RE: asynDriver, problem with asynInt32Array read routine

["Mark Rivers" <rivers@cars.uchicago.edu>]

Hi Janez,

 

I think your problem probably does not have to do directly with asyn, but rather is something with the waveform record.  The error you are getting:

> When I run episc database in which I have some ai record type everything working fine but when I add waveform record type I get following error:

> fatal error: dbScanLock. Bad case choice.

Here is the code in EPICS base (dbLock.c) that is generating that error:

    while(1) {
        epicsMutexMustLock(lockSetModifyLock);
        plockSet = plockRecord->plockSet;
        if(!plockSet) goto getGlobalLock;
        switch(plockSet->state) {
            case lockSetStateFree:
                status = epicsMutexTryLock(plockSet->lock);
                assert(status==epicsMutexLockOK);
                plockSet->nRecursion = 1;
                plockSet->thread_id = idSelf;
                plockSet->precord = precord;
                plockSet->state = lockSetStateScanLock;
                epicsMutexUnlock(lockSetModifyLock);
                return;
            case lockSetStateScanLock:
                if(plockSet->thread_id!=idSelf) {
                    plockSet->nWaiting +=1;
                    epicsMutexUnlock(lockSetModifyLock);
                    epicsMutexMustLock(plockSet->lock);
                    epicsMutexMustLock(lockSetModifyLock);
                    plockSet->nWaiting -=1;
                    if(plockSet->state==lockSetStateRecordLock) {
                       epicsMutexUnlock(plockSet->lock);
                       goto getGlobalLock;
                    }
                    assert(plockSet->state==lockSetStateScanLock);
                    plockSet->nRecursion = 1;
                    plockSet->thread_id = idSelf;
                    plockSet->precord = precord;
                } else {
                    plockSet->nRecursion += 1;
                }
                epicsMutexUnlock(lockSetModifyLock);
                return;
            case lockSetStateRecordLock:
                /*Only recursive locking is permitted*/
                if((plockSet->nRecursion==0) || (plockSet->thread_id!=idSelf))
                    goto getGlobalLock;
                plockSet->nRecursion += 1;
                epicsMutexUnlock(lockSetModifyLock);
                return;
            default:
                cantProceed("dbScanLock. Bad case choice");
        }

So for some reason pLockSet->state has an invalid value.  I suspect there is a memory corruption problem, perhaps the waveform record support is going beyond array bounds?

Mark

---

From: Janez Golob [mailto:jgolob15@gmail.com]
Sent: Thursday, February 22, 2007 1:26 PM
To: tech-talk@aps.anl.gov
Subject: asynDriver, problem with asynInt32Array read routine

Hello,

I am writing portDriver for aio card. I read some data from asynInt32 and asynInt32Array interface type:

static asynInt32 drvAio16Int32 = {

NULL,

readInt32,

getBounds,

NULL,

NULL

};

static asynInt32Array drvAio16Int32Array = {

NULL,

readInt32Array,

NULL,

NULL

};

...

Ok interfaces inicialization and ....

...

pPvt->common.interfaceType = asynCommonType;

pPvt->common.pinterface = (void *)&aio16Common;

pPvt->common.drvPvt = pPvt;

pPvt->int32.interfaceType = asynInt32Type;

pPvt->int32.pinterface = (void *)&drvAio16Int32;

pPvt->int32.drvPvt = pPvt;

pPvt->int32Array.interfaceType = asynInt32ArrayType;

pPvt->int32Array.pinterface = (void *)&drvAio16Int32Array;

pPvt->int32Array.drvPvt = pPvt;

...

And asynInt32Array adn asynInt32 read methods are as follows

...

static asynStatus readInt32( void *drvPvt, asynUser *pasynUser, epicsInt32 *value )

{

drvAio16Pvt *pPvt = (drvAio16Pvt *)drvPvt;

rtems_unsigned8 channel;

rtems_unsigned16 data, size;

pasynManager->getAddr(pasynUser, &channel);

....

some code (calculating...)

....

*value = ( epicsInt32 )data_buffer[0];

asynPrint(pasynUser, ASYN_TRACEIO_DRIVER,

"drvAio16, port %s, read %d from channel %d\n",

pPvt->port_name, *value, channel);

return asynSucces;

}

/* Asyn Waveform read routine */

static asynStatus readInt32Array(void *drvPvt, asynUser *pasynUser, epicsInt32 *value, size_t nelements, size_t *nIn)

{

drvAio16Pvt *pPvt = (drvAio16Pvt *)drvPvt;

rtems_unsigned32 read_samples, i;

pasynManager->getAddr(pasynUser, &channel);

....

some code (calculating...)

....

return asynSucces;

}

When I run episc database in which I have some ai record type everything working fine but when I add waveform record type I get following error:

fatal error: dbScanLock. Bad case choice.

So what it's different between asynInt32 and asynInt32array read routine? Why asynInt32 read routine work but asynInt32Array read routine crash?

Episc version: 3.14.8.2

Asyn version: 4.3

Build target RTEMS-pc386

Thanks in advance!