Linux Device Drivers Development

Implementing mmap in the kernel

Since user space code cannot access kernel memory, the purpose of the mmap() function is to derive one or more protected kernel page table entries (which correspond to the memory to be mapped) and duplicate the user space page tables, remove the kernel flag protection, and set permission flags that will allow the user to access the same memory as the kernel without needing special privileges.

The steps to write a mmap file operation are as follows:

1. Get the mapping offset and check whether it is beyond our buffer size or not:

unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;  
if (offset >= buffer_size) 
         return -EINVAL; 

2. Check whether the mapping size is bigger than our buffer size:

unsigned long size = vma->vm_end ...