[PATCH] erofs: fix a potential NULL dereference of alloc_pages()

xkernel.wang@foxmail.com posted 1 patch 4 years, 3 months ago
fs/erofs/zdata.c | 14 +++++++++-----
1 file changed, 9 insertions(+), 5 deletions(-)
[PATCH] erofs: fix a potential NULL dereference of alloc_pages()
Posted by xkernel.wang@foxmail.com 4 years, 3 months ago
From: Xiaoke Wang <xkernel.wang@foxmail.com>

alloc_pages() returns the page on success or NULL if allocation fails,
while set_page_private() will dereference `newpage`. So it is better to
catch the memory error in case other errors happen.

Signed-off-by: Xiaoke Wang <xkernel.wang@foxmail.com>
---
 fs/erofs/zdata.c | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 11c7a1a..36a5421 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -735,11 +735,15 @@ static int z_erofs_do_read_page(struct z_erofs_decompress_frontend *fe,
 		struct page *const newpage =
 				alloc_page(GFP_NOFS | __GFP_NOFAIL);
 
-		set_page_private(newpage, Z_EROFS_SHORTLIVED_PAGE);
-		err = z_erofs_attach_page(clt, newpage,
-					  Z_EROFS_PAGE_TYPE_EXCLUSIVE);
-		if (!err)
-			goto retry;
+		if (!newpage) {
+			err = -ENOMEM;
+		} else {
+			set_page_private(newpage, Z_EROFS_SHORTLIVED_PAGE);
+			err = z_erofs_attach_page(clt, newpage,
+						Z_EROFS_PAGE_TYPE_EXCLUSIVE);
+			if (!err)
+				goto retry;
+		}
 	}
 
 	if (err)
--
Re: [PATCH] erofs: fix a potential NULL dereference of alloc_pages()
Posted by Gao Xiang 4 years, 3 months ago
On Tue, Mar 22, 2022 at 04:08:12PM +0800, xkernel.wang@foxmail.com wrote:
> From: Xiaoke Wang <xkernel.wang@foxmail.com>
> 
> alloc_pages() returns the page on success or NULL if allocation fails,
> while set_page_private() will dereference `newpage`. So it is better to
> catch the memory error in case other errors happen.
> 
> Signed-off-by: Xiaoke Wang <xkernel.wang@foxmail.com>
> ---
>  fs/erofs/zdata.c | 14 +++++++++-----
>  1 file changed, 9 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
> index 11c7a1a..36a5421 100644
> --- a/fs/erofs/zdata.c
> +++ b/fs/erofs/zdata.c
> @@ -735,11 +735,15 @@ static int z_erofs_do_read_page(struct z_erofs_decompress_frontend *fe,
>  		struct page *const newpage =
>  				alloc_page(GFP_NOFS | __GFP_NOFAIL);
>  

It's really a nofail allocation, am I missing something?

Thanks,
Gao Xiang

> -		set_page_private(newpage, Z_EROFS_SHORTLIVED_PAGE);
> -		err = z_erofs_attach_page(clt, newpage,
> -					  Z_EROFS_PAGE_TYPE_EXCLUSIVE);
> -		if (!err)
> -			goto retry;
> +		if (!newpage) {
> +			err = -ENOMEM;
> +		} else {
> +			set_page_private(newpage, Z_EROFS_SHORTLIVED_PAGE);
> +			err = z_erofs_attach_page(clt, newpage,
> +						Z_EROFS_PAGE_TYPE_EXCLUSIVE);
> +			if (!err)
> +				goto retry;
> +		}
>  	}
>  
>  	if (err)
> --
Re: [PATCH] erofs: fix a potential NULL dereference of alloc_pages()
Posted by Xiaoke Wang 4 years, 3 months ago
On Tue, 22 Mar 2022 16:28:09 +0800, hsiangkao@linux.alibaba.com wrote:

&gt; It's really a nofail allocation, am I missing something?

Sorry, I did not carefully notice the flag `__GFP_NOFAIL`.
Please ignore this patch if this does not need to deal with failure.

Regards,
Xiaoke Wang