Iam trying to understand the mmap entry point for a driver.
This is how the mmap() is implemented
himem_buf_allocated = 0;
int xxx_mmap(struct file *filp,
struct vm_area_struct *vma)
{
unsigned long size;
char * virt_addr;
int index;
size = vma->vm_end - vma->vm_start;
if ((size % PAGE_SIZE) != 0){
size = (size / PAGE_SIZE) * PAGE_SIZE + PAGE_SIZE;
}
/* himem_buf_size is 0x80000000 */
if (size + himem_buf_allocated >= himem_buf_size){
return -ENOMEM;
}
/* himem_buf is calculated as high_memory - PAGE_OFFSET */
umem_addr = himem_buf + himem_buf_allocated;
if (umem_addr == 0){
return -ENOMEM;
}
himem_buf_allocated += size;
virt_addr = ioremap((unsigned long)umem_addr, PAGE_SIZE);
if (virt_addr == 0){
return -ENOMEM;
}
/* write the index into the first 4 bytes */
writel(index, (uint32_t *)virt_addr);
/* the values of index and *(virt_addr) do not match */
/* *(virt_addr) is always -1 */
/* Is something wrong here */
dbg_printf(0,"index is %d, *(virt_addr) is %d\n", index,
(int)readl(virt_addr));
iounmap(virt_addr);
remap_page_range(vma->vm_start, (ulong)umem_addr,
vma->vm_end - vma->vm_start, vma->vm_page_prot);
return 0;
}
Can you help me in understanding what exactly is the mmap() doing here and
if its doing it right.