Re: [PATCH linux-next] mm/folio-compat: fix potential NULL pointer access

[Matthew Wilcox <willy@infradead.org>]

On Mon, Sep 13, 2021 at 08:11:13AM +0000, cgel.zte@gmail.com wrote:
> From: xu xin <xu.xin16@zte.com.cn>
> 
> The pointer 'folio' might be NULL, but the structure it points to is accessed.
> Accordingly, we add a check of NULL pointer by 'if (!folio)'.
> Secondly, there is no need to check if folio is pointer or value with 'xa_is_value(folio)' 
> because folio is alwayse pointer.
> 
> Reported-by: Zeal Robot <zealci@zte.com.cn>

Your robot is garbage.

>  	folio = __filemap_get_folio(mapping, index, fgp_flags, gfp);
> -	if ((fgp_flags & FGP_HEAD) || !folio || xa_is_value(folio))
> +	if (!folio)
> +		return NULL;
> +	if ((fgp_flags & FGP_HEAD))
>  		return &folio->page;

&folio->page *does not access folio*.  It does pointer arithmetic on
folio.  Specifically, it adds zero to folio, because page is the first
element of folio.

Secondly, __filemap_get_folio() absolutely can return an xa_is_value()
result if FGP_ENTRY entry is set.  It's right there in the first few
lines:

        folio = mapping_get_entry(mapping, index);
        if (xa_is_value(folio)) {
                if (fgp_flags & FGP_ENTRY)
                        return folio;