
* continuation's outcome exactly once, dispatched off the calling thread via
|
|
* vtask_enq_classified (VTASK_LIVE_BLOCK), never inline on the ACK/iomgr
|
|
* path in the real (pool-wired) case. Covers the four Observable triggers
|
|
* from tasks.md 1.2 -- {majority ack, timeout, demotion, truncation} -- plus
|
|
* idempotency, unregistered-lsn safety, admission-slot release, and the
|
|
* fail-loud inline fallback when no vtaskmgr pool is wired.
|
|
*
|
|
* On "timeout"/"demotion"/"truncation": commit_resolve() is a generic
|
|
* outcome+rc delivery primitive (design §Commit Engine Resolve) -- it does
|
|
* not itself detect any of these three conditions. Tracing the existing
|
|
* codebase (persist.c truncate path, ha.c raft_cb_become_follower demotion
|
|
* path, repl_wait_majority_ack's error vocabulary) found NO existing
|
|
* distinct rc for demotion or truncation anywhere today: commit_submit's
|
|
* ONLY negative "did not commit" outcome is -ETIMEDOUT, because the current
|
|
* blocking-WAIT model doesn't distinguish these causes either -- a
|
|
* demoted/truncated pending entry simply never satisfies its predicate and
|
|
* dies at the caller's own deadline. Wiring an ACTUAL demotion/truncation
|
|
* detector to call commit_resolve() belongs to later tasks that touch
|
|
* ha.c/persist.c/repl.c (out of this task's commit.c-only boundary; see
|
|
* task 3.2 in tasks.md). These tests therefore "force" each condition the
|
|
* way a future detector will: by calling commit_resolve(FAILED, rc)
|
|
* directly, using representative existing rc vocabulary (-ETIMEDOUT for
|
|
* timeout, -XCE_ERR_STANDBY -- "op is primary-only" -- for demotion, -ENOENT
|
|
* for a log-truncated-away entry) to prove commit_resolve delivers ANY
|
|
* caller-supplied outcome/rc faithfully, exactly once, off-thread.
|
|
*
|
|
* Requirements (task 1.2): 1.3, 9.3, 9.4
|
|
*
|
|
* Timeout-sweep half (task 1.3): commit_submit_async() no longer parks a
|
|
* thread that would otherwise notice its own pthread_cond_timedwait deadline
|
|
* expire, so a dedicated timeout-sweep thread (private to commit_engine_t,
|
|
* started in commit_engine_init / stopped+joined in commit_engine_fini) is
|
|
* the only remaining detector of "this registered lsn has been pending too
|
|
* long". The cases below prove: (a) a registered continuation for an lsn
|
|
* that never reaches majority and is NEVER manually resolved still resolves
|
|
* FAILED/-ETIMEDOUT on its own, near the existing DEFAULT_SYNC_TIMEOUT_MS
|
|
* default (Req 11.1 -- identical failure timing/vocabulary to today's
|
|
* blocking path); and (b) the majority-satisfied outcome always wins a race
|
|
* against an already-expired deadline, mirroring commit_submit's own
|
|
* last-look-before-timeout tie-break.
|
|
*
|
|
* Requirements (task 1.3): 1.3, 1.4, 9.3, 9.4
|
|
*/
|
|
#include <errno.h>
|
|
#include <pthread.h>
|
|
#include <stdatomic.h>
|
|
#include <stdint.h>
|
|
#include <string.h>
|
|
#include <time.h>
|
|
#include <unistd.h>
|
|
|
|
#include <unity/unity.h>
|
|
|
|
#include "xce/errnox.h"
|
|
|
|
#include "commit.h"
|
|
#include "repl.h"
|
|
#include "vtaskmgr.h"
|
|
|
|
/* ── Fixture ──────────────────────────────────────────────────────────────
|
|
* Static: persist_t embeds large path buffers, and statics survive Unity's
|
|
* longjmp on assertion failure (pattern from test_commit_engine.c /
|
|
* test_commit_admission_pool.c). */
|
|
static persist_t g_p;
|
|
static commit_engine_t g_ce;
|
|
static int g_eps[MAX_STANDBYS]; /* dummy endpoints — pointer-only */
|
|
|
|
/* Arrange a fresh engine over a stub persist_t + replicator_t, mirroring
|
|
* test_commit_engine.c's engine_setup_vcpus. n_vcpus drives the local
|
|
