database/postgres
database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-04-25 21:42:33 +03:00
Code

Fix race leading to incorrect conflict cause in InvalidatePossiblyObsoleteSlot()

Browse Source 
The invalidation of an active slot is done in two steps:
- Termination of the backend holding it, if any.
- Report that the slot is obsolete, with a conflict cause depending on
the slot's data.

This can be racy because between these two steps the slot mutex would be
released while doing system calls, which means that the effective_xmin
and effective_catalog_xmin could advance during that time, detecting a
conflict cause different than the one originally wanted before the
process owning a slot is terminated.

Holding the mutex longer is not an option, so this commit changes the
code to record the LSNs stored in the slot during the termination of the
process owning the slot.

Bonus thanks to Alexander Lakhin for the various tests and the analysis.

Author: Bertrand Drouvot
Reviewed-by: Michael Paquier, Bharath Rupireddy
Discussion: https://postgr.es/m/ZaTjW2Xh+TQUCOH0@ip-10-97-1-34.eu-west-3.compute.internal
Backpatch-through: 16
This commit is contained in:
Michael Paquier 2024-02-20 13:43:51 +09:00
parent 01ec4d89b9
commit 818fefd8fd
1 changed files with 33 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

39
src/backend/replication/slot.c
View File

@ -1454,6 +1454,11 @@ InvalidatePossiblyObsoleteSlot(ReplicationSlotInvalidationCause cause,
{
int last_signaled_pid = 0;
bool released_lock = false;
bool terminated = false;
XLogRecPtr initial_effective_xmin = InvalidXLogRecPtr;
XLogRecPtr initial_catalog_effective_xmin = InvalidXLogRecPtr;
XLogRecPtr initial_restart_lsn = InvalidXLogRecPtr;
ReplicationSlotInvalidationCause conflict_prev PG_USED_FOR_ASSERTS_ONLY = RS_INVAL_NONE;
for (;;)
{
@ -1488,11 +1493,24 @@ InvalidatePossiblyObsoleteSlot(ReplicationSlotInvalidationCause cause,
*/
if (s->data.invalidated == RS_INVAL_NONE)
{
/*
* The slot's mutex will be released soon, and it is possible that
* those values change since the process holding the slot has been
* terminated (if any), so record them here to ensure that we
* would report the correct conflict cause.
*/
if (!terminated)
{
initial_restart_lsn = s->data.restart_lsn;
initial_effective_xmin = s->effective_xmin;
initial_catalog_effective_xmin = s->effective_catalog_xmin;
}
switch (cause)
{
case RS_INVAL_WAL_REMOVED:
if (s->data.restart_lsn != InvalidXLogRecPtr &&
s->data.restart_lsn < oldestLSN)
if (initial_restart_lsn != InvalidXLogRecPtr &&
initial_restart_lsn < oldestLSN)
conflict = cause;
break;
case RS_INVAL_HORIZON:
@ -1501,12 +1519,12 @@ InvalidatePossiblyObsoleteSlot(ReplicationSlotInvalidationCause cause,
/* invalid DB oid signals a shared relation */
if (dboid != InvalidOid && dboid != s->data.database)
break;
if (TransactionIdIsValid(s->effective_xmin) &&
TransactionIdPrecedesOrEquals(s->effective_xmin,
if (TransactionIdIsValid(initial_effective_xmin) &&
TransactionIdPrecedesOrEquals(initial_effective_xmin,
snapshotConflictHorizon))
conflict = cause;
else if (TransactionIdIsValid(s->effective_catalog_xmin) &&
TransactionIdPrecedesOrEquals(s->effective_catalog_xmin,
else if (TransactionIdIsValid(initial_catalog_effective_xmin) &&
TransactionIdPrecedesOrEquals(initial_catalog_effective_xmin,
snapshotConflictHorizon))
conflict = cause;
break;
@ -1519,6 +1537,13 @@ InvalidatePossiblyObsoleteSlot(ReplicationSlotInvalidationCause cause,
}
}
/*
* The conflict cause recorded previously should not change while the
* process owning the slot (if any) has been terminated.
*/
Assert(!(conflict_prev != RS_INVAL_NONE && terminated &&
conflict_prev != conflict));
/* if there's no conflict, we're done */
if (conflict == RS_INVAL_NONE)
{
@ -1601,6 +1626,8 @@ InvalidatePossiblyObsoleteSlot(ReplicationSlotInvalidationCause cause,
(void) kill(active_pid, SIGTERM);
last_signaled_pid = active_pid;
terminated = true;
conflict_prev = conflict;
}
/* Wait until the slot is released. */