RE: [RFC][PATCH 3/3] audit: Audit proc cmdline value

[William Roberts]

-----Original Message-----
From: Mateusz Guzik [mailto:mguzik@xxxxxxxxxx] 
Sent: Monday, January 06, 2014 9:09 AM
To: William Roberts
Cc: linux-audit@xxxxxxxxxx; linux-mm@xxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx; rgb@xxxxxxxxxx; viro@xxxxxxxxxxxxxxxxxx; akpm@xxxxxxxxxxxxxxxxxxxx; sds@xxxxxxxxxxxxx; William Roberts
Subject: Re: [RFC][PATCH 3/3] audit: Audit proc cmdline value

I can't comment on the concept, but have one nit.

On Mon, Jan 06, 2014 at 07:30:30AM -0800, William Roberts wrote:
> +static void audit_log_cmdline(struct audit_buffer *ab, struct task_struct *tsk,
> +			 struct audit_context *context)
> +{
> +	int res;
> +	char *buf;
> +	char *msg = "(null)";
> +	audit_log_format(ab, " cmdline=");
> +
> +	/* Not  cached */
> +	if (!context->cmdline) {
> +		buf = kmalloc(PATH_MAX, GFP_KERNEL);
> +		if (!buf)
> +			goto out;
> +		res = get_cmdline(tsk, buf, PATH_MAX);
> +		/* Ensure NULL terminated */
> +		if (buf[res-1] != '\0')
> +			buf[res-1] = '\0';

This accesses memory below the buffer if get_cmdline returned 0, which I believe will be the case when someone jokingly unmaps the area (all maybe when it is swapped out but can't be swapped in due to I/O errors).
[William Roberts] 
Sorry for the weird inline posting (Thanks MS Outlook of doom). Anyways, this isnât a nit. This is a major issue that should be dealt with. Thanks.

Also since you are just putting 0 in there anyway I don't see much point in testing for it.

> +		context->cmdline = buf;
> +	}
> +	msg = context->cmdline;
> +out:
> +	audit_log_untrustedstring(ab, msg);
> +}
> +



--
Mateusz Guzik
N‹§²æìr¸›yúèšØb²X¬¶ÇvØ^–)Þ{.nÇ+‰·¥Š{±‘êçzX§¶›¡Ü}©ž²ÆzÚ&j:+v‰¨¾«‘êçzZ+€Ê+zf£¢·hšˆ§~†­†Ûiÿûàz¹®w¥¢¸?™¨è­Ú&¢)ßf”ù^jÇy§m…á@A«a¶Úÿ0¶ìh®å’i