Question about swap_in() in 2.2.16 ....

Question about swap_in() in 2.2.16 ....

[Shuvabrata Ganguly]

hi,

after the missing page has been swapped in this bit of code is
executed:-

if (!write_access || is_page_shared(page_map)) {
      set_pte(page_table, mk_pte(page, vma->vm_page_prot));
      return 1;
 }

Now this creates a read-only mapping  even if the access was a "write
acess"  ( if the page is shared ). Doesnt this mean that an additional
"write-protect" fault will be taken immediately when the process tries
to write again ? Or am i missing something here ?

joy

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

Re: Question about swap_in() in 2.2.16 ....

[Stephen C. Tweedie]

Hi,

On Wed, Nov 08, 2000 at 01:32:26AM -0500, Shuvabrata Ganguly wrote:
> 
> after the missing page has been swapped in this bit of code is
> executed:-
> 
> if (!write_access || is_page_shared(page_map)) {
>       set_pte(page_table, mk_pte(page, vma->vm_page_prot));
>       return 1;
>  }
> 
> Now this creates a read-only mapping  even if the access was a "write
> acess"  ( if the page is shared ). Doesnt this mean that an additional
> "write-protect" fault will be taken immediately when the process tries
> to write again ?

Yes.

Cheers,
 Stephen
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

Re: Question about swap_in() in 2.2.16 ....

[Shuvabrata Ganguly]

"Stephen C. Tweedie" wrote:

> Hi,
>
> On Wed, Nov 08, 2000 at 01:32:26AM -0500, Shuvabrata Ganguly wrote:
> >
> > after the missing page has been swapped in this bit of code is
> > executed:-
> >
> > if (!write_access || is_page_shared(page_map)) {
> >       set_pte(page_table, mk_pte(page, vma->vm_page_prot));
> >       return 1;
> >  }
> >
> > Now this creates a read-only mapping  even if the access was a "write
> > acess"  ( if the page is shared ). Doesnt this mean that an additional
> > "write-protect" fault will be taken immediately when the process tries
> > to write again ?
>
> Yes.
>

Then why dont we give it a private page in the first place ?

Cheers
Joy

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/

Re: Question about swap_in() in 2.2.16 ....

[Stephen C. Tweedie]

Hi,

On Thu, Nov 09, 2000 at 11:23:05PM -0500, Shuvabrata Ganguly wrote:
> "Stephen C. Tweedie" wrote:
> > > Now this creates a read-only mapping  even if the access was a "write
> > > acess"  ( if the page is shared ). Doesnt this mean that an additional
> > > "write-protect" fault will be taken immediately when the process tries
> > > to write again ?
> >
> > Yes.
> 
> Then why dont we give it a private page in the first place ?

Normal copy-on-write is an extremely performance-critical code path.
It's really not worth the trouble of adding extra code to it to make
the swapin page fault do the same copy-on-write immediately, because
swapin simply is not that important for performance.

--Stephen
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux.eu.org/Linux-MM/