sbp2.c

Armlinux ieee1394接口驱动

/*
 * sbp2.c - SBP-2 protocol driver for IEEE-1394
 *
 * Copyright (C) 2000 James Goodwin, Filanet Corporation (www.filanet.com)
 * [e-mail]
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software Foundation,
 * Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
 */

/*
 * Brief Description:
 *
 * This driver implements the Serial Bus Protocol 2 (SBP-2) over IEEE-1394
 * under Linux. The SBP-2 driver is implemented as an IEEE-1394 high-level
 * driver. It also registers as a SCSI lower-level driver in order to accept
 * SCSI commands for transport using SBP-2.
 *
 * Driver Loading:
 *
 * Currently, the SBP-2 driver is supported only as a module. Because the 
 * Linux SCSI stack is not Plug-N-Play aware, module load order is 
 * important. Assuming the SCSI core drivers are either built into the 
 * kernel or already loaded as modules, you should load the IEEE-1394 modules 
 * in the following order:
 *
 * 	ieee1394 (e.g. insmod ieee1394)
 *	ohci1394 (e.g. insmod ohci1394)
 *	sbp2 (e.g. insmod sbp2)
 *
 * The SBP-2 driver will attempt to discover any attached SBP-2 devices when first
 * loaded, or after any IEEE-1394 bus reset (e.g. a hot-plug). It will then print 
 * out a debug message indicating if it was able to discover a SBP-2 device.
 *
 * Currently, the SBP-2 driver will catch any attached SBP-2 devices during the
 * initial scsi bus scan (when the driver is first loaded). To add or remove
 * SBP-2 devices after this initial scan (i.e. if you plug-in or un-plug a 
 * device after the SBP-2 driver is loaded), you must either use the scsi procfs
 * add-single-device, remove-single-device, or a shell script such as 
 * rescan-scsi-bus.sh.
 *
 * The easiest way to add/detect new SBP-2 devices is to run the shell script
 * rescan-scsi-bus.sh (or re-load the SBP-2 driver). This script may be 
 * found at:
 * http://www.garloff.de/kurt/linux/rescan-scsi-bus.sh
 *
 * As an alternative, you may manually add/remove SBP-2 devices via the procfs with
 * add-single-device <h> <b> <t> <l> or remove-single-device <h> <b> <t> <l>, where:
 *	<h> = host (starting at zero for first SCSI adapter)
 *	<b> = bus (normally zero)
 *	<t> = target (starting at zero for first SBP-2 device)
 *	<l> = lun (normally zero)
 *
 * e.g. To manually add/detect a new SBP-2 device
 *	echo "scsi add-single-device 0 0 0 0" > /proc/scsi/scsi
 *
 * e.g. To manually remove a SBP-2 device after it's been unplugged
 *	echo "scsi remove-single-device 0 0 0 0" > /proc/scsi/scsi
 *
 * e.g. To check to see which SBP-2/SCSI devices are currently registered
 * 	cat /proc/scsi/scsi
 *
 * After scanning for new SCSI devices (above), you may access any attached 
 * SBP-2 storage devices as if they were SCSI devices (e.g. mount /dev/sda1, 
 * fdisk, mkfs, etc.).
 *
 *
 * Current Support:
 *
 * The SBP-2 driver is still in an early state, but supports a variety of devices.
 * I have read/written many gigabytes of data from/to SBP-2 drives, and have seen 
 * performance of more than 16 MBytes/s on individual drives (limit of the media 
 * transfer rate).
 *
 * Following are the devices that have been tested successfully:
 *
 *	- Western Digital IEEE-1394 hard drives
 *	- Maxtor IEEE-1394 hard drives
 *	- VST (SmartDisk) IEEE-1394 hard drives and Zip drives (several flavors)
 *	- LaCie IEEE-1394 hard drives (several flavors)
 *	- QPS IEEE-1394 CD-RW/DVD drives and hard drives
 *	- BusLink IEEE-1394 hard drives
 *	- Iomega IEEE-1394 Zip/Jazz drives
 *	- ClubMac IEEE-1394 hard drives
 *	- FirePower IEEE-1394 hard drives
 *	- EzQuest IEEE-1394 hard drives and CD-RW drives
 *	- Castlewood/ADS IEEE-1394 ORB drives
 *	- Evergreen IEEE-1394 hard drives and CD-RW drives
 *	- Addonics IEEE-1394 CD-RW drives
 *	- Bellstor IEEE-1394 hard drives and CD-RW drives
 *	- APDrives IEEE-1394 hard drives
 *	- Fujitsu IEEE-1394 MO drives
 *	- Sony IEEE-1394 CD-RW drives
 *	- Epson IEEE-1394 scanner
 *	- ADS IEEE-1394 memory stick and compact flash readers 
 *	  (e.g. "insmod sbp2 mode_sense_hack=1" for mem stick and flash readers))
 *	- SBP-2 bridge-based devices (LSI, Oxford Semiconductor, Indigita bridges)
