linux-kernel - Re: [PATCH] drivers: staging: lustre: lustre: include: add "__attribute_

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: <52DD0EFF.2010305@imgtec.com>
Date:	Mon, 20 Jan 2014 11:56:47 +0000
From:	James Hogan <james.hogan@...tec.com>
To:	Chen Gang <gang.chen.5i5j@...il.com>
CC:	Dan Carpenter <dan.carpenter@...cle.com>,
	<devel@...verdev.osuosl.org>, <andreas.dilger@...el.com>,
	Greg KH <gregkh@...uxfoundation.org>, <bergwolf@...il.com>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	<oleg.drokin@...el.com>, <jacques-charles.lafoucriere@....fr>,
	<jinshan.xiong@...el.com>, <linux-metag@...r.kernel.org>
Subject: Re: [PATCH] drivers: staging: lustre: lustre: include: add "__attribute__((packed))"
 for the related union

Hi Chen,

On 19/01/14 10:07, Chen Gang wrote:
> BTW: this patch is related with another patch which is discussing (so I
> have cc that patch to you and Greg too): "if we could sure that it is a
> compiler's feature issue, we will skip this patch".

If you're referring to the #pragma pack portability issue then this
issue is unrelated since it doesn't use #pragma pack.

The issue is *not* that the compiler is defectively failing to pack
nested structs. Doing that would be utterly broken since it would change
the layout of the same struct depending on where it is placed in the
program, Consider this example (which uses a nested struct rather than
union to demonstrate the point):

struct a {
	struct b {
		unsigned int x;
		unsigned short y;
	} x;
	unsigned short y;
} __packed;

Both ABIs behave the same here:

Arch	sizeof(struct b)	sizeof(struct a)
x86_64	8			10
metag	8			10

If struct b is made __packed, again both ABIs behave differently in the
same way:

Arch	sizeof(struct b)	sizeof(struct a)
x86_64	6			8
metag	6			8

The issue is that C compiler ABIs may (and unfortunately metag ABI does)
pack structs and unions to at least 4 bytes, even if no members of the
struct or union are that large, which means that the nested struct/union
should be __packed too to portably ensure it's the expected size.

Cheers
James

--
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/