MDEV-23855: Shrink fil_space_t

Merge n_pending_ios, n_pending_ops to std::atomic<uint32_t> n_pending.
Change some more fil_space_t members to uint32_t to reduce
the memory footprint.

fil_space_t::add(), fil_ibd_create(): Attach the already opened
handle to the tablespace, and enforce the fil_system.n_open limit.

dict_boot(): Initialize fil_system.max_assigned_id.

srv_boot(): Call srv_thread_pool_init() before anything else,
so that files should be opened in the correct mode on Windows.

fil_ibd_create(): Create the file in OS_FILE_AIO mode, just like
fil_node_open_file_low() does it.

dict_table_t::is_accessible(): Replaces fil_table_accessible().

Reviewed by: Vladislav Vaintroub

5 files changed, 52 insertions, 86 deletions

diff --git a/storage/innobase/buf/buf0buf.cc b/storage/innobase/buf/buf0buf.cc
index 2046ffd4273..2cdd36a8b60 100644
--- a/storage/innobase/buf/buf0buf.cc
+++ b/storage/innobase/buf/buf0buf.cc
@@ -415,7 +415,7 @@ static bool buf_tmp_page_decrypt(byte* tmp_frame, byte* src_frame)
 static bool buf_page_decrypt_after_read(buf_page_t *bpage,
                                         const fil_node_t &node)
 {
-	ut_ad(node.space->pending_io());
+	ut_ad(node.space->referenced());
 	ut_ad(node.space->id == bpage->id().space());
 	const auto flags = node.space->flags;
 
@@ -475,7 +475,7 @@ decompress_with_slot:
 		slot->release();
 		ut_ad(!write_size
 		      || fil_page_type_validate(node.space, dst_frame));
-		ut_ad(node.space->pending_io());
+		ut_ad(node.space->referenced());
 		return write_size != 0;
 	}
 
@@ -516,7 +516,7 @@ decrypt_failed:
 		goto decompress;
 	}
 
-	ut_ad(node.space->pending_io());
+	ut_ad(node.space->referenced());
 	return true;
 }
 #endif /* !UNIV_INNOCHECKSUM */
@@ -2768,7 +2768,7 @@ buf_zip_decompress(
 	ulint		size = page_zip_get_size(&block->page.zip);
 	/* The tablespace will not be found if this function is called
 	during IMPORT. */
-	fil_space_t* space= fil_space_t::get_for_io(block->page.id().space());
+	fil_space_t* space= fil_space_t::get(block->page.id().space());
 	const unsigned key_version = mach_read_from_4(
 		frame + FIL_PAGE_FILE_FLUSH_LSN_OR_KEY_VERSION);
 	fil_space_crypt_t* crypt_data = space ? space->crypt_data : NULL;
@@ -2805,7 +2805,7 @@ buf_zip_decompress(
 		if (page_zip_decompress(&block->page.zip,
 					block->frame, TRUE)) {
 			if (space) {
-				space->release_for_io();
+				space->release();
 			}
 			return(TRUE);
 		}
@@ -2824,7 +2824,7 @@ buf_zip_decompress(
 		/* Copy to uncompressed storage. */
 		memcpy(block->frame, frame, block->zip_size());
 		if (space) {
-			space->release_for_io();
+			space->release();
 		}
 
 		return(TRUE);
@@ -2848,7 +2848,7 @@ err_exit:
 			dict_set_corrupted_by_space(space);
 		}
 
-		space->release_for_io();
+		space->release();
 	}
 
 	return(FALSE);
@@ -3160,10 +3160,10 @@ lookup:
 			asserting. */
 			if (page_id.space() == TRX_SYS_SPACE) {
 			} else if (page_id.space() == SRV_TMP_SPACE_ID) {
-			} else if (fil_space_t* space= fil_space_t::get_for_io(
+			} else if (fil_space_t* space= fil_space_t::get(
 					   page_id.space())) {
 				bool set = dict_set_corrupted_by_space(space);
-				space->release_for_io();
+				space->release();
 				if (set) {
 					return NULL;
 				}
@@ -3374,8 +3374,7 @@ re_evict:
 	if (mode != BUF_GET_IF_IN_POOL
 	    && mode != BUF_GET_IF_IN_POOL_OR_WATCH) {
 	} else if (!ibuf_debug) {
-	} else if (fil_space_t* space
-		   = fil_space_t::get_for_io(page_id.space())) {
+	} else if (fil_space_t* space = fil_space_t::get(page_id.space())) {
 		/* Try to evict the block from the buffer pool, to use the
 		insert buffer (change buffer) as much as possible. */
 
