medley.c

linux-2.4.29操作系统的源码

/*
 * MEDLEY SOFTWARE RAID DRIVER (Silicon Image 3112 and others)
 *
 * Copyright (C) 2003 Thomas Horsten <[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
 * Copyright (C) 2003 Thomas Horsten <[e-mail]>
 * All Rights Reserved.
 *
 * This driver uses the ATA RAID driver framework and is based on
 * code from Arjan van de Ven's silraid.c and hptraid.c.
 *
 * It is a driver for the Medley software RAID, which is used by
 * some IDE controllers, including the Silicon Image 3112 SATA
 * controller found onboard many modern motherboards, and the
 * CMD680 stand-alone PCI RAID controller.
 *
 * The author has only tested this on the Silicon Image SiI3112.
 * If you have any luck using more than 2 drives, and/or more
 * than one RAID set, and/or any other chip than the SiI3112,
 * please let me know by sending me mail at the above address.
 *
 * Currently, only striped mode is supported for these RAIDs.
 *
 * You are welcome to contact me if you have any questions or
 * suggestions for improvement.
 *
 * Change history:
 *
 * 20040310 - [e-mail]
 *   Removed C99 style variable declarations that confused gcc-2.x
 *   Fixed a bug where more than one RAID set would not be detected correctly
 *   General cleanup for submission to kernel
 *
 * 20031012 - [e-mail]
 *   Added support for BLKRRPART ioctl to re-read partition table
 *
 * 20030801 - [e-mail]
 *   First test release
 *
 */

#include <linux/version.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/smp_lock.h>
#include <linux/blkdev.h>
#include <linux/blkpg.h>
#include <linux/genhd.h>
#include <linux/ioctl.h>

#include <linux/ide.h>
#include <asm/uaccess.h>

#include "ataraid.h"

/*
 * These options can be tuned if the need should occur.
 *
 * Even better, this driver could be changed to allocate the structures
 * dynamically.
 */
#define MAX_DRIVES_PER_SET 8
#define MAX_MEDLEY_ARRAYS 4

/*
 * Set to 1 only if you are debugging the driver, or if it doesn't work
 * the way you expect and you want to to report it.
 *
 * This will produce lots of kernel messages, some of which might
 * help me figure out what is going wrong).
 */
#define DEBUGGING_MEDLEY 0

#if DEBUGGING_MEDLEY
#define dprintk(fmt, args...) printk(fmt, ##args)
#else
#define dprintk(fmt, args...)
#endif

/*
 * Medley RAID metadata structure.
 *
 * The metadata structure is based on the ATA drive ID from the drive itself,
 * with the RAID information in the vendor specific regions.
 *
 * We do not use all the fields, since we only support Striped Sets.
 */
struct medley_metadata {
	u8  driveid0[46];
	u8  ascii_version[8];
	u8  driveid1[52];
	u32 total_sectors_low;
	u32 total_sectors_high;
	u16 reserved0;
	u8  driveid2[142];
	u16 product_id;
	u16 vendor_id;
	u16 minor_ver;
	u16 major_ver;
	u16 creation_timestamp[3];
	u16 chunk_size;
	u16 reserved1;
	u8  drive_number;
	u8  raid_type;
	u8  drives_per_striped_set;
	u8  striped_set_number;
	u8  drives_per_mirrored_set;
	u8  mirrored_set_number;
	u32 rebuild_ptr_low;
	u32 rebuild_ptr_high;
	u32 incarnation_no;
	u8  member_status;
	u8  mirrored_set_state;
	u8  reported_device_location;
	u8  member_location;
	u8  auto_rebuild;
	u8  reserved3[17];
	u16 checksum;
};

/*
 * This struct holds the information about a Medley array
 */
struct medley_array {
	u8       drives;
	u16      chunk_size;
	u32      sectors_per_row;
	u8       chunk_size_log;
	u16      present;
	u16      timestamp[3];
	u32      sectors;
	int      registered;
	atomic_t valid;
	int      access;

	kdev_t   members[MAX_DRIVES_PER_SET];
	struct   block_device *bdev[MAX_DRIVES_PER_SET];
};

static struct medley_array raid[MAX_MEDLEY_ARRAYS];

