ISAR/libmdbx
3
0
Fork 0
mirror of https://github.com/isar/libmdbx.git synced 2025-01-02 01:04:13 +08:00
Code Issues Packages Projects Releases Wiki Activity

mdbx: доработка подготовки резерва перед обновлением GC при включенном BigFoot.

Browse Source
This commit is contained in:
Леонид Юрьев (Leonid Yuriev) 2022-11-18 15:41:15 +03:00
parent 3757eb72f7
commit 07674ada47
1 changed files with 125 additions and 63 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

188
src/core.c
View File

@ -6761,6 +6761,7 @@ static pgr_t page_alloc_slowpath(const MDBX_cursor *mc, const size_t num,
MDBX_cursor *const gc =
(MDBX_cursor *)((char *)env->me_txn0 + sizeof(MDBX_txn));
eASSERT(env, mc != gc && gc->mc_next == nullptr);
gc->mc_txn = txn;
gc->mc_flags = 0;
@ -7188,8 +7189,16 @@ done:
}
if (LOG_ENABLED(level))
debug_log(level, __func__, __LINE__,
"unable alloc %zu %s, flags 0x%x, errcode %d\n", num, what,
flags, ret.err);
"unable alloc %zu %s, alloc-flags 0x%x, err %d, txn-flags "
"0x%x, re-list-len %zu, loose-count %zu, gc: height %u, "
"branch %zu, leaf %zu, large %zu, entries %zu\n",
num, what, flags, ret.err, txn->mt_flags,
MDBX_PNL_GETSIZE(txn->tw.relist), txn->tw.loose_count,
txn->mt_dbs[FREE_DBI].md_depth,
(size_t)txn->mt_dbs[FREE_DBI].md_branch_pages,
(size_t)txn->mt_dbs[FREE_DBI].md_leaf_pages,
(size_t)txn->mt_dbs[FREE_DBI].md_overflow_pages,
(size_t)txn->mt_dbs[FREE_DBI].md_entries);
ret.page = NULL;
}
} else {
@ -9508,7 +9517,13 @@ static __always_inline size_t gcu_backlog_size(MDBX_txn *txn) {
static int gcu_clean_stored_retired(MDBX_txn *txn, gcu_context_t *ctx) {
int err = MDBX_SUCCESS;
if (ctx->retired_stored)
if (ctx->retired_stored) {
MDBX_cursor *const gc = (MDBX_cursor *)((char *)txn + sizeof(MDBX_txn));
tASSERT(txn, txn == txn->mt_env->me_txn0 && gc->mc_next == nullptr);
gc->mc_txn = txn;
gc->mc_flags = 0;
gc->mc_next = txn->mt_cursors[FREE_DBI];
txn->mt_cursors[FREE_DBI] = gc;
do {
MDBX_val key, val;
#if MDBX_ENABLE_BIGFOOT
@ -9517,11 +9532,10 @@ static int gcu_clean_stored_retired(MDBX_txn *txn, gcu_context_t *ctx) {
key.iov_base = &txn->mt_txnid;
#endif /* MDBX_ENABLE_BIGFOOT */
key.iov_len = sizeof(txnid_t);
const struct cursor_set_result csr =
cursor_set(&ctx->cursor, &key, &val, MDBX_SET);
const struct cursor_set_result csr = cursor_set(gc, &key, &val, MDBX_SET);
if (csr.err == MDBX_SUCCESS && csr.exact) {
ctx->retired_stored = 0;
err = mdbx_cursor_del(&ctx->cursor, 0);
err = mdbx_cursor_del(gc, 0);
TRACE("== clear-4linear, backlog %zu, err %d", gcu_backlog_size(txn),
err);
}
@ -9531,6 +9545,9 @@ static int gcu_clean_stored_retired(MDBX_txn *txn, gcu_context_t *ctx) {
#else
while (0);
#endif /* MDBX_ENABLE_BIGFOOT */
txn->mt_cursors[FREE_DBI] = gc->mc_next;
gc->mc_next = nullptr;
}
return err;
}
@ -9544,28 +9561,45 @@ static int gcu_touch(gcu_context_t *ctx) {
/* Prepare a backlog of pages to modify GC itself, while reclaiming is
* prohibited. It should be enough to prevent search in page_alloc_slowpath()
* during a deleting, when GC tree is unbalanced. */
static int gcu_prepare_backlog(MDBX_txn *txn, gcu_context_t *ctx,
const bool reserve4retired) {
const size_t pages4retiredlist =
reserve4retired ? number_of_ovpages(
