[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <202309251109.5AEBF4369@keescook>
Date:   Mon, 25 Sep 2023 11:19:20 -0700
From:   Kees Cook <keescook@...omium.org>
To:     Justin Stitt <justinstitt@...gle.com>
Cc:     Alasdair Kergon <agk@...hat.com>,
        Mike Snitzer <snitzer@...nel.org>, dm-devel@...hat.com,
        linux-kernel@...r.kernel.org, linux-hardening@...r.kernel.org
Subject: Re: [PATCH] dm ioctl: replace deprecated strncpy with strscpy_pad
On Mon, Sep 25, 2023 at 06:54:51AM +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 expect `spec->target_type` to be NUL-terminated based on its use with
> a format string after `dm_table_add_target()` is called
> | 	r = dm_table_add_target(table, spec->target_type,
> | 				(sector_t) spec->sector_start,
> | 				(sector_t) spec->length,
> | 				target_params);
> ... wherein `spec->target_type` is passed as parameter `type` and later
> printed with DMERR:
> |       DMERR("%s: %s: unknown target type", dm_device_name(t->md), type);
> 
> It appears that `spec` is not zero-allocated and thus NUL-padding may be
> required in this ioctl context.
Yeah... this one was a bit odd. I see the memory being filled from
userspace in copy_params(), so it's unlikely there was any kernel memory
content leaking, but I think the old code may have been buggy:
		strncpy(spec->target_type, ti->type->name,
			sizeof(spec->target_type) - 1);
I can't find any reason to think that
spec->target_type[sizeof(spec->target_type) - 1] would be %NUL after
this code runs.
get_result_buffer() is basically:
        return ((void *) param) + param->data_start;
then:
        outptr = outbuf = get_result_buffer(param, param_size, &len);
	...
                spec = (struct dm_target_spec *) outptr;
and before that:
        r = copy_params(user, ¶m_kernel, ioctl_flags, ¶m, ¶m_flags);
        if (r)
                return r;
        input_param_size = param->data_size;
        r = validate_params(cmd, param);
        if (r)
                goto out;
        param->data_size = offsetof(struct dm_ioctl, data);
        r = fn(file, param, input_param_size);
Notably validate_params() %NUL-terminates some _other_ strings, but
doesn't clear the param->data area.
So I think this is actually fixing a bug too, but I can't quite tell.
Regardless, the change looks correct:
Reviewed-by: Kees Cook <keescook@...omium.org>
> 
> Considering the above, a suitable replacement is `strscpy_pad` due to
> the fact that it guarantees NUL-termination whilst maintaining the
> NUL-padding behavior that strncpy provides.
> 
> Link: https://www.kernel.org/doc/html/latest/process/deprecated.html#strncpy-on-nul-terminated-strings [1]
> Link: https://github.com/KSPP/linux/issues/90
> Cc: linux-hardening@...r.kernel.org
> Signed-off-by: Justin Stitt <justinstitt@...gle.com>
> ---
> Note: build-tested only.
> ---
>  drivers/md/dm-ioctl.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c
> index 21ebb6c39394..e65058e0ed06 100644
> --- a/drivers/md/dm-ioctl.c
> +++ b/drivers/md/dm-ioctl.c
> @@ -1295,8 +1295,8 @@ static void retrieve_status(struct dm_table *table,
>  		spec->status = 0;
>  		spec->sector_start = ti->begin;
>  		spec->length = ti->len;
> -		strncpy(spec->target_type, ti->type->name,
> -			sizeof(spec->target_type) - 1);
> +		strscpy_pad(spec->target_type, ti->type->name,
> +			sizeof(spec->target_type));
>  
>  		outptr += sizeof(struct dm_target_spec);
>  		remaining = len - (outptr - outbuf);
> 
> ---
> base-commit: 6465e260f48790807eef06b583b38ca9789b6072
> change-id: 20230925-strncpy-drivers-md-dm-ioctl-c-ea5c10e77981
> 
> Best regards,
> --
> Justin Stitt <justinstitt@...gle.com>
> 
> 
-- 
Kees Cook
Powered by blists - more mailing lists