Add additional sanity checks when reading a blkreftable.
Code elsewhere in the system assumes that fork numbers and chunk sizes
are within bounds, so the code that reads those quantities from disk
should validate that they are. Without these additional checks, a
corrupted file can cause us to index off the end of fork number or chunk
entry arrays, potentially resulting in a crash.
Reported-by: oxsignal <awo@kakao.com> (chunk sizes)
Reported-by: Robert Haas <rhaas@postgresql.org> (fork numbers)
Reviewed-by: Daniel Gustafsson <daniel@yesql.se>
Discussion: http://postgr.es/m/CA+TgmoYP8RKoBGosS7C6Fdr-GNCfyz_W1zmK=Tx1Fe0ZvzGh0g@mail.gmail.com
Backpatch-through: 17
Branch
------
REL_17_STABLE
Details
-------
https://git.postgresql.org/pg/commitdiff/0c5516a0df97e1e707ea5e9e08712e29ce4ee047
Modified Files
--------------
src/common/blkreftable.c | 22 ++++++++++++++++++++++
1 file changed, 22 insertions(+)
