udf: KASAN: slab-out-of-bounds in udf_readdir

[PATCH] udf: KASAN: slab-out-of-bounds in udf_readdir

Posted by Vladislav Efanov 2 years, 7 months ago

The KASAN report is:
[ 1922.586560] BUG: KASAN: slab-out-of-bounds in udf_readdir+0xe00/0x19e0
[ 1922.586922] Write of size 89 at addr ffff888000cd9ea6 by task rm/18493

udf_readdir() tries to write file name out of allocated memory
buffer bounds. The UDF_NAME_LEN_CS0 (255) is used as max length
for file name in udf_put_filename(). But UDF_NAME_LEN (254) is
used as the size for buffer allocation in udf_readdir(). As the
result out-of-bounds write happened.

Found by Linux Verification Center (linuxtesting.org) with xfstests

Fixes: 066b9cded00b ("udf: Use separate buffer for copying split names")
Signed-off-by: Vladislav Efanov <VEfanov@ispras.ru>
---
 fs/udf/dir.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/udf/dir.c b/fs/udf/dir.c
index be640f4b2f2c..b95607c42ad4 100644
--- a/fs/udf/dir.c
+++ b/fs/udf/dir.c
@@ -169,7 +169,7 @@ static int udf_readdir(struct file *file, struct dir_context *ctx)
 				nameptr = (char *)(fibh.ebh->b_data + poffset - lfi);
 			} else {
 				if (!copy_name) {
-					copy_name = kmalloc(UDF_NAME_LEN,
+					copy_name = kmalloc(UDF_NAME_LEN_CS0,
 							    GFP_NOFS);
 					if (!copy_name) {
 						ret = -ENOMEM;
-- 
2.34.1

Re: [PATCH] udf: KASAN: slab-out-of-bounds in udf_readdir

Posted by Jan Kara 2 years, 7 months ago

On Mon 06-02-23 19:22:06, Vladislav Efanov wrote:
> The KASAN report is:
> [ 1922.586560] BUG: KASAN: slab-out-of-bounds in udf_readdir+0xe00/0x19e0
> [ 1922.586922] Write of size 89 at addr ffff888000cd9ea6 by task rm/18493
> 
> udf_readdir() tries to write file name out of allocated memory
> buffer bounds. The UDF_NAME_LEN_CS0 (255) is used as max length
> for file name in udf_put_filename(). But UDF_NAME_LEN (254) is
> used as the size for buffer allocation in udf_readdir(). As the
> result out-of-bounds write happened.
> 
> Found by Linux Verification Center (linuxtesting.org) with xfstests
> 
> Fixes: 066b9cded00b ("udf: Use separate buffer for copying split names")
> Signed-off-by: Vladislav Efanov <VEfanov@ispras.ru>

Thanks for the fix Vladislav! But this code is not there anymore in current
version of UDF code. It got fixed as part of the directory code rewrite -
you can check current code state in my tree:

git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs.git for_next

								Honza

> ---
>  fs/udf/dir.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/fs/udf/dir.c b/fs/udf/dir.c
> index be640f4b2f2c..b95607c42ad4 100644
> --- a/fs/udf/dir.c
> +++ b/fs/udf/dir.c
> @@ -169,7 +169,7 @@ static int udf_readdir(struct file *file, struct dir_context *ctx)
>  				nameptr = (char *)(fibh.ebh->b_data + poffset - lfi);
>  			} else {
>  				if (!copy_name) {
> -					copy_name = kmalloc(UDF_NAME_LEN,
> +					copy_name = kmalloc(UDF_NAME_LEN_CS0,
>  							    GFP_NOFS);
>  					if (!copy_name) {
>  						ret = -ENOMEM;
> -- 
> 2.34.1
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR

Re: [PATCH] udf: KASAN: slab-out-of-bounds in udf_readdir

Posted by Ефанов Владислав Александрович 2 years, 7 months ago

Thank you for the information. Will check this repository in the future.

Vlad.

On 07.02.2023 15:44, Jan Kara wrote:
> On Mon 06-02-23 19:22:06, Vladislav Efanov wrote:
>> The KASAN report is:
>> [ 1922.586560] BUG: KASAN: slab-out-of-bounds in udf_readdir+0xe00/0x19e0
>> [ 1922.586922] Write of size 89 at addr ffff888000cd9ea6 by task rm/18493
>>
>> udf_readdir() tries to write file name out of allocated memory
>> buffer bounds. The UDF_NAME_LEN_CS0 (255) is used as max length
>> for file name in udf_put_filename(). But UDF_NAME_LEN (254) is
>> used as the size for buffer allocation in udf_readdir(). As the
>> result out-of-bounds write happened.
>>
>> Found by Linux Verification Center (linuxtesting.org) with xfstests
>>
>> Fixes: 066b9cded00b ("udf: Use separate buffer for copying split names")
>> Signed-off-by: Vladislav Efanov <VEfanov@ispras.ru>
> Thanks for the fix Vladislav! But this code is not there anymore in current
> version of UDF code. It got fixed as part of the directory code rewrite -
> you can check current code state in my tree:
>
> git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs.git for_next
>
> 								Honza
>
>> ---
>>   fs/udf/dir.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/fs/udf/dir.c b/fs/udf/dir.c
>> index be640f4b2f2c..b95607c42ad4 100644
>> --- a/fs/udf/dir.c
>> +++ b/fs/udf/dir.c
>> @@ -169,7 +169,7 @@ static int udf_readdir(struct file *file, struct dir_context *ctx)
>>   				nameptr = (char *)(fibh.ebh->b_data + poffset - lfi);
>>   			} else {
>>   				if (!copy_name) {
>> -					copy_name = kmalloc(UDF_NAME_LEN,
>> +					copy_name = kmalloc(UDF_NAME_LEN_CS0,
>>   							    GFP_NOFS);
>>   					if (!copy_name) {
>>   						ret = -ENOMEM;
>> -- 
>> 2.34.1
>>