viodasd.c

Linux块设备驱动源码

/* -*- linux-c -*-
 * viodasd.c
 *  Authors: Dave Boutcher <[e-mail]>
 *           Ryan Arnold <[e-mail]>
 *           Colin Devilbiss <[e-mail]>
 *           Stephen Rothwell <[e-mail]>
 *
 * (C) Copyright 2000-2004 IBM Corporation
 *
 * 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
 *
 * This routine provides access to disk space (termed "DASD" in historical
 * IBM terms) owned and managed by an OS/400 partition running on the
 * same box as this Linux partition.
 *
 * All disk operations are performed by sending messages back and forth to
 * the OS/400 partition.
 */
#include <linux/major.h>
#include <linux/fs.h>
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/blkdev.h>
#include <linux/genhd.h>
#include <linux/hdreg.h>
#include <linux/errno.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/dma-mapping.h>
#include <linux/completion.h>
#include <linux/device.h>
#include <linux/kernel.h>

#include <asm/uaccess.h>
#include <asm/vio.h>
#include <asm/iSeries/HvTypes.h>
#include <asm/iSeries/HvLpEvent.h>
#include <asm/iSeries/HvLpConfig.h>
#include <asm/iSeries/vio.h>

MODULE_DESCRIPTION("iSeries Virtual DASD");
MODULE_AUTHOR("Dave Boutcher");
MODULE_LICENSE("GPL");

/*
 * We only support 7 partitions per physical disk....so with minor
 * numbers 0-255 we get a maximum of 32 disks.
 */
#define VIOD_GENHD_NAME		"iseries/vd"
#define VIOD_GENHD_DEVFS_NAME	"iseries/disc"

#define VIOD_VERS		"1.64"

#define VIOD_KERN_WARNING	KERN_WARNING "viod: "
#define VIOD_KERN_INFO		KERN_INFO "viod: "

enum {
	PARTITION_SHIFT = 3,
	MAX_DISKNO = HVMAXARCHITECTEDVIRTUALDISKS,
	MAX_DISK_NAME = sizeof(((struct gendisk *)0)->disk_name)
};

static DEFINE_SPINLOCK(viodasd_spinlock);

#define VIOMAXREQ		16
#define VIOMAXBLOCKDMA		12

#define DEVICE_NO(cell)	((struct viodasd_device *)(cell) - &viodasd_devices[0])

struct open_data {
	u64	disk_size;
	u16	max_disk;
	u16	cylinders;
	u16	tracks;
	u16	sectors;
	u16	bytes_per_sector;
};

struct rw_data {
	u64	offset;
	struct {
		u32	token;
		u32	reserved;
		u64	len;
	} dma_info[VIOMAXBLOCKDMA];
};

struct vioblocklpevent {
	struct HvLpEvent	event;
	u32			reserved;
	u16			version;
	u16			sub_result;
	u16			disk;
	u16			flags;
	union {
		struct open_data	open_data;
		struct rw_data		rw_data;
		u64			changed;
	} u;
};

#define vioblockflags_ro   0x0001

enum vioblocksubtype {
	vioblockopen = 0x0001,
	vioblockclose = 0x0002,
	vioblockread = 0x0003,
	vioblockwrite = 0x0004,
	vioblockflush = 0x0005,
	vioblockcheck = 0x0007
};

struct viodasd_waitevent {
	struct completion	com;
	int			rc;
	u16			sub_result;
	int			max_disk;	/* open */
};

static const struct vio_error_entry viodasd_err_table[] = {
	{ 0x0201, EINVAL, "Invalid Range" },
	{ 0x0202, EINVAL, "Invalid Token" },
	{ 0x0203, EIO, "DMA Error" },
	{ 0x0204, EIO, "Use Error" },
	{ 0x0205, EIO, "Release Error" },
	{ 0x0206, EINVAL, "Invalid Disk" },
	{ 0x0207, EBUSY, "Cant Lock" },
	{ 0x0208, EIO, "Already Locked" },
	{ 0x0209, EIO, "Already Unlocked" },
	{ 0x020A, EIO, "Invalid Arg" },
	{ 0x020B, EIO, "Bad IFS File" },
	{ 0x020C, EROFS, "Read Only Device" },
	{ 0x02FF, EIO, "Internal Error" },
	{ 0x0000, 0, NULL },
};

/*
 * Figure out the biggest I/O request (in sectors) we can accept
 */
#define VIODASD_MAXSECTORS (4096 / 512 * VIOMAXBLOCKDMA)

/*
 * Number of disk I/O requests we've sent to OS/400
 */
static int num_req_outstanding;