@@ -3386,7 +3385,7 @@ re_evict:
 		/* Blocks cannot be relocated or enter or exit the
 		buf_pool while we are holding the buf_pool.mutex. */
 		const bool evicted = buf_LRU_free_page(&fix_block->page, true);
-		space->release_for_io();
+		space->release();
 
 		if (evicted) {
 			hash_lock = buf_pool.page_hash.lock_get(fold);
@@ -4108,7 +4107,7 @@ after decryption normal page checksum does not match.
 static dberr_t buf_page_check_corrupt(buf_page_t *bpage,
                                       const fil_node_t &node)
 {
-	ut_ad(node.space->pending_io());
+	ut_ad(node.space->referenced());
 
 	byte* dst_frame = (bpage->zip.data) ? bpage->zip.data :
 		((buf_block_t*) bpage)->frame;
diff --git a/storage/innobase/buf/buf0dblwr.cc b/storage/innobase/buf/buf0dblwr.cc
index 6b1a32d8930..eb460af2de2 100644
--- a/storage/innobase/buf/buf0dblwr.cc
+++ b/storage/innobase/buf/buf0dblwr.cc
@@ -363,7 +363,7 @@ void buf_dblwr_t::recover()
       continue;
     }
 
-    fil_space_t *space= fil_space_t::get_for_io(space_id);
+    fil_space_t *space= fil_space_t::get(space_id);
 
     if (!space)
       /* The tablespace that this page once belonged to does not exist */
@@ -379,7 +379,7 @@ void buf_dblwr_t::recover()
                    << " is beyond the end of tablespace " << space->name
                    << " (" << space->size << " pages)";
 next_page:
-      space->release_for_io();
+      space->release();
       continue;
     }
 
@@ -420,7 +420,7 @@ next_page:
 
     /* Write the good page from the doublewrite buffer to the intended
     position. */
-    space->reacquire_for_io();
+    space->reacquire();
     fio= space->io(IORequestWrite,
                    os_offset_t{page_id.page_no()} * physical_size,
                    physical_size, page);
@@ -506,10 +506,10 @@ static void buf_dblwr_check_page_lsn(const page_t* page, const fil_space_t& s)
 
 static void buf_dblwr_check_page_lsn(const buf_page_t &b, const byte *page)
 {
-  if (fil_space_t *space= fil_space_t::get_for_io(b.id().space()))
+  if (fil_space_t *space= fil_space_t::get(b.id().space()))
   {
     buf_dblwr_check_page_lsn(page, *space);
-    space->release_for_io();
+    space->release();
   }
 }
 
@@ -583,7 +583,7 @@ bool buf_dblwr_t::flush_buffered_writes(const ulint size)
   }
 #endif /* UNIV_DEBUG */
   /* Write out the first block of the doublewrite buffer */
-  ut_a(fil_system.sys_space->acquire_for_io());
+  ut_a(fil_system.sys_space->acquire());
   fil_system.sys_space->io(IORequestWrite,
                            os_offset_t{block1.page_no()} <<
                            srv_page_size_shift,
@@ -593,7 +593,7 @@ bool buf_dblwr_t::flush_buffered_writes(const ulint size)
   if (old_first_free > size)
   {
     /* Write out the second block of the doublewrite buffer. */
-    ut_a(fil_system.sys_space->acquire_for_io());
+    ut_a(fil_system.sys_space->acquire());
     fil_system.sys_space->io(IORequestWrite,
                              os_offset_t{block2.page_no()} <<
                              srv_page_size_shift,
@@ -687,7 +687,7 @@ void buf_dblwr_t::add_to_batch(fil_space_t *space, const IORequest &request,
   ut_ad(request.bpage);
   ut_ad(request.bpage->in_file());
   ut_ad(space->id == request.bpage->id().space());
-  ut_ad(space->pending_io());
+  ut_ad(space->referenced());
   ut_ad(!srv_read_only_mode);
 
