Re: Tracking adresses of write and read

[Sourav Sen <sourav@csa.iisc.ernet.in>]

> That's what i'm trying to see: Where does the kernel physically write
> when my program write in the mmaped area. I don't want the user space program
> to handle physical adresses. It just care about it for debugging
> purposes, I just want the physical adres as information.
	
	In your program u are not write() ing but storing some value in
those mapped addresses, isn't it? So kernel does not come into picture
there unless u fault (i.e., page fault, then kernel sees that address and
tries to recover etc. etc..). 

	You can only get the physical address of the pages if u r inside
kernel. U may perhaps try implementing a syscall that does that and
returns the physical address. In that syscall u have to do a page table
lookup to get the address(the user address u supply should be page
address). In 2.4 one such function is given, I think, the following one is
similar  to that :

/* returns the page address of a page with u_addr = user v_addr
 */

inline unsigned long get_page_addr(unsigned long u_addr)
{
	pgd_t *u_pgd;
	pmd_t *u_pmd;
	pte_t *u_pte;

	u_pgd = pgd_offset(current->mm, u_addr);
	u_pmd = pmd_offset(u_pgd, u_addr);
	if (u_pmd) {
		u_pte = pte_offset(u_pmd, u_addr);
		if (u_pte && pte_present(*u_pte))
			return pte_page(*u_pte);
	}
	return 0;
}

The address returns is page address and then u call virt_to_phys on it
to get the physical address. You can't call  virt_to_phys() on a user
address.

HTH
sourav

-
Kernelnewbies: Help each other learn about the Linux kernel.
Archive:       http://mail.nl.linux.org/
IRC Channel:   irc.openprojects.net / #kernelnewbies
Web Page:      http://www.kernelnewbies.org/