OSDev.org

Posted: **Thu Apr 09, 2009 4:44 am**

I know that JamesM has said his heap isn't the best around and he has written a better one, but still I wanted to ask you a question regarding it. There's something in the code that confuses me and I've only noticed it after countless hours of debugging my code. While testing my memory, I noticed that after a while I get a page-fault. It occurs after the heap starts expanding itself because it's too small. What I've noticed in the code is this (my version of the code):

Code: Select all

void *HeapAlloc(unsigned size, bool pageAlign, Heap *heap)
{
	/* Size including the header and footer. */
	unsigned RealSize = size + sizeof(HeapHeader) + sizeof(HeapFooter);

	int i = GetSmallestHole(RealSize, pageAlign, heap);

	/* No hole found? */
	if(i == -1)
	{
		unsigned OldLength = heap->EndAddress - heap->StartAddress;
		unsigned OldEndAddress = heap->EndAddress;

		/* Make the heap larger, return 0 if failed to prevent going past the maximum address. */
		if(ExpandHeap(heap, OldLength + RealSize) == -1)
			return 0;

        //...

And so on. Okay, the heap needs to expand itself and that's what the call there does. But if we enter the function that expands the heap:

Code: Select all

static int ExpandHeap(Heap *heap, unsigned newSize)
{
   //...
   
    /* Make sure there are pages present at the new addresses (start at the heap ending address). */
    for(unsigned i(heap->EndAddress); i < heap->StartAddress + newSize; i += HEAP_PAGESIZE)
	AllocateFrame(GetPage(i, true, KernelDir), heap->Supervisor? 1 : 0, heap->ReadOnly? 0 : 1);

   //...
}

Okay, so we need to make sure the expansion to the heap is also page-aligned. But what does 'GetPage' do when it needs to create a page?

Code: Select all

directory->PageTables[TableID] = (PageTable *) MM->AllocAlignGetPhys(sizeof(PageTable), &PhysAddr);

It calls allocate all over again! Allocation will go through the memory manager and call HeapAlloc to get that memory (and because the heap has already been created, it won't use a placement address but the HeapAlloc function). The call to HeapAlloc will result in the heap trying to expand itself again, AGAIN going through the ExpandHeap function, GetPage, MM->AllocAlignGetPhys, ...

Am I making a mistake somewhere, or is this an infinite loop? Setting debug log prints only acknowledges what I'm thinking.

Thanks for your advice,
Creature

OSDev.org

Confusion Regarding JamesM's Heap

Confusion Regarding JamesM's Heap