   const ulint buf_size= 2 * block_size();
diff --git a/storage/innobase/buf/buf0dump.cc b/storage/innobase/buf/buf0dump.cc
index 19a9e09e4a1..c62c0112e92 100644
--- a/storage/innobase/buf/buf0dump.cc
+++ b/storage/innobase/buf/buf0dump.cc
@@ -621,19 +621,11 @@ buf_load()
 	ulint		last_check_time = 0;
 	ulint		last_activity_cnt = 0;
 
-	/* Avoid calling the expensive fil_space_acquire_silent() for each
+	/* Avoid calling the expensive fil_space_t::get() for each
 	page within the same tablespace. dump[] is sorted by (space, page),
 	so all pages from a given tablespace are consecutive. */
 	ulint		cur_space_id = dump[0].space();
-	fil_space_t*	space = fil_space_acquire_silent(cur_space_id);
-	if (space) {
-		bool ok = space->acquire_for_io();
-		space->release();
-		if (!ok) {
-			space = nullptr;
-		}
-	}
-
+	fil_space_t*	space = fil_space_t::get(cur_space_id);
 	ulint		zip_size = space ? space->zip_size() : 0;
 
 	PSI_stage_progress*	pfs_stage_progress __attribute__((unused))
@@ -653,24 +645,16 @@ buf_load()
 
 		if (this_space_id != cur_space_id) {
 			if (space) {
-				space->release_for_io();
+				space->release();
 			}
 
 			cur_space_id = this_space_id;
-			space = fil_space_acquire_silent(cur_space_id);
+			space = fil_space_t::get(cur_space_id);
 
 			if (!space) {
 				continue;
 			}
 
-			bool ok = space->acquire_for_io();
-			space->release();
-
-			if (!ok) {
-				space = nullptr;
-				continue;
-			}
-
 			zip_size = space->zip_size();
 		}
 
@@ -684,17 +668,17 @@ buf_load()
 		}
 
 		if (space->is_stopping()) {
-			space->release_for_io();
+			space->release();
 			space = nullptr;
 			continue;
 		}
 
-		space->reacquire_for_io();
+		space->reacquire();
 		buf_read_page_background(space, dump[i], zip_size, true);
 
 		if (buf_load_abort_flag) {
 			if (space) {
-				space->release_for_io();
+				space->release();
 			}
 			buf_load_abort_flag = false;
 			ut_free(dump);
@@ -722,7 +706,7 @@ buf_load()
 	}
 
 	if (space) {
-		space->release_for_io();
+		space->release();
 	}
 
 	ut_free(dump);
diff --git a/storage/innobase/buf/buf0flu.cc b/storage/innobase/buf/buf0flu.cc
index 25523ab53f1..f671aadeb4e 100644
--- a/storage/innobase/buf/buf0flu.cc
+++ b/storage/innobase/buf/buf0flu.cc
@@ -786,7 +786,7 @@ static bool buf_flush_page(buf_page_t *bpage, bool lru, fil_space_t *space)
         (space == fil_system.temp_space));
   ut_ad(space->purpose == FIL_TYPE_TABLESPACE ||
         space->atomic_write_supported);
-  ut_ad(space->pending_io());
+  ut_ad(space->referenced());
 
   rw_lock_t *rw_lock;
 
@@ -854,7 +854,7 @@ static bool buf_flush_page(buf_page_t *bpage, bool lru, fil_space_t *space)
     buf_release_freed_page(&block->page);
   else
   {
-    space->reacquire_for_io();
+    space->reacquire();
     ut_ad(status == buf_page_t::NORMAL || status == buf_page_t::INIT_ON_FLUSH);
     size_t size, orig_size;
     IORequest::Type type= lru ? IORequest::WRITE_LRU : IORequest::WRITE_ASYNC;
@@ -1029,7 +1029,7 @@ static void buf_flush_freed_pages(fil_space_t *space)
 