 *	- Various other standard IEEE-1394 hard drives and enclosures
 *
 *
 * Performance Issues:
 *
 *	- Make sure you are "not" running fat/fat32 on your attached SBP-2 drives. You'll
 *	  get much better performance formatting the drive ext2 (but you will lose the
 *	  ability to easily move the drive between Windows/Linux).
 *
 *
 * Current Issues:
 *
 *	- Currently, all I/O from the scsi stack is serialized by default, as there
 *	  are some stress issues under investigation with deserialized I/O. To enable
 *	  deserialized I/O for testing, do "insmod sbp2 serialize_io=0"
 *
 *	- Error Handling: SCSI aborts and bus reset requests are handled somewhat
 *	  but the code needs additional debugging.
 *
 *	- The SBP-2 driver is currently only supported as a module. It would not take
 *	  much work to allow it to be compiled into the kernel, but you'd have to 
 *	  add some init code to the kernel to support this... and modules are much
 *	  more flexible anyway.   ;-)
 *
 *     - The scsi stack in recent kernels pass down the data transfer
 *       direction as scsicmd->sc_data_direction, which we should use
 *       instead of the sbp2scsi_direction_table.
 *
 *
 * History:
 *
 *	07/25/00 - Initial revision (JSG)
 *	08/11/00 - Following changes/bug fixes were made (JSG):
 *		   * Bug fix to SCSI procfs code (still needs to be synched with 2.4 kernel).
 *		   * Bug fix where request sense commands were actually sent on the bus.
 *		   * Changed bus reset/abort code to deal with devices that spin up quite
 *		     slowly (which result in SCSI time-outs).
 *		   * "More" properly pull information from device's config rom, for enumeration
 *		     of SBP-2 devices, and determining SBP-2 register offsets.
 *		   * Change Simplified Direct Access Device type to Direct Access Device type in
 *		     returned inquiry data, in order to make the SCSI stack happy.
 *		   * Modified driver to register with the SCSI stack "before" enumerating any attached
 *		     SBP-2 devices. This means that you'll have to use procfs scsi-add-device or 
 *		     some sort of script to discover new SBP-2 devices.
 *		   * Minor re-write of some code and other minor changes.
 *	08/28/00 - Following changes/bug fixes were made (JSG):
 *		   * Bug fixes to scatter/gather support (case of one s/g element)
 *		   * Updated direction table for scsi commands (mostly DVD commands)
 *		   * Retries when trying to detect SBP-2 devices (for slow devices)
 *		   * Slightly better error handling (previously none) when commands time-out.
 *		   * Misc. other bug fixes and code reorganization.
 *	09/13/00 - Following changes/bug fixes were made (JSG)
 *		   * Moved detection/enumeration code to a kernel thread which is woken up when IEEE-1394
 *		     bus resets occur.
 *		   * Added code to handle bus resets and hot-plugging while devices are mounted, but full
 *		     hot-plug support is not quite there yet.
 *		   * Now use speed map to determine speed and max payload sizes for ORBs
 *		   * Clean-up of code and reorganization 
 *	09/19/00 - Added better hot-plug support and other minor changes (JSG)
 *	10/15/00 - Fixes for latest 2.4.0 test kernel, minor fix for hot-plug race. (JSG)
 *	12/03/00 - Created pool of request packet structures for use in sending out sbp2 command
 *		   and agent reset requests. This removes the kmallocs/kfrees in the critical I/O paths,
 *		   and also deals with some subtle race conditions related to allocating and freeing
 *		   packets. (JSG)
 *      12/09/00 - Improved the sbp2 device detection by actually reading the root and unit 
 *		   directory ([e-mail])
 *	12/23/00 - Following changes/enhancements were made (JSG)
 *		   * Only do SCSI to RBC command conversion for Direct Access and Simplified
 *		     Direct Access Devices (this is pulled from the config rom root directory).
 *		     This is needed because doing the conversion for all device types broke the
 *		     Epson scanner. Still looking for a better way of determining when to convert
 *		     commands (for RBC devices). Thanks to khk for helping on this!
 *		   * Added ability to "emulate" physical dma support, for host adapters such as TILynx.
