static analysis bug report: staging r8712u memcpy of uninitialized variable

static analysis bug report: staging r8712u memcpy of uninitialized variable

[Colin Ian King]

Hi,

Static analysis with cppcheck found a couple of interesting issues with
memcpy'ing of an uninitialized variable. Two occurrences of the same
issue are found in drivers/staging/rtl8712/rtl8712_cmd.c in functions
read_bbreg_hdl and read_rfreg_hdl.

For example:

static u8 read_bbreg_hdl(struct _adapter *padapter, u8 *pbuf)
{
        u32 val;
        void (*pcmd_callback)(struct _adapter *dev, struct cmd_obj
*pcmd);
        struct cmd_obj *pcmd  = (struct cmd_obj *)pbuf;

        if (pcmd->rsp && pcmd->rspsz > 0)
                memcpy(pcmd->rsp, (u8 *)&val, pcmd->rspsz);

	....

}

I don't understand why the contents of val is being memcpy'd to
pcmd->rsp, especially when val is uninitialized and hence contains
garbage. Any ideas?

Colin

Re: static analysis bug report: staging r8712u memcpy of uninitialized variable

[Dan Carpenter]

On Mon, Mar 18, 2019 at 11:20:51AM +0000, Colin Ian King wrote:
> Hi,
> 
> Static analysis with cppcheck found a couple of interesting issues with
> memcpy'ing of an uninitialized variable. Two occurrences of the same
> issue are found in drivers/staging/rtl8712/rtl8712_cmd.c in functions
> read_bbreg_hdl and read_rfreg_hdl.
> 
> For example:
> 
> static u8 read_bbreg_hdl(struct _adapter *padapter, u8 *pbuf)
> {
>         u32 val;
>         void (*pcmd_callback)(struct _adapter *dev, struct cmd_obj
> *pcmd);
>         struct cmd_obj *pcmd  = (struct cmd_obj *)pbuf;
> 
>         if (pcmd->rsp && pcmd->rspsz > 0)
>                 memcpy(pcmd->rsp, (u8 *)&val, pcmd->rspsz);
> 
> 	....
> 
> }
> 
> I don't understand why the contents of val is being memcpy'd to
> pcmd->rsp, especially when val is uninitialized and hence contains
> garbage. Any ideas?
> 

The concern would be that it's reading a user specified amount of
stack memory to pcmd->rsp and that sounds like an information leak.

The pcmd_callback function pointer is always r8712_getbbrfreg_cmdrsp_callback()
which frees pcmd->parmbuf and pcmd but leaks pcmd->rsp.  I don't see
a way for anyone to access the ->rsp memory so probably there aren't any
security implications.

Anyway, let me send a patch.

regards,
dan carpenter

[PATCH] staging: rtl8712: uninitialized memory in read_bbreg_hdl()

[Dan Carpenter]

Colin King reported a bug in read_bbreg_hdl():

	memcpy(pcmd->rsp, (u8 *)&val, pcmd->rspsz);

The problem is that "val" is uninitialized.

This code is obviously not useful, but so far as I can tell
"pcmd->cmdcode" is never GEN_CMD_CODE(_Read_BBREG) so it's not harmful
either.  For now the easiest fix is to just call r8712_free_cmd_obj()
and return.

Fixes: 2865d42c78a9 ("staging: r8712u: Add the new driver to the mainline kernel")
Reported-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
---
 drivers/staging/rtl8712/rtl8712_cmd.c | 10 +---------
 drivers/staging/rtl8712/rtl8712_cmd.h |  2 +-
 2 files changed, 2 insertions(+), 10 deletions(-)

diff --git a/drivers/staging/rtl8712/rtl8712_cmd.h b/drivers/staging/rtl8712/rtl8712_cmd.h
index 92fb77666d44..1ef86b8c592f 100644
--- a/drivers/staging/rtl8712/rtl8712_cmd.h
+++ b/drivers/staging/rtl8712/rtl8712_cmd.h
@@ -140,7 +140,7 @@ enum rtl8712_h2c_cmd {
 static struct _cmd_callback	cmd_callback[] = {
 	{GEN_CMD_CODE(_Read_MACREG), NULL}, /*0*/
 	{GEN_CMD_CODE(_Write_MACREG), NULL},
-	{GEN_CMD_CODE(_Read_BBREG), &r8712_getbbrfreg_cmdrsp_callback},
+	{GEN_CMD_CODE(_Read_BBREG), NULL},
 	{GEN_CMD_CODE(_Write_BBREG), NULL},
 	{GEN_CMD_CODE(_Read_RFREG), &r8712_getbbrfreg_cmdrsp_callback},
 	{GEN_CMD_CODE(_Write_RFREG), NULL}, /*5*/



