linux-kernel - Re: [PATCH] lib/sha1: remove memsets and allocate workspace on the stack

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: <20110809055226.GE6103@google.com>
Date:	Mon, 8 Aug 2011 22:52:27 -0700
From:	Mandeep Singh Baines <msb@...omium.org>
To:	Joe Perches <joe@...ches.com>
Cc:	Mandeep Singh Baines <msb@...omium.org>,
	linux-kernel@...r.kernel.org,
	Ramsay Jones <ramsay@...say1.demon.co.uk>,
	Nicolas Pitre <nico@...xnic.net>,
	Joachim Eastwood <manabian@...il.com>,
	Andreas Schwab <schwab@...ux-m68k.org>,
	Herbert Xu <herbert@...dor.hengli.com.au>,
	"David S. Miller" <davem@...emloft.net>,
	linux-crypto@...r.kernel.org,
	Linus Torvalds <torvalds@...ux-foundation.org>
Subject: Re: [PATCH] lib/sha1: remove memsets and allocate workspace on the
 stack

Joe Perches (joe@...ches.com) wrote:
> On Mon, 2011-08-08 at 16:07 -0700, Mandeep Singh Baines wrote:
> > The previous implementation required the workspace to be passed in as
> > a parameter. This prevents the compiler from being able to store the
> > workspace in registers. I've also removed the memset since that also
> > prevents the compiler from storing the workspace in registers.
> 
> I did a similar patch locally.
> 
> > There is no loss of security due to removing the memset. It would be a
> > bug for the stack to leak to userspace. However, a defence-in-depth
> > argument could be made for keeping the clearing of the workspace.
> 
> You should add #include <linux/crypthash.h> to
> lib/sha1.c and perhaps rationalize the use of __u8
> and char for the second argument to sha_transform in
> the definition and uses.
> 
> For defense in depth, a bool could be added to sha_transform
> like:
> 
> void sha_transform(__u32 *digest, const char *data, bool wipe);
> 

This seems reasonable to me. In our case, there is no secrecy
issue. We hash the blocks of the RO root filesystem. There is nothing
secret about the rootfs blocks or their hash. We use the hashes to verify
that the blocks haven't been modified by an attacker.

We don't call sha_tranform directly. We use crypto_hash_digest. So maybe
add a wipe param there. I'm happy to work on or test such a patch if folks
think its interesting. Its saves me 190 ms on a 6 second boot. I suspect
there may be other hash intense applications that also don't need secracy.

> and the internal workspace memset after use if wipe set though
> perhaps the memset could be a compile time option like
> CONFIG_CRYPTO_DEFENSE_IN_DEPTH or such instead.
> 

Maybe instead use wipe=0 in code paths where you don't need secracy.

> > diff --git a/drivers/char/random.c b/drivers/char/random.c
> []
> > @@ -816,13 +816,13 @@ static size_t account(struct entropy_store *r, size_t nbytes, int min,
> >  static void extract_buf(struct entropy_store *r, __u8 *out)
> []
> > -		sha_transform(hash, (__u8 *)(r->pool + i), workspace);
> > +		sha_transform(hash, (__u8 *)(r->pool + i));
> []
> > diff --git a/include/linux/cryptohash.h b/include/linux/cryptohash.h
> []
> > @@ -3,10 +3,9 @@
> []
> > -void sha_transform(__u32 *digest, const char *data, __u32 *W);
> > +void sha_transform(__u32 *digest, const char *data);
> []
> > diff --git a/lib/sha1.c b/lib/sha1.c
> []
> > +void sha_transform(__u32 *digest, const char *data)
> []
> > diff --git a/net/ipv4/syncookies.c b/net/ipv4/syncookies.c
> []
> > @@ -50,7 +49,7 @@ static u32 cookie_hash(__be32 saddr, __be32 daddr, __be16 sport, __be16 dport,
> []
> > -	sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
> > +	sha_transform(tmp + 16, (__u8 *)tmp);
> []
> > diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
> []
> > @@ -2511,8 +2510,7 @@ struct sk_buff *tcp_make_synack(struct sock *sk, struct dst_entry *dst,
> []
> >  			sha_transform((__u32 *)&xvp->cookie_bakery[0],
> > -				      (char *)mess,
> > -				      &workspace[0]);
> > +				      (char *)mess);
> []
> > diff --git a/net/ipv6/syncookies.c b/net/ipv6/syncookies.c
> []
> > @@ -81,7 +80,7 @@ static u32 cookie_hash(const struct in6_addr *saddr, const struct in6_addr *dadd
> []
> > -	sha_transform(tmp + 16, (__u8 *)tmp, tmp + 16 + 5);
> > +	sha_transform(tmp + 16, (__u8 *)tmp);
> 
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/