 *		   * Determine max payload and speed by also looking at the host adapter's max_rec field.
 *	01/19/01 - Added checks to sbp2 login and made the login time-out longer. Also fixed a compile 
 *		   problem for 2.4.0. (JSG)
 *	01/24/01 - Fixed problem when individual s/g elements are 64KB or larger. Needed to break
 *		   up these larger elements, since the sbp2 page table element size is only 16 bits. (JSG)
 *	01/29/01 - Minor byteswap fix for login response (used for reconnect and log out).
 *	03/07/01 - Following changes/enhancements were made (JSG)
 *		   * Changes to allow us to catch the initial scsi bus scan (for detecting sbp2
 *		     devices when first loading sbp2.o). To disable this, un-define 
 *		     SBP2_SUPPORT_INITIAL_BUS_SCAN.
 *		   * Temporary fix to deal with many sbp2 devices that do not support individual
 *		     transfers of greater than 128KB in size. 
 *		   * Mode sense conversion from 6 byte to 10 byte versions for CDRW/DVD devices. (Mark Burton)
 *		   * Define allowing support for goofy sbp2 devices that do not support mode
 *		     sense command at all, allowing them to be mounted rw (such as 1394 memory
 *		     stick and compact flash readers). Define SBP2_MODE_SENSE_WRITE_PROTECT_HACK
 *		     if you need this fix.
 *	03/29/01 - Major performance enhancements and misc. other changes. Thanks to Daniel Berlin for many of
 *		   changes and suggestions for change:
 *		   * Now use sbp2 doorbell and link commands on the fly (instead of serializing requests)
 *		   * Removed all bit fields in an attempt to run on PPC machines (still needs a little more work)
 *		   * Added large request break-up/linking support for sbp2 chipsets that do not support transfers 
 *		     greater than 128KB in size.
 *		   * Bumped up max commands per lun to two, and max total outstanding commands to eight.
 *	04/03/01 - Minor clean-up. Write orb pointer directly if no outstanding commands (saves one 1394 bus
 *		   transaction). Added module load options (bus scan, mode sense hack, max speed, serialize_io,
 *		   no_large_transfers). Better bus reset handling while I/O pending. Set serialize_io to 1 by 
 *		   default (debugging of deserialized I/O in progress).
 *	04/04/01 - Added workaround for PPC Pismo firewire chipset. See #define below. (Daniel Berlin)
 *	04/20/01 - Minor clean-up. Allocate more orb structures when running with sbp2 target chipsets with
 *		   128KB max transfer limit.
 *	06/16/01 - Converted DMA interfaces to pci_dma - Ben Collins
 *							 <[e-mail]
 *	07/22/01 - Use NodeMngr to get info about the local host and
 *		   attached devices. Ben Collins
 *
 *      09/15/01 - Remove detection code, instead subscribe to the nodemgr
 *                 driver management interface.  This also removes the
 *                 initial bus scan stuff since the nodemgr calls
 *                 sbp2_probe for each sbp2 device already on the bus,
 *                 when we register our driver.  This change 
 *                 automtically adds hotplug support to the driver.
 *                                 Kristian Hogsberg <[e-mail]>
 *
 *      11/17/01 - Various bugfixes/cleanups:
 *                 * Remember to logout of device in sbp2_disconnect.
 *                 * If we fail to reconnect to a device after bus reset
 *                   remember to release unit directory, so the ieee1394
 *                   knows we no longer manage it.
 *                 * Unregister scsi hosts in sbp2_remove_host when a
 *                   hpsb_host goes away.
 *                 * Remove stupid hack in sbp2_remove_host.
 *                 * Switched to "manual" module initialization
 *                   (i.e. not scsi_module.c) and moved sbp2_cleanup
 *                   moved sbp2scsi_release to sbp2_module_ext.  The
 *                   release function is called once pr. registered
 *                   scsi host, but sbp2_cleanup should only be called
 *                   upon module unload.  Moved much initialization
 *                   from sbp2scsi_detect to sbp2_module_init.
 *                                 Kristian Hogsberg <[e-mail]>
 */



