Respect hpa_min_purge_interval_ms option

Currently, hugepages aware allocator backend works together with classic
one as a fallback for not yet supported allocations. When background
threads are enabled wake up time for classic interfere with hpa as there
were no checks inside hpa purging logic to check if we are not purging too
frequently. If background thread is running and `hpa_should_purge`
returns true, then we will purge, even if we purged less than
hpa_min_purge_interval_ms ago.
This commit is contained in:
Dmitry Ilvokhin 2024-05-31 06:35:48 -07:00 committed by Qi Wang
parent 90c627edb7
commit 91a6d230db
2 changed files with 48 additions and 14 deletions

View file

@ -378,6 +378,16 @@ static bool
hpa_try_purge(tsdn_t *tsdn, hpa_shard_t *shard) {
malloc_mutex_assert_owner(tsdn, &shard->mtx);
/*
* Make sure we respect purge interval setting and don't purge
* too frequently.
*/
uint64_t since_last_purge_ms = shard->central->hooks.ms_since(
&shard->last_purge);
if (since_last_purge_ms < shard->opts.min_purge_interval_ms) {
return false;
}
hpdata_t *to_purge = psset_pick_purge(&shard->psset);
if (to_purge == NULL) {
return false;

View file

@ -83,7 +83,36 @@ wait_until_thread_is_enabled(unsigned arena_id) {
}
static void
expect_purging(unsigned arena_ind, bool expect_deferred) {
expect_purging(unsigned arena_ind) {
size_t empty_ndirty = get_empty_ndirty(arena_ind);
expect_zu_eq(0, empty_ndirty, "Expected arena to start unused.");
void *ptrs[2];
ptrs[0] = mallocx(PAGE,
MALLOCX_TCACHE_NONE | MALLOCX_ARENA(arena_ind));
ptrs[1] = mallocx(PAGE,
MALLOCX_TCACHE_NONE | MALLOCX_ARENA(arena_ind));
empty_ndirty = get_empty_ndirty(arena_ind);
expect_zu_eq(0, empty_ndirty, "All pages should be active");
dallocx(ptrs[0], MALLOCX_TCACHE_NONE);
expect_true(empty_ndirty == 0 || empty_ndirty == 1,
"Unexpected extra dirty page count: %zu", empty_ndirty);
/*
* Wait for at least hpa_min_purge_interval_ms to trigger purge on next
* deallocation.
*/
sleep_for_background_thread_interval();
dallocx(ptrs[1], MALLOCX_TCACHE_NONE);
empty_ndirty = get_empty_ndirty(arena_ind);
expect_zu_eq(0, empty_ndirty, "There are should be no dirty pages");
}
static void
expect_deferred_purging(unsigned arena_ind) {
size_t empty_ndirty;
empty_ndirty = get_empty_ndirty(arena_ind);
@ -103,20 +132,15 @@ expect_purging(unsigned arena_ind, bool expect_deferred) {
expect_zu_eq(0, empty_ndirty, "All pages should be active");
dallocx(ptr, MALLOCX_TCACHE_NONE);
empty_ndirty = get_empty_ndirty(arena_ind);
if (expect_deferred) {
expect_true(empty_ndirty == 0 || empty_ndirty == 1 ||
opt_prof, "Unexpected extra dirty page count: %zu",
empty_ndirty);
} else {
assert_zu_eq(0, empty_ndirty,
"Saw dirty pages without deferred purging");
}
expect_true(empty_ndirty == 0 || empty_ndirty == 1 ||
opt_prof, "Unexpected extra dirty page count: %zu",
empty_ndirty);
if (empty_ndirty > 0) {
observed_dirty_page = true;
break;
}
}
expect_b_eq(expect_deferred, observed_dirty_page, "");
expect_true(observed_dirty_page, "");
/*
* Under high concurrency / heavy test load (e.g. using run_test.sh),
@ -125,7 +149,7 @@ expect_purging(unsigned arena_ind, bool expect_deferred) {
*/
unsigned retry = 0;
while ((empty_ndirty = get_empty_ndirty(arena_ind)) > 0 &&
expect_deferred && (retry++ < 100)) {
(retry++ < 100)) {
sleep_for_background_thread_interval();
}
@ -144,7 +168,7 @@ TEST_BEGIN(test_hpa_background_thread_purges) {
* Our .sh sets dirty mult to 0, so all dirty pages should get purged
* any time any thread frees.
*/
expect_purging(arena_ind, /* expect_deferred */ true);
expect_deferred_purging(arena_ind);
}
TEST_END
@ -158,11 +182,11 @@ TEST_BEGIN(test_hpa_background_thread_enable_disable) {
unsigned arena_ind = create_arena();
set_background_thread_enabled(false);
expect_purging(arena_ind, false);
expect_purging(arena_ind);
set_background_thread_enabled(true);
wait_until_thread_is_enabled(arena_ind);
expect_purging(arena_ind, true);
expect_deferred_purging(arena_ind);
}
TEST_END