semop() hanging forever on HP-UX 11.23 itanium

[Goutham YJ]

Hi All,

The semop() call is hanging forever in my code. Please find the code below.

Can anyone make out what's wrong? The same code works well on PA-RISC systems with HP-UX 11.11.


if (!getWriteLock())
{
OSCMon().outputAlarm(SIAALRMRTGTBLERR,
ALARMCRITICAL,
"Error in NpaNxxRTWriter::update: semaphore failure.");
throw exception("Error in NpaNxxRTWriter::update -> semaphore failure.");
}

---------------------------------------------


bool NpaNxxRT::getWriteLock()
{
OSCMon().debugMsg("",
OSCMONTRACEMSGLOWPRIORITY,
"NpaNxxRT: getWriteLock: lock semaphore\n");

struct sembuf sop;

sop.sem_num = 0; // the write lock is the first one
sop.sem_op = -1; // locked
sop.sem_flg = SEM_UNDO; // in case I die

cout << "Before semop" << endl;
int temp = semop(_semid, &sop, 1);
cout << "After semop=" << temp << endl;

return semop(_semid, &sop, 1) == 0;
}

[Dennis Handly]

Did you really want to call semop(2) twice in getWriteLock?

Perhaps you want: return temp >= 0;

[Goutham YJ]

Hi Dennis,

I added first semop() for debugging purposes... Plz ignore.. The actual code is

bool NpaNxxRT::getWriteLock()
{
OSCMon().debugMsg("",
OSCMONTRACEMSGLOWPRIORITY,
"NpaNxxRT: getWriteLock: lock semaphore\n");

struct sembuf sop;

sop.sem_num = 0; // the write lock is the first one
sop.sem_op = -1; // locked
sop.sem_flg = SEM_UNDO; // in case I die


return semop(_semid, &sop, 1) == 0;
}

[Dennis Handly · Accepted solution]

>return semop(_semid, &sop, 1) == 0;

semop(2) says the correct check is ">=".
Also, how are you calling semctl(2)? There is a common programming mistake you can make on Integrity.

[Goutham YJ]

The semctl() is as below:

if (semid > 0)
{
semctl(semid, 0, SETVAL, 1); // start out unlocked
semctl(semid, 1, SETVAL, 0); // shared mem not set
return semid;
}

We are not using the semun union.. Is that causing the issue?

[Goutham YJ]

Also semop() call is hanging and not returning anything. So ">=" check is not getting performed.

[Dennis Handly]

>We are not using the semun union. Is that causing the issue?

Exactly, you must use that union. The IPF calling conventions require strict adherence to what's documented.

[Goutham YJ]

Thanks a lot dennis. You are great help. The problem is resolved after using the semun union for semctl() system calls.

Thanks,
Goutham

[Goutham YJ]

Problem resolved as explained above.