pktcdvd.c

Linux块设备驱动源码

/*
 * Copyright (C) 2000 Jens Axboe <[e-mail]>
 * Copyright (C) 2001-2004 Peter Osterlund <[e-mail]>
 *
 * May be copied or modified under the terms of the GNU General Public
 * License.  See linux/COPYING for more information.
 *
 * Packet writing layer for ATAPI and SCSI CD-RW, DVD+RW, DVD-RW and
 * DVD-RAM devices.
 *
 * Theory of operation:
 *
 * At the lowest level, there is the standard driver for the CD/DVD device,
 * typically ide-cd.c or sr.c. This driver can handle read and write requests,
 * but it doesn't know anything about the special restrictions that apply to
 * packet writing. One restriction is that write requests must be aligned to
 * packet boundaries on the physical media, and the size of a write request
 * must be equal to the packet size. Another restriction is that a
 * GPCMD_FLUSH_CACHE command has to be issued to the drive before a read
 * command, if the previous command was a write.
 *
 * The purpose of the packet writing driver is to hide these restrictions from
 * higher layers, such as file systems, and present a block device that can be
 * randomly read and written using 2kB-sized blocks.
 *
 * The lowest layer in the packet writing driver is the packet I/O scheduler.
 * Its data is defined by the struct packet_iosched and includes two bio
 * queues with pending read and write requests. These queues are processed
 * by the pkt_iosched_process_queue() function. The write requests in this
 * queue are already properly aligned and sized. This layer is responsible for
 * issuing the flush cache commands and scheduling the I/O in a good order.
 *
 * The next layer transforms unaligned write requests to aligned writes. This
 * transformation requires reading missing pieces of data from the underlying
 * block device, assembling the pieces to full packets and queuing them to the
 * packet I/O scheduler.
 *
 * At the top layer there is a custom make_request_fn function that forwards
 * read requests directly to the iosched queue and puts write requests in the
 * unaligned write queue. A kernel thread performs the necessary read
 * gathering to convert the unaligned writes to aligned writes and then feeds
 * them to the packet I/O scheduler.
 *
 *************************************************************************/

#define VERSION_CODE	"v0.2.0a 2004-07-14 Jens Axboe ([e-mail]) and [e-mail]"

#include <linux/pktcdvd.h>
#include <linux/config.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/kthread.h>
#include <linux/errno.h>
#include <linux/spinlock.h>
#include <linux/file.h>
#include <linux/proc_fs.h>
#include <linux/seq_file.h>
#include <linux/miscdevice.h>
#include <linux/suspend.h>
#include <scsi/scsi_cmnd.h>
#include <scsi/scsi_ioctl.h>

#include <asm/uaccess.h>

#if PACKET_DEBUG
#define DPRINTK(fmt, args...) printk(KERN_NOTICE fmt, ##args)
#else
#define DPRINTK(fmt, args...)
#endif

#if PACKET_DEBUG > 1
#define VPRINTK(fmt, args...) printk(KERN_NOTICE fmt, ##args)
#else
#define VPRINTK(fmt, args...)
#endif

#define MAX_SPEED 0xffff

#define ZONE(sector, pd) (((sector) + (pd)->offset) & ~((pd)->settings.size - 1))

static struct pktcdvd_device *pkt_devs[MAX_WRITERS];
static struct proc_dir_entry *pkt_proc;
static int pkt_major;
static struct semaphore ctl_mutex;	/* Serialize open/close/setup/teardown */
static mempool_t *psd_pool;


static void pkt_bio_finished(struct pktcdvd_device *pd)
{
	BUG_ON(atomic_read(&pd->cdrw.pending_bios) <= 0);
	if (atomic_dec_and_test(&pd->cdrw.pending_bios)) {
		VPRINTK("pktcdvd: queue empty\n");
		atomic_set(&pd->iosched.attention, 1);
		wake_up(&pd->wqueue);
	}
}

static void pkt_bio_destructor(struct bio *bio)
{
	kfree(bio->bi_io_vec);
	kfree(bio);
}

static struct bio *pkt_bio_alloc(int nr_iovecs)
{
	struct bio_vec *bvl = NULL;
	struct bio *bio;

	bio = kmalloc(sizeof(struct bio), GFP_KERNEL);
	if (!bio)
		goto no_bio;
	bio_init(bio);

