nodemgmt.c

linux 内核源代码

/*
 * JFFS2 -- Journalling Flash File System, Version 2.
 *
 * Copyright © 2001-2007 Red Hat, Inc.
 *
 * Created by David Woodhouse <[e-mail]>
 *
 * For licensing information, see the file 'LICENCE' in this directory.
 *
 */

#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/mtd/mtd.h>
#include <linux/compiler.h>
#include <linux/sched.h> /* For cond_resched() */
#include "nodelist.h"
#include "debug.h"

/**
 *	jffs2_reserve_space - request physical space to write nodes to flash
 *	@c: superblock info
 *	@minsize: Minimum acceptable size of allocation
 *	@len: Returned value of allocation length
 *	@prio: Allocation type - ALLOC_{NORMAL,DELETION}
 *
 *	Requests a block of physical space on the flash. Returns zero for success
 *	and puts 'len' into the appropriate place, or returns -ENOSPC or other 
 *	error if appropriate. Doesn't return len since that's 
 *
 *	If it returns zero, jffs2_reserve_space() also downs the per-filesystem
 *	allocation semaphore, to prevent more than one allocation from being
 *	active at any time. The semaphore is later released by jffs2_commit_allocation()
 *
 *	jffs2_reserve_space() may trigger garbage collection in order to make room
 *	for the requested allocation.
 */

static int jffs2_do_reserve_space(struct jffs2_sb_info *c,  uint32_t minsize,
				  uint32_t *len, uint32_t sumsize);

int jffs2_reserve_space(struct jffs2_sb_info *c, uint32_t minsize,
			uint32_t *len, int prio, uint32_t sumsize)
{
	int ret = -EAGAIN;
	int blocksneeded = c->resv_blocks_write;
	/* align it */
	minsize = PAD(minsize);

	D1(printk(KERN_DEBUG "jffs2_reserve_space(): Requested 0x%x bytes\n", minsize));
	down(&c->alloc_sem);

	D1(printk(KERN_DEBUG "jffs2_reserve_space(): alloc sem got\n"));

	spin_lock(&c->erase_completion_lock);

	/* this needs a little more thought (true <tglx> :)) */
	while(ret == -EAGAIN) {
		while(c->nr_free_blocks + c->nr_erasing_blocks < blocksneeded) {
			int ret;
			uint32_t dirty, avail;

			/* calculate real dirty size
			 * dirty_size contains blocks on erase_pending_list
			 * those blocks are counted in c->nr_erasing_blocks.
			 * If one block is actually erased, it is not longer counted as dirty_space
			 * but it is counted in c->nr_erasing_blocks, so we add it and subtract it
			 * with c->nr_erasing_blocks * c->sector_size again.
			 * Blocks on erasable_list are counted as dirty_size, but not in c->nr_erasing_blocks
			 * This helps us to force gc and pick eventually a clean block to spread the load.
			 * We add unchecked_size here, as we hopefully will find some space to use.
			 * This will affect the sum only once, as gc first finishes checking
			 * of nodes.
			 */
			dirty = c->dirty_size + c->erasing_size - c->nr_erasing_blocks * c->sector_size + c->unchecked_size;
			if (dirty < c->nospc_dirty_size) {
				if (prio == ALLOC_DELETION && c->nr_free_blocks + c->nr_erasing_blocks >= c->resv_blocks_deletion) {
					D1(printk(KERN_NOTICE "jffs2_reserve_space(): Low on dirty space to GC, but it's a deletion. Allowing...\n"));
					break;
				}
				D1(printk(KERN_DEBUG "dirty size 0x%08x + unchecked_size 0x%08x < nospc_dirty_size 0x%08x, returning -ENOSPC\n",
					  dirty, c->unchecked_size, c->sector_size));

				spin_unlock(&c->erase_completion_lock);
				up(&c->alloc_sem);
				return -ENOSPC;
			}

			/* Calc possibly available space. Possibly available means that we
			 * don't know, if unchecked size contains obsoleted nodes, which could give us some
			 * more usable space. This will affect the sum only once, as gc first finishes checking
			 * of nodes.
			 + Return -ENOSPC, if the maximum possibly available space is less or equal than
			 * blocksneeded * sector_size.
			 * This blocks endless gc looping on a filesystem, which is nearly full, even if
			 * the check above passes.
			 */
			avail = c->free_size + c->dirty_size + c->erasing_size + c->unchecked_size;
			if ( (avail / c->sector_size) <= blocksneeded) {
				if (prio == ALLOC_DELETION && c->nr_free_blocks + c->nr_erasing_blocks >= c->resv_blocks_deletion) {
					D1(printk(KERN_NOTICE "jffs2_reserve_space(): Low on possibly available space, but it's a deletion. Allowing...\n"));
					break;
				}

				D1(printk(KERN_DEBUG "max. available size 0x%08x  < blocksneeded * sector_size 0x%08x, returning -ENOSPC\n",
					  avail, blocksneeded * c->sector_size));
				spin_unlock(&c->erase_completion_lock);
				up(&c->alloc_sem);
				return -ENOSPC;
			}

			up(&c->alloc_sem);