* admission cap: max(1, n_vcpus - COMMIT_WAIT_RESERVED). */
|
|
static void resolve_setup(int n_peers, int n_standbys, uint64_t appended_lsn,
|
|
int n_vcpus)
|
|
{
|
|
char name[16];
|
|
|
|
memset(&g_p, 0, sizeof(g_p));
|
|
TEST_ASSERT_EQUAL_INT(0, repl_init(&g_p.repl));
|
|
for (int i = 0; i < n_standbys; i++) {
|
|
snprintf(name, sizeof(name), "sb%d", i);
|
|
TEST_ASSERT_EQUAL_INT(
|
|
i, repl_add_standby(&g_p.repl, &g_eps[i], name, 7000 + (uint32_t)i));
|
|
}
|
|
g_p.current_lsn = appended_lsn + 1;
|
|
atomic_store(&g_p.committed_lsn, 0);
|
|
atomic_store(&g_p.durable_lsn, appended_lsn);
|
|
|
|
TEST_ASSERT_EQUAL_INT(0, commit_engine_init(&g_ce, &g_p, n_peers, n_vcpus));
|
|
}
|
|
|
|
static void resolve_teardown(void)
|
|
{
|
|
commit_engine_fini(&g_ce);
|
|
repl_fini(&g_p.repl);
|
|
}
|
|
|
|
/* ── Pool-wired fixture (task 1.2) ───────────────────────────────────────────
|
|
* A real, RUNNING vtaskmgr_t wired via commit_set_pool, mirroring
|
|
* test_commit_admission_pool.c's pool_setup but additionally calling
|
|
* vtaskmgr_start so enqueued vtasks are actually picked up and executed —
|
|
* required to observe commit_resolve's real (non-fallback) off-thread
|
|
* dispatch, not just the admission-accounting seam. */
|
|
static vtaskmgr_t g_mgr;
|
|
|
|
static void resolve_pool_setup(int n_peers, int n_standbys,
|
|
uint64_t appended_lsn, int n_vcpus)
|
|
{
|
|
resolve_setup(n_peers, n_standbys, appended_lsn, n_vcpus);
|
|
TEST_ASSERT_EQUAL_INT(0, vtaskmgr_init(&g_mgr, n_vcpus));
|
|
TEST_ASSERT_EQUAL_INT(0, vtaskmgr_start(&g_mgr));
|
|
commit_set_pool(&g_ce, &g_mgr);
|
|
}
|
|
|
|
static void resolve_pool_teardown(void)
|
|
{
|
|
commit_engine_fini(&g_ce);
|
|
vtaskmgr_fini(&g_mgr);
|
|
repl_fini(&g_p.repl);
|
|
}
|
|
|
|
static uint64_t mono_ms_r(void)
|
|
{
|
|
struct timespec ts;
|
|
|
|
clock_gettime(CLOCK_MONOTONIC, &ts);
|
|
return (uint64_t)ts.tv_sec * 1000u + (uint64_t)ts.tv_nsec / 1000000u;
|
|
}
|
|
|
|
/* Spin-poll (bounded) until *calls reaches want; avoids correctness-by-sleep
|
|
* on the assertion path (pattern from test_vtask_admission.c wait_atomic_ge /
|
|
* test_commit_engine.c poll_inflight). */
|
|
static int poll_calls(_Atomic int* calls, int want, uint32_t deadline_ms)
|
|
{
|
|
uint64_t t0 = mono_ms_r();
|
|
int v;
|
|
|
|
for (;;) {
|
|
v = atomic_load(calls);
|
|
if (v >= want || mono_ms_r() - t0 >= deadline_ms) {
|
|
return v;
|
|
}
|
|
usleep(500);
|
|
}
|
|
}
|
|
|
|
/* Continuation recorder — proves (non-)invocation and captures the args a
|
|
* commit_resolve() delivers, including the thread it ran on (task 1.2: proves
|
|
* dispatch is off the calling thread, never inline on the real pool-wired
|
|
* path). */
|
|
typedef struct {
|
|
_Atomic int calls;
|
|
uint64_t last_lsn;
|
|
commit_outcome_t last_outcome;
|
|
int last_rc;
|
|
void* last_ctx;
|
|
pthread_t last_thread;
|
|
/* raft-apply-decoupling task 1.3 lost-wakeup regression: when nonzero,
|
|
* recorder_fn sleeps this long BEFORE recording. Zero (the default
|
|
* zero-initializer every existing test uses) preserves today's
|
|
* immediate-recording behavior. */
|
|
unsigned sleep_us;
|
|
} continuation_recorder_t;
|
|
|
|
static void recorder_fn(void* ctx, uint64_t lsn, commit_outcome_t outcome,
|
|
int rc)
|
|
{
|
|
continuation_recorder_t* r = (continuation_recorder_t*)ctx;
|
|
|
|
if (r->sleep_us > 0) {
|
|
usleep(r->sleep_us);
|
|
}
|
|
r->last_lsn = lsn;
|
|
r->last_outcome = outcome;
|
|
r->last_rc = rc;
|
|
r->last_ctx = ctx;
|
|
r->last_thread = pthread_self();
|
|
atomic_fetch_add(&r->calls, 1);
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (1) Registration returns immediately, before the LSN commits, and *
|
|
