eba.c

来自「linux 内核源代码」· C语言 代码 · 共 1,228 行 · 第 1/3 页

/*
 * Copyright (c) International Business Machines Corp., 2006
 *
 * 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
 *
 * Author: Artem Bityutskiy (Битюцкий Артём)
 */

/*
 * The UBI Eraseblock Association (EBA) unit.
 *
 * This unit is responsible for I/O to/from logical eraseblock.
 *
 * Although in this implementation the EBA table is fully kept and managed in
 * RAM, which assumes poor scalability, it might be (partially) maintained on
 * flash in future implementations.
 *
 * The EBA unit implements per-logical eraseblock locking. Before accessing a
 * logical eraseblock it is locked for reading or writing. The per-logical
 * eraseblock locking is implemented by means of the lock tree. The lock tree
 * is an RB-tree which refers all the currently locked logical eraseblocks. The
 * lock tree elements are &struct ltree_entry objects. They are indexed by
 * (@vol_id, @lnum) pairs.
 *
 * EBA also maintains the global sequence counter which is incremented each
 * time a logical eraseblock is mapped to a physical eraseblock and it is
 * stored in the volume identifier header. This means that each VID header has
 * a unique sequence number. The sequence number is only increased an we assume
 * 64 bits is enough to never overflow.
 */

#include <linux/slab.h>
#include <linux/crc32.h>
#include <linux/err.h>
#include "ubi.h"

/* Number of physical eraseblocks reserved for atomic LEB change operation */
#define EBA_RESERVED_PEBS 1

/**
 * struct ltree_entry - an entry in the lock tree.
 * @rb: links RB-tree nodes
 * @vol_id: volume ID of the locked logical eraseblock
 * @lnum: locked logical eraseblock number
 * @users: how many tasks are using this logical eraseblock or wait for it
 * @mutex: read/write mutex to implement read/write access serialization to
 * the (@vol_id, @lnum) logical eraseblock
 *
 * When a logical eraseblock is being locked - corresponding &struct ltree_entry
 * object is inserted to the lock tree (@ubi->ltree).
 */
struct ltree_entry {
	struct rb_node rb;
	int vol_id;
	int lnum;
	int users;
	struct rw_semaphore mutex;
};

/* Slab cache for lock-tree entries */
static struct kmem_cache *ltree_slab;

/**
 * next_sqnum - get next sequence number.
 * @ubi: UBI device description object
 *
 * This function returns next sequence number to use, which is just the current
 * global sequence counter value. It also increases the global sequence
 * counter.
 */
static unsigned long long next_sqnum(struct ubi_device *ubi)
{
	unsigned long long sqnum;

	spin_lock(&ubi->ltree_lock);
	sqnum = ubi->global_sqnum++;
	spin_unlock(&ubi->ltree_lock);

	return sqnum;
}

/**
 * ubi_get_compat - get compatibility flags of a volume.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 *
 * This function returns compatibility flags for an internal volume. User
 * volumes have no compatibility flags, so %0 is returned.
 */
static int ubi_get_compat(const struct ubi_device *ubi, int vol_id)
{
	if (vol_id == UBI_LAYOUT_VOL_ID)
		return UBI_LAYOUT_VOLUME_COMPAT;
	return 0;
}

/**
 * ltree_lookup - look up the lock tree.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 * @lnum: logical eraseblock number
 *
 * This function returns a pointer to the corresponding &struct ltree_entry
 * object if the logical eraseblock is locked and %NULL if it is not.
 * @ubi->ltree_lock has to be locked.
 */
static struct ltree_entry *ltree_lookup(struct ubi_device *ubi, int vol_id,
					int lnum)
{
	struct rb_node *p;

	p = ubi->ltree.rb_node;
	while (p) {
		struct ltree_entry *le;

		le = rb_entry(p, struct ltree_entry, rb);