     if (punch_hole)
     {
-      space->reacquire_for_io();
+      space->reacquire();
       space->io(IORequest(IORequest::PUNCH_RANGE),
                           os_offset_t{range.first} * physical_size,
                           (range.last - range.first + 1) * physical_size,
@@ -1039,7 +1039,7 @@ static void buf_flush_freed_pages(fil_space_t *space)
     {
       for (os_offset_t i= range.first; i <= range.last; i++)
       {
-        space->reacquire_for_io();
+        space->reacquire();
         space->io(IORequest(IORequest::WRITE_ASYNC),
                   i * physical_size, physical_size,
                   const_cast<byte*>(field_ref_zero));
@@ -1170,7 +1170,7 @@ static ulint buf_free_from_unzip_LRU_list_batch(ulint max)
 @retval nullptr if the pages for this tablespace should be discarded */
 static fil_space_t *buf_flush_space(const uint32_t id)
 {
-  fil_space_t *space= fil_space_t::get_for_io(id);
+  fil_space_t *space= fil_space_t::get(id);
   if (space)
     buf_flush_freed_pages(space);
   return space;
@@ -1261,7 +1261,7 @@ static void buf_flush_LRU_list_batch(ulint max, flush_counters_t *n)
         if (last_space_id != space_id)
         {
           if (space)
-            space->release_for_io();
+            space->release();
           space= buf_flush_space(space_id);
           last_space_id= space_id;
         }
@@ -1270,7 +1270,7 @@ static void buf_flush_LRU_list_batch(ulint max, flush_counters_t *n)
       }
       else if (space->is_stopping())
       {
-        space->release_for_io();
+        space->release();
         space= nullptr;
       }
 
@@ -1298,7 +1298,7 @@ reacquire_mutex:
   buf_pool.lru_hp.set(nullptr);
 
   if (space)
-    space->release_for_io();
+    space->release();
 
   /* We keep track of all flushes happening as part of LRU flush. When
   estimating the desired rate at which flush_list should be flushed,
@@ -1393,7 +1393,7 @@ static ulint buf_do_flush_list_batch(ulint max_n, lsn_t lsn)
         if (last_space_id != space_id)
         {
           if (space)
-            space->release_for_io();
+            space->release();
           space= buf_flush_space(space_id);
           last_space_id= space_id;
         }
@@ -1402,7 +1402,7 @@ static ulint buf_do_flush_list_batch(ulint max_n, lsn_t lsn)
       }
       else if (space->is_stopping())
       {
-        space->release_for_io();
+        space->release();
         space= nullptr;
       }
 
@@ -1431,7 +1431,7 @@ reacquire_mutex:
   mysql_mutex_unlock(&buf_pool.flush_list_mutex);
 
   if (space)
-    space->release_for_io();
+    space->release();
 
   if (scanned)
     MONITOR_INC_VALUE_CUMULATIVE(MONITOR_FLUSH_BATCH_SCANNED,
diff --git a/storage/innobase/buf/buf0rea.cc b/storage/innobase/buf/buf0rea.cc
index daea53ec130..2f59f1ae4d5 100644
--- a/storage/innobase/buf/buf0rea.cc
+++ b/storage/innobase/buf/buf0rea.cc
@@ -288,7 +288,7 @@ buf_read_page_low(
 			<< page_id;
 		ut_ad(0);
 nothing_read:
-		space->release_for_io();
+		space->release();
 		return false;
 	}
 
@@ -358,7 +358,7 @@ nothing_read:
 
 		/* The i/o was already completed in space->io() */
 		*err = buf_page_read_complete(bpage, *fio.node);
-		space->release_for_io();
+		space->release();
 
 		if (*err != DB_SUCCESS) {
 			return false;
@@ -402,7 +402,7 @@ buf_read_ahead_random(const page_id_t page_id, ulint zip_size, bool ibuf)
   if (buf_pool.n_pend_reads > buf_pool.curr_size / BUF_READ_AHEAD_PEND_LIMIT)
     return 0;
 
-  fil_space_t* space= fil_space_acquire(page_id.space());
+  fil_space_t* space= fil_space_t::get(page_id.space());
   if (!space)
     return 0;
 
@@ -431,7 +431,7 @@ no_read_ahead:
   return 0;
 
 read_ahead:
-  if (!space->acquire_for_io())
+  if (!space->acquire_if_not_stopped())
     goto no_read_ahead;
 