txn->mt_env, MDBX_PNL_SIZEOF(txn->tw.retired_pages))
: 0;
const size_t backlog4cow = txn->mt_dbs[FREE_DBI].md_depth;
const size_t backlog4rebalance = backlog4cow + 1;
static int gcu_prepare_backlog(MDBX_txn *txn, gcu_context_t *ctx) {
const size_t for_cow = txn->mt_dbs[FREE_DBI].md_depth;
const size_t for_rebalance = for_cow + 1;
size_t for_split = ctx->retired_stored == 0;
if (likely(pages4retiredlist < 2 &&
gcu_backlog_size(txn) > (reserve4retired
? backlog4rebalance
: (backlog4cow + backlog4rebalance))))
const intptr_t retired_left =
MDBX_PNL_SIZEOF(txn->tw.retired_pages) - ctx->retired_stored;
size_t for_retiredlist = 0;
if (MDBX_ENABLE_BIGFOOT && retired_left > 0) {
for_retiredlist = (retired_left + txn->mt_env->me_maxgc_ov1page - 1) /
txn->mt_env->me_maxgc_ov1page;
const size_t per_branch_page =
(txn->mt_env->me_psize - PAGEHDRSZ) /
(sizeof(indx_t) + sizeof(MDBX_node) + sizeof(txnid_t));
for (size_t entries = for_retiredlist; entries > 1; for_split += entries)
entries = (entries + per_branch_page - 1) / per_branch_page;
} else if (!MDBX_ENABLE_BIGFOOT && retired_left != 0) {
for_retiredlist =
number_of_ovpages(txn->mt_env, MDBX_PNL_SIZEOF(txn->tw.retired_pages));
}
const size_t for_tree_before_touch = for_cow + for_rebalance + for_split;
const size_t for_tree_after_touch = for_rebalance + for_split;
const size_t for_data = for_retiredlist;
const size_t for_all_before_touch = for_data + for_tree_before_touch;
const size_t for_all_after_touch = for_data + for_tree_after_touch;
if (likely(for_data < 2 && gcu_backlog_size(txn) > for_all_before_touch))
return MDBX_SUCCESS;
TRACE(
">> reserve4retired %c, backlog %zu, 4list %zu, 4cow %zu, 4rebalance %zu",
reserve4retired ? 'Y' : 'N', gcu_backlog_size(txn), pages4retiredlist,
backlog4cow, backlog4rebalance);
TRACE(">> retired-stored %zu, left %zi, backlog %zu, need %zu (4list %zu, "
"4split %zu, "
"4cow %zu, 4tree %zu)",
ctx->retired_stored, retired_left, gcu_backlog_size(txn),
for_all_before_touch, for_data, for_split, for_cow,
for_tree_before_touch);
int err;
if (unlikely(pages4retiredlist > 2)) {
if (unlikely(for_data > 2)) {
MDBX_val key, val;
key.iov_base = val.iov_base = nullptr;
key.iov_len = sizeof(txnid_t);
@ -9578,28 +9612,34 @@ static int gcu_prepare_backlog(MDBX_txn *txn, gcu_context_t *ctx,
err = gcu_touch(ctx);
TRACE("== after-touch, backlog %zu, err %d", gcu_backlog_size(txn), err);
if (unlikely(pages4retiredlist > 1) &&
if (!MDBX_ENABLE_BIGFOOT && unlikely(for_data > 1) &&
MDBX_PNL_GETSIZE(txn->tw.retired_pages) != ctx->retired_stored &&
err == MDBX_SUCCESS) {
tASSERT(txn, reserve4retired);
err = gcu_clean_stored_retired(txn, ctx);
if (unlikely(err != MDBX_SUCCESS))
return err;
err =
page_alloc_slowpath(&ctx->cursor, pages4retiredlist, MDBX_ALLOC_RESERVE)
.err;
if (unlikely(ctx->retired_stored)) {
err = gcu_clean_stored_retired(txn, ctx);
if (unlikely(err != MDBX_SUCCESS))
return err;
if (!ctx->retired_stored)
return /* restart by tail-recursion */ gcu_prepare_backlog(txn, ctx);
}
err = page_alloc_slowpath(&ctx->cursor, for_data, MDBX_ALLOC_RESERVE).err;
TRACE("== after-4linear, backlog %zu, err %d", gcu_backlog_size(txn), err);
cASSERT(&ctx->cursor,