		if (vol_id < le->vol_id)
			p = p->rb_left;
		else if (vol_id > le->vol_id)
			p = p->rb_right;
		else {
			if (lnum < le->lnum)
				p = p->rb_left;
			else if (lnum > le->lnum)
				p = p->rb_right;
			else
				return le;
		}
	}

	return NULL;
}

/**
 * ltree_add_entry - add new entry to the lock tree.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 * @lnum: logical eraseblock number
 *
 * This function adds new entry for logical eraseblock (@vol_id, @lnum) to the
 * lock tree. If such entry is already there, its usage counter is increased.
 * Returns pointer to the lock tree entry or %-ENOMEM if memory allocation
 * failed.
 */
static struct ltree_entry *ltree_add_entry(struct ubi_device *ubi, int vol_id,
					   int lnum)
{
	struct ltree_entry *le, *le1, *le_free;

	le = kmem_cache_alloc(ltree_slab, GFP_NOFS);
	if (!le)
		return ERR_PTR(-ENOMEM);

	le->vol_id = vol_id;
	le->lnum = lnum;

	spin_lock(&ubi->ltree_lock);
	le1 = ltree_lookup(ubi, vol_id, lnum);

	if (le1) {
		/*
		 * This logical eraseblock is already locked. The newly
		 * allocated lock entry is not needed.
		 */
		le_free = le;
		le = le1;
	} else {
		struct rb_node **p, *parent = NULL;

		/*
		 * No lock entry, add the newly allocated one to the
		 * @ubi->ltree RB-tree.
		 */
		le_free = NULL;

		p = &ubi->ltree.rb_node;
		while (*p) {
			parent = *p;
			le1 = rb_entry(parent, struct ltree_entry, rb);

			if (vol_id < le1->vol_id)
				p = &(*p)->rb_left;
			else if (vol_id > le1->vol_id)
				p = &(*p)->rb_right;
			else {
				ubi_assert(lnum != le1->lnum);
				if (lnum < le1->lnum)
					p = &(*p)->rb_left;
				else
					p = &(*p)->rb_right;
			}
		}

		rb_link_node(&le->rb, parent, p);
		rb_insert_color(&le->rb, &ubi->ltree);
	}
	le->users += 1;
	spin_unlock(&ubi->ltree_lock);

	if (le_free)
		kmem_cache_free(ltree_slab, le_free);

	return le;
}

/**
 * leb_read_lock - lock logical eraseblock for reading.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 * @lnum: logical eraseblock number
 *
 * This function locks a logical eraseblock for reading. Returns zero in case
 * of success and a negative error code in case of failure.
 */
static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
{
	struct ltree_entry *le;

	le = ltree_add_entry(ubi, vol_id, lnum);
	if (IS_ERR(le))
		return PTR_ERR(le);
	down_read(&le->mutex);
	return 0;
}

/**
 * leb_read_unlock - unlock logical eraseblock.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 * @lnum: logical eraseblock number
 */
static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
{
	int free = 0;
	struct ltree_entry *le;

	spin_lock(&ubi->ltree_lock);
	le = ltree_lookup(ubi, vol_id, lnum);
	le->users -= 1;
	ubi_assert(le->users >= 0);
	if (le->users == 0) {
		rb_erase(&le->rb, &ubi->ltree);
		free = 1;
	}
	spin_unlock(&ubi->ltree_lock);

	up_read(&le->mutex);
	if (free)
		kmem_cache_free(ltree_slab, le);
}

/**
 * leb_write_lock - lock logical eraseblock for writing.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 * @lnum: logical eraseblock number
 *
 * This function locks a logical eraseblock for writing. Returns zero in case
 * of success and a negative error code in case of failure.
 */
static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
{
	struct ltree_entry *le;

	le = ltree_add_entry(ubi, vol_id, lnum);
	if (IS_ERR(le))
		return PTR_ERR(le);
	down_write(&le->mutex);
	return 0;
}