	bvl = kcalloc(nr_iovecs, sizeof(struct bio_vec), GFP_KERNEL);
	if (!bvl)
		goto no_bvl;

	bio->bi_max_vecs = nr_iovecs;
	bio->bi_io_vec = bvl;
	bio->bi_destructor = pkt_bio_destructor;

	return bio;

 no_bvl:
	kfree(bio);
 no_bio:
	return NULL;
}

/*
 * Allocate a packet_data struct
 */
static struct packet_data *pkt_alloc_packet_data(void)
{
	int i;
	struct packet_data *pkt;

	pkt = kzalloc(sizeof(struct packet_data), GFP_KERNEL);
	if (!pkt)
		goto no_pkt;

	pkt->w_bio = pkt_bio_alloc(PACKET_MAX_SIZE);
	if (!pkt->w_bio)
		goto no_bio;

	for (i = 0; i < PAGES_PER_PACKET; i++) {
		pkt->pages[i] = alloc_page(GFP_KERNEL|__GFP_ZERO);
		if (!pkt->pages[i])
			goto no_page;
	}

	spin_lock_init(&pkt->lock);

	for (i = 0; i < PACKET_MAX_SIZE; i++) {
		struct bio *bio = pkt_bio_alloc(1);
		if (!bio)
			goto no_rd_bio;
		pkt->r_bios[i] = bio;
	}

	return pkt;

no_rd_bio:
	for (i = 0; i < PACKET_MAX_SIZE; i++) {
		struct bio *bio = pkt->r_bios[i];
		if (bio)
			bio_put(bio);
	}

no_page:
	for (i = 0; i < PAGES_PER_PACKET; i++)
		if (pkt->pages[i])
			__free_page(pkt->pages[i]);
	bio_put(pkt->w_bio);
no_bio:
	kfree(pkt);
no_pkt:
	return NULL;
}

/*
 * Free a packet_data struct
 */
static void pkt_free_packet_data(struct packet_data *pkt)
{
	int i;

	for (i = 0; i < PACKET_MAX_SIZE; i++) {
		struct bio *bio = pkt->r_bios[i];
		if (bio)
			bio_put(bio);
	}
	for (i = 0; i < PAGES_PER_PACKET; i++)
		__free_page(pkt->pages[i]);
	bio_put(pkt->w_bio);
	kfree(pkt);
}

static void pkt_shrink_pktlist(struct pktcdvd_device *pd)
{
	struct packet_data *pkt, *next;

	BUG_ON(!list_empty(&pd->cdrw.pkt_active_list));

	list_for_each_entry_safe(pkt, next, &pd->cdrw.pkt_free_list, list) {
		pkt_free_packet_data(pkt);
	}
}

static int pkt_grow_pktlist(struct pktcdvd_device *pd, int nr_packets)
{
	struct packet_data *pkt;

	INIT_LIST_HEAD(&pd->cdrw.pkt_free_list);
	INIT_LIST_HEAD(&pd->cdrw.pkt_active_list);
	spin_lock_init(&pd->cdrw.active_list_lock);
	while (nr_packets > 0) {
		pkt = pkt_alloc_packet_data();
		if (!pkt) {
			pkt_shrink_pktlist(pd);
			return 0;
		}
		pkt->id = nr_packets;
		pkt->pd = pd;
		list_add(&pkt->list, &pd->cdrw.pkt_free_list);
		nr_packets--;
	}
	return 1;
}

static void *pkt_rb_alloc(gfp_t gfp_mask, void *data)
{
	return kmalloc(sizeof(struct pkt_rb_node), gfp_mask);
}

static void pkt_rb_free(void *ptr, void *data)
{
	kfree(ptr);
}

static inline struct pkt_rb_node *pkt_rbtree_next(struct pkt_rb_node *node)
{
	struct rb_node *n = rb_next(&node->rb_node);
	if (!n)
		return NULL;
	return rb_entry(n, struct pkt_rb_node, rb_node);
}

static inline void pkt_rbtree_erase(struct pktcdvd_device *pd, struct pkt_rb_node *node)
{
	rb_erase(&node->rb_node, &pd->bio_queue);
	mempool_free(node, pd->rb_pool);
	pd->bio_queue_size--;
	BUG_ON(pd->bio_queue_size < 0);
}