   /* Read all the suitable blocks within the area */
@@ -444,7 +444,7 @@ read_ahead:
     if (space->is_stopping())
       break;
     dberr_t err;
-    space->reacquire_for_io();
+    space->reacquire();
     if (buf_read_page_low(&err, space, false, ibuf_mode, i, zip_size, false))
       count++;
   }
@@ -453,7 +453,6 @@ read_ahead:
     DBUG_PRINT("ib_buf", ("random read-ahead %zu pages from %s: %u",
 			  count, space->chain.start->name,
 			  low.page_no()));
-  space->release_for_io();
   space->release();
 
   /* Read ahead is considered one I/O operation for the purpose of
@@ -478,22 +477,13 @@ after decryption normal page checksum does not match.
 @retval DB_TABLESPACE_DELETED if tablespace .ibd file is missing */
 dberr_t buf_read_page(const page_id_t page_id, ulint zip_size)
 {
-  fil_space_t *space= fil_space_acquire(page_id.space());
+  fil_space_t *space= fil_space_t::get(page_id.space());
   if (!space)
   {
     ib::info() << "trying to read page " << page_id
                << " in nonexisting or being-dropped tablespace";
     return DB_TABLESPACE_DELETED;
   }
-  else if (!space->acquire_for_io())
-  {
-    ib::warn() << "unable to read " << page_id << " from tablespace "
-               << space->name;
-    space->release();
-    return DB_PAGE_CORRUPTED;
-  }
-
-  space->release();
 
   dberr_t err;
   if (buf_read_page_low(&err, space, true, BUF_READ_ANY_PAGE,
@@ -607,22 +597,15 @@ buf_read_ahead_linear(const page_id_t page_id, ulint zip_size, bool ibuf)
     read-ahead, as that could break the ibuf page access order */
     return 0;
 
-  fil_space_t *space= fil_space_acquire(page_id.space());
+  fil_space_t *space= fil_space_t::get(page_id.space());
   if (!space)
     return 0;
-  else
-  {
-    bool ok= space->acquire_for_io();
-    space->release();
-    if (!ok)
-      return 0;
-  }
 
   if (high_1.page_no() > space->last_page_number())
   {
     /* The area is not whole. */
 fail:
-    space->release_for_io();
+    space->release();
     return 0;
   }
 
@@ -721,7 +704,7 @@ failed:
     if (space->is_stopping())
       break;
     dberr_t err;
-    space->reacquire_for_io();
+    space->reacquire();
     count+= buf_read_page_low(&err, space, false, ibuf_mode, new_low, zip_size,
                               false);
   }
@@ -730,7 +713,7 @@ failed:
     DBUG_PRINT("ib_buf", ("random read-ahead %zu pages from %s: %u",
                           count, space->chain.start->name,
                           new_low.page_no()));
-  space->release_for_io();
+  space->release();
 
   /* Read ahead is considered one I/O operation for the purpose of
   LRU policy decision. */
@@ -747,7 +730,7 @@ highest page number the last in the array
 @param[in]	n		number of page numbers in the array */
 void buf_read_recv_pages(ulint space_id, const uint32_t* page_nos, ulint n)
 {
-	fil_space_t* space = fil_space_t::get_for_io(space_id);
+	fil_space_t* space = fil_space_t::get(space_id);
 
 	if (!space) {
 		/* The tablespace is missing or unreadable: do nothing */
@@ -784,7 +767,7 @@ void buf_read_recv_pages(ulint space_id, const uint32_t* page_nos, ulint n)
 		}
 
 		dberr_t err;
-		space->reacquire_for_io();
+		space->reacquire();
 		buf_read_page_low(&err, space, false,
 				  BUF_READ_ANY_PAGE, cur_page_id, zip_size,
 				  true);
@@ -798,5 +781,5 @@ void buf_read_recv_pages(ulint space_id, const uint32_t* page_nos, ulint n)
 
         DBUG_PRINT("ib_buf", ("recovery read (%u pages) for %s", n,
 			      space->chain.start->name));
-	space->release_for_io();
+	space->release();
 }