* never invokes the continuation inline. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_submit_async_returns_immediately_before_commit(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
|
|
/* n_peers=3, no standby ACKs delivered: LSN 1 can never commit on its
|
|
* own (majority unreachable), so a blocking WAIT would park here. */
|
|
resolve_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/8);
|
|
|
|
int rc = commit_submit_async(&g_ce, /*lsn=*/1, recorder_fn, &rec);
|
|
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
0, rc, "registration must succeed and return without blocking");
|
|
TEST_ASSERT_EQUAL_UINT64_MESSAGE(
|
|
0, atomic_load(&g_p.committed_lsn),
|
|
"the LSN must still be uncommitted at the point of registration");
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
0, atomic_load(&rec.calls),
|
|
"the continuation must NOT be invoked by registration alone");
|
|
|
|
resolve_teardown();
|
|
printf("test_commit_submit_async_returns_immediately_before_commit: "
|
|
"PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (2) Bad-argument rejection. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_submit_async_rejects_invalid_args(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
|
|
resolve_setup(/*n_peers=*/1, /*n_standbys=*/0, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/8);
|
|
|
|
TEST_ASSERT_EQUAL_INT(-EINVAL,
|
|
commit_submit_async(NULL, 1, recorder_fn, &rec));
|
|
TEST_ASSERT_EQUAL_INT(
|
|
-EINVAL, commit_submit_async(&g_ce, /*lsn=*/0, recorder_fn, &rec));
|
|
TEST_ASSERT_EQUAL_INT(-EINVAL,
|
|
commit_submit_async(&g_ce, /*lsn=*/1, NULL, &rec));
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
0, atomic_load(&rec.calls),
|
|
"no rejected registration may invoke the continuation");
|
|
|
|
resolve_teardown();
|
|
printf("test_commit_submit_async_rejects_invalid_args: PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (3) Admission cap governs registration exactly as it governs WAIT: *
|
|
* n_vcpus=2, COMMIT_WAIT_RESERVED=2 -> local cap = max(1,0) = 1. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_submit_async_respects_admission_cap(void)
|
|
{
|
|
continuation_recorder_t rec1 = {0}, rec2 = {0};
|
|
|
|
resolve_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/2,
|
|
/*n_vcpus=*/2);
|
|
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec1));
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(1, g_ce.inflight_waits,
|
|
"registration consumes an admission slot");
|
|
|
|
int rc = commit_submit_async(&g_ce, 2, recorder_fn, &rec2);
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
-XCE_ERR_NOT_READY, rc,
|
|
"over-cap registration must fast-reject, never block");
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, g_ce.inflight_waits,
|
|
"a rejected registration must not consume a slot");
|
|
TEST_ASSERT_EQUAL_INT(0, atomic_load(&rec1.calls));
|
|
TEST_ASSERT_EQUAL_INT(0, atomic_load(&rec2.calls));
|
|
|
|
resolve_teardown();
|
|
printf("test_commit_submit_async_respects_admission_cap: PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (4) The admission budget is SHARED with COMMIT_MODE_WAIT: a parked *
|
|
* blocking waiter occupies the same pool an async registration *
|
|
* draws from, and releasing it frees the slot for either caller. *
|
|
* ------------------------------------------------------------------ */
|
|
typedef struct {
|
|
uint64_t lsn;
|
|
uint32_t timeout_ms;
|
|
int rc;
|
|
} waiter_arg_t;
|
|
|
|
static void* waiter_thread(void* a)
|
|
{
|
|
waiter_arg_t* w = (waiter_arg_t*)a;
|
|
w->rc = commit_submit(&g_ce, w->lsn, COMMIT_MODE_WAIT,
|
|
w->timeout_ms);
|
|
return NULL;
|
|
}
|
|
|
|
static uint64_t mono_ms(void)
|
|
{
|
|
struct timespec ts;
|
|
clock_gettime(CLOCK_MONOTONIC, &ts);
|
|
return (uint64_t)ts.tv_sec * 1000 + ts.tv_nsec / 1000000;
|
|
}
