Hi!
On Sep 06, Sergey Vojtovich wrote:
> ChangeSet@stripped, 2007-09-06 17:05:07+05:00, svoj@stripped +3 -0
> BUG#30583 - Partition on DOUBLE key + INNODB + count(*) == crash
>
> Issuing SELECT COUNT(*) against partitioned InnoDB table may cause
> server crash.
>
> Fixed that not all required fields were included into read_set.
>
> diff -Nrup a/sql/ha_partition.cc b/sql/ha_partition.cc
> --- a/sql/ha_partition.cc 2007-08-02 09:50:52 +05:00
> +++ b/sql/ha_partition.cc 2007-09-06 17:05:05 +05:00
> @@ -3324,6 +3324,24 @@ int ha_partition::index_init(uint inx, b
> */
> if (m_lock_type == F_WRLCK)
> bitmap_union(table->read_set, &m_part_info->full_part_field_set);
> + else if (!(m_table_flags & HA_STATS_RECORDS_IS_EXACT) &&
> + m_table_flags & HA_PARTIAL_COLUMN_READ)
1. Why HA_STATS_RECORDS_IS_EXACT ? Your test case uses COUNT(*) but I
doubt it's the only scenario when ha_partition will decide to do an
ordered scan and read_set will be lacking necessary columns.
2. You need to check for 'sorted' - if it's false, ha_partition will do
unordered scan, and as far as I understand there will be no need to
modify read_set.
> + {
> + /*
> + We may get empty read_set for engines that support partial column
> + read and ::info doesn't return exact row count. This may happen
> + e.g. with SELECT COUNT(*) FROM t1. We must ensure that all columns
> + of current key are included into read_set, as partitioning requires
> + them for sorting (see ha_partition::handle_ordered_index_scan).
Too specific. It should not mention COUNT(*) and ::info(), but describe
a generic situation - a ordered scan is requested and necessary fields
aren't in read_set (you can add "for example this can happen in COUNT(*)
queries, see the test for bug#...").
> + TODO: optimize this workaround away from here, as we actually do not
> + need to perform ordered index scan in this case.
Wrong comment. I agree that COUNT(*) should be optimized not to request
ordered scan, but this fix should not be removed anyway as - see above -
there could be other cases when an ordered scan is requested and
necessary fields aren't in read_set.
> + */
> + uint i;
> + for (i= 0; i < m_curr_key_info->key_parts; i++)
> + bitmap_set_bit(table->read_set,
> + m_curr_key_info->key_part[i].field->field_index);
> + }
> file= m_file;
> do
> {
>
Regards / Mit vielen Grüssen,
Sergei
--
__ ___ ___ ____ __
/ |/ /_ __/ __/ __ \/ / Sergei Golubchik <serg@stripped>
/ /|_/ / // /\ \/ /_/ / /__ Principal Software Developer
/_/ /_/\_, /___/\___\_\___/ MySQL GmbH, Dachauer Str. 37, D-80335 München
<___/ Geschäftsführer: Kaj Arnö - HRB
München 162140