/*
 * Here we keep the offset of the ATARAID device ID's compared to our
 * own (this will normally be 0, unless another ATARAID driver has
 * registered some arrays before us).
 */
static int medley_devid_offset = 0;

/*
 * This holds the number of detected arrays.
 */
static int medley_arrays = 0;

/*
 * Wait queue for opening device (used when re-reading partition table)
 */
static DECLARE_WAIT_QUEUE_HEAD(medley_wait_open);

/*
 * The interface functions used by the ataraid framework.
 */
static int medley_open(struct inode *inode, struct file *filp);
static int medley_release(struct inode *inode, struct file *filp);
static int medley_ioctl(struct inode *inode, struct file *file,
			unsigned int cmd, unsigned long arg);
static int medley_make_request(request_queue_t * q, int rw,
			       struct buffer_head *bh);

static struct raid_device_operations medley_ops = {
	open:         medley_open,
	release:      medley_release,
	ioctl:        medley_ioctl,
	make_request: medley_make_request
};

/*
 * This is the list of devices to probe.
 */
static const kdev_t probelist[] = {
	MKDEV(IDE0_MAJOR, 0),
	MKDEV(IDE0_MAJOR, 64),
	MKDEV(IDE1_MAJOR, 0),
	MKDEV(IDE1_MAJOR, 64),
	MKDEV(IDE2_MAJOR, 0),
	MKDEV(IDE2_MAJOR, 64),
	MKDEV(IDE3_MAJOR, 0),
	MKDEV(IDE3_MAJOR, 64),
	MKDEV(IDE4_MAJOR, 0),
	MKDEV(IDE4_MAJOR, 64),
	MKDEV(IDE5_MAJOR, 0),
	MKDEV(IDE5_MAJOR, 64),
	MKDEV(IDE6_MAJOR, 0),
	MKDEV(IDE6_MAJOR, 64),
	MKDEV(0, 0)
};

/*
 * Handler for ioctl calls to the virtual device
 */
static int medley_ioctl(struct inode *inode, struct file *file,
			unsigned int cmd, unsigned long arg)
{
	unsigned int minor;
	unsigned long sectors;
	int devminor = (inode->i_rdev >> SHIFT) & MAJOR_MASK;
	int device = devminor - medley_devid_offset;
	int partition;

	dprintk("medley_ioctl\n");

	minor = MINOR(inode->i_rdev) >> SHIFT;

	switch (cmd) {

	case BLKGETSIZE:	/* Return device size */
		if (!arg)
			return -EINVAL;
		sectors =
		    ataraid_gendisk.part[MINOR(inode->i_rdev)].nr_sects;
		dprintk("medley_ioctl: BLKGETSIZE\n");
		if (MINOR(inode->i_rdev) & 15)
			return put_user(sectors, (unsigned long *) arg);
		return put_user(raid[minor - medley_devid_offset].sectors,
				(unsigned long *) arg);
		break;

	case HDIO_GETGEO: {
			struct hd_geometry *loc =
			    (struct hd_geometry *) arg;
			unsigned short bios_cyl = (unsigned short)
			    (raid[minor].sectors / 255 / 63);	/* truncate */

			dprintk("medley_ioctl: HDIO_GETGEO\n");
			if (!loc)
				return -EINVAL;
			if (put_user(255, (byte *) & loc->heads))
				return -EFAULT;
			if (put_user(63, (byte *) & loc->sectors))
				return -EFAULT;
			if (put_user
			    (bios_cyl, (unsigned short *) &loc->cylinders))
				return -EFAULT;
			if (put_user
			    ((unsigned) ataraid_gendisk.
			     part[MINOR(inode->i_rdev)].start_sect,
			     (unsigned long *) &loc->start))
				return -EFAULT;
			return 0;
		}

