From mboxrd@z Thu Jan  1 00:00:00 1970
From: Mark Salyzyn <mark_salyzyn@xyratex.com>
Subject: Re: [PATCH] pm8001: support HDA (flashless) mode
Date: Mon, 30 Apr 2012 12:56:30 -0400
Message-ID: <E53424C2-2CAE-43AA-81EE-0E7B44586C6F@xyratex.com>
References: <420864BE-8542-4BD8-B3DB-FDA403D193F9@xyratex.com> <A60D2B04CB1E487D9D0371139484E1C2@usish.com.cn> <73464C75-47D9-4409-87C2-8DDD0CBD5C67@xyratex.com>
Mime-Version: 1.0 (Apple Message framework v1257)
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 8BIT
Return-path: <linux-scsi-owner@vger.kernel.org>
Received: from xyratex197.xyratex.com ([194.131.166.197]:45284 "EHLO
	XY01EX21.xy01.xyratex.com" rhost-flags-OK-OK-OK-FAIL)
	by vger.kernel.org with ESMTP id S1755517Ab2D3Q4f convert rfc822-to-8bit
	(ORCPT <rfc822;linux-scsi@vger.kernel.org>);
	Mon, 30 Apr 2012 12:56:35 -0400
In-Reply-To: <73464C75-47D9-4409-87C2-8DDD0CBD5C67@xyratex.com>
Sender: linux-scsi-owner@vger.kernel.org
List-Id: linux-scsi@vger.kernel.org
To: Jack Wang <jack_wang@usish.com>
Cc: Mark Salyzyn <mark_salyzyn@xyratex.com>, linux-scsi@vger.kernel.org, 'James Bottomley' <JBottomley@parallels.com>, 'lindar_liu' <lindar_liu@usish.com>, =?utf-8?B?J+S6jueIseWNjic=?= <crystal_yu@usish.com>, 'john_gong' <john_gong@usish.com>

Upon further investigation of the documentation, bit [2] (0x04) is the Force HDA mode, but the flashless operations are bit [3] (0x08) meaning Firmware to be loaded in HDA mode. I expect either SEEPROM bit set is enough.

Xyratex is NOT setting the SEEPROM bits so the flags are 'incorrect' and a value of Zero, yet the hardware is in-fact in HDA-mode because of the lack of flash, so the mpi_uninit_check() still makes loads of sense to capture this state and engineer ourselves out of that problem.

Jack, would the following trailing fragment to replace the six lines of the one you noted below resolve your concern? :

+	return pm8001_ha->main_cfg_tbl.hda_mode_flag & 0xC;

If so, I will 'take2' the patch with this change as I have not seen any other reviews.

Thanks for the excellent catch.

Sincerely -- Mark Salyzyn

On Apr 30, 2012, at 9:49 AM, Mark Salyzyn wrote:

> You are right ... however this code is 'never reached' since mpi_uninit_check() generally fails before this in the chip_in_hda_mode set of tests.
> 
> I instrumented up, on our HDA mode chip, we are seeing a value of ZERO in MAIN_HDA_FLAGS_OFFSET ... Ahh, the world of initializing hardware that is in an indeterminate (code not running) state ...
> 
> -- Mark
> 
> On Apr 28, 2012, at 2:07 AM, Jack Wang wrote:
> 
>> 
>> +	data = pm8001_mr32(
>> +		pm8001_ha->main_cfg_tbl_addr,
>> +		pm8001_ha->main_cfg_tbl.hda_mode_flag);
>> +	PM8001_INIT_DBG(pm8001_ha,
>> +		pm8001_printk("*MAIN_HDA_FLAGS = 0x%x\n", data));
>> +	return data & 0x4;
>> 
>> Here seems wrong, you may want to check the force hda bit? So you need to
>> read offset #define MAIN_HDA_FLAGS_OFFSET		0x84/* DWORD 0x21 */
>> 
>> Jack
>> 
>> 
>>> 
>>> The pm8001 can be delivered as a standalone product with flash-programmed
>>> firmware images, or without the flash present requiring the driver to
>> upload
>>> the images into the chip's RAM and then run. This is called HDA mode.
>>> 
>>> We add support for this firmware upload in the enclosed patch. We try some
>>> basic initialization checks of the Firmware, and if it appears dead, we
>> make
>>> the assumption the adapter must in-fact be halted in this HDA mode. The
>>> Firmware images themselves have not been cleared for open-release by PMC,
>> but
>>> they are available in OpenSolaris <hint hint>. PMC's rationalization for
>> not
>>> wanting an open-release of the Firmware Images is that they do not want to
>>> take support calls except from paying OEMs (such as Xyratex) that are
>> embedding
>>> PMC product into the motherboards and thus may have a tested combination
>> of
>>> Firmware and Hardware. Please respect this sentiment. Images are expected
>> in:
>>> 
>>> /lib/firmware/aap1img.bin
>>> /lib/firmware/ilaimg.bin
>>> /lib/firmware/iopimg.bin
>>> /lib/firmware/istrimg.bin
>>> 
>>> using the exact same naming convention as PMC and in OpenSolaris (and its
>>> followon children) for these image files.
>>> 
>>> Signed-off-by: Mark Salyzyn <mark_salyzyn@xyratex.com>
>>> 
>>> drivers/scsi/pm8001/pm8001_hwi.c  |  584
>>> +++++++++++++++++++++++++++++++++++---
>>> drivers/scsi/pm8001/pm8001_hwi.h  |   37 ++
>>> drivers/scsi/pm8001/pm8001_init.c |   30 +
>>> drivers/scsi/pm8001/pm8001_sas.h  |    3
>>> 4 files changed, 613 insertions(+), 41 deletions(-)
>>> 
>>> Please see enclosed attachment
>> 
>> 
>