/*
 * Find the first node in the pd->bio_queue rb tree with a starting sector >= s.
 */
static struct pkt_rb_node *pkt_rbtree_find(struct pktcdvd_device *pd, sector_t s)
{
	struct rb_node *n = pd->bio_queue.rb_node;
	struct rb_node *next;
	struct pkt_rb_node *tmp;

	if (!n) {
		BUG_ON(pd->bio_queue_size > 0);
		return NULL;
	}

	for (;;) {
		tmp = rb_entry(n, struct pkt_rb_node, rb_node);
		if (s <= tmp->bio->bi_sector)
			next = n->rb_left;
		else
			next = n->rb_right;
		if (!next)
			break;
		n = next;
	}

	if (s > tmp->bio->bi_sector) {
		tmp = pkt_rbtree_next(tmp);
		if (!tmp)
			return NULL;
	}
	BUG_ON(s > tmp->bio->bi_sector);
	return tmp;
}

/*
 * Insert a node into the pd->bio_queue rb tree.
 */
static void pkt_rbtree_insert(struct pktcdvd_device *pd, struct pkt_rb_node *node)
{
	struct rb_node **p = &pd->bio_queue.rb_node;
	struct rb_node *parent = NULL;
	sector_t s = node->bio->bi_sector;
	struct pkt_rb_node *tmp;

	while (*p) {
		parent = *p;
		tmp = rb_entry(parent, struct pkt_rb_node, rb_node);
		if (s < tmp->bio->bi_sector)
			p = &(*p)->rb_left;
		else
			p = &(*p)->rb_right;
	}
	rb_link_node(&node->rb_node, parent, p);
	rb_insert_color(&node->rb_node, &pd->bio_queue);
	pd->bio_queue_size++;
}

/*
 * Add a bio to a single linked list defined by its head and tail pointers.
 */
static inline void pkt_add_list_last(struct bio *bio, struct bio **list_head, struct bio **list_tail)
{
	bio->bi_next = NULL;
	if (*list_tail) {
		BUG_ON((*list_head) == NULL);
		(*list_tail)->bi_next = bio;
		(*list_tail) = bio;
	} else {
		BUG_ON((*list_head) != NULL);
		(*list_head) = bio;
		(*list_tail) = bio;
	}
}

/*
 * Remove and return the first bio from a single linked list defined by its
 * head and tail pointers.
 */
static inline struct bio *pkt_get_list_first(struct bio **list_head, struct bio **list_tail)
{
	struct bio *bio;

	if (*list_head == NULL)
		return NULL;

	bio = *list_head;
	*list_head = bio->bi_next;
	if (*list_head == NULL)
		*list_tail = NULL;

	bio->bi_next = NULL;
	return bio;
}

/*
 * Send a packet_command to the underlying block device and
 * wait for completion.
 */
static int pkt_generic_packet(struct pktcdvd_device *pd, struct packet_command *cgc)
{
	char sense[SCSI_SENSE_BUFFERSIZE];
	request_queue_t *q;
	struct request *rq;
	DECLARE_COMPLETION(wait);
	int err = 0;

	q = bdev_get_queue(pd->bdev);

	rq = blk_get_request(q, (cgc->data_direction == CGC_DATA_WRITE) ? WRITE : READ,
			     __GFP_WAIT);
	rq->errors = 0;
	rq->rq_disk = pd->bdev->bd_disk;
	rq->bio = NULL;
	rq->buffer = NULL;
	rq->timeout = 60*HZ;
	rq->data = cgc->buffer;
	rq->data_len = cgc->buflen;
	rq->sense = sense;
	memset(sense, 0, sizeof(sense));
	rq->sense_len = 0;
	rq->flags |= REQ_BLOCK_PC | REQ_HARDBARRIER;
	if (cgc->quiet)
		rq->flags |= REQ_QUIET;
	memcpy(rq->cmd, cgc->cmd, CDROM_PACKET_SIZE);
	if (sizeof(rq->cmd) > CDROM_PACKET_SIZE)
		memset(rq->cmd + CDROM_PACKET_SIZE, 0, sizeof(rq->cmd) - CDROM_PACKET_SIZE);

	rq->ref_count++;
	rq->flags |= REQ_NOMERGE;
	rq->waiting = &wait;
	rq->end_io = blk_end_sync_rq;
	elv_add_request(q, rq, ELEVATOR_INSERT_BACK, 1);
	generic_unplug_device(q);
	wait_for_completion(&wait);

