If a message is subscribed, then unsubscribed, additional unsubscribes do not raise error events

[johnphamngc]

Describe the bug
If a message is subscribed, then unsubscribed, additional unsubscribes do not raise error events

To Reproduce
Have SB subscribe to a message
Have SB unsubscribe to that message
Have SB unsubscribe again to that message

Expected behavior
Event message should be raised

Code snips

cFE/fsw/cfe-core/src/sb/cfe_sb_api.c

Lines 1059 to 1112 in 08f6eab

	/* get index into routing table */
	MsgKey = CFE_SB_ConvertMsgIdtoMsgKey(MsgId);
	RouteIdx = CFE_SB_GetRoutingTblIdx(MsgKey);
	/* if there are no subscriptions for this message id... */
	if(!CFE_SB_IsValidRouteIdx(RouteIdx)){
	char PipeName[OS_MAX_API_NAME] = {'\0'};
	CFE_SB_UnlockSharedData(__func__,__LINE__);
	CFE_SB_GetPipeName(PipeName, sizeof(PipeName), PipeId);
	CFE_EVS_SendEventWithAppID(CFE_SB_UNSUB_NO_SUBS_EID,CFE_EVS_EventType_INFORMATION,CFE_SB.AppId,
	"Unsubscribe Err:No subs for Msg 0x%x on %s,app %s",
	(unsigned int)CFE_SB_MsgIdToValue(MsgId),
	PipeName,CFE_SB_GetAppTskName(TskId,FullName));
	return CFE_SUCCESS;
	}/* end if */
	/* At this point, there must be at least one destination. */
	/* So the value of 'ListHeadPtr' will not be NULL by design */
	RoutePtr = CFE_SB_GetRoutePtrFromIdx(RouteIdx);
	/* search the list for a matching pipe id */
	DestPtr = RoutePtr->ListHeadPtr;
	do{
	if(DestPtr->PipeId == PipeId){
	/* match found, remove node from list */
	CFE_SB_RemoveDest(RoutePtr,DestPtr);
	/* return node to memory pool */
	CFE_SB_PutDestinationBlk(DestPtr);
	RoutePtr->Destinations--;
	CFE_SB.StatTlmMsg.Payload.SubscriptionsInUse--;
	MatchFound = true;
	}/* end if */
	DestPtr = DestPtr->Next;
	}while((MatchFound == false)&&(DestPtr != NULL));
	CFE_SB_UnlockSharedData(__func__,__LINE__);
	CFE_EVS_SendEventWithAppID(CFE_SB_SUBSCRIPTION_REMOVED_EID,CFE_EVS_EventType_DEBUG,CFE_SB.AppId,
	"Subscription Removed:Msg 0x%x on pipe %d,app %s",
	(unsigned int)CFE_SB_MsgIdToValue(MsgId),
	(int)PipeId,CFE_SB_GetAppTskName(TskId,FullName));
	return CFE_SUCCESS;

If a message was previously subscribed to, and all pipes subscribing to it were subsequently unsubscribed, and an additional unsubscribe is issued, a valid routing index would still exist in the msgmap table, and thus an error event would not be raised.

System observed on:
N/A, discovered via code inspection

Additional context
N/A

Reporter Info
John N Pham, Northrop Grumman

[CDKnightNASA]

This is likely a side-effect of a fix I implemented in #101

[skliper]

Yeah, actually looks to me like this will seg fault (tries to access NULL pointer). Note I'd like to add this as a test scenario to the new functional test framework (part of subscribe/unsubscribe API testing).

[CDKnightNASA]

Yeah, actually looks to me like this will seg fault (tries to access NULL pointer). Note I'd like to add this as a test scenario to the new functional test framework (part of subscribe/unsubscribe API testing).

I'm re-designing; can you point at the line you expect a seg fault?

[skliper]

I don't think 1078 is true now, so 1083 will == NULL, and any of the following references to it would fault. Didn't test, but that's what it looks like to me upon inspection.

cFE/fsw/cfe-core/src/sb/cfe_sb_api.c

Lines 1078 to 1083 in 08f6eab

	/* At this point, there must be at least one destination. */
	/* So the value of 'ListHeadPtr' will not be NULL by design */
	RoutePtr = CFE_SB_GetRoutePtrFromIdx(RouteIdx);
	/* search the list for a matching pipe id */
	DestPtr = RoutePtr->ListHeadPtr;

[CDKnightNASA]

I don't think 1078 is true now, so 1083 will == NULL, and any of the following references to it would fault. Didn't test, but that's what it looks like to me upon inspection.

cFE/fsw/cfe-core/src/sb/cfe_sb_api.c

Lines 1078 to 1083 in 08f6eab

	/* At this point, there must be at least one destination. */
	/* So the value of 'ListHeadPtr' will not be NULL by design */
	RoutePtr = CFE_SB_GetRoutePtrFromIdx(RouteIdx);
	/* search the list for a matching pipe id */
	DestPtr = RoutePtr->ListHeadPtr;

Ah, yes...I am guessing RoutePtr will always be != NULL, but ListHeadPtr could be NULL and yes line 1087 would segfault. My redesign avoids this.