|
|
|
|
/* Poll ce->inflight_waits until it reaches `want` or the deadline elapses;
|
|
* returns the last observed value (pattern from test_commit_admission_pool.c
|
|
* poll_inflight — avoids a fixed sleep race against the background thread's
|
|
* admission acquire). */
|
|
static int poll_inflight(int want, uint32_t deadline_ms)
|
|
{
|
|
uint64_t t0 = mono_ms();
|
|
int v;
|
|
|
|
for (;;) {
|
|
pthread_mutex_lock(&g_ce.lock);
|
|
v = g_ce.inflight_waits;
|
|
pthread_mutex_unlock(&g_ce.lock);
|
|
if (v >= want || mono_ms() - t0 >= deadline_ms) {
|
|
return v;
|
|
}
|
|
usleep(500);
|
|
}
|
|
}
|
|
|
|
static void test_commit_submit_async_shares_admission_budget_with_wait(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
pthread_t tid;
|
|
waiter_arg_t w = {.lsn = 1, .timeout_ms = 4000, .rc = -9999};
|
|
|
|
/* n_vcpus=2 -> local cap = 1 (same sizing as test 3 above). */
|
|
resolve_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/2,
|
|
/*n_vcpus=*/2);
|
|
|
|
TEST_ASSERT_EQUAL_INT(0, pthread_create(&tid, NULL, waiter_thread, &w));
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, poll_inflight(1, 3000),
|
|
"the blocking WAIT must occupy the sole admission slot");
|
|
|
|
int rc = commit_submit_async(&g_ce, /*lsn=*/2, recorder_fn, &rec);
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
-XCE_ERR_NOT_READY, rc,
|
|
"async registration must fast-reject while a blocking WAIT holds "
|
|
"the shared admission slot");
|
|
|
|
/* Release the blocking waiter via a covering majority ACK (self + 1
|
|
* standby = 2 > 3/2), freeing the slot for the pending registration. */
|
|
repl_on_ack(&g_p.repl, &g_eps[0], 1);
|
|
commit_on_ack(&g_ce, 0, 1);
|
|
TEST_ASSERT_EQUAL_INT(0, pthread_join(tid, NULL));
|
|
TEST_ASSERT_EQUAL_INT(0, w.rc);
|
|
|
|
rc = commit_submit_async(&g_ce, /*lsn=*/2, recorder_fn, &rec);
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
0, rc, "registration must succeed once the shared slot is freed");
|
|
|
|
resolve_teardown();
|
|
printf("test_commit_submit_async_shares_admission_budget_with_wait: "
|
|
"PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (5) Distinct concurrent registrations occupy distinct per-LSN slots. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_submit_async_tracks_distinct_lsns(void)
|
|
{
|
|
continuation_recorder_t rec1 = {0}, rec2 = {0};
|
|
|
|
resolve_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/2,
|
|
/*n_vcpus=*/8);
|
|
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec1));
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 2, recorder_fn, &rec2));
|
|
TEST_ASSERT_EQUAL_INT(2, g_ce.inflight_waits);
|
|
|
|
/* White-box: the two registrations must land in distinct slots, each
|
|
* carrying its own lsn/fn/ctx (the per-LSN bookkeeping the task
|
|
* requires, not a single-shot overwrite). */
|
|
int found_lsn1 = 0, found_lsn2 = 0;
|
|
for (int i = 0; i < g_ce.n_continuations; i++) {
|
|
if (g_ce.continuations[i].lsn == 1) {
|
|
found_lsn1 = 1;
|
|
TEST_ASSERT_EQUAL_PTR(recorder_fn, g_ce.continuations[i].fn);
|
|
TEST_ASSERT_EQUAL_PTR(&rec1, g_ce.continuations[i].ctx);
|
|
} else if (g_ce.continuations[i].lsn == 2) {
|
|
found_lsn2 = 1;
|
|
TEST_ASSERT_EQUAL_PTR(recorder_fn, g_ce.continuations[i].fn);
|
|
TEST_ASSERT_EQUAL_PTR(&rec2, g_ce.continuations[i].ctx);
|
|
}
|
|
}
|
|
TEST_ASSERT_TRUE_MESSAGE(found_lsn1, "lsn=1 registration must be tracked");
|
|
TEST_ASSERT_TRUE_MESSAGE(found_lsn2, "lsn=2 registration must be tracked");
|
|
|
|
resolve_teardown();
|
|
printf("test_commit_submit_async_tracks_distinct_lsns: PASS\n");
|
|
}
|
|
|
|
/* ════════════════════════════════════════════════════════════════════════
|
|
* Task 1.2 — commit_resolve(): majority-commit and failure-path resolution.
