Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.33 as permitted sender) client-ip=23.128.96.33;
Date:   Thu, 30 Nov 2023 13:25:33 -0800
From:   Kees Cook <keescook@chromium.org>
To:     Justin Stitt <justinstitt@google.com>
Cc:     Michael Cyr <mikecyr@linux.ibm.com>,
        "James E.J. Bottomley" <jejb@linux.ibm.com>,
        "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: ibmvscsi_tgt: replace deprecated strncpy with
 strscpy
Message-ID: <202311301315.BAB096926@keescook>
References: <20231030-strncpy-drivers-scsi-ibmvscsi_tgt-ibmvscsi_tgt-c-v1-1-859b5ce257fd@google.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20231030-strncpy-drivers-scsi-ibmvscsi_tgt-ibmvscsi_tgt-c-v1-1-859b5ce257fd@google.com>
Precedence: bulk

On Mon, Oct 30, 2023 at 09:43:20PM +0000, Justin Stitt wrote:
> strncpy() is deprecated for use on NUL-terminated destination strings
> [1] and as such we should prefer more robust and less ambiguous string
> interfaces.
> 
> We don't need the NUL-padding behavior that strncpy() provides as vscsi
> is NUL-allocated in ibmvscsis_probe() which proceeds to call
> ibmvscsis_adapter_info():
> |       vscsi = kzalloc(sizeof(*vscsi), GFP_KERNEL);
> 
> ibmvscsis_probe() -> ibmvscsis_handle_crq() -> ibmvscsis_parse_command()
> -> ibmvscsis_mad() -> ibmvscsis_process_mad() -> ibmvscsis_adapter_info()
> 
> Following the same idea, `partition_name` is defiend as:
> |       static char partition_name[PARTITION_NAMELEN] = "UNKNOWN";
> 
> ... which is NUL-padded already, meaning strscpy() is the best option.
> 
> Considering the above, a suitable replacement is `strscpy` [2] due to
> the fact that it guarantees NUL-termination on the destination buffer
> without unnecessarily NUL-padding.

My only worry here is that I don't see if %NUL termination is _required_
for these variables. (i.e. do we run the risk of truncating these by 1
byte if they're right at the limit?) Are they __nonstring?

I *think* they're %NUL terminated since they follow the same sizing as
the global "partition_name", but I'm not sure.

Can any of the SCSI authors comment on this?

> 
> However, for cap->name let's use strscpy_pad as cap is allocated via
> dma_alloc_coherent():
> |       cap = dma_alloc_coherent(&vscsi->dma_dev->dev, olen, &token,
> |                                GFP_ATOMIC);

This is also true for the "info" allocation (it comes out of DMA).

> 
> Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1]
> Link: https://manpages.debian.org/testing/linux-manual-4.8/strscpy.9.en.html [2]
> Link: https://github.com/KSPP/linux/issues/90
> Cc: linux-hardening@vger.kernel.org
> Signed-off-by: Justin Stitt <justinstitt@google.com>
> ---
> Note: build-tested only.
> 
> Found with: $ rg "strncpy\("
> ---
>  drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c | 14 +++++++-------
>  1 file changed, 7 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c b/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
> index 385f812b8793..cd223ef696e5 100644
> --- a/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
> +++ b/drivers/scsi/ibmvscsi_tgt/ibmvscsi_tgt.c
> @@ -1551,17 +1551,17 @@ static long ibmvscsis_adapter_info(struct scsi_info *vscsi,
>  	if (vscsi->client_data.partition_number == 0)
>  		vscsi->client_data.partition_number =
>  			be32_to_cpu(info->partition_number);
> -	strncpy(vscsi->client_data.srp_version, info->srp_version,
> +	strscpy(vscsi->client_data.srp_version, info->srp_version,
>  		sizeof(vscsi->client_data.srp_version));
> -	strncpy(vscsi->client_data.partition_name, info->partition_name,
> +	strscpy(vscsi->client_data.partition_name, info->partition_name,
>  		sizeof(vscsi->client_data.partition_name));
>  	vscsi->client_data.mad_version = be32_to_cpu(info->mad_version);
>  	vscsi->client_data.os_type = be32_to_cpu(info->os_type);
>  
>  	/* Copy our info */
> -	strncpy(info->srp_version, SRP_VERSION,
> +	strscpy(info->srp_version, SRP_VERSION,
>  		sizeof(info->srp_version));
> -	strncpy(info->partition_name, vscsi->dds.partition_name,
> +	strscpy(info->partition_name, vscsi->dds.partition_name,
>  		sizeof(info->partition_name));

Since "info" is from DMA, let's use the _pad variant here just to be
safe.

>  	info->partition_number = cpu_to_be32(vscsi->dds.partition_num);
>  	info->mad_version = cpu_to_be32(MAD_VERSION_1);
> @@ -1645,8 +1645,8 @@ static int ibmvscsis_cap_mad(struct scsi_info *vscsi, struct iu_entry *iue)
>  			 be64_to_cpu(mad->buffer),
>  			 vscsi->dds.window[LOCAL].liobn, token);
>  	if (rc == H_SUCCESS) {
> -		strncpy(cap->name, dev_name(&vscsi->dma_dev->dev),
> -			SRP_MAX_LOC_LEN);
> +		strscpy_pad(cap->name, dev_name(&vscsi->dma_dev->dev),
> +			sizeof(cap->name));

And this is a safe conversion to sizeof():

struct capabilities {
	...
        char name[SRP_MAX_LOC_LEN];


If we can convince ourselves that non of these are __nonstring types,
then I think with the "info" change above, this should be good.

-Kees

-- 
Kees Cook