Received-SPF: pass (google.com: domain of linux-kernel+bounces-108374-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Date: Tue, 19 Mar 2024 19:42:31 -0700
From: Kees Cook <keescook@chromium.org>
To: Justin Stitt <justinstitt@google.com>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>,
	linux-scsi@vger.kernel.org, target-devel@vger.kernel.org,
	linux-kernel@vger.kernel.org, linux-hardening@vger.kernel.org
Subject: Re: [PATCH] scsi: target: core: replace deprecated strncpy with
 strscpy
Message-ID: <202403191941.D2818DB@keescook>
References: <20240318-strncpy-drivers-target-target_core_transport-c-v1-1-a086b9a0d639@google.com>
 <202403191908.1B6BBA3@keescook>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <202403191908.1B6BBA3@keescook>

On Tue, Mar 19, 2024 at 07:23:43PM -0700, Kees Cook wrote:
> Hm, this actually fixes potential over-reads and potential memory content
> exposures (in the face of malicious/weird hardware) since p_buf_len
> appears to always be sizeof(p_buf) in callers, which means the old use
> of strncpy() could have left the string unterminated.
> 
> In practice I would assume it's not a problem, but, for example, here's
> a place where the 254 p_buf_len could run out when doing the sprintf:
> 
> #define VPD_TMP_BUF_SIZE                      254
> ...
> #define INQUIRY_VPD_DEVICE_IDENTIFIER_LEN       254
> ...
> struct t10_vpd {
>         unsigned char device_identifier[INQUIRY_VPD_DEVICE_IDENTIFIER_LEN];
> 	...
> };
> ...
> int transport_dump_vpd_ident(..., unsigned char *p_buf, int p_buf_len)
> {
> 	...
>         unsigned char buf[VPD_TMP_BUF_SIZE];
> 	...
>                 snprintf(buf, sizeof(buf),
>                         "T10 VPD ASCII Device Identifier: %s\n",
>                         &vpd->device_identifier[0]);
> 	...
>         if (p_buf)
>                 strncpy(p_buf, buf, p_buf_len);	// may write 254 chars and no NUL

Wait, no, it's safe; I got confused by the double buffers. The
snprintf() will always NUL-terminate, so "buf" copied into p_buf will
always be NUL terminated.

So, nevermind! Regardless, still a good conversion. Thank you!

-Kees

-- 
Kees Cook