|
|
* ════════════════════════════════════════════════════════════════════════ */
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (6) Majority ack: a registered continuation is resolved COMMITTED, *
|
|
* exactly once, dispatched OFF the ACK-driving thread (real, *
|
|
* pool-wired path — never inline). *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_resolve_majority_ack_dispatches_committed_off_thread(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
pthread_t self = pthread_self();
|
|
|
|
/* 3-member cluster, entry 1 appended+durable, no ACKs yet. */
|
|
resolve_pool_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec));
|
|
TEST_ASSERT_EQUAL_INT(1, g_ce.inflight_waits);
|
|
|
|
/* Majority ACK arrives on THIS thread (mirrors commit_on_ack running on
|
|
* the iomgr thread in production). */
|
|
repl_on_ack(&g_p.repl, &g_eps[0], 1);
|
|
commit_on_ack(&g_ce, 0, 1);
|
|
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, poll_calls(&rec.calls, 1, 2000),
|
|
"commit_resolve must deliver COMMITTED once the majority ack lands");
|
|
TEST_ASSERT_EQUAL_UINT64(1, rec.last_lsn);
|
|
TEST_ASSERT_EQUAL_INT(COMMIT_OUTCOME_COMMITTED, rec.last_outcome);
|
|
TEST_ASSERT_EQUAL_INT(0, rec.last_rc);
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, atomic_load(&rec.calls),
|
|
"the continuation must be invoked exactly once");
|
|
TEST_ASSERT_FALSE_MESSAGE(
|
|
pthread_equal(self, rec.last_thread),
|
|
"the continuation must run off the calling (ACK-driving) thread, "
|
|
"never inline");
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
0, g_ce.inflight_waits,
|
|
"resolve must release the continuation's admission slot");
|
|
|
|
resolve_pool_teardown();
|
|
printf("test_commit_resolve_majority_ack_dispatches_committed_off_thread: "
|
|
"PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (7) Failure outcomes: commit_resolve delivers ANY caller-supplied *
|
|
* outcome/rc faithfully, off-thread, for the three failure causes *
|
|
* tasks.md 1.2 names — timeout, demotion, truncation. commit.c has *
|
|
* no existing distinct detector for demotion/truncation (see file *
|
|
* header); these calls simulate a future detector invoking the *
|
|
* generic resolve primitive directly. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_resolve_delivers_failed_outcome_for_timeout(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
pthread_t self = pthread_self();
|
|
|
|
resolve_pool_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec));
|
|
|
|
commit_resolve(&g_ce, 1, COMMIT_OUTCOME_FAILED, -ETIMEDOUT);
|
|
|
|
TEST_ASSERT_EQUAL_INT(1, poll_calls(&rec.calls, 1, 2000));
|
|
TEST_ASSERT_EQUAL_UINT64(1, rec.last_lsn);
|
|
TEST_ASSERT_EQUAL_INT(COMMIT_OUTCOME_FAILED, rec.last_outcome);
|
|
TEST_ASSERT_EQUAL_INT(-ETIMEDOUT, rec.last_rc);
|
|
TEST_ASSERT_EQUAL_INT(1, atomic_load(&rec.calls));
|
|
TEST_ASSERT_FALSE(pthread_equal(self, rec.last_thread));
|
|
TEST_ASSERT_EQUAL_INT(0, g_ce.inflight_waits);
|
|
|
|
resolve_pool_teardown();
|
|
printf("test_commit_resolve_delivers_failed_outcome_for_timeout: PASS\n");
|
|
}
|
|
|
|
static void test_commit_resolve_delivers_failed_outcome_for_demotion(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
|
|
resolve_pool_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec));
|
|
|
|
/* -XCE_ERR_STANDBY ("op is primary-only") is the existing vocabulary code
|
|
* closest to "this controller is no longer primary" — a future
|
|
* ha.c demotion detector's natural choice of rc. */
|
|
commit_resolve(&g_ce, 1, COMMIT_OUTCOME_FAILED, -XCE_ERR_STANDBY);
|
|
|
|
TEST_ASSERT_EQUAL_INT(1, poll_calls(&rec.calls, 1, 2000));
|
|
TEST_ASSERT_EQUAL_INT(COMMIT_OUTCOME_FAILED, rec.last_outcome);
|
|
TEST_ASSERT_EQUAL_INT(-XCE_ERR_STANDBY, rec.last_rc);
|
|
TEST_ASSERT_EQUAL_INT(1, atomic_load(&rec.calls));
|
|
|
|
resolve_pool_teardown();
|
|
printf("test_commit_resolve_delivers_failed_outcome_for_demotion: PASS\n");
|
|
}
|
|
|
|
static void test_commit_resolve_delivers_failed_outcome_for_truncation(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
|
|
resolve_pool_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec));
|
|
|
|
/* -ENOENT ("no such entry") represents a pending lsn whose log tail was
|
|
* truncated out from under it — a future persist.c truncation detector's
|
|
* natural choice of rc. */
|
|
commit_resolve(&g_ce, 1, COMMIT_OUTCOME_FAILED, -ENOENT);
|
|
|
|
TEST_ASSERT_EQUAL_INT(1, poll_calls(&rec.calls, 1, 2000));
|
|
TEST_ASSERT_EQUAL_INT(COMMIT_OUTCOME_FAILED, rec.last_outcome);
