separate doublewrite buffer should use SRV_EXTRA_SYS_SPACE_FIRST_ID instead of SRV_LOG_SPACE_FIRST_ID more

[Laurynas Biveinis]

Stock InnoDB uses SRV_LOG_SPACE_FIRST_ID as the upper bound for the allowed table tablespace ids. Since the separate doublewrite buffer allocates a high tablespace id as reserver for the separate buffer, it introduces SRV_EXTRA_SYS_SPACE_FIRST_ID that is less than SRV_LOG_SPACE_FIRST_ID.

But not all the code has been converted to use the proper boundary in checks:
1) fil_mutex_enter_and_prepare_for_io
2) fil_assign_new_space_id
3) fil_set_max_space_id_if_bigger - it is not clear, callers must be analyzed. Perhaps SRV_EXTRA_SYS_SPACE_FIRST_ID should be already fatal.
4) fil_create_new_single_table_tablespace
5) _fil_io
6) fil_aio_wait
7) dict_table_set_corrupt_by_space
8) buf_page_io_complete

[Laurynas Biveinis (laurynas-biveinis)]

Benign in regular operation, might become fatal or even data corrupting when actual tablespace count approaches the upper bound.

[Laurynas Biveinis (laurynas-biveinis)]

Alternative fix (for 5.6 perhaps) is to allocate low space id instead, after the undo seg space ids.

[Shahriyar Rzayev (rzayev-sehriyar)]

Percona now uses JIRA for bug reports so this bug report is migrated to: https://jira.percona.com/browse/PS-2833