elf_map: why offset of vma need to subtract eppnt->p_vaddr ?

Prabhu nath gprabhunath at gmail.com
Fri Mar 29 04:56:29 EDT 2013


On Fri, Mar 29, 2013 at 12:45 PM, Jacky <jackyclivia at 163.com> wrote:

>  Hi All,
>
> When mmaping elf image into memory, why offset vma need to subtract
> eppnt->p_vaddr as the following code ?
>
> static unsigned long elf_map()
> {
>     ...
>     unsigned long off = eppnt->p_offset - ELF_PAGEOFFSET(eppnt->p_vaddr);
>     ...
> }
>
      Assume you have logically divided the contents of the ELF file into
chunks of PAGE_SIZE (Typically 4K on x86). Let us name each chunk as *file
page*-synonym to *page frame* in the physical address space and *page* in
the virtual address space respectively.
Here variable *off* is the *file page* base offset - synonym to page base
address or page frame base address which are typically multiple of
PAGE_SIZE.

      off >> PAGE_SHIFT is the value stored in *vm_pgoff* in struct
vm_area_struct
In a nutshell vm_pgoff is the page number (*file page *number) in the ELF
file where the corresponding PT_LOAD segment starts.

Suppose if a file offset (eppnt->p_offset) of a PT_LOAD segment is 9560
(0x2558) bytes into the file and eppnt->p_vaddr is 0x08048558. Then the
file page number base address *off* is calculated as

               off = 0x2558 - 0x558; // ELF_PAGEOFFSET(eppnt->p_vaddr) will
expand to 0x558.
Then off is 0x2000 and vm_pgoff is (off >> PAGE_SHIFT) = 2. This means that
the PT_LOAD segment starts at 2nd page or 2nd file page in the ELF file.


Thanks in advance.
>
> Jacky
>

-- 
Regards,
Prabhunath G
Linux Trainer
Bangalore
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20130329/630ae2ff/attachment.html 


More information about the Kernelnewbies mailing list