MDEV-29593 Purge misses a chance to free not-yet-reused undo pages

trx_purge_truncate_rseg_history(): If all other conditions for
invoking trx_purge_remove_log_hdr() hold, but the state is
TRX_UNDO_CACHED instead of TRX_UNDO_TO_PURGE, detach and free it.

Tested by: Matthias Leich

storage/innobase/include/mtr0mtr.h

@@ -343,7 +343,7 @@ struct mtr_t {
   {
     mtr_memo_slot_t &slot= m_memo[savepoint];
     ut_ad(slot.type <= MTR_MEMO_BUF_FIX);
-    ut_ad(type <= MTR_MEMO_BUF_FIX);
+    ut_ad(type < MTR_MEMO_S_LOCK);
     slot.type= type;
   }
 

storage/innobase/trx/trx0purge.cc

@@ -275,7 +275,7 @@ trx_purge_add_undo_to_history(const trx_t* trx, trx_undo_t*& undo, mtr_t* mtr)
 	if (undo->state != TRX_UNDO_CACHED) {
 		/* The undo log segment will not be reused */
 		ut_a(undo->id < TRX_RSEG_N_SLOTS);
-		compile_time_assert(FIL_NULL == 0xffffffff);
+		static_assert(FIL_NULL == 0xffffffff, "");
 		mtr->memset(rseg_header,
 			    TRX_RSEG + TRX_RSEG_UNDO_SLOTS
 			    + undo->id * TRX_RSEG_SLOT_SIZE, 4, 0xff);
@@ -385,45 +385,11 @@ static dberr_t trx_purge_remove_log_hdr(buf_block_t *rseg, buf_block_t* log,
                      uint16_t(offset + TRX_UNDO_HISTORY_NODE), mtr);
 }
 
-MY_ATTRIBUTE((nonnull, warn_unused_result))
-/** Free an undo log segment, and remove the header from the history list.
-@param[in,out]	mtr		mini-transaction
-@param[in,out]	rseg		rollback segment
-@param[in]	hdr_addr	file address of log_hdr
-@return error code */
-static dberr_t
-trx_purge_free_segment(mtr_t &mtr, trx_rseg_t* rseg, fil_addr_t hdr_addr)
+/** Free an undo log segment.
+@param block     rollback segment header page
+@param mtr       mini-transaction */
+static void trx_purge_free_segment(buf_block_t *block, mtr_t &mtr)
 {
-  mtr.commit();
-  log_free_check();
-  mtr.start();
-
-  const page_id_t hdr_page_id{rseg->space->id, hdr_addr.page};
-  dberr_t err;
-  buf_block_t *rseg_hdr= rseg->get(&mtr, &err);
-  if (!rseg_hdr)
-    return err;
-  buf_block_t *block= buf_page_get_gen(hdr_page_id, 0, RW_X_LATCH,
-                                       nullptr, BUF_GET_POSSIBLY_FREED,
-                                       &mtr, &err);
-  if (!block)
-    return err;
-
-  const uint32_t seg_size=
-    flst_get_len(TRX_UNDO_SEG_HDR + TRX_UNDO_PAGE_LIST + block->page.frame);
-
-  err= trx_purge_remove_log_hdr(rseg_hdr, block, hdr_addr.boffset, &mtr);
-  if (UNIV_UNLIKELY(err != DB_SUCCESS))
-    return err;
-
-  ut_ad(rseg->curr_size >= seg_size);
-  rseg->curr_size-= seg_size;
-  rseg->history_size--;
-
-  byte *hist= TRX_RSEG + TRX_RSEG_HISTORY_SIZE + rseg_hdr->page.frame;
-  ut_ad(mach_read_from_4(hist) >= seg_size);
-  mtr.write<4>(*rseg_hdr, hist, mach_read_from_4(hist) - seg_size);
-
   while (!fseg_free_step_not_header(TRX_UNDO_SEG_HDR + TRX_UNDO_FSEG_HEADER +
                                     block->page.frame, &mtr))
   {
@@ -444,9 +410,9 @@ trx_purge_free_segment(mtr_t &mtr, trx_rseg_t* rseg, fil_addr_t hdr_addr)
     {
       block->unfix();
       block->page.lock.x_unlock();
-      block= buf_page_get_gen(id, 0, RW_X_LATCH, nullptr, BUF_GET, &mtr, &err);
+      block= buf_page_get_gen(id, 0, RW_X_LATCH, nullptr, BUF_GET, &mtr);
       if (!block)
-        return err;
+        return;
     }
     else
       mtr.memo_push(block, MTR_MEMO_PAGE_X_MODIFY);
@@ -454,100 +420,133 @@ trx_purge_free_segment(mtr_t &mtr, trx_rseg_t* rseg, fil_addr_t hdr_addr)
 
   while (!fseg_free_step(TRX_UNDO_SEG_HDR + TRX_UNDO_FSEG_HEADER +
                          block->page.frame, &mtr));
-  return DB_SUCCESS;
 }
 
