Assertion `mode == 16 || mode == 12 || fix_block->page.status != buf_page_t::FREED' failed in buf_page_get_low

Problem found during RQG testing

mysqld: /Server/bb-10.6-MDEV-24731/storage/innobase/buf/buf0buf.cc:3230: buf_block_t* buf_page_get_low(page_id_t, ulint, ulint, buf_block_t*, ulint, mtr_t*, dberr_t*, bool): Assertion `mode == 16 || mode == 12 || fix_block->page.status != buf_page_t::FREED' failed.

...

Query (0x62b000111238): ALTER TABLE t1 ADD PRIMARY KEY ( col4(10), col1 )

Status: KILL_TIMEOUT

#0  0x0000559ccfa3e27e in buf_page_free (space=0x613000122158, page=6, mtr=0x2d4943f35140)

    at /Server/bb-10.6-MDEV-24731/storage/innobase/buf/buf0buf.cc:2510

#1  0x0000559ccfb9d7dd in fseg_free_step (header=0x538310f5304a "", mtr=0x2d4943f35140, ahi=false)

    at /Server/bb-10.6-MDEV-24731/storage/innobase/fsp/fsp0fsp.cc:2751

#2  0x0000559ccf98ecdf in btr_free_but_not_root (block=0x538310ba77a0, log_mode=MTR_LOG_ALL, ahi=false)

    at /Server/bb-10.6-MDEV-24731/storage/innobase/btr/btr0btr.cc:1153

#3  0x0000559ccf98f284 in dict_index_t::clear (this=0x616000bf8c20, thr=0x6170001715d8)

    at /Server/bb-10.6-MDEV-24731/storage/innobase/btr/btr0btr.cc:1206

#4  0x0000559ccf82f73a in dict_table_t::clear (this=0x618000093d20, thr=0x6170001715d8)

    at /Server/bb-10.6-MDEV-24731/storage/innobase/row/row0log.cc:4108

#5  0x0000559ccfc76a9b in row_undo_ins (node=0x61b00004ed20, thr=0x6170001715d8)

    at /Server/bb-10.6-MDEV-24731/storage/innobase/row/row0uins.cc:607

#6  0x0000559ccfa00225 in btr_copy_blob_prefix (

    buf=0x621000352288 '\276' <repeats 200 times>..., len=2049, id=..., 

    offset=38)

    at /Server/bb-10.6-MDEV-24731/storage/innobase/btr/btr0cur.cc:7867

…

#10 0x0000559ccf81836a in row_log_table_apply_convert_mrec (

    mrec=0x334a4fcb5005 "", index=0x616000bf8c20, offsets=0x60600007e758, 

    log=0x7fc12f71d818, heap=0x621000352118, error=0x26202e839990)

    at /Server/bb-10.6-MDEV-24731/storage/innobase/row/row0log.cc:1620

Per short analysis of Marko:

This is a online table rebuild and its claimed that the BLOB is already purged.

It looks like a MDEV-515 related problem.

watch -l *(unsigned*)0x538310ba8fb4

reverse-continue

helps a lot.

sdp:/home/mleich/RQG_O/storage/1612370745/TBR-776/dev/shm/vardir/1612370745/50/1/rr

_RR_TRACE_DIR="." rr replay --mark-stdio