/*
 * Includes
 */

#include <linux/config.h>
#include <linux/kernel.h>
#include <linux/list.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/fs.h>
#include <linux/poll.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/delay.h>
#include <linux/sched.h>
#include <linux/proc_fs.h>
#include <linux/blk.h>
#include <linux/smp_lock.h>
#include <linux/init.h>
#include <asm/current.h>
#include <asm/uaccess.h>
#include <asm/io.h>
#include <asm/byteorder.h>
#include <asm/system.h>
#include <asm/io.h>
#include <asm/scatterlist.h>

#include "ieee1394.h"
#include "ieee1394_types.h"
#include "ieee1394_core.h"
#include "hosts.h"
#include "nodemgr.h"
#include "highlevel.h"
#include "ieee1394_transactions.h"
#include "ieee1394_hotplug.h"
#include "../scsi/scsi.h"
#include "../scsi/hosts.h"
#include "../scsi/sd.h"
#include "sbp2.h"

/*
 * Module load parameter definitions
 */

/*
 * Set mode_sense_hack to 1 if you have some sort of unusual sbp2 device,
 * like a 1394 memory stick reader, compact flash reader, or MO drive that
 * does not support mode sense. Allows you to mount the media rw instead
 * of ro.
 */
MODULE_PARM(mode_sense_hack,"i");
MODULE_PARM_DESC(mode_sense_hack, "Emulate mode sense for devices like 1394 memory stick readers");
static int mode_sense_hack = 0;

/*
 * Change max_speed on module load if you have a bad IEEE-1394 controller
 * that has trouble running 2KB packets at 400mb.
 *
 * NOTE: On certain OHCI parts I have seen short packets on async transmit
 * (probably due to PCI latency/throughput issues with the part). You can
 * bump down the speed if you are running into problems.
 *
 * Valid values:
 * max_speed = 2 (default: max speed 400mb)
 * max_speed = 1 (max speed 200mb)
 * max_speed = 0 (max speed 100mb)
 */
MODULE_PARM(max_speed,"i");
MODULE_PARM_DESC(max_speed, "Force down max speed (2 = 400mb default, 1 = 200mb, 0 = 100mb)");
static int max_speed = SPEED_400;

/*
 * Set serialize_io to 1 if you'd like only one scsi command sent down to
 * us at a time (debugging).
 */
MODULE_PARM(serialize_io,"i");
MODULE_PARM_DESC(serialize_io, "Serialize all I/O coming down from the scsi drivers (debugging)");
static int serialize_io = 1;	/* serialize I/O until stress issues are resolved */

/*
 * Set no_large_packets to 1 if you'd like to limit the size of requests
 * sent down to us (normally the sbp2 driver will break up any requests to
 * any individual devices with 128KB transfer size limits).  Sets max s/g
 * list elements to 0x1f in size and disables s/g clustering.
 */
MODULE_PARM(no_large_packets,"i");
MODULE_PARM_DESC(no_large_packets, "Do not allow large transfers from scsi drivers (debugging)");
static int no_large_packets = 0;