|
|
TEST_ASSERT_EQUAL_INT(-ENOENT, rec.last_rc);
|
|
TEST_ASSERT_EQUAL_INT(1, atomic_load(&rec.calls));
|
|
|
|
resolve_pool_teardown();
|
|
printf("test_commit_resolve_delivers_failed_outcome_for_truncation: PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (8) Idempotency: a second resolve() for an already-resolved lsn is a *
|
|
* no-op — the FIRST outcome/rc stands, no second invocation. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_resolve_second_resolve_is_noop(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
|
|
resolve_pool_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec));
|
|
|
|
commit_resolve(&g_ce, 1, COMMIT_OUTCOME_FAILED, -ETIMEDOUT);
|
|
TEST_ASSERT_EQUAL_INT(1, poll_calls(&rec.calls, 1, 2000));
|
|
|
|
/* A second resolve for the SAME lsn, with a DIFFERENT outcome/rc, must
|
|
* be a silent no-op: no second invocation, and the recorded outcome/rc
|
|
* from the first (and only) delivery must stand unchanged. */
|
|
commit_resolve(&g_ce, 1, COMMIT_OUTCOME_COMMITTED, 0);
|
|
usleep(50 * 1000); /* give a wrongly-dispatched second call a chance to land */
|
|
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, atomic_load(&rec.calls),
|
|
"a second resolve for an already-resolved lsn must not re-invoke");
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
COMMIT_OUTCOME_FAILED, rec.last_outcome,
|
|
"the first delivery's outcome must stand, not be overwritten");
|
|
TEST_ASSERT_EQUAL_INT(-ETIMEDOUT, rec.last_rc);
|
|
|
|
resolve_pool_teardown();
|
|
printf("test_commit_resolve_second_resolve_is_noop: PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (9) Resolving an lsn with no registered continuation is a safe *
|
|
* no-op — never crashes, never invokes anything. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_resolve_unregistered_lsn_is_safe_noop(void)
|
|
{
|
|
resolve_pool_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
|
|
commit_resolve(&g_ce, /*lsn=*/1, COMMIT_OUTCOME_COMMITTED, 0);
|
|
commit_resolve(NULL, 1, COMMIT_OUTCOME_FAILED, -ETIMEDOUT);
|
|
commit_resolve(&g_ce, /*lsn=*/0, COMMIT_OUTCOME_FAILED, -ETIMEDOUT);
|
|
|
|
/* No crash is the primary assertion; give any errant dispatch a moment
|
|
* then confirm nothing was admitted/released unexpectedly. */
|
|
usleep(50 * 1000);
|
|
TEST_ASSERT_EQUAL_INT(0, g_ce.inflight_waits);
|
|
|
|
resolve_pool_teardown();
|
|
printf("test_commit_resolve_unregistered_lsn_is_safe_noop: PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (10) Fail-loud inline fallback: with no vtaskmgr pool wired (the *
|
|
* "commit-only unit fixtures" case, commit.h field comment), the *
|
|
* continuation must STILL be resolved — never silently dropped — *
|
|
* via the documented inline fallback. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_resolve_falls_back_to_inline_without_pool(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
|
|
/* Plain resolve_setup: no commit_set_pool call, so g_ce.pool == NULL. */
|
|
resolve_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
TEST_ASSERT_NULL(g_ce.pool);
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec));
|
|
|
|
commit_resolve(&g_ce, 1, COMMIT_OUTCOME_FAILED, -ETIMEDOUT);
|
|
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, atomic_load(&rec.calls),
|
|
"without a pool the continuation must still resolve, via the "
|
|
"fail-loud inline fallback (never silently dropped)");
|
|
TEST_ASSERT_EQUAL_INT(COMMIT_OUTCOME_FAILED, rec.last_outcome);
|
|
TEST_ASSERT_EQUAL_INT(-ETIMEDOUT, rec.last_rc);
|
|
TEST_ASSERT_EQUAL_INT(0, g_ce.inflight_waits);
|
|
|
|
resolve_teardown();
|
|
printf("test_commit_resolve_falls_back_to_inline_without_pool: PASS\n");
|
|
}
|
|
|
|
/* ════════════════════════════════════════════════════════════════════════
|
|
* Task 1.3 — timeout-sweep: async continuations resolve FAILED/-ETIMEDOUT
|
|
* on their own, with no manual commit_resolve() call.
|
|
* ════════════════════════════════════════════════════════════════════════ */
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (11) A registered continuation for an lsn that never reaches *
|
|
* majority resolves FAILED/-ETIMEDOUT on its own, near the *
|
|
* existing DEFAULT_SYNC_TIMEOUT_MS default -- no manual *
|
|
* commit_resolve() call, no thread parked to detect it. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_submit_async_resolves_failed_on_timeout(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
uint64_t t0;
|
|
uint64_t elapsed;
|
|
|
|
/* n_peers=3, no standby ACKs ever delivered: lsn=1 can never commit on