 /** Remove unnecessary history data from a rollback segment.
 @param[in,out]	rseg		rollback segment
 @param[in]	limit		truncate anything before this
 @return error code */
-static
-dberr_t
-trx_purge_truncate_rseg_history(
-	trx_rseg_t&			rseg,
+static dberr_t
+trx_purge_truncate_rseg_history(trx_rseg_t& rseg,
                                 const purge_sys_t::iterator& limit)
 {
   fil_addr_t hdr_addr;
   mtr_t mtr;
 
+  log_free_check();
   mtr.start();
 
   dberr_t err;
-	buf_block_t* rseg_hdr = rseg.get(&mtr, &err);
-	if (!rseg_hdr) {
-		goto func_exit;
-	}
-
-	hdr_addr = flst_get_last(TRX_RSEG + TRX_RSEG_HISTORY
-				 + rseg_hdr->page.frame);
-	hdr_addr.boffset = static_cast<uint16_t>(hdr_addr.boffset
-						 - TRX_UNDO_HISTORY_NODE);
-
-loop:
-	if (hdr_addr.page == FIL_NULL) {
+reget:
+  buf_block_t *rseg_hdr= rseg.get(&mtr, &err);
+  if (!rseg_hdr)
+  {
 func_exit:
     mtr.commit();
     return err;
   }
 
-	buf_block_t* block = buf_page_get_gen(page_id_t(rseg.space->id,
-							hdr_addr.page),
-					      0, RW_X_LATCH, nullptr,
-					      BUF_GET_POSSIBLY_FREED,
+  hdr_addr= flst_get_last(TRX_RSEG + TRX_RSEG_HISTORY + rseg_hdr->page.frame);
+  hdr_addr.boffset= static_cast<uint16_t>(hdr_addr.boffset -
+                                          TRX_UNDO_HISTORY_NODE);
+
+loop:
+  if (hdr_addr.page == FIL_NULL)
+    goto func_exit;
+
+  buf_block_t *b=
+    buf_page_get_gen(page_id_t(rseg.space->id, hdr_addr.page),
+                     0, RW_X_LATCH, nullptr, BUF_GET_POSSIBLY_FREED,
                      &mtr, &err);
-	if (!block) {
+  if (!b)
     goto func_exit;
-	}
 
-	const trx_id_t undo_trx_no = mach_read_from_8(
-		block->page.frame + hdr_addr.boffset + TRX_UNDO_TRX_NO);
+  const trx_id_t undo_trx_no=
+    mach_read_from_8(b->page.frame + hdr_addr.boffset + TRX_UNDO_TRX_NO);
 
-	if (undo_trx_no >= limit.trx_no) {
-		if (undo_trx_no == limit.trx_no) {
-			err = trx_undo_truncate_start(
-				&rseg, hdr_addr.page,
+  if (undo_trx_no >= limit.trx_no)
+  {
+    if (undo_trx_no == limit.trx_no)
+      err = trx_undo_truncate_start(&rseg, hdr_addr.page,
                                     hdr_addr.boffset, limit.undo_no);
+    goto func_exit;
   }
 
+  fil_addr_t prev_hdr_addr=
+    flst_get_prev_addr(b->page.frame + hdr_addr.boffset +
+                       TRX_UNDO_HISTORY_NODE);
+  prev_hdr_addr.boffset= static_cast<uint16_t>(prev_hdr_addr.boffset -
+                                               TRX_UNDO_HISTORY_NODE);
+  err= trx_purge_remove_log_hdr(rseg_hdr, b, hdr_addr.boffset, &mtr);
+  if (UNIV_UNLIKELY(err != DB_SUCCESS))
     goto func_exit;
-	}
 
-	fil_addr_t prev_hdr_addr = flst_get_prev_addr(
-		block->page.frame + hdr_addr.boffset + TRX_UNDO_HISTORY_NODE);
-	prev_hdr_addr.boffset = static_cast<uint16_t>(prev_hdr_addr.boffset
-						      - TRX_UNDO_HISTORY_NODE);
+  rseg_hdr->fix();
 
