[
https://issues.apache.org/jira/browse/HAWQ-1722?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ruilong Huo updated HAWQ-1722:
------------------------------
Description:
since the lock is not released before reporting errors, it will leads to panic
during transaction abort since the lock will be acquired again. A RWLock is
acquired twice in one process will leads to panic.
There are two occurrences of this bug in this function, one for AO, one for
parquet.
{code:c}
if (id == NEXT_END_OF_LIST)
{
pfree(allfsinfoParquet);
ereport(ERROR, (errmsg("cannot open more than
%d "
"append-only table segment "
"files cocurrently",
MaxAORelSegFileStatus)));
return false;
}
{code}
was:
since the lock is not released before reporting errors, it will leads to panic
during transaction abort since the lock will be acquired again. A RWLock is
acquired twice in one process will leads to panic.
There are two occurrences of this bug in this function, one for AO, one for
parquet.
> Core dump due to lock is not released before reporting errors when exceeding
> MaxAORelSegFileStatus
> --------------------------------------------------------------------------------------------------
>
> Key: HAWQ-1722
> URL: https://issues.apache.org/jira/browse/HAWQ-1722
> Project: Apache HAWQ
> Issue Type: Bug
> Components: Core
> Affects Versions: 2.4.0.0
> Reporter: Ruilong Huo
> Assignee: Ruilong Huo
> Priority: Major
>
> since the lock is not released before reporting errors, it will leads to
> panic during transaction abort since the lock will be acquired again. A
> RWLock is acquired twice in one process will leads to panic.
> There are two occurrences of this bug in this function, one for AO, one for
> parquet.
> {code:c}
> if (id == NEXT_END_OF_LIST)
> {
> pfree(allfsinfoParquet);
> ereport(ERROR, (errmsg("cannot open more than
> %d "
> "append-only table segment "
> "files cocurrently",
> MaxAORelSegFileStatus)));
> return false;
> }
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
