Re: [PATCH v2 3/7] mm/mremap: introduce and use vma_remap_struct threaded state

[Lorenzo Stoakes <lorenzo.stoakes@oracle.com>]

On Mon, Mar 10, 2025 at 10:19:17AM +0000, Lorenzo Stoakes wrote:
> On Thu, Mar 06, 2025 at 10:33:59AM +0000, Lorenzo Stoakes wrote:
> > A number of mremap() calls both pass around and modify a large number of
> > parameters, making the code less readable and often repeatedly having to
> > determine things such as VMA, size delta, and more.
>
> [snip]
>
> Hi Andrew,
>
> I have a small fixpatch to address a corner case below, please apply!

Actually, will respin so I can address Vlastimil's comments to make life
easier!

>
> I have checked locally and this applies cleanly at this patch and causes no
> merge conflicts when subsequent patches are applied on top.
>
> Thanks, Lorenzo
>
> ----8<----
> From 14a22fa872e19b468b6d06c15205efe77e460ce4 Mon Sep 17 00:00:00 2001
> From: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
> Date: Mon, 10 Mar 2025 10:13:17 +0000
> Subject: [PATCH] always check if !vma
>
> We perform VMA lookup in a couple more places which, in theory, can
> unexpectedly fail to find a VMA.
>
> In these cases, correctly indicate failure.
>
> Signed-off-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
> Reported-by: kernel test robot <oliver.sang@intel.com>
> Closes: https://lore.kernel.org/oe-lkp/202503101328.442cc724-lkp@intel.com
> ---
>  mm/mremap.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/mm/mremap.c b/mm/mremap.c
> index 0ab0c88072a0..af022e3b89e2 100644
> --- a/mm/mremap.c
> +++ b/mm/mremap.c
> @@ -1066,10 +1066,13 @@ static unsigned long shrink_vma(struct vma_remap_struct *vrm,
>  	 * replace the invalidated VMA with the one that may have now been
>  	 * split.
>  	 */
> -	if (drop_lock)
> +	if (drop_lock) {
>  		vrm->mmap_locked = false;
> -	else
> +	} else {
>  		vrm->vma = vma_lookup(mm, vrm->addr);
> +		if (!vrm->vma)
> +			return -EFAULT;
> +	}
>
>  	return 0;
>  }
> @@ -1108,6 +1111,8 @@ static unsigned long mremap_to(struct vma_remap_struct *vrm)
>  		 * this can invalidate the old VMA. Reset.
>  		 */
>  		vrm->vma = vma_lookup(mm, vrm->addr);
> +		if (!vrm->vma)
> +			return -EFAULT;
>  	}
>
>  	if (vrm->remap_type == MREMAP_SHRINK) {
> --
> 2.48.1