Postgres Pro
Downloads
Home   >   mailing lists

Thread: Problem with locks

Problem with locks

From
Gregory Stark
Date:
I'm debugging this problem with stalled transactions waiting on locks which
have already been granted and I'm puzzled by something. What is this PGPROC
entry from?

It's not a real process (pid is 0) and it's not garbage either (prev and next
both point to a real PGPROC entry, namely MyProc which as you can see is
waiting for this PGPROC).

(gdb)  p *(PGPROC*)(ShmemBase + MyProc->links.prev)
$2 = {links = {prev = 2520941752, next = 2520941752}, sem = {semId = 1, semNum = 0}, waitStatus = 0, xid = 0, xmin = 0,
pid= 0,  databaseId = 0, roleId = 0, inCommit = 0 '\0', inVacuum = 0 '\0', isAutovacuum = 0 '\0', lwWaiting = 0 '\0',
lwExclusive= 2 '\002',  lwWaitLink = 0x0, waitLock = 0x2, waitProcLock = 0x0, waitLockMode = 0, heldLocks = 0,
myProcLocks= {{prev = 0, next = 1}, {     prev = 4294967296, next = 0}, {prev = 3028785127, next = 360654}, {prev =
73183493944770560,next = 137438953600}, {     prev = 2520929504, next = 2520929072}, {prev = 2520940536, next =
2520940536},{prev = 1, next = 0}, {prev = 0,      next = 4294967296}, {prev = 4294967296, next = 0}, {prev = 2, next =
0},{prev = 0, next = 0}, {prev = 1, next = 4294967296}, {     prev = 0, next = 4000475635}, {prev = 360707, next =
73183493944770560},{prev = 128, next = 2520933680}, {prev = 2520933680,      next = 2520914616}}, subxids = {overflowed
=-72 '�', nxids = 0, xids = {0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0,      0, 0, 1, 0, 0, 1, 0, 0,
1626942462,0, 16384, 16430, 0, 16777216, 2, 0, 2520930944, 0, 2520930944, 0, 2520914784, 0, 2520914784,      0, 0, 0,
0,1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0}}} 
(gdb) p MyProc-ShmemBase
$3 = (PGPROC *) 0xff9ac36e36e264b8
(gdb) p (unsigned long)MyProc-ShmemBase
$4 = 2520941752


--  Gregory Stark EnterpriseDB          http://www.enterprisedb.com

Re: Problem with locks

From
Alvaro Herrera
Date:
Gregory Stark wrote:
> 
> I'm debugging this problem with stalled transactions waiting on locks which
> have already been granted and I'm puzzled by something. What is this PGPROC
> entry from? 
> 
> It's not a real process (pid is 0) and it's not garbage either (prev and next
> both point to a real PGPROC entry, namely MyProc which as you can see is
> waiting for this PGPROC).

Doesn't this look like your pointer arithmetic being borked?  There are
too many values that look invalid.

> (gdb)  p *(PGPROC*)(ShmemBase + MyProc->links.prev)
> $2 = {links = {prev = 2520941752, next = 2520941752}, sem = {semId = 1, semNum = 0}, waitStatus = 0, xid = 0, xmin =
0,pid = 0, 
 
>   databaseId = 0, roleId = 0, inCommit = 0 '\0', inVacuum = 0 '\0', isAutovacuum = 0 '\0', lwWaiting = 0 '\0',
lwExclusive= 2 '\002', 
 
>   lwWaitLink = 0x0, waitLock = 0x2, waitProcLock = 0x0, waitLockMode = 0, heldLocks = 0, myProcLocks = {{prev = 0,
next= 1}, {
 
>       prev = 4294967296, next = 0}, {prev = 3028785127, next = 360654}, {prev = 73183493944770560, next =
137438953600},{
 
>       prev = 2520929504, next = 2520929072}, {prev = 2520940536, next = 2520940536}, {prev = 1, next = 0}, {prev = 0,

>       next = 4294967296}, {prev = 4294967296, next = 0}, {prev = 2, next = 0}, {prev = 0, next = 0}, {prev = 1, next
=4294967296}, {
 
>       prev = 0, next = 4000475635}, {prev = 360707, next = 73183493944770560}, {prev = 128, next = 2520933680}, {prev
=2520933680, 
 
>       next = 2520914616}}, subxids = {overflowed = -72 '�', nxids = 0, xids = {0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,
1,0, 0, 0, 0, 0, 
 
>       0, 0, 1, 0, 0, 1, 0, 0, 1626942462, 0, 16384, 16430, 0, 16777216, 2, 0, 2520930944, 0, 2520930944, 0,
2520914784,0, 2520914784, 
 
>       0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0}}}
> (gdb) p MyProc-ShmemBase
> $3 = (PGPROC *) 0xff9ac36e36e264b8
> (gdb) p (unsigned long)MyProc-ShmemBase
> $4 = 2520941752
> 
> 
> -- 
>   Gregory Stark
>   EnterpriseDB          http://www.enterprisedb.com
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Have you searched our list archives?
> 
>                http://archives.postgresql.org


-- 
Alvaro Herrera                                http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: Problem with locks

From
Tom Lane
Date:
Gregory Stark <stark@enterprisedb.com> writes:
> I'm debugging this problem with stalled transactions waiting on locks which
> have already been granted and I'm puzzled by something. What is this PGPROC
> entry from?

I seem to recall that there's a dummy sentinel header in proc lists ...
is that what you're looking at?
        regards, tom lane