Invalid result from hash_page_items function - Mailing list pgsql-hackers

From Masahiko Sawada
Subject Invalid result from hash_page_items function
Date
Msg-id CAD21AoBc5Asx7pXdUWu6NqU_g=Ysn95EGL9SMeYhLLduYoO_OA@mail.gmail.com
Whole thread Raw
Responses Re: Invalid result from hash_page_items function
List pgsql-hackers
Hi,

While researching hash index, I found that hash_page_items could
return an invalid result as follows.

postgres(1:1056)=# create table hash_test (c int);
CREATE TABLE
postgres(1:1056)=# insert into hash_test select generate_series(1,50) % 5;
INSERT 0 50
postgres(1:1056)=# create index hash_idx on hash_test using hash (c);
CREATE INDEX
postgres(1:1056)=# create extension pageinspect ;
CREATE EXTENSION
postgres(1:1056)=# select * from hash_page_items(get_raw_page('hash_idx', 4));
itemoffset | ctid | data
------------+--------------------+------------
1 | (0,49) | 3283889963
2 | (2139062143,32639) | 2139062143
3 | (2139062143,32639) | 2139062143
4 | (2139062143,32639) | 2139062143
5 | (2139062143,32639) | 2139062143
6 | (2139062143,32639) | 2139062143
7 | (2139062143,32639) | 2139062143
8 | (2139062143,32639) | 2139062143
9 | (2139062143,32639) | 2139062143
10 | (2139062143,32639) | 2139062143
(10 rows)

This appears at PostgreSQL 10 and current HEAD. The cause of this
seems that hash_page_items allocates the memory space for the page
before switching memory context. AFAICS there is no similar problem in
pageinspect contrib module. Attached patch fixes it.

Regards,

--
Masahiko Sawada
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center
Attachment

pgsql-hackers by date:

Previous
From: Bruce Momjian
Date:
Subject: Re: After dropping the rule - Not able to insert / server crash (onetime ONLY)
Next
From: Alvaro Herrera
Date:
Subject: Re: [HACKERS] Patch: Add --no-comments to skip COMMENTs with pg_dump