/*
 * Export information about protocols/devices supported by this driver.
 */
static struct ieee1394_device_id sbp2_id_table[] = {
	{
		match_flags:  IEEE1394_MATCH_SPECIFIER_ID |
		              IEEE1394_MATCH_VERSION,
		specifier_id: SBP2_UNIT_SPEC_ID_ENTRY & 0xffffff,
		version:      SBP2_SW_VERSION_ENTRY & 0xffffff
	},
	{ }
};

MODULE_DEVICE_TABLE(ieee1394, sbp2_id_table);

/*
 * Debug levels, configured via kernel config.
 */

#ifdef CONFIG_IEEE1394_SBP2_DEBUG_ORBS
#define SBP2_ORB_DEBUG(fmt, args...)	HPSB_ERR("sbp2("__FUNCTION__"): "fmt, ## args)
static u32 global_outstanding_command_orbs = 0;
#define outstanding_orb_incr global_outstanding_command_orbs++
#define outstanding_orb_decr global_outstanding_command_orbs--
#else
#define SBP2_ORB_DEBUG(fmt, args...)
#define outstanding_orb_incr
#define outstanding_orb_decr
#endif

#ifdef CONFIG_IEEE1394_SBP2_DEBUG_DMA
#define SBP2_DMA_ALLOC(fmt, args...) \
	HPSB_ERR("sbp2("__FUNCTION__")alloc(%d): "fmt, \
		 ++global_outstanding_dmas, ## args)
#define SBP2_DMA_FREE(fmt, args...) \
	HPSB_ERR("sbp2("__FUNCTION__")free(%d): "fmt, \
		 --global_outstanding_dmas, ## args)
static u32 global_outstanding_dmas = 0;
#else
#define SBP2_DMA_ALLOC(fmt, args...)
#define SBP2_DMA_FREE(fmt, args...)
#endif


#if CONFIG_IEEE1394_SBP2_DEBUG >= 2
#define SBP2_DEBUG(fmt, args...)	HPSB_ERR("sbp2: "fmt, ## args)
#define SBP2_INFO(fmt, args...)		HPSB_ERR("sbp2: "fmt, ## args)
#define SBP2_NOTICE(fmt, args...)	HPSB_ERR("sbp2: "fmt, ## args)
#define SBP2_WARN(fmt, args...)		HPSB_ERR("sbp2: "fmt, ## args)
#elif CONFIG_IEEE1394_SBP2_DEBUG == 1
#define SBP2_DEBUG(fmt, args...)	HPSB_DEBUG("sbp2: "fmt, ## args)
#define SBP2_INFO(fmt, args...)		HPSB_INFO("sbp2: "fmt, ## args)
#define SBP2_NOTICE(fmt, args...)	HPSB_NOTICE("sbp2: "fmt, ## args)
#define SBP2_WARN(fmt, args...)		HPSB_WARN("sbp2: "fmt, ## args)
#else 
#define SBP2_DEBUG(fmt, args...)
#define SBP2_INFO(fmt, args...)
#define SBP2_NOTICE(fmt, args...)
#define SBP2_WARN(fmt, args...)
#endif

#define SBP2_ERR(fmt, args...)		HPSB_ERR("sbp2: "fmt, ## args)

/*
 * Spinlock debugging stuff. I'm playing it safe until the driver has been
 * debugged on SMP. (JSG)
 */
/* #define SBP2_USE_REAL_SPINLOCKS */
#ifdef SBP2_USE_REAL_SPINLOCKS
#define sbp2_spin_lock(lock, flags)	spin_lock_irqsave(lock, flags)	
#define sbp2_spin_unlock(lock, flags)	spin_unlock_irqrestore(lock, flags);
static spinlock_t sbp2_host_info_lock = SPIN_LOCK_UNLOCKED;
#else
#define sbp2_spin_lock(lock, flags)	do {save_flags(flags); cli();} while (0)