-	if (!rseg.is_referenced()
-	    && rseg.needs_purge <= (purge_sys.head.trx_no
+  if (mach_read_from_2(b->page.frame + hdr_addr.boffset + TRX_UNDO_NEXT_LOG) ||
+      rseg.is_referenced() ||
+      rseg.needs_purge > (purge_sys.head.trx_no
                           ? purge_sys.head.trx_no
-				    : purge_sys.tail.trx_no)
-	    && mach_read_from_2(TRX_UNDO_SEG_HDR + TRX_UNDO_STATE
-				+ block->page.frame)
-	    == TRX_UNDO_TO_PURGE
-	    && !mach_read_from_2(block->page.frame + hdr_addr.boffset
-				 + TRX_UNDO_NEXT_LOG)) {
-		/* We can free the whole log segment.
-		This will call trx_purge_remove_log_hdr(). */
-		err = trx_purge_free_segment(mtr, &rseg, hdr_addr);
-	} else {
-		/* Remove the log hdr from the rseg history. */
-		rseg.history_size--;
-		err = trx_purge_remove_log_hdr(rseg_hdr, block,
-					       hdr_addr.boffset, &mtr);
+                          : purge_sys.tail.trx_no))
+    /* We cannot free the entire undo page. */;
+  else
+  {
+    const uint32_t seg_size=
+      flst_get_len(TRX_UNDO_SEG_HDR + TRX_UNDO_PAGE_LIST + b->page.frame);
+    switch (mach_read_from_2(TRX_UNDO_SEG_HDR + TRX_UNDO_STATE +
+                             b->page.frame)) {
+    case TRX_UNDO_TO_PURGE:
+      {
+        byte *hist= TRX_RSEG + TRX_RSEG_HISTORY_SIZE + rseg_hdr->page.frame;
+        ut_ad(mach_read_from_4(hist) >= seg_size);
+        mtr.write<4>(*rseg_hdr, hist, mach_read_from_4(hist) - seg_size);
+      }
+    free_segment:
+      ut_ad(rseg.curr_size >= seg_size);
+      rseg.curr_size-= seg_size;
+      trx_purge_free_segment(b, mtr);
+      break;
+    case TRX_UNDO_CACHED:
+      /* rseg.undo_cached must point to this page */
+      trx_undo_t *undo= UT_LIST_GET_FIRST(rseg.undo_cached);
+      for (; undo; undo= UT_LIST_GET_NEXT(undo_list, undo))
+        if (undo->hdr_page_no == hdr_addr.page)
+          goto found_cached;
+      ut_ad("inconsistent undo logs" == 0);
+      break;
+    found_cached:
+      UT_LIST_REMOVE(rseg.undo_cached, undo);
+      static_assert(FIL_NULL == 0xffffffff, "");
+      if (UNIV_UNLIKELY(mach_read_from_4(TRX_RSEG + TRX_RSEG_FORMAT +
+                                         rseg_hdr->page.frame)))
+        trx_rseg_format_upgrade(rseg_hdr, &mtr);
+      mtr.memset(rseg_hdr, TRX_RSEG + TRX_RSEG_UNDO_SLOTS +
+                 undo->id * TRX_RSEG_SLOT_SIZE, 4, 0xff);
+      ut_free(undo);
+      mtr.write<8,mtr_t::MAYBE_NOP>(*rseg_hdr, TRX_RSEG + TRX_RSEG_MAX_TRX_ID +
+                                    rseg_hdr->page.frame,
+                                    trx_sys.get_max_trx_id() - 1);
+      MONITOR_DEC(MONITOR_NUM_UNDO_SLOT_CACHED);
+      MONITOR_DEC(MONITOR_NUM_UNDO_SLOT_USED);
+      goto free_segment;
     }
-
-	mtr.commit();
-	if (err != DB_SUCCESS) {
-		return err;
   }
-	mtr.start();
 
-	hdr_addr = prev_hdr_addr;
+  hdr_addr= prev_hdr_addr;
 
-	rseg_hdr = rseg.get(&mtr, &err);
-	if (!rseg_hdr) {
-		goto func_exit;
+  mtr.commit();
+  ut_ad(rseg.history_size > 0);
+  rseg.history_size--;
+  log_free_check();
+  mtr.start();
+  rseg_hdr->page.lock.x_lock();
+  if (UNIV_UNLIKELY(rseg_hdr->page.id() != rseg.page_id()))
+  {
+    rseg_hdr->unfix();
+    rseg_hdr->page.lock.x_unlock();
+    goto reget;
   }
+  mtr.memo_push(rseg_hdr, MTR_MEMO_PAGE_X_MODIFY);
 
   goto loop;
 }