gcu_backlog_size(txn) >= pages4retiredlist || err != MDBX_SUCCESS);
gcu_backlog_size(txn) >= for_data || err != MDBX_SUCCESS);
}
while (gcu_backlog_size(txn) < backlog4cow + pages4retiredlist &&
err == MDBX_SUCCESS)
while (gcu_backlog_size(txn) < for_all_after_touch && err == MDBX_SUCCESS)
err = page_alloc_slowpath(&ctx->cursor, 0,
MDBX_ALLOC_RESERVE | MDBX_ALLOC_UNIMPORTANT)
.err;
TRACE("<< backlog %zu, err %d", gcu_backlog_size(txn), err);
TRACE("<< backlog %zu, err %d, gc: height %u, branch %zu, leaf %zu, large "
"%zu, entries %zu",
gcu_backlog_size(txn), err, txn->mt_dbs[FREE_DBI].md_depth,
(size_t)txn->mt_dbs[FREE_DBI].md_branch_pages,
(size_t)txn->mt_dbs[FREE_DBI].md_leaf_pages,
(size_t)txn->mt_dbs[FREE_DBI].md_overflow_pages,
(size_t)txn->mt_dbs[FREE_DBI].md_entries);
return (err != MDBX_NOTFOUND) ? err : MDBX_SUCCESS;
}
@ -9665,9 +9705,10 @@ retry:
TRACE("%s", " >> continue");
if (ctx->retired_stored != MDBX_PNL_GETSIZE(txn->tw.retired_pages) &&
(MDBX_PNL_GETSIZE(txn->tw.retired_pages) > env->me_maxgc_ov1page ||
(ctx->loop == 1 ||
MDBX_PNL_GETSIZE(txn->tw.retired_pages) > env->me_maxgc_ov1page ||
ctx->retired_stored > env->me_maxgc_ov1page)) {
rc = gcu_prepare_backlog(txn, ctx, true);
rc = gcu_prepare_backlog(txn, ctx);
if (unlikely(rc != MDBX_SUCCESS))
goto bailout;
}
@ -9696,7 +9737,7 @@ retry:
if (unlikely(rc != MDBX_SUCCESS))
goto bailout;
if (likely(!ctx->dense)) {
rc = gcu_prepare_backlog(txn, ctx, false);
rc = gcu_prepare_backlog(txn, ctx);
if (unlikely(rc != MDBX_SUCCESS))
goto bailout;
}
@ -9730,7 +9771,7 @@ retry:
if (ctx->cleaned_id > txn->tw.last_reclaimed)
break;
if (likely(!ctx->dense)) {
rc = gcu_prepare_backlog(txn, ctx, false);
rc = gcu_prepare_backlog(txn, ctx);
if (unlikely(rc != MDBX_SUCCESS))
goto bailout;
}
@ -9767,30 +9808,29 @@ retry:
/* handle loose pages - put ones into the reclaimed- or retired-list */
if (txn->tw.loose_pages) {
tASSERT(txn, txn->tw.loose_count > 0);
/* Return loose page numbers to tw.relist,
* though usually none are left at this point.
* The pages themselves remain in dirtylist. */
if (unlikely(!txn->tw.lifo_reclaimed && txn->tw.last_reclaimed < 1)) {
if (txn->tw.loose_count > 0) {
TRACE("%s: try allocate gc-slot for %zu loose-pages", dbg_prefix_mode,
txn->tw.loose_count);
rc = page_alloc_slowpath(&ctx->cursor, 0, MDBX_ALLOC_RESERVE).err;
if (rc == MDBX_SUCCESS) {
TRACE("%s: retry since gc-slot for %zu loose-pages available",
dbg_prefix_mode, txn->tw.loose_count);
continue;
}
/* Put loose page numbers in tw.retired_pages,
* since unable to return them to tw.relist. */
if (unlikely((rc = pnl_need(&txn->tw.retired_pages,
txn->tw.loose_count)) != 0))
goto bailout;
for (MDBX_page *mp = txn->tw.loose_pages; mp; mp = mp->mp_next)
pnl_xappend(txn->tw.retired_pages, mp->mp_pgno);
TRACE("%s: append %zu loose-pages to retired-pages", dbg_prefix_mode,
txn->tw.loose_count);
TRACE("%s: try allocate gc-slot for %zu loose-pages", dbg_prefix_mode,
txn->tw.loose_count);
rc = page_alloc_slowpath(&ctx->cursor, 0, MDBX_ALLOC_RESERVE).err;
if (rc == MDBX_SUCCESS) {
TRACE("%s: retry since gc-slot for %zu loose-pages available",
dbg_prefix_mode, txn->tw.loose_count);
continue;
}
/* Put loose page numbers in tw.retired_pages,