	case HDIO_GETGEO_BIG: {
			struct hd_big_geometry *loc =
			    (struct hd_big_geometry *) arg;

			dprintk("medley_ioctl: HDIO_GETGEO_BIG\n");
			if (!loc)
				return -EINVAL;
			if (put_user(255, (byte *) & loc->heads))
				return -EFAULT;
			if (put_user(63, (byte *) & loc->sectors))
				return -EFAULT;
			if (put_user
			    (raid[minor - medley_devid_offset].sectors /
			     255 / 63, (unsigned int *) &loc->cylinders))
				return -EFAULT;
			if (put_user
			    ((unsigned) ataraid_gendisk.
			     part[MINOR(inode->i_rdev)].start_sect,
			     (unsigned long *) &loc->start))
				return -EFAULT;
			return 0;
		}

	case BLKROSET:
	case BLKROGET:
	case BLKSSZGET:
		dprintk("medley_ioctl: BLK*\n");
		return blk_ioctl(inode->i_rdev, cmd, arg);

	case BLKRRPART:	/* Re-read partition tables */
		if (!capable(CAP_SYS_ADMIN))
			return -EACCES;
		if (minor != 0)
			return -EINVAL;
		if (atomic_read(&(raid[device].valid)) == 0)
			return -EINVAL;

		atomic_set(&(raid[device].valid), 0);
		if (raid[device].access != 1) {
			atomic_set(&(raid[device].valid), 1);
			return -EBUSY;
		}

		for (partition = 15; partition >= 0; partition--) {
			invalidate_device(MKDEV(ATARAID_MAJOR,
						partition + devminor), 1);
			ataraid_gendisk.part[partition +
					     devminor].start_sect = 0;
			ataraid_gendisk.part[partition +
					     devminor].nr_sects = 0;
		}
		ataraid_register_disk(device, raid[device].sectors);
		atomic_set(&(raid[device].valid), 1);
		wake_up(&medley_wait_open);
		return 0;

	default:
		return -EINVAL;
	}

	return 0;
}

/*
 * Handler to map a request to the real device.
 * If the request cannot be made because it spans multiple disks,
 * we return -1, otherwise we modify the request and return 1.
 */
static int medley_make_request(request_queue_t * q, int rw,
			       struct buffer_head *bh)
{
	u8 disk;
	u32 rsect = bh->b_rsector;
	int device =
	    ((bh->b_rdev >> SHIFT) & MAJOR_MASK) - medley_devid_offset;
	struct medley_array *r = raid + device;

	/* Add the partition offset */
	rsect = rsect + ataraid_gendisk.part[MINOR(bh->b_rdev)].start_sect;

	dprintk("medley_make_request, rsect=%ul\n", rsect);

	/* Detect if the request crosses a chunk barrier */
	if (r->chunk_size_log) {
		if (((rsect & (r->chunk_size - 1)) +
		     (bh->b_size / 512)) > (1 << r->chunk_size_log)) {
			return -1;
		}
	} else {
		if ((rsect / r->chunk_size) !=
		    ((rsect + (bh->b_size / 512) - 1) / r->chunk_size)) {
			return -1;
		}
	}

	/*
	 * Medley arrays are simple enough, since the smallest disk decides the
	 * number of sectors used per disk. So there is no need for the cutoff
	 * magic found in other drivers like hptraid.
	 */
	if (r->chunk_size_log) {
		/* We save some expensive operations (1 div, 1 mul, 1 mod),
		 * if the chunk size is a power of 2, which is true in most
		 * cases (at least with my version of the RAID BIOS).
		 */
		disk = (rsect >> r->chunk_size_log) % r->drives;
		rsect = ((rsect / r->sectors_per_row) <<
			 r->chunk_size_log) + (rsect & (r->chunk_size -
							1));
	} else {
		disk = (rsect / r->chunk_size) % r->drives;
		rsect = rsect / r->sectors_per_row * r->chunk_size +