Re: 230-objrmap fixes for 2.6.3-mjb2

[Andrea Arcangeli <andrea@suse.de>]

On Wed, Mar 03, 2004 at 08:09:33AM +0100, Andrea Arcangeli wrote:
> --- sles-objrmap/mm/mmap.c.~1~	2004-03-03 06:45:38.980596736 +0100
> +++ sles-objrmap/mm/mmap.c	2004-03-03 06:53:46.945414808 +0100
> @@ -1284,8 +1284,8 @@ int do_munmap(struct mm_struct *mm, unsi
>  	/*
>  	 * Remove the vma's, and unmap the actual pages
>  	 */
> -	detach_vmas_to_be_unmapped(mm, mpnt, prev, end);
>  	spin_lock(&mm->page_table_lock);
> +	detach_vmas_to_be_unmapped(mm, mpnt, prev, end);
>  	unmap_region(mm, mpnt, prev, start, end);
>  	spin_unlock(&mm->page_table_lock);
>  
> --- sles-objrmap/mm/swapfile.c.~1~	2004-03-03 06:45:39.023590200 +0100
> +++ sles-objrmap/mm/swapfile.c	2004-03-03 07:03:33.128301464 +0100
> @@ -499,7 +499,6 @@ static int unuse_process(struct mm_struc
>  	/*
>  	 * Go through process' page directory.
>  	 */
> -	down_read(&mm->mmap_sem);
>  	spin_lock(&mm->page_table_lock);
>  	for (vma = mm->mmap; vma; vma = vma->vm_next) {
>  		pgd_t * pgd = pgd_offset(mm, vma->vm_start);
> @@ -507,7 +506,6 @@ static int unuse_process(struct mm_struc
>  			break;
>  	}
>  	spin_unlock(&mm->page_table_lock);
> -	up_read(&mm->mmap_sem);
>  	pte_chain_free(pte_chain);
>  	return 0;
>  }

Martin, I just found I was wrong about the above, I would been right in
the 2.4 VM, but in 2.6 the above is not needed. So you can delete the
above part from your tree too. it seems 2.6 really splitted the
page_table_lock from the mmap_sem and the only way to touch vmas is to
get the mmap_sem, the page_table_lock is unrelated to the vmas. This
wasn't the case in 2.4 where readers were allowed to take the
page_table_lock only.