* since unable to return them to tw.relist. */
if (unlikely((rc = pnl_need(&txn->tw.retired_pages,
txn->tw.loose_count)) != 0))
goto bailout;
for (MDBX_page *mp = txn->tw.loose_pages; mp; mp = mp->mp_next)
pnl_xappend(txn->tw.retired_pages, mp->mp_pgno);
TRACE("%s: append %zu loose-pages to retired-pages", dbg_prefix_mode,
txn->tw.loose_count);
} else {
/* Room for loose pages + temp PNL with same */
rc = pnl_need(&txn->tw.relist, 2 * txn->tw.loose_count + 2);
@ -9874,14 +9914,31 @@ retry:
}
retired_pages_before = MDBX_PNL_GETSIZE(txn->tw.retired_pages);
rc = gcu_prepare_backlog(txn, ctx, true);
rc = gcu_prepare_backlog(txn, ctx);
if (unlikely(rc != MDBX_SUCCESS))
goto bailout;
if (retired_pages_before != MDBX_PNL_GETSIZE(txn->tw.retired_pages)) {
TRACE("%s: retired-list changed (%zu -> %zu), retry", dbg_prefix_mode,
retired_pages_before, MDBX_PNL_GETSIZE(txn->tw.retired_pages));
break;
}
pnl_sort(txn->tw.retired_pages, txn->mt_next_pgno);
ctx->retired_stored = 0;
ctx->bigfoot = txn->mt_txnid;
do {
if (ctx->retired_stored) {
rc = gcu_prepare_backlog(txn, ctx);
if (unlikely(rc != MDBX_SUCCESS))
goto bailout;
if (ctx->retired_stored >=
MDBX_PNL_GETSIZE(txn->tw.retired_pages)) {
TRACE("%s: retired-list changed (%zu -> %zu), retry",
dbg_prefix_mode, retired_pages_before,
MDBX_PNL_GETSIZE(txn->tw.retired_pages));
break;
}
}
key.iov_len = sizeof(txnid_t);
key.iov_base = &ctx->bigfoot;
const size_t left =
@ -9927,7 +9984,7 @@ retry:
key.iov_len = sizeof(txnid_t);
key.iov_base = &txn->mt_txnid;
do {
gcu_prepare_backlog(txn, ctx, true);
gcu_prepare_backlog(txn, ctx);
data.iov_len = MDBX_PNL_SIZEOF(txn->tw.retired_pages);
rc = mdbx_cursor_put(&ctx->cursor, &key, &data, MDBX_RESERVE);
if (unlikely(rc != MDBX_SUCCESS))
@ -9940,7 +9997,7 @@ retry:
eASSERT(env, data.iov_len == MDBX_PNL_SIZEOF(txn->tw.retired_pages));
memcpy(data.iov_base, txn->tw.retired_pages, data.iov_len);
TRACE("%s: put-retired #%u @ %" PRIaTXN, dbg_prefix_mode,
TRACE("%s: put-retired #%zu @ %" PRIaTXN, dbg_prefix_mode,
ctx->retired_stored, txn->mt_txnid);
#endif /* MDBX_ENABLE_BIGFOOT */
if (LOG_ENABLED(MDBX_LOG_EXTRA)) {
@ -9982,7 +10039,7 @@ retry:
if (0 >= (intptr_t)left)
break;
const size_t prefer_max_scatter = 257;
const size_t prefer_max_scatter = MDBX_ENABLE_BIGFOOT ? MDBX_TXL_MAX : 257;
txnid_t reservation_gc_id;
if (ctx->lifo) {
if (txn->tw.lifo_reclaimed == nullptr) {
@ -10154,6 +10211,10 @@ retry:
: (ctx->rid < INT16_MAX) ? (size_t)ctx->rid
: INT16_MAX;
if (avail_gc_slots > 1) {
#if MDBX_ENABLE_BIGFOOT
chunk = (chunk < env->me_maxgc_ov1page * 2) ? chunk / 2
: env->me_maxgc_ov1page;
#else
if (chunk < env->me_maxgc_ov1page * 2)
chunk /= 2;
else {
@ -10190,6 +10251,7 @@ retry:
: tail;
}
}
#endif /* MDBX_ENABLE_BIGFOOT */
}
}
tASSERT(txn, chunk > 0);
@ -10217,7 +10279,7 @@ retry:
data.iov_len = (chunk + 1) * sizeof(pgno_t);
TRACE("%s: reserve %zu [%zu...%zu) @%" PRIaTXN, dbg_prefix_mode, chunk,
ctx->settled + 1, ctx->settled + chunk + 1, reservation_gc_id);
gcu_prepare_backlog(txn, ctx, true);
gcu_prepare_backlog(txn, ctx);
rc = mdbx_cursor_put(&ctx->cursor, &key, &data,
MDBX_RESERVE | MDBX_NOOVERWRITE);
tASSERT(txn, pnl_check_allocated(txn->tw.relist,