Re: [PATCH v2 1/2] mm/userfaultfd: Support WP on multiple VMAs

[Peter Xu <peterx@redhat.com>]

On Mon, Feb 13, 2023 at 09:31:23PM +0500, Muhammad Usama Anjum wrote:
> mwriteprotect_range() errors out if [start, end) doesn't fall in one
> VMA. We are facing a use case where multiple VMAs are present in one
> range of interest. For example, the following pseudocode reproduces the
> error which we are trying to fix:
> 
> - Allocate memory of size 16 pages with PROT_NONE with mmap
> - Register userfaultfd
> - Change protection of the first half (1 to 8 pages) of memory to
>   PROT_READ | PROT_WRITE. This breaks the memory area in two VMAs.
> - Now UFFDIO_WRITEPROTECT_MODE_WP on the whole memory of 16 pages errors
>   out.
> 
> This is a simple use case where user may or may not know if the memory
> area has been divided into multiple VMAs.
> 
> Reported-by: Paul Gofman <pgofman@codeweavers.com>
> Signed-off-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
> ---
> Changes since v1:
> - Correct the start and ending values passed to uffd_wp_range()
> ---
>  mm/userfaultfd.c | 38 ++++++++++++++++++++++----------------
>  1 file changed, 22 insertions(+), 16 deletions(-)
> 
> diff --git a/mm/userfaultfd.c b/mm/userfaultfd.c
> index 65ad172add27..bccea08005a8 100644
> --- a/mm/userfaultfd.c
> +++ b/mm/userfaultfd.c
> @@ -738,9 +738,12 @@ int mwriteprotect_range(struct mm_struct *dst_mm, unsigned long start,
>  			unsigned long len, bool enable_wp,
>  			atomic_t *mmap_changing)
>  {
> +	unsigned long end = start + len;
> +	unsigned long _start, _end;
>  	struct vm_area_struct *dst_vma;
>  	unsigned long page_mask;
>  	int err;

I think this needs to be initialized or it can return anything when range
not mapped.

> +	VMA_ITERATOR(vmi, dst_mm, start);
>  
>  	/*
>  	 * Sanitize the command parameters:
> @@ -762,26 +765,29 @@ int mwriteprotect_range(struct mm_struct *dst_mm, unsigned long start,
>  	if (mmap_changing && atomic_read(mmap_changing))
>  		goto out_unlock;
>  
> -	err = -ENOENT;
> -	dst_vma = find_dst_vma(dst_mm, start, len);
> +	for_each_vma_range(vmi, dst_vma, end) {
> +		err = -ENOENT;
>  
> -	if (!dst_vma)
> -		goto out_unlock;
> -	if (!userfaultfd_wp(dst_vma))
> -		goto out_unlock;
> -	if (!vma_can_userfault(dst_vma, dst_vma->vm_flags))
> -		goto out_unlock;
> +		if (!dst_vma->vm_userfaultfd_ctx.ctx)
> +			break;
> +		if (!userfaultfd_wp(dst_vma))
> +			break;
> +		if (!vma_can_userfault(dst_vma, dst_vma->vm_flags))
> +			break;
>  
> -	if (is_vm_hugetlb_page(dst_vma)) {
> -		err = -EINVAL;
> -		page_mask = vma_kernel_pagesize(dst_vma) - 1;
> -		if ((start & page_mask) || (len & page_mask))
> -			goto out_unlock;
> -	}
> +		if (is_vm_hugetlb_page(dst_vma)) {
> +			err = -EINVAL;
> +			page_mask = vma_kernel_pagesize(dst_vma) - 1;
> +			if ((start & page_mask) || (len & page_mask))
> +				break;
> +		}
>  
> -	uffd_wp_range(dst_mm, dst_vma, start, len, enable_wp);
> +		_start = (dst_vma->vm_start > start) ? dst_vma->vm_start : start;
> +		_end = (dst_vma->vm_end < end) ? dst_vma->vm_end : end;
>  
> -	err = 0;
> +		uffd_wp_range(dst_mm, dst_vma, _start, _end - _start, enable_wp);
> +		err = 0;
> +	}
>  out_unlock:
>  	mmap_read_unlock(dst_mm);
>  	return err;

This whole patch also changes the abi, so I'm worried whether there can be
app that relies on the existing behavior.

Is this for the new pagemap effort?  Can this just be done in the new
interface rather than changing the old?

Side note: in your other pagemap series, you can optimize "WP_ENGAGE &&
!GET" to not do generic pgtable walk at all, but use what it does in this
patch for the initial round or wr-protect.

Thanks,

-- 
Peter Xu