Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 14 Jul 2022 18:04:18 -0700
From:   Andrew Morton <akpm@linux-foundation.org>
To:     Charan Teja Kalla <quic_charante@quicinc.com>
Cc:     <pasha.tatashin@soleen.com>, <sjpark@amazon.de>,
        <sieberf@amazon.com>, <shakeelb@google.com>, <dhowells@redhat.com>,
        <willy@infradead.org>, <mhocko@suse.com>, <vbabka@suse.cz>,
        <david@redhat.com>, <minchan@kernel.org>,
        <linux-kernel@vger.kernel.org>, <linux-mm@kvack.org>
Subject: Re: [PATCH] mm: fix use-after free of page_ext after race with
 memory-offline
Message-Id: <20220714180418.6d546650b3e5ae745f09814d@linux-foundation.org>
In-Reply-To: <1657810063-28938-1-git-send-email-quic_charante@quicinc.com>
References: <1657810063-28938-1-git-send-email-quic_charante@quicinc.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Thu, 14 Jul 2022 20:17:43 +0530 Charan Teja Kalla <quic_charante@quicinc.com> wrote:

> The below is one path where race between page_ext and  offline of the
> respective memory blocks will cause use-after-free on the access of
> page_ext structure.
> 
> ...
>
> --- a/include/linux/page_ext.h
> +++ b/include/linux/page_ext.h
> @@ -64,6 +64,25 @@ static inline struct page_ext *page_ext_next(struct page_ext *curr)
>  	return next;
>  }
>  
> +static inline struct page_ext *get_page_ext(struct page *page)
> +{
> +	struct page_ext *page_ext;
> +
> +	rcu_read_lock();

If page_ext.h is to call rcu functions then it will need to include
appropriate header files.

> +	page_ext = lookup_page_ext(page);
> +	if (!page_ext) {
> +		rcu_read_unlock();
> +		return NULL;
> +	}
> +
> +	return page_ext;
> +}
> +
> +static inline void put_page_ext(void)
> +{
> +	rcu_read_unlock();
> +}

Better names would be page_ext_get() and page_ext_put().  The rest of
the page_ext API appears to have got this right, so let's not mess that
up.

Also, these aren't really get and put functions - page_ext doesn't have
a refcount.  But I can't immediately think of a name that better
communicates what we're doing here so I guess get and put will do.

And are we able to add some comments here explaining why these
functions exist and what they do?

>  #else /* !CONFIG_PAGE_EXTENSION */
>  struct page_ext;

Are you sure we didn't need CONFIG_PAGE_EXTENSION=n stubs for these two
functions?