	if (rq->errors)
		err = -EIO;

	blk_put_request(rq);
	return err;
}

/*
 * A generic sense dump / resolve mechanism should be implemented across
 * all ATAPI + SCSI devices.
 */
static void pkt_dump_sense(struct packet_command *cgc)
{
	static char *info[9] = { "No sense", "Recovered error", "Not ready",
				 "Medium error", "Hardware error", "Illegal request",
				 "Unit attention", "Data protect", "Blank check" };
	int i;
	struct request_sense *sense = cgc->sense;

	printk("pktcdvd:");
	for (i = 0; i < CDROM_PACKET_SIZE; i++)
		printk(" %02x", cgc->cmd[i]);
	printk(" - ");

	if (sense == NULL) {
		printk("no sense\n");
		return;
	}

	printk("sense %02x.%02x.%02x", sense->sense_key, sense->asc, sense->ascq);

	if (sense->sense_key > 8) {
		printk(" (INVALID)\n");
		return;
	}

	printk(" (%s)\n", info[sense->sense_key]);
}

/*
 * flush the drive cache to media
 */
static int pkt_flush_cache(struct pktcdvd_device *pd)
{
	struct packet_command cgc;

	init_cdrom_command(&cgc, NULL, 0, CGC_DATA_NONE);
	cgc.cmd[0] = GPCMD_FLUSH_CACHE;
	cgc.quiet = 1;

	/*
	 * the IMMED bit -- we default to not setting it, although that
	 * would allow a much faster close, this is safer
	 */
#if 0
	cgc.cmd[1] = 1 << 1;
#endif
	return pkt_generic_packet(pd, &cgc);
}

/*
 * speed is given as the normal factor, e.g. 4 for 4x
 */
static int pkt_set_speed(struct pktcdvd_device *pd, unsigned write_speed, unsigned read_speed)
{
	struct packet_command cgc;
	struct request_sense sense;
	int ret;

	init_cdrom_command(&cgc, NULL, 0, CGC_DATA_NONE);
	cgc.sense = &sense;
	cgc.cmd[0] = GPCMD_SET_SPEED;
	cgc.cmd[2] = (read_speed >> 8) & 0xff;
	cgc.cmd[3] = read_speed & 0xff;
	cgc.cmd[4] = (write_speed >> 8) & 0xff;
	cgc.cmd[5] = write_speed & 0xff;

	if ((ret = pkt_generic_packet(pd, &cgc)))
		pkt_dump_sense(&cgc);

	return ret;
}

/*
 * Queue a bio for processing by the low-level CD device. Must be called
 * from process context.
 */
static void pkt_queue_bio(struct pktcdvd_device *pd, struct bio *bio)
{
	spin_lock(&pd->iosched.lock);
	if (bio_data_dir(bio) == READ) {
		pkt_add_list_last(bio, &pd->iosched.read_queue,
				  &pd->iosched.read_queue_tail);
	} else {
		pkt_add_list_last(bio, &pd->iosched.write_queue,
				  &pd->iosched.write_queue_tail);
	}
	spin_unlock(&pd->iosched.lock);

	atomic_set(&pd->iosched.attention, 1);
	wake_up(&pd->wqueue);
}

/*
 * Process the queued read/write requests. This function handles special
 * requirements for CDRW drives:
 * - A cache flush command must be inserted before a read request if the
 *   previous request was a write.
 * - Switching between reading and writing is slow, so don't do it more often
 *   than necessary.
 * - Optimize for throughput at the expense of latency. This means that streaming
 *   writes will never be interrupted by a read, but if the drive has to seek
 *   before the next write, switch to reading instead if there are any pending
 *   read requests.
 * - Set the read speed according to current usage pattern. When only reading
 *   from the device, it's best to use the highest possible read speed, but
 *   when switching often between reading and writing, it's better to have the
 *   same read and write speeds.
 */
static void pkt_iosched_process_queue(struct pktcdvd_device *pd)
{
	request_queue_t *q;

	if (atomic_read(&pd->iosched.attention) == 0)
		return;
	atomic_set(&pd->iosched.attention, 0);

	q = bdev_get_queue(pd->bdev);

	for (;;) {
		struct bio *bio;
		int reads_queued, writes_queued;

		spin_lock(&pd->iosched.lock);