Re: [hid-picolcd] Avoid compile warning/error triggered bycopy_from_user()

[Jiri Kosina]

On Wed, 4 May 2011, Bruno PrÃmont wrote:

> With CONFIG_DEBUG_STRICT_USER_COPY_CHECKS=y compilation of PicoLCD
> driver fails on copy_from_user(), without it a warning is generated:
> 
>   CC [M]  drivers/hid/hid-picolcd.o
> In file included from /usr/src/linux-2.6/arch/x86/include/asm/uaccess.h:571,
>                  from /usr/src/linux-2.6/arch/x86/include/asm/sections.h:5,
>                  from /usr/src/linux-2.6/arch/x86/include/asm/hw_irq.h:26,
>                  from /usr/src/linux-2.6/include/linux/irq.h:359,
>                  from /usr/src/linux-2.6/arch/x86/include/asm/hardirq.h:5,
>                  from /usr/src/linux-2.6/include/linux/hardirq.h:7,
>                  from /usr/src/linux-2.6/include/linux/interrupt.h:12,
>                  from /usr/src/linux-2.6/include/linux/usb.h:15,
>                  from /usr/src/linux-2.6/drivers/hid/hid-picolcd.c:25:
> In function 'copy_from_user',
>     inlined from 'picolcd_debug_eeprom_write' at drivers/hid/hid-picolcd.c:1592:
> arch/x86/include/asm/uaccess_32.h:212: error: call to 'copy_from_user_overflow' declared with attribute error: copy_from_user() buffer size is not provably correct
> 
> gcc-4.4.5 is not able to track size calculation when it is stored into
> a variable, thus tell copy_from_user() maximum size via
> min(*max-size*, *effective-size*) explicitly and inline how much to copy
> at most.
> 
> Signed-off-by: Bruno PrÃmont <bonbons@xxxxxxxxxxxxxxxxx>
> --
> 
> diff --git a/drivers/hid/hid-picolcd.c b/drivers/hid/hid-picolcd.c
> index b2f56a1..9d8710f 100644
> --- a/drivers/hid/hid-picolcd.c
> +++ b/drivers/hid/hid-picolcd.c
> @@ -1585,11 +1585,11 @@ static ssize_t picolcd_debug_eeprom_write(struct file *f, const char __user *u,
>  	memset(raw_data, 0, sizeof(raw_data));
>  	raw_data[0] = *off & 0xff;
>  	raw_data[1] = (*off >> 8) & 0xff;
> -	raw_data[2] = s < 20 ? s : 20;
> +	raw_data[2] = min((size_t)20, s);
>  	if (*off + raw_data[2] > 0xff)
>  		raw_data[2] = 0x100 - *off;
>  
> -	if (copy_from_user(raw_data+3, u, raw_data[2]))
> +	if (copy_from_user(raw_data+3, u, min((u8)20, raw_data[2])))

Hmm ... this is quite an obfuscation just for the sake of making gcc 
happy.

Do other versions of gcc get this right? (i.e. is this gcc bug?)

Don't we have similar problems all over the place in the kernel?

>  		return -EFAULT;
>  	resp = picolcd_send_and_wait(data->hdev, REPORT_EE_WRITE, raw_data,
>  			sizeof(raw_data));
> 

-- 
Jiri Kosina
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/