|
|
* its own. Nothing in this test ever calls commit_resolve() -- the ONLY
|
|
* thing that can resolve this continuation is the timeout-sweep
|
|
* mechanism built by this task. */
|
|
resolve_pool_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
|
|
t0 = mono_ms_r();
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec));
|
|
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, poll_calls(&rec.calls, 1, DEFAULT_SYNC_TIMEOUT_MS + 1500),
|
|
"a registered continuation that never reaches majority must resolve "
|
|
"FAILED/-ETIMEDOUT on its own within the configured timeout window");
|
|
elapsed = mono_ms_r() - t0;
|
|
|
|
TEST_ASSERT_EQUAL_UINT64(1, rec.last_lsn);
|
|
TEST_ASSERT_EQUAL_INT(COMMIT_OUTCOME_FAILED, rec.last_outcome);
|
|
TEST_ASSERT_EQUAL_INT(-ETIMEDOUT, rec.last_rc);
|
|
TEST_ASSERT_EQUAL_INT(1, atomic_load(&rec.calls));
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
0, g_ce.inflight_waits,
|
|
"the timeout resolve must release the continuation's admission slot");
|
|
|
|
/* Req 11.1 -- identical failure timing to today's blocking path:
|
|
* generous bounds absorb scheduler jitter and the sweep thread's own
|
|
* bounded scan overhead without masking a mistimed implementation. */
|
|
TEST_ASSERT_TRUE_MESSAGE(
|
|
elapsed + 50 >= DEFAULT_SYNC_TIMEOUT_MS,
|
|
"must not resolve meaningfully before the configured default "
|
|
"timeout");
|
|
TEST_ASSERT_TRUE_MESSAGE(
|
|
elapsed < DEFAULT_SYNC_TIMEOUT_MS + 1000,
|
|
"must resolve within the configured timeout window, not drastically "
|
|
"late");
|
|
|
|
resolve_pool_teardown();
|
|
printf("test_commit_submit_async_resolves_failed_on_timeout: PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (12) Race: majority-satisfied always wins over an already-expired *
|
|
* deadline observed in the SAME sweep pass -- mirrors *
|
|
* commit_submit's own last-look-before-timeout tie-break. White- *
|
|
* box: force the deadline into the past and satisfy the majority *
|
|
* predicate directly (bypassing commit_on_ack / the already- *
|
|
* tested commit_resolve_pending ack path) so ONLY the timeout- *
|
|
* sweep thread's own scan discovers both conditions at once. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_submit_async_majority_wins_race_at_expired_deadline(void)
|
|
{
|
|
continuation_recorder_t rec = {0};
|
|
int found = 0;
|
|
|
|
resolve_pool_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/4);
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &rec));
|
|
|
|
pthread_mutex_lock(&g_ce.lock);
|
|
for (int i = 0; i < g_ce.n_continuations; i++) {
|
|
if (g_ce.continuations[i].lsn == 1) {
|
|
g_ce.continuations[i].deadline.tv_sec = 0;
|
|
g_ce.continuations[i].deadline.tv_nsec = 0;
|
|
found = 1;
|
|
}
|
|
}
|
|
/* Satisfy wait_done_locked directly -- no ack path involved. */
|
|
atomic_store(&g_p.committed_lsn, 1);
|
|
pthread_cond_broadcast(&g_ce.cond); /* wake the sweep thread promptly */
|
|
pthread_mutex_unlock(&g_ce.lock);
|
|
TEST_ASSERT_TRUE_MESSAGE(found, "lsn=1 registration must be tracked");
|
|
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, poll_calls(&rec.calls, 1, 2000),
|
|
"the sweep thread must independently notice and resolve this lsn");
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
COMMIT_OUTCOME_COMMITTED, rec.last_outcome,
|
|
"majority-satisfied must win over an already-expired deadline "
|
|
"observed in the same scan");
|
|
TEST_ASSERT_EQUAL_INT(0, rec.last_rc);
|
|
TEST_ASSERT_EQUAL_INT(1, atomic_load(&rec.calls));
|
|
|
|
resolve_pool_teardown();
|
|
printf("test_commit_submit_async_majority_wins_race_at_expired_deadline: "
|
|
"PASS\n");
|
|
}
|
|
|
|
/* ------------------------------------------------------------------ *
|
|
* (13) Lost-wakeup regression: a registration landing WHILE the sweep *
|
|
* thread is mid-dispatch for a previous continuation must not be *
|
|
* missed. commit_sweep_scan_locked()'s classification runs under *
|
|
* ce->lock, but the actual commit_resolve() dispatch for whatever *
|
|
* it found must run UNLOCKED (lock-ordering rule) -- a *
|
|
* commit_submit_async() broadcast landing in that unlocked window *
|
|
* finds no thread parked in cond_wait to receive it (POSIX *
|
|
* condvars do not latch a signal sent to no waiter). If the sweep *
|
|
* worker then trusted the have_next/next_deadline it computed *
|
|
* BEFORE that dispatch, it would commit to an unbounded *
|
|
* pthread_cond_wait() (or a stale deadline) and never notice the *
|
|
