Re: [PATCH v4] ntfs: handle a failed size rollback in the write error path

[Hyunchul Lee]

> diff --git a/fs/ntfs/file.c b/fs/ntfs/file.c
> index d810ca5904bd..ad032cccfb15 100644
> --- a/fs/ntfs/file.c
> +++ b/fs/ntfs/file.c
> @@ -934,23 +934,46 @@ ssize_t ntfs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
>  				&ntfs_iomap_folio_ops, NULL);
>  out:
>  	if (ret < 0 && ret != -EIOCBQUEUED) {
> -		if (ni->initialized_size != old_init_size) {
> -			struct ntfs_inode *mrec_ni =
> -				ntfs_base_inode(ni);
> +		int init_err = 0;
> +		int data_err = 0;
>  
> +		if (ni->data_size != old_data_size) {
> +			struct ntfs_inode *mrec_ni = ntfs_base_inode(ni);
> +
> +			truncate_setsize(vi, old_data_size);
> +			/*
> +			 * ntfs_attr_truncate() would reject a compressed
> +			 * attribute.
> +			 */
>  			mutex_lock(&mrec_ni->mrec_lock);
> -			ntfs_attr_set_initialized_size(ni, old_init_size);
> +			data_err = __ntfs_attr_truncate_vfs(ni, old_data_size,
> +					old_data_size);
>  			mutex_unlock(&mrec_ni->mrec_lock);
> +			/* Like ntfs_setattr_size(): stay on the durable size. */
> +			if (data_err)
> +				i_size_write(vi, ni->data_size);
>  		}
> -		if (ni->data_size != old_data_size) {
> -			struct ntfs_inode *mrec_ni =
> -				ntfs_base_inode(ni);
> +		/*
> +		 * The shrink above may leave the initialized size above
> +		 * the old one; ntfs_attr_set_initialized_size() rejects
> +		 * a resident attribute.
> +		 */
> +		if (ni->initialized_size != old_init_size &&
> +		    NInoNonResident(ni)) {
> +			struct ntfs_inode *mrec_ni = ntfs_base_inode(ni);
>  
> -			truncate_setsize(vi, old_data_size);
>  			mutex_lock(&mrec_ni->mrec_lock);
> -			ntfs_attr_truncate(ni, old_data_size);
> +			init_err = ntfs_attr_set_initialized_size(ni,
> +					old_init_size);

The shrink above can clamp initialzied_size to old_data_size. If
ntfs_attr_set_initialized_size() fails, the initialized_size remains
instead of being restored to old_init_size. For a compressed file,
this can expose stale data.

Could you restore the initialized_size first and perform the shink only
if that rollback succeeds?

-- 
Thanks,
Hyunchul