/*
 * This is our internal structure for keeping track of disk devices
 */
struct viodasd_device {
	u16		cylinders;
	u16		tracks;
	u16		sectors;
	u16		bytes_per_sector;
	u64		size;
	int		read_only;
	spinlock_t	q_lock;
	struct gendisk	*disk;
	struct device	*dev;
} viodasd_devices[MAX_DISKNO];

/*
 * External open entry point.
 */
static int viodasd_open(struct inode *ino, struct file *fil)
{
	struct viodasd_device *d = ino->i_bdev->bd_disk->private_data;
	HvLpEvent_Rc hvrc;
	struct viodasd_waitevent we;
	u16 flags = 0;

	if (d->read_only) {
		if ((fil != NULL) && (fil->f_mode & FMODE_WRITE))
			return -EROFS;
		flags = vioblockflags_ro;
	}

	init_completion(&we.com);

	/* Send the open event to OS/400 */
	hvrc = HvCallEvent_signalLpEventFast(viopath_hostLp,
			HvLpEvent_Type_VirtualIo,
			viomajorsubtype_blockio | vioblockopen,
			HvLpEvent_AckInd_DoAck, HvLpEvent_AckType_ImmediateAck,
			viopath_sourceinst(viopath_hostLp),
			viopath_targetinst(viopath_hostLp),
			(u64)(unsigned long)&we, VIOVERSION << 16,
			((u64)DEVICE_NO(d) << 48) | ((u64)flags << 32),
			0, 0, 0);
	if (hvrc != 0) {
		printk(VIOD_KERN_WARNING "HV open failed %d\n", (int)hvrc);
		return -EIO;
	}

	wait_for_completion(&we.com);

	/* Check the return code */
	if (we.rc != 0) {
		const struct vio_error_entry *err =
			vio_lookup_rc(viodasd_err_table, we.sub_result);

		printk(VIOD_KERN_WARNING
				"bad rc opening disk: %d:0x%04x (%s)\n",
				(int)we.rc, we.sub_result, err->msg);
		return -EIO;
	}

	return 0;
}

/*
 * External release entry point.
 */
static int viodasd_release(struct inode *ino, struct file *fil)
{
	struct viodasd_device *d = ino->i_bdev->bd_disk->private_data;
	HvLpEvent_Rc hvrc;

	/* Send the event to OS/400.  We DON'T expect a response */
	hvrc = HvCallEvent_signalLpEventFast(viopath_hostLp,
			HvLpEvent_Type_VirtualIo,
			viomajorsubtype_blockio | vioblockclose,
			HvLpEvent_AckInd_NoAck, HvLpEvent_AckType_ImmediateAck,
			viopath_sourceinst(viopath_hostLp),
			viopath_targetinst(viopath_hostLp),
			0, VIOVERSION << 16,
			((u64)DEVICE_NO(d) << 48) /* | ((u64)flags << 32) */,
			0, 0, 0);
	if (hvrc != 0)
		printk(VIOD_KERN_WARNING "HV close call failed %d\n",
				(int)hvrc);
	return 0;
}


/* External ioctl entry point.
 */
static int viodasd_ioctl(struct inode *ino, struct file *fil,
			 unsigned int cmd, unsigned long arg)
{
	unsigned char sectors;
	unsigned char heads;
	unsigned short cylinders;
	struct hd_geometry *geo;
	struct gendisk *gendisk;
	struct viodasd_device *d;

	switch (cmd) {
	case HDIO_GETGEO:
		geo = (struct hd_geometry *)arg;
		if (geo == NULL)
			return -EINVAL;
		if (!access_ok(VERIFY_WRITE, geo, sizeof(*geo)))
			return -EFAULT;
		gendisk = ino->i_bdev->bd_disk;
		d = gendisk->private_data;
		sectors = d->sectors;
		if (sectors == 0)
			sectors = 32;
		heads = d->tracks;
		if (heads == 0)
			heads = 64;
		cylinders = d->cylinders;
		if (cylinders == 0)
			cylinders = get_capacity(gendisk) / (sectors * heads);
		if (__put_user(sectors, &geo->sectors) ||
		    __put_user(heads, &geo->heads) ||
		    __put_user(cylinders, &geo->cylinders) ||
		    __put_user(get_start_sect(ino->i_bdev), &geo->start))
			return -EFAULT;
		return 0;
	}

	return -EINVAL;
}

/*
 * Our file operations table
 */
static struct block_device_operations viodasd_fops = {
	.owner = THIS_MODULE,
	.open = viodasd_open,
	.release = viodasd_release,
	.ioctl = viodasd_ioctl,
};