* new registration until some UNRELATED future broadcast -- *
|
|
* unbounded delay instead of DEFAULT_SYNC_TIMEOUT_MS. *
|
|
* *
|
|
* Deterministic reproduction: use the no-pool inline-dispatch *
|
|
* fallback, whose continuation function runs SYNCHRONOUSLY on the *
|
|
* thread that called commit_resolve() -- the sweep thread itself, *
|
|
* for a timeout-triggered resolve. Continuation A's recorder *
|
|
* deliberately sleeps, pinning the sweep thread outside ce->lock *
|
|
* for a wide, fully-controlled window (this IS the dispatch *
|
|
* window the bug lived in, just stretched from microseconds to *
|
|
* ~200ms so a concurrent registration can land inside it with *
|
|
* certainty instead of chasing a race). Continuation B is *
|
|
* registered mid-sleep and must still resolve within *
|
|
* DEFAULT_SYNC_TIMEOUT_MS of ITS OWN registration, not stall on *
|
|
* whatever unrelated broadcast happens to come next. *
|
|
* ------------------------------------------------------------------ */
|
|
static void test_commit_submit_async_registration_during_dispatch_not_lost(void)
|
|
{
|
|
continuation_recorder_t recA = {0};
|
|
continuation_recorder_t recB = {0};
|
|
|
|
/* Plain (no-pool) setup: dispatch_continuation's fail-loud inline
|
|
* fallback invokes the continuation function directly, unlocked, on
|
|
* the calling thread -- the sweep thread, here. */
|
|
resolve_setup(/*n_peers=*/3, /*n_standbys=*/2, /*appended_lsn=*/1,
|
|
/*n_vcpus=*/8);
|
|
TEST_ASSERT_NULL(g_ce.pool);
|
|
|
|
recA.sleep_us = 200 * 1000; /* 200ms: sweep thread is unlocked, "busy
|
|
* dispatching" A for this long */
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 1, recorder_fn, &recA));
|
|
|
|
/* Let A's DEFAULT_SYNC_TIMEOUT_MS deadline pass and the sweep thread
|
|
* begin dispatching it (enter recorder_fn's sleep). Comfortably past
|
|
* the deadline, comfortably before the 200ms sleep would elapse on its
|
|
* own. */
|
|
usleep((DEFAULT_SYNC_TIMEOUT_MS + 100) * 1000);
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
0, atomic_load(&recA.calls),
|
|
"sanity: A must still be inside its 200ms sleep, not yet recorded, "
|
|
"when B is registered below -- otherwise this test isn't actually "
|
|
"landing inside the dispatch window");
|
|
|
|
/* Register B while the sweep thread is (with certainty, given the
|
|
* 200ms window) unlocked and blocked inside recorder_fn(A)'s sleep --
|
|
* exactly the window a real registration could race in production,
|
|
* just widened for a deterministic test. */
|
|
TEST_ASSERT_EQUAL_INT(0, commit_submit_async(&g_ce, 2, recorder_fn, &recB));
|
|
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(1, poll_calls(&recA.calls, 1, 2000),
|
|
"continuation A must still resolve");
|
|
TEST_ASSERT_EQUAL_INT_MESSAGE(
|
|
1, poll_calls(&recB.calls, 1, DEFAULT_SYNC_TIMEOUT_MS + 2000),
|
|
"continuation B, registered while the sweep thread was mid-dispatch "
|
|
"for A, must still resolve within the configured timeout window -- "
|
|
"not delayed indefinitely waiting for some unrelated future "
|
|
"broadcast (the lost-wakeup regression)");
|
|
TEST_ASSERT_EQUAL_UINT64(2, recB.last_lsn);
|
|
TEST_ASSERT_EQUAL_INT(COMMIT_OUTCOME_FAILED, recB.last_outcome);
|
|
TEST_ASSERT_EQUAL_INT(-ETIMEDOUT, recB.last_rc);
|
|
TEST_ASSERT_EQUAL_INT(0, g_ce.inflight_waits);
|
|
|
|
resolve_teardown();
|
|
printf("test_commit_submit_async_registration_during_dispatch_not_lost: "
|
|
"PASS\n");
|
|
}
|
|
|
|
void test_commit_resolve(void)
|
|
{
|
|
RUN_TEST(test_commit_submit_async_returns_immediately_before_commit);
|
|
RUN_TEST(test_commit_submit_async_rejects_invalid_args);
|
|
RUN_TEST(test_commit_submit_async_respects_admission_cap);
|
|
RUN_TEST(test_commit_submit_async_shares_admission_budget_with_wait);
|
|
RUN_TEST(test_commit_submit_async_tracks_distinct_lsns);
|
|
RUN_TEST(test_commit_resolve_majority_ack_dispatches_committed_off_thread);
|
|
RUN_TEST(test_commit_resolve_delivers_failed_outcome_for_timeout);
|
|
RUN_TEST(test_commit_resolve_delivers_failed_outcome_for_demotion);
|
|
RUN_TEST(test_commit_resolve_delivers_failed_outcome_for_truncation);
|
|
RUN_TEST(test_commit_resolve_second_resolve_is_noop);
|
|
RUN_TEST(test_commit_resolve_unregistered_lsn_is_safe_noop);
|
|
RUN_TEST(test_commit_resolve_falls_back_to_inline_without_pool);
|
|
RUN_TEST(test_commit_submit_async_resolves_failed_on_timeout);
|
|
RUN_TEST(test_commit_submit_async_majority_wins_race_at_expired_deadline);
|
|
RUN_TEST(test_commit_submit_async_registration_during_dispatch_not_lost);
|
|
}
|