[bug report] rtc: m48t59: use generic nvmem

[bug report] rtc: m48t59: use generic nvmem

[Dan Carpenter]

Hello Alexandre Belloni,

This is a semi-automatic email about new static checker warnings.

The patch ce0e0076da3e: "rtc: m48t59: use generic nvmem" from Feb 12, 
2018, leads to the following Smatch complaint:

    drivers/rtc/rtc-m48t59.c:409 m48t59_rtc_probe()
    warn: variable dereferenced before check 'pdata' (see line 391)

drivers/rtc/rtc-m48t59.c
   390			.stride = 1,
   391			.size = pdata->offset,
                                ^^^^^^^^^^^^^
New dereference.

   392			.reg_read = m48t59_nvram_read,
   393			.reg_write = m48t59_nvram_write,
   394			.priv = pdev,
   395		};
   396	
   397		/* This chip could be memory-mapped or I/O-mapped */
   398		res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
   399		if (!res) {
   400			res = platform_get_resource(pdev, IORESOURCE_IO, 0);
   401			if (!res)
   402				return -EINVAL;
   403		}
   404	
   405		if (res->flags & IORESOURCE_IO) {
   406			/* If we are I/O-mapped, the platform should provide
   407			 * the operations accessing chip registers.
   408			 */
   409			if (!pdata || !pdata->write_byte || !pdata->read_byte)
                             ^^^^^
The existing code assumed "pdata" could be NULL.

   410				return -EINVAL;
   411		} else if (res->flags & IORESOURCE_MEM) {

regards,
dan carpenter

Re: [bug report] rtc: m48t59: use generic nvmem

[Alexandre Belloni]

Hi Dan,

Thanks for the report, I'm fixing up the patch (it uncovered another
issue).

On 20/02/2018 at 12:00:25 +0300, Dan Carpenter wrote:
> Hello Alexandre Belloni,
> 
> This is a semi-automatic email about new static checker warnings.
> 
> The patch ce0e0076da3e: "rtc: m48t59: use generic nvmem" from Feb 12, 
> 2018, leads to the following Smatch complaint:
> 
>     drivers/rtc/rtc-m48t59.c:409 m48t59_rtc_probe()
>     warn: variable dereferenced before check 'pdata' (see line 391)
> 
> drivers/rtc/rtc-m48t59.c
>    390			.stride = 1,
>    391			.size = pdata->offset,
>                                 ^^^^^^^^^^^^^
> New dereference.
> 
>    392			.reg_read = m48t59_nvram_read,
>    393			.reg_write = m48t59_nvram_write,
>    394			.priv = pdev,
>    395		};
>    396	
>    397		/* This chip could be memory-mapped or I/O-mapped */
>    398		res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>    399		if (!res) {
>    400			res = platform_get_resource(pdev, IORESOURCE_IO, 0);
>    401			if (!res)
>    402				return -EINVAL;
>    403		}
>    404	
>    405		if (res->flags & IORESOURCE_IO) {
>    406			/* If we are I/O-mapped, the platform should provide
>    407			 * the operations accessing chip registers.
>    408			 */
>    409			if (!pdata || !pdata->write_byte || !pdata->read_byte)
>                              ^^^^^
> The existing code assumed "pdata" could be NULL.
> 
>    410				return -EINVAL;
>    411		} else if (res->flags & IORESOURCE_MEM) {
> 
> regards,
> dan carpenter

-- 
Alexandre Belloni, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com