dm-exception-store.c

linux 内核源代码

/*
 * dm-exception-store.c
 *
 * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
 * Copyright (C) 2006 Red Hat GmbH
 *
 * This file is released under the GPL.
 */

#include "dm.h"
#include "dm-snap.h"
#include "dm-io.h"
#include "kcopyd.h"

#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/vmalloc.h>
#include <linux/slab.h>

#define DM_MSG_PREFIX "snapshots"
#define DM_CHUNK_SIZE_DEFAULT_SECTORS 32	/* 16KB */

/*-----------------------------------------------------------------
 * Persistent snapshots, by persistent we mean that the snapshot
 * will survive a reboot.
 *---------------------------------------------------------------*/

/*
 * We need to store a record of which parts of the origin have
 * been copied to the snapshot device.  The snapshot code
 * requires that we copy exception chunks to chunk aligned areas
 * of the COW store.  It makes sense therefore, to store the
 * metadata in chunk size blocks.
 *
 * There is no backward or forward compatibility implemented,
 * snapshots with different disk versions than the kernel will
 * not be usable.  It is expected that "lvcreate" will blank out
 * the start of a fresh COW device before calling the snapshot
 * constructor.
 *
 * The first chunk of the COW device just contains the header.
 * After this there is a chunk filled with exception metadata,
 * followed by as many exception chunks as can fit in the
 * metadata areas.
 *
 * All on disk structures are in little-endian format.  The end
 * of the exceptions info is indicated by an exception with a
 * new_chunk of 0, which is invalid since it would point to the
 * header chunk.
 */

/*
 * Magic for persistent snapshots: "SnAp" - Feeble isn't it.
 */
#define SNAP_MAGIC 0x70416e53

/*
 * The on-disk version of the metadata.
 */
#define SNAPSHOT_DISK_VERSION 1

struct disk_header {
	uint32_t magic;

	/*
	 * Is this snapshot valid.  There is no way of recovering
	 * an invalid snapshot.
	 */
	uint32_t valid;

	/*
	 * Simple, incrementing version. no backward
	 * compatibility.
	 */
	uint32_t version;

	/* In sectors */
	uint32_t chunk_size;
};

struct disk_exception {
	uint64_t old_chunk;
	uint64_t new_chunk;
};

struct commit_callback {
	void (*callback)(void *, int success);
	void *context;
};

/*
 * The top level structure for a persistent exception store.
 */
struct pstore {
	struct dm_snapshot *snap;	/* up pointer to my snapshot */
	int version;
	int valid;
	uint32_t exceptions_per_area;

	/*
	 * Now that we have an asynchronous kcopyd there is no
	 * need for large chunk sizes, so it wont hurt to have a
	 * whole chunks worth of metadata in memory at once.
	 */
	void *area;

	/*
	 * Used to keep track of which metadata area the data in
	 * 'chunk' refers to.
	 */
	uint32_t current_area;

	/*
	 * The next free chunk for an exception.
	 */
	uint32_t next_free;

	/*
	 * The index of next free exception in the current
	 * metadata area.
	 */
	uint32_t current_committed;

	atomic_t pending_count;
	uint32_t callback_count;
	struct commit_callback *callbacks;
	struct dm_io_client *io_client;

	struct workqueue_struct *metadata_wq;
};

static unsigned sectors_to_pages(unsigned sectors)
{
	return sectors / (PAGE_SIZE >> 9);
}

static int alloc_area(struct pstore *ps)
{
	int r = -ENOMEM;
	size_t len;

	len = ps->snap->chunk_size << SECTOR_SHIFT;

	/*
	 * Allocate the chunk_size block of memory that will hold
	 * a single metadata area.
	 */
	ps->area = vmalloc(len);
	if (!ps->area)
		return r;

	return 0;
}

static void free_area(struct pstore *ps)
{
	vfree(ps->area);
	ps->area = NULL;
}

struct mdata_req {
	struct io_region *where;
	struct dm_io_request *io_req;
	struct work_struct work;
	int result;
};

static void do_metadata(struct work_struct *work)
{
	struct mdata_req *req = container_of(work, struct mdata_req, work);

	req->result = dm_io(req->io_req, 1, req->where, NULL);
}

/*
 * Read or write a chunk aligned and sized block of data from a device.
 */
static int chunk_io(struct pstore *ps, uint32_t chunk, int rw, int metadata)
{
	struct io_region where = {
		.bdev = ps->snap->cow->bdev,
		.sector = ps->snap->chunk_size * chunk,
		.count = ps->snap->chunk_size,
	};
	struct dm_io_request io_req = {
		.bi_rw = rw,
		.mem.type = DM_IO_VMA,
		.mem.ptr.vma = ps->area,
		.client = ps->io_client,
		.notify.fn = NULL,
	};
	struct mdata_req req;

	if (!metadata)
		return dm_io(&io_req, 1, &where, NULL);

	req.where = &where;
	req.io_req = &io_req;

