linux-kernel - Re: [PATCH 03/11] staging: replace weird strncpy() with memcpy()

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Thu, 28 Mar 2024 19:35:27 +0300
From: Dan Carpenter <dan.carpenter@...aro.org>
To: Arnd Bergmann <arnd@...nel.org>
Cc: linux-kernel@...r.kernel.org,
	Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
	Kees Cook <keescook@...omium.org>,
	Daniel Micay <danielmicay@...il.com>, Arnd Bergmann <arnd@...db.de>,
	linux-staging@...ts.linux.dev
Subject: Re: [PATCH 03/11] staging: replace weird strncpy() with memcpy()

On Thu, Mar 28, 2024 at 03:04:47PM +0100, Arnd Bergmann wrote:
> From: Arnd Bergmann <arnd@...db.de>
> 
> When -Wstringop-truncation is enabled, gcc finds a function that
> always does a short copy:
> 
> In function 'inquiry',
>     inlined from 'rtsx_scsi_handler' at drivers/staging/rts5208/rtsx_scsi.c:3210:12:
> drivers/staging/rts5208/rtsx_scsi.c:526:17: error: 'strncpy' output truncated copying between 1 and 28 bytes from a string of length 28 [-Werror=stringop-truncation]
>   526 |                 strncpy(buf + 8, inquiry_string, sendbytes - 8);
>       |                 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> 
> Since the actual size of the copy is already known at this point, just
> copy the bytes directly and skip the length check and zero-padding.
> 
> This partially reverts an earlier bugfix that replaced the original
> incorrect memcpy() with a less bad strncpy(), but it now also avoids
> the original overflow.
> 
> Fixes: 88a5b39b69ab ("staging/rts5208: Fix read overflow in memcpy")

I don't see a problem with this commit.  The "sendbytes - 8" prevents
a write overflow to buf, and the strncpy() prevents read overflow from
inquiry_string.

> Signed-off-by: Arnd Bergmann <arnd@...db.de>
> ---
>  drivers/staging/rts5208/rtsx_scsi.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/staging/rts5208/rtsx_scsi.c b/drivers/staging/rts5208/rtsx_scsi.c
> index 08bd768ad34d..a73b0959f5a9 100644
> --- a/drivers/staging/rts5208/rtsx_scsi.c
> +++ b/drivers/staging/rts5208/rtsx_scsi.c
> @@ -523,7 +523,7 @@ static int inquiry(struct scsi_cmnd *srb, struct rtsx_chip *chip)
>  
>  	if (sendbytes > 8) {
>  		memcpy(buf, inquiry_buf, 8);
> -		strncpy(buf + 8, inquiry_string, sendbytes - 8);
> +		memcpy(buf + 8, inquiry_string, min(sendbytes, 36) - 8);

I think your math is off.  The string is 29 characters + NUL.  So it
should be "min(sendbytes, 38) - 8".  You're chopping off the space and
the NUL terminator.

This only affects pro_formatter_flag code...

This code is such a mess.  I'm not sure your fix is the complete fix.
When I see code that's clearly buggy like this and it's not sure the fix
is complete then I generally prefer to leave the static checker warning
as is so that we are reminded of the bug occasionally.  How close are
you to removing all these -Wstringop-truncation warnings?  Maybe we just
add a comment or a TODO item in the drivers/staging/rts5208/TODO file.

regards,
dan carpenter