From mboxrd@z Thu Jan  1 00:00:00 1970
From: =?iso-8859-1?Q?Robert_Deli=EBn?= <robert.delien@nxp.com>
Date: Thu, 14 Dec 2006 17:52:21 +0100
Subject: [U-Boot-Users] Bug in MIPS linker scripts?
Message-ID: <000c01c71fa0$38e775f0$dfdd9182@code1.emi.philips.com>
List-Id: <u-boot.lists.denx.de>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: u-boot@lists.denx.de

Hi,

I believe I have found a nasty bug in my target's linker script. And since
all MIPS targets use copies of more-or-less the same linker script, I think
all MIPS targets suffer from this bug.

It's alignment dependent whether the bug will cause a problem or not. The
symptoms can vary from hang-ups in find_cmd for any command to a hang-up if
the first command in the command table is used (usually bdinfo).

The cause of this problem is the 'Update GOT' part in the function relocate
in cpu/mips/start.S. Depending on alignment, this update part will update a
number of entries too many, corrupting whatever is next in memory: The
command table.

The cause of the problem is a misalignment between the _gp linker variable
and the actual '.'. This is what's in board/mipstarget/u-boot.lds now:

	. = ALIGN(4);
	.sdata  : { *(.sdata) }

	_gp = ALIGN(16);

	__got_start = .;
	.got  : { *(.got) }
	__got_end = .;

	.sdata  : { *(.sdata) }

	__u_boot_cmd_start = .;

And this is what it's supposed to be:
	. = ALIGN(4);
	.sdata  : { *(.sdata) }

	. = ALIGN(16);
	_gp = .;
 
	__got_start = .;
	.got  : { *(.got) }
	__got_end = .;

	.sdata  : { *(.sdata) }

	__u_boot_cmd_start = .;

Who can give me any comments?

With kind regards,

	Robert.