Re: Page Physical Address and pte

["Kuas (gmane)" <ku4s@xxxxxxxxxxxxxxxxxxxxx>]

> > But I am in a kernel mode. How do I use MMU? And I want to scan the 
> > whole pages not just accessing a particular address.
>
> you might want to check vmscan to see how walkthrough of pages are done.
> i believe what he meant is to walk through/scan the linear/virtual
> addresses of the pages. that is how you access those pages.

I look in the file, it is doing swapping and scanning through pages but 
does not look into the bytes in the page. It involves moving pages to 
different page pools. What I need is to read the frame bytes of a page 
that I have the pte_t and page struct.

> you don't intentionally "use" MMU. the kernel does that for you. you
> just use the kernel api when necessary.

Right, I am already in the kernel, particularly in: mm/memory.c at 
do_no_page() function which reads the block of file and put it into the 
the frame which corresponds to a new_page. So now I have the pte_t of 
the new page and the page struct of the new pages. Since I'm already in 
kernel mode, I think I have the physical address of the fram from the 
pte_t, am I correct?

The pte_t if I'm not mistaken is the physical base address of the page, 
so I just have to run a pointer starting from that address to PAGE_SIZE. 
Can I do that? Or I have to call read syscall with fd pointing to 
/dev/kmem? Can I make a system call in the kernel mode?

I am confused of what kind of address pte_t is, and if the kernel 
operates in physical address or still virtual address. I just want to 
read the bytes in that page frame memory.

> the size of a x86 page is 4kb. whether it is filled or not, depends on
> the type of page it is. if it is a file/device backed page, then it is
> usually filled with driver data. if the page is an anonymous page, and 
> you are writing to it, then the page is allocated, zeroed, before it is 
> used/reclaimed to be used.

Right now the page that I have is from a block device (disk file 
system). I believe the interesting data from file is brought in by 
blocks (a multiple of 512 bytes frame). Which means the page might only 
have a few frames, for example the page that is at the end of the file. 
How can I detect this? That's why I am thinking if I have to stop after 
the end of  the data and not to read more eventhough it's not the end of 
the page.

Thanks a lot for the info.

Kuas


--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive:       http://mail.nl.linux.org/kernelnewbies/
FAQ:           http://kernelnewbies.org/faq/