linux-kernel - Re: [PATCH v5 13/21] x86, boot: Report overlap failures in 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]

Message-ID: <20160415174231.7425bc7a@tukaani.org>
Date:	Fri, 15 Apr 2016 17:42:31 +0300
From:	Lasse Collin <lasse.collin@...aani.org>
To:	Kees Cook <keescook@...omium.org>
Cc:	Ingo Molnar <mingo@...nel.org>, Yinghai Lu <yinghai@...nel.org>,
	Baoquan He <bhe@...hat.com>,
	Ard Biesheuvel <ard.biesheuvel@...aro.org>,
	Matt Redfearn <matt.redfearn@...tec.com>, x86@...nel.org,
	"H. Peter Anvin" <hpa@...or.com>, Ingo Molnar <mingo@...hat.com>,
	Borislav Petkov <bp@...en8.de>,
	Vivek Goyal <vgoyal@...hat.com>,
	Andy Lutomirski <luto@...nel.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Dave Young <dyoung@...hat.com>,
	kernel-hardening@...ts.openwall.com,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v5 13/21] x86, boot: Report overlap failures in memcpy

On 2016-04-14 Kees Cook wrote:
> From: Yinghai Lu <yinghai@...nel.org>
> 
> parse_elf is using a local memcpy to move sections to their final
> running position. However, this memcpy only supports non-overlapping
> arguments (or dest < src).

The same copy of memcpy is used by the decompressors too.

> To avoid future hard-to-debug surprises, this adds checking in memcpy
> to detect the unhandled condition (which should not be happening
> currently).

It's already a minor surprise that memcpy is expected to work for
overlapping buffers at all. It could be good to have a comment about
it because "scroll" and parse_elf seem to rely on it.

On the other hand, the new code and error message take quite a few bytes
of space, so a complete memmove can be smaller:

void *memmove(void *dest, const void *src, size_t n)
{
	unsigned char *d = dest;
	const unsigned char *s = src;

	if (d <= s || d - s >= n)
		return __memcpy(dest, src, n);

	while (n-- > 0)
		d[n] = s[n];

	return dest;
}

Note that memmove is needed by lib/decompress_unxz.c. It contains its
own small version inside a "#ifndef memmove" block. That #ifndef should
be taken into account when adding a memmove symbol. Changing
decompress_unxz.c is fine but then one needs to think about other
archs too.

-- 
Lasse Collin  |  IRC: Larhzu @ IRCnet & Freenode