On Sun, 27 Jul 2003, Eugene Teo wrote:
> <quote sender="Raghu R. Arur">
> > On Sat, 26 Jul 2003, Eugene Teo wrote:
> > > With reference to 2.4.20, I have a few questions:
> > >
> > > [1] I was looking at mm/memory.c. I noticed that there is a
> > > difference between minor, and major faults. My guess is that
> > > when a major fault occurs, the mm performs a page-in from the
> > > swap to the memory, whilst a minor fault doesn't? No?
> >
> > not necessarily from swap area. it can read a page from the disk
> > when faults on a memory mapped file's address. So if u have to wait for a
> > disk operation when you fault then it is a major fault else if the fault
> > is just because of protection reasons or something else which does not
> > require to wait for a disk operation then it is a minor fault.
>
> I don't really understand the part about memory-mapped files. When will
> it be used?
>
The simplest example is the code of the program. The object code
on the hard disk is mapped to the process address space.
> > > [2] I understand that for the handle_pte_fault routine, the
> > > if structure basically handles page-in. I am trying to figure
> > > out where I can find the code for page-out which I understand
> > > it can be involved the swap, etc. Can someone give me some
> > > pointers as to where I can look for:
> > >
> > > - what causes page-outs,
> > > - where in the kernel can i look for them?
> > >
> > look at try_to_free_pages in vmscan.. it will tell when a page is
> > swapped out. the whole of vmscan.c tells you about swapping out.
>
> oh, the code that does the page eviction. i also noticed that there
> is some code in the block.c that does page in and out. i point it to
> you when i get to copy and paste in X.
>
I think block.c does the actual read and write from the disk. This
is not necessarily for page faults only. when u have to swap in and swap
out u use the VFS routines that does read and write. So these are
implemented in block.c
> > > [3] in mm/memory.c, in do_wp_page, I am not sure what the
> > > portion of code is about:
> > >
> > > // If old_page bit is not set, set it, and test.
> > > if (!TryLockPage(old_page) {
> > >
> > > // [QN:] I don't understand what can_share_swap_page() do
> > > // I tried tracing, but i still don't quite get it.
> > > int reuse = can_share_swap_page(old_page);
> >
> > can_share_swap_page checks whether the page is in the swap cache
> > and is it an exclusive swap page ie no other process is sharing it. if it
> > is so then it establishes the pte entry.
>
> 1) can i say that like shared memory, swap pages can be swapped? or is
> it because it is shared memory among processes, and that pages happened
> to be in the swap cache, so checking if those pages are exclusive swap
> pages will fail because it is shared memory, and processes are sharing
> it?
>
think u are right....
> 2) establishes the pte entry - i noticed that it flushes the tlb with
> the previous old pte entry. set_pte() simples assigns the new entry to
> the page table?
>
yes it does flush the tlb entry so that the new mapping is visible
to the CPU now.
> 3) any idea why this is done? check it is exclusive swap page, and if it
> is in the swap cache, then establish the pte entry, then return to the
> caller. Any idea why mm->rss is not increased? oh, is it because there
> is no new pages page-in, and out, and it is merely a replacement of the
> old page with a new page?
>
if the page is an exclusive swap page then it is the only process
that is using the page that is on the swap cache. so it can get that page
and add it to its address space. if it is a shared page it cannot do as
other processes' pte entries will be pointing to the swap cache.
am not sure why the rss value is not increased over here...
please correct me if I am wrong anywhere...
HTH,
Raghu
> >
> > --
> > Kernelnewbies: Help each other learn about the Linux kernel.
> > Archive: http://mail.nl.linux.org/kernelnewbies/
> > FAQ: http://kernelnewbies.org/faq/
> >
>
--
Kernelnewbies: Help each other learn about the Linux kernel.
Archive: http://mail.nl.linux.org/kernelnewbies/
FAQ: http://kernelnewbies.org/faq/
