linux-kernel - Re: [PATCH] Avoid truncating to 'long' in ALIGN() macro

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: <20061126111703.33247a84.akpm@osdl.org>
Date:	Sun, 26 Nov 2006 11:17:03 -0800
From:	Andrew Morton <akpm@...l.org>
To:	Roland Dreier <rdreier@...co.com>
Cc:	David Miller <davem@...emloft.net>, linux-kernel@...r.kernel.org,
	openib-general@...nib.org, tom@...ngridcomputing.com,
	Al Viro <viro@...iv.linux.org.uk>
Subject: Re: [PATCH] Avoid truncating to 'long' in ALIGN() macro

On Sun, 26 Nov 2006 11:10:23 -0800
Roland Dreier <rdreier@...co.com> wrote:

> Commit 4c8bd7ee ("Do not truncate to 'int' in ALIGN() macro.") was
> merged to fix the case of code like the following:
> 
> 	unsigned long addr;
> 	unsigned int alignment;
> 	addr = ALIGN(addr, alignment);
> 
> The original ALIGN macro calculated a mask as ~(alignment - 1), and
> when alignment is just an int, this creates an int mask.  If alignment
> is also unsigned, then this mask will not be sign extended when
> promoted to a long, which leads to the code above chopping off the top
> half of addr when long is 64 bits.
> 
> However, the changed ALIGN macro, which computes the mask as
> ~(alignment - 1UL) actually breaks code like the following when long
> is 32 bits:
> 
> 	u64 addr;
>         int alignment;
>         addr = ALIGN(addr, alignment);
> 
> The reason this breaks is pretty much the same as the original bug
> that the change was supposed to fix: ~(alignment - 1UL) creates a mask
> that is an unsigned long, which is not sign extended when promoted to
> u64 (if long is 32 bits).
> 
> As suggested by Dave Miller and Al Viro, I fixed this by having the
> ALIGN macro make sure the alignment is promoted to the same type as
> the value being aligned before doing the negation.
> 
> This second construct is actually used in the amso1100 driver, so that
> driver does not work on 32-bit architectures right now.  Unfortunately
> almost everyone using it runs 64-bit kernels, so this regression was
> not noticed until now.
> 
> Signed-off-by: Roland Dreier <rolandd@...co.com>
> 
> ---
> Patch updated as suggested by Dave Miller and Al Viro...
> 
> can we merge this for 2.6.20?
> 
> diff --git a/include/linux/kernel.h b/include/linux/kernel.h
> index 24b6111..80955b3 100644
> --- a/include/linux/kernel.h
> +++ b/include/linux/kernel.h
> @@ -31,7 +31,7 @@ #define ULLONG_MAX	(~0ULL)
>  #define STACK_MAGIC	0xdeadbeef
>  
>  #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0]))
> -#define ALIGN(x,a) (((x)+(a)-1UL)&~((a)-1UL))
> +#define ALIGN(x,a) ((typeof(x)) (((x) + (a) - 1) & ~((typeof(x)) (a) - 1)))
>  #define FIELD_SIZEOF(t, f) (sizeof(((t*)0)->f))
>  #define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
>  #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))

I'd be inclined to merge it for 2.6.19.  Is everyone OK with it?
-
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/