On Tue, Jul 07, 2015 at 10:28:00PM +0800, Nicolas Iooss wrote:
> Some modules call config_item_init_type_name() and
> config_group_init_type_name() with parameter "name" directly controlled
> by userspace.  These two functions call config_item_set_name() with this
> name used as a format string, which can be used to leak information such
> as content of the stack to userspace.
> 
> For example, make_netconsole_target() in netconsole module calls
> config_item_init_type_name() with the name of a newly-created directory.
> This means that the following commands give some unexpected output, with
> configfs mounted in /sys/kernel/config/ and on a system with a
> configured eth0 ethernet interface:
> 
>     # modprobe netconsole
>     # mkdir /sys/kernel/config/netconsole/target_%lx
>     # echo eth0 > /sys/kernel/config/netconsole/target_%lx/dev_name
>     # echo 1 > /sys/kernel/config/netconsole/target_%lx/enabled
>     # echo eth0 > /sys/kernel/config/netconsole/target_%lx/dev_name
>     # dmesg |tail -n1
>     [  142.697668] netconsole: target (target_ffffffffc0ae8080) is
>     enabled, disable to update parameters
> 
> The directory name is correct but %lx has been interpreted in the
> internal item name, displayed here in the error message used by
> store_dev_name() in drivers/net/netconsole.c.
> 
> To fix this, update every caller of config_item_set_name to use "%s"
> when operating on untrusted input.
> 
> This issue was found using -Wformat-security gcc flag, once a __printf
> attribute has been added to config_item_set_name().
> 
> Signed-off-by: Nicolas Iooss <[email protected]>
> ---
>  drivers/usb/gadget/configfs.c | 2 +-
>  fs/configfs/item.c            | 4 ++--
>  2 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/usb/gadget/configfs.c b/drivers/usb/gadget/configfs.c
> index 0495c94a23d7..289e20119fea 100644
> --- a/drivers/usb/gadget/configfs.c
> +++ b/drivers/usb/gadget/configfs.c
> @@ -571,7 +571,7 @@ static struct config_group *function_make(
>       if (IS_ERR(fi))
>               return ERR_CAST(fi);
>  
> -     ret = config_item_set_name(&fi->group.cg_item, name);
> +     ret = config_item_set_name(&fi->group.cg_item, "%s", name);

No objections from me:

Acked-by: Felipe Balbi <[email protected]>


>       if (ret) {
>               usb_put_function_instance(fi);
>               return ERR_PTR(ret);
> diff --git a/fs/configfs/item.c b/fs/configfs/item.c
> index 4d6a30e76168..b863a09cd2f1 100644
> --- a/fs/configfs/item.c
> +++ b/fs/configfs/item.c
> @@ -115,7 +115,7 @@ void config_item_init_type_name(struct config_item *item,
>                               const char *name,
>                               struct config_item_type *type)
>  {
> -     config_item_set_name(item, name);
> +     config_item_set_name(item, "%s", name);
>       item->ci_type = type;
>       config_item_init(item);
>  }
> @@ -124,7 +124,7 @@ EXPORT_SYMBOL(config_item_init_type_name);
>  void config_group_init_type_name(struct config_group *group, const char 
> *name,
>                        struct config_item_type *type)
>  {
> -     config_item_set_name(&group->cg_item, name);
> +     config_item_set_name(&group->cg_item, "%s", name);
>       group->cg_item.ci_type = type;
>       config_group_init(group);
>  }
> -- 
> 2.4.5
> 

-- 
balbi

Attachment: signature.asc
Description: Digital signature

Reply via email to