diff --git a/drivers/staging/rtl8712/rtl8712_cmd.c b/drivers/staging/rtl8712/rtl8712_cmd.c
index 1920d02f7c9f..8c36acedf507 100644
--- a/drivers/staging/rtl8712/rtl8712_cmd.c
+++ b/drivers/staging/rtl8712/rtl8712_cmd.c
@@ -147,17 +147,9 @@ static u8 write_macreg_hdl(struct _adapter *padapter, u8 *pbuf)
 
 static u8 read_bbreg_hdl(struct _adapter *padapter, u8 *pbuf)
 {
-	u32 val;
-	void (*pcmd_callback)(struct _adapter *dev, struct cmd_obj	*pcmd);
 	struct cmd_obj *pcmd  = (struct cmd_obj *)pbuf;
 
-	if (pcmd->rsp && pcmd->rspsz > 0)
-		memcpy(pcmd->rsp, (u8 *)&val, pcmd->rspsz);
-	pcmd_callback = cmd_callback[pcmd->cmdcode].callback;
-	if (!pcmd_callback)
-		r8712_free_cmd_obj(pcmd);
-	else
-		pcmd_callback(padapter, pcmd);
+	r8712_free_cmd_obj(pcmd);
 	return H2C_SUCCESS;
 }
 
-- 
2.17.1

Re: [PATCH] staging: rtl8712: uninitialized memory in read_bbreg_hdl()

[Dan Carpenter]

On Thu, Mar 21, 2019 at 09:26:38AM +0300, Dan Carpenter wrote:
> Colin King reported a bug in read_bbreg_hdl():
> 
> 	memcpy(pcmd->rsp, (u8 *)&val, pcmd->rspsz);
> 
> The problem is that "val" is uninitialized.
> 
> This code is obviously not useful, but so far as I can tell
> "pcmd->cmdcode" is never GEN_CMD_CODE(_Read_BBREG) so it's not harmful
> either.  For now the easiest fix is to just call r8712_free_cmd_obj()
> and return.
> 
> Fixes: 2865d42c78a9 ("staging: r8712u: Add the new driver to the mainline kernel")
> Reported-by: Colin Ian King <colin.king@canonical.com>
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> ---
>  drivers/staging/rtl8712/rtl8712_cmd.c | 10 +---------
>  drivers/staging/rtl8712/rtl8712_cmd.h |  2 +-
>  2 files changed, 2 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/staging/rtl8712/rtl8712_cmd.h b/drivers/staging/rtl8712/rtl8712_cmd.h
> index 92fb77666d44..1ef86b8c592f 100644
> --- a/drivers/staging/rtl8712/rtl8712_cmd.h
> +++ b/drivers/staging/rtl8712/rtl8712_cmd.h
> @@ -140,7 +140,7 @@ enum rtl8712_h2c_cmd {
>  static struct _cmd_callback	cmd_callback[] = {
>  	{GEN_CMD_CODE(_Read_MACREG), NULL}, /*0*/
>  	{GEN_CMD_CODE(_Write_MACREG), NULL},
> -	{GEN_CMD_CODE(_Read_BBREG), &r8712_getbbrfreg_cmdrsp_callback},
> +	{GEN_CMD_CODE(_Read_BBREG), NULL},
>  	{GEN_CMD_CODE(_Write_BBREG), NULL},
>  	{GEN_CMD_CODE(_Read_RFREG), &r8712_getbbrfreg_cmdrsp_callback},
                                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>  	{GEN_CMD_CODE(_Write_RFREG), NULL}, /*5*/

The other place that calls r8712_getbbrfreg_cmdrsp_callback() is
read_rfreg_hdl().  For GEN_CMD_CODE(_Read_RFREG) we don't allocate
the ->rsp pointer so we can't kfree() it.  The read_rfreg_hdl()
functions calls r8712_free_cmd_obj() which kfrees it.  But fortunately
that is dead code.

This code is obviously staging code...

It would be fairly straight forward to get rid of the cmd_callback[]
array.

regards,
dan carpenter