/**
 * leb_write_unlock - unlock logical eraseblock.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 * @lnum: logical eraseblock number
 */
static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
{
	int free;
	struct ltree_entry *le;

	spin_lock(&ubi->ltree_lock);
	le = ltree_lookup(ubi, vol_id, lnum);
	le->users -= 1;
	ubi_assert(le->users >= 0);
	if (le->users == 0) {
		rb_erase(&le->rb, &ubi->ltree);
		free = 1;
	} else
		free = 0;
	spin_unlock(&ubi->ltree_lock);

	up_write(&le->mutex);
	if (free)
		kmem_cache_free(ltree_slab, le);
}

/**
 * ubi_eba_unmap_leb - un-map logical eraseblock.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 * @lnum: logical eraseblock number
 *
 * This function un-maps logical eraseblock @lnum and schedules corresponding
 * physical eraseblock for erasure. Returns zero in case of success and a
 * negative error code in case of failure.
 */
int ubi_eba_unmap_leb(struct ubi_device *ubi, int vol_id, int lnum)
{
	int idx = vol_id2idx(ubi, vol_id), err, pnum;
	struct ubi_volume *vol = ubi->volumes[idx];

	if (ubi->ro_mode)
		return -EROFS;

	err = leb_write_lock(ubi, vol_id, lnum);
	if (err)
		return err;

	pnum = vol->eba_tbl[lnum];
	if (pnum < 0)
		/* This logical eraseblock is already unmapped */
		goto out_unlock;

	dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);

	vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
	err = ubi_wl_put_peb(ubi, pnum, 0);

out_unlock:
	leb_write_unlock(ubi, vol_id, lnum);
	return err;
}

/**
 * ubi_eba_read_leb - read data.
 * @ubi: UBI device description object
 * @vol_id: volume ID
 * @lnum: logical eraseblock number
 * @buf: buffer to store the read data
 * @offset: offset from where to read
 * @len: how many bytes to read
 * @check: data CRC check flag
 *
 * If the logical eraseblock @lnum is unmapped, @buf is filled with 0xFF
 * bytes. The @check flag only makes sense for static volumes and forces
 * eraseblock data CRC checking.
 *
 * In case of success this function returns zero. In case of a static volume,
 * if data CRC mismatches - %-EBADMSG is returned. %-EBADMSG may also be
 * returned for any volume type if an ECC error was detected by the MTD device
 * driver. Other negative error cored may be returned in case of other errors.
 */
int ubi_eba_read_leb(struct ubi_device *ubi, int vol_id, int lnum, void *buf,
		     int offset, int len, int check)
{
	int err, pnum, scrub = 0, idx = vol_id2idx(ubi, vol_id);
	struct ubi_vid_hdr *vid_hdr;
	struct ubi_volume *vol = ubi->volumes[idx];
	uint32_t uninitialized_var(crc);

	err = leb_read_lock(ubi, vol_id, lnum);
	if (err)
		return err;

	pnum = vol->eba_tbl[lnum];
	if (pnum < 0) {
		/*
		 * The logical eraseblock is not mapped, fill the whole buffer
		 * with 0xFF bytes. The exception is static volumes for which
		 * it is an error to read unmapped logical eraseblocks.
		 */
		dbg_eba("read %d bytes from offset %d of LEB %d:%d (unmapped)",
			len, offset, vol_id, lnum);
		leb_read_unlock(ubi, vol_id, lnum);
		ubi_assert(vol->vol_type != UBI_STATIC_VOLUME);
		memset(buf, 0xFF, len);
		return 0;
	}

	dbg_eba("read %d bytes from offset %d of LEB %d:%d, PEB %d",
		len, offset, vol_id, lnum, pnum);

	if (vol->vol_type == UBI_DYNAMIC_VOLUME)
		check = 0;

retry:
	if (check) {
		vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
		if (!vid_hdr) {
			err = -ENOMEM;
			goto out_unlock;
		}

		err = ubi_io_read_vid_hdr(ubi, pnum, vid_hdr, 1);
		if (err && err != UBI_IO_BITFLIPS) {