/*
 * End a request
 */
static void viodasd_end_request(struct request *req, int uptodate,
		int num_sectors)
{
	if (end_that_request_first(req, uptodate, num_sectors))
		return;
	add_disk_randomness(req->rq_disk);
	end_that_request_last(req);
}

/*
 * Send an actual I/O request to OS/400
 */
static int send_request(struct request *req)
{
	u64 start;
	int direction;
	int nsg;
	u16 viocmd;
	HvLpEvent_Rc hvrc;
	struct vioblocklpevent *bevent;
	struct scatterlist sg[VIOMAXBLOCKDMA];
	int sgindex;
	int statindex;
	struct viodasd_device *d;
	unsigned long flags;

	start = (u64)req->sector << 9;

	if (rq_data_dir(req) == READ) {
		direction = DMA_FROM_DEVICE;
		viocmd = viomajorsubtype_blockio | vioblockread;
		statindex = 0;
	} else {
		direction = DMA_TO_DEVICE;
		viocmd = viomajorsubtype_blockio | vioblockwrite;
		statindex = 1;
	}

        d = req->rq_disk->private_data;

	/* Now build the scatter-gather list */
	nsg = blk_rq_map_sg(req->q, req, sg);
	nsg = dma_map_sg(d->dev, sg, nsg, direction);

	spin_lock_irqsave(&viodasd_spinlock, flags);
	num_req_outstanding++;

	/* This optimization handles a single DMA block */
	if (nsg == 1)
		hvrc = HvCallEvent_signalLpEventFast(viopath_hostLp,
				HvLpEvent_Type_VirtualIo, viocmd,
				HvLpEvent_AckInd_DoAck,
				HvLpEvent_AckType_ImmediateAck,
				viopath_sourceinst(viopath_hostLp),
				viopath_targetinst(viopath_hostLp),
				(u64)(unsigned long)req, VIOVERSION << 16,
				((u64)DEVICE_NO(d) << 48), start,
				((u64)sg_dma_address(&sg[0])) << 32,
				sg_dma_len(&sg[0]));
	else {
		bevent = (struct vioblocklpevent *)
			vio_get_event_buffer(viomajorsubtype_blockio);
		if (bevent == NULL) {
			printk(VIOD_KERN_WARNING
			       "error allocating disk event buffer\n");
			goto error_ret;
		}

		/*
		 * Now build up the actual request.  Note that we store
		 * the pointer to the request in the correlation
		 * token so we can match the response up later
		 */
		memset(bevent, 0, sizeof(struct vioblocklpevent));
		bevent->event.xFlags.xValid = 1;
		bevent->event.xFlags.xFunction = HvLpEvent_Function_Int;
		bevent->event.xFlags.xAckInd = HvLpEvent_AckInd_DoAck;
		bevent->event.xFlags.xAckType = HvLpEvent_AckType_ImmediateAck;
		bevent->event.xType = HvLpEvent_Type_VirtualIo;
		bevent->event.xSubtype = viocmd;
		bevent->event.xSourceLp = HvLpConfig_getLpIndex();
		bevent->event.xTargetLp = viopath_hostLp;
		bevent->event.xSizeMinus1 =
			offsetof(struct vioblocklpevent, u.rw_data.dma_info) +
			(sizeof(bevent->u.rw_data.dma_info[0]) * nsg) - 1;
		bevent->event.xSourceInstanceId =
			viopath_sourceinst(viopath_hostLp);
		bevent->event.xTargetInstanceId =
			viopath_targetinst(viopath_hostLp);
		bevent->event.xCorrelationToken = (u64)req;
		bevent->version = VIOVERSION;
		bevent->disk = DEVICE_NO(d);
		bevent->u.rw_data.offset = start;

		/*
		 * Copy just the dma information from the sg list
		 * into the request
		 */
		for (sgindex = 0; sgindex < nsg; sgindex++) {
			bevent->u.rw_data.dma_info[sgindex].token =
				sg_dma_address(&sg[sgindex]);
			bevent->u.rw_data.dma_info[sgindex].len =
				sg_dma_len(&sg[sgindex]);
		}

		/* Send the request */
		hvrc = HvCallEvent_signalLpEvent(&bevent->event);
		vio_free_event_buffer(viomajorsubtype_blockio, bevent);
	}

	if (hvrc != HvLpEvent_Rc_Good) {
		printk(VIOD_KERN_WARNING
		       "error sending disk event to OS/400 (rc %d)\n",
		       (int)hvrc);
		goto error_ret;
	}
	spin_unlock_irqrestore(&viodasd_spinlock, flags);
	return 0;

error_ret:
	num_req_outstanding--;
	spin_unlock_irqrestore(&viodasd_spinlock, flags);
	dma_unmap_sg(d->dev, sg, nsg, direction);