	/*
	 * Issue the synchronous I/O from a different thread
	 * to avoid generic_make_request recursion.
	 */
	INIT_WORK(&req.work, do_metadata);
	queue_work(ps->metadata_wq, &req.work);
	flush_workqueue(ps->metadata_wq);

	return req.result;
}

/*
 * Read or write a metadata area.  Remembering to skip the first
 * chunk which holds the header.
 */
static int area_io(struct pstore *ps, uint32_t area, int rw)
{
	int r;
	uint32_t chunk;

	/* convert a metadata area index to a chunk index */
	chunk = 1 + ((ps->exceptions_per_area + 1) * area);

	r = chunk_io(ps, chunk, rw, 0);
	if (r)
		return r;

	ps->current_area = area;
	return 0;
}

static int zero_area(struct pstore *ps, uint32_t area)
{
	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
	return area_io(ps, area, WRITE);
}

static int read_header(struct pstore *ps, int *new_snapshot)
{
	int r;
	struct disk_header *dh;
	chunk_t chunk_size;
	int chunk_size_supplied = 1;

	/*
	 * Use default chunk size (or hardsect_size, if larger) if none supplied
	 */
	if (!ps->snap->chunk_size) {
        	ps->snap->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
		    bdev_hardsect_size(ps->snap->cow->bdev) >> 9);
		ps->snap->chunk_mask = ps->snap->chunk_size - 1;
		ps->snap->chunk_shift = ffs(ps->snap->chunk_size) - 1;
		chunk_size_supplied = 0;
	}

	ps->io_client = dm_io_client_create(sectors_to_pages(ps->snap->
							     chunk_size));
	if (IS_ERR(ps->io_client))
		return PTR_ERR(ps->io_client);

	r = alloc_area(ps);
	if (r)
		return r;

	r = chunk_io(ps, 0, READ, 1);
	if (r)
		goto bad;

	dh = (struct disk_header *) ps->area;

	if (le32_to_cpu(dh->magic) == 0) {
		*new_snapshot = 1;
		return 0;
	}

	if (le32_to_cpu(dh->magic) != SNAP_MAGIC) {
		DMWARN("Invalid or corrupt snapshot");
		r = -ENXIO;
		goto bad;
	}

	*new_snapshot = 0;
	ps->valid = le32_to_cpu(dh->valid);
	ps->version = le32_to_cpu(dh->version);
	chunk_size = le32_to_cpu(dh->chunk_size);

	if (!chunk_size_supplied || ps->snap->chunk_size == chunk_size)
		return 0;

	DMWARN("chunk size %llu in device metadata overrides "
	       "table chunk size of %llu.",
	       (unsigned long long)chunk_size,
	       (unsigned long long)ps->snap->chunk_size);

	/* We had a bogus chunk_size. Fix stuff up. */
	free_area(ps);

	ps->snap->chunk_size = chunk_size;
	ps->snap->chunk_mask = chunk_size - 1;
	ps->snap->chunk_shift = ffs(chunk_size) - 1;

	r = dm_io_client_resize(sectors_to_pages(ps->snap->chunk_size),
				ps->io_client);
	if (r)
		return r;

	r = alloc_area(ps);
	return r;

bad:
	free_area(ps);
	return r;
}

static int write_header(struct pstore *ps)
{
	struct disk_header *dh;

	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);

	dh = (struct disk_header *) ps->area;
	dh->magic = cpu_to_le32(SNAP_MAGIC);
	dh->valid = cpu_to_le32(ps->valid);
	dh->version = cpu_to_le32(ps->version);
	dh->chunk_size = cpu_to_le32(ps->snap->chunk_size);

	return chunk_io(ps, 0, WRITE, 1);
}

/*
 * Access functions for the disk exceptions, these do the endian conversions.
 */
static struct disk_exception *get_exception(struct pstore *ps, uint32_t index)
{
	BUG_ON(index >= ps->exceptions_per_area);

	return ((struct disk_exception *) ps->area) + index;
}

static void read_exception(struct pstore *ps,
			   uint32_t index, struct disk_exception *result)
{
	struct disk_exception *e = get_exception(ps, index);

	/* copy it */
	result->old_chunk = le64_to_cpu(e->old_chunk);
	result->new_chunk = le64_to_cpu(e->new_chunk);
}

static void write_exception(struct pstore *ps,
			    uint32_t index, struct disk_exception *de)
{
	struct disk_exception *e = get_exception(ps, index);

	/* copy it */
	e->old_chunk = cpu_to_le64(de->old_chunk);