			D1(printk(KERN_DEBUG "Triggering GC pass. nr_free_blocks %d, nr_erasing_blocks %d, free_size 0x%08x, dirty_size 0x%08x, wasted_size 0x%08x, used_size 0x%08x, erasing_size 0x%08x, bad_size 0x%08x (total 0x%08x of 0x%08x)\n",
				  c->nr_free_blocks, c->nr_erasing_blocks, c->free_size, c->dirty_size, c->wasted_size, c->used_size, c->erasing_size, c->bad_size,
				  c->free_size + c->dirty_size + c->wasted_size + c->used_size + c->erasing_size + c->bad_size, c->flash_size));
			spin_unlock(&c->erase_completion_lock);

			ret = jffs2_garbage_collect_pass(c);
			if (ret)
				return ret;

			cond_resched();

			if (signal_pending(current))
				return -EINTR;

			down(&c->alloc_sem);
			spin_lock(&c->erase_completion_lock);
		}

		ret = jffs2_do_reserve_space(c, minsize, len, sumsize);
		if (ret) {
			D1(printk(KERN_DEBUG "jffs2_reserve_space: ret is %d\n", ret));
		}
	}
	spin_unlock(&c->erase_completion_lock);
	if (!ret)
		ret = jffs2_prealloc_raw_node_refs(c, c->nextblock, 1);
	if (ret)
		up(&c->alloc_sem);
	return ret;
}

int jffs2_reserve_space_gc(struct jffs2_sb_info *c, uint32_t minsize,
			   uint32_t *len, uint32_t sumsize)
{
	int ret = -EAGAIN;
	minsize = PAD(minsize);

	D1(printk(KERN_DEBUG "jffs2_reserve_space_gc(): Requested 0x%x bytes\n", minsize));

	spin_lock(&c->erase_completion_lock);
	while(ret == -EAGAIN) {
		ret = jffs2_do_reserve_space(c, minsize, len, sumsize);
		if (ret) {
			D1(printk(KERN_DEBUG "jffs2_reserve_space_gc: looping, ret is %d\n", ret));
		}
	}
	spin_unlock(&c->erase_completion_lock);
	if (!ret)
		ret = jffs2_prealloc_raw_node_refs(c, c->nextblock, 1);

	return ret;
}


/* Classify nextblock (clean, dirty of verydirty) and force to select an other one */

static void jffs2_close_nextblock(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb)
{

	if (c->nextblock == NULL) {
		D1(printk(KERN_DEBUG "jffs2_close_nextblock: Erase block at 0x%08x has already been placed in a list\n",
		  jeb->offset));
		return;
	}
	/* Check, if we have a dirty block now, or if it was dirty already */
	if (ISDIRTY (jeb->wasted_size + jeb->dirty_size)) {
		c->dirty_size += jeb->wasted_size;
		c->wasted_size -= jeb->wasted_size;
		jeb->dirty_size += jeb->wasted_size;
		jeb->wasted_size = 0;
		if (VERYDIRTY(c, jeb->dirty_size)) {
			D1(printk(KERN_DEBUG "Adding full erase block at 0x%08x to very_dirty_list (free 0x%08x, dirty 0x%08x, used 0x%08x\n",
			  jeb->offset, jeb->free_size, jeb->dirty_size, jeb->used_size));
			list_add_tail(&jeb->list, &c->very_dirty_list);
		} else {
			D1(printk(KERN_DEBUG "Adding full erase block at 0x%08x to dirty_list (free 0x%08x, dirty 0x%08x, used 0x%08x\n",
			  jeb->offset, jeb->free_size, jeb->dirty_size, jeb->used_size));
			list_add_tail(&jeb->list, &c->dirty_list);
		}
	} else {
		D1(printk(KERN_DEBUG "Adding full erase block at 0x%08x to clean_list (free 0x%08x, dirty 0x%08x, used 0x%08x\n",
		  jeb->offset, jeb->free_size, jeb->dirty_size, jeb->used_size));
		list_add_tail(&jeb->list, &c->clean_list);
	}
	c->nextblock = NULL;

}

/* Select a new jeb for nextblock */

static int jffs2_find_nextblock(struct jffs2_sb_info *c)
{
	struct list_head *next;

	/* Take the next block off the 'free' list */

	if (list_empty(&c->free_list)) {

		if (!c->nr_erasing_blocks &&
			!list_empty(&c->erasable_list)) {
			struct jffs2_eraseblock *ejeb;

			ejeb = list_entry(c->erasable_list.next, struct jffs2_eraseblock, list);
			list_move_tail(&ejeb->list, &c->erase_pending_list);
			c->nr_erasing_blocks++;
			jffs2_erase_pending_trigger(c);
			D1(printk(KERN_DEBUG "jffs2_find_nextblock: Triggering erase of erasable block at 0x%08x\n",
				  ejeb->offset));
		}

		if (!c->nr_erasing_blocks &&
			!list_empty(&c->erasable_pending_wbuf_list)) {
			D1(printk(KERN_DEBUG "jffs2_find_nextblock: Flushing write buffer\n"));
			/* c->nextblock is NULL, no update to c->nextblock allowed */
			spin_unlock(&c->erase_completion_lock);
			jffs2_flush_wbuf_pad(c);
			spin_lock(&c->erase_completion_lock);
			/* Have another go. It'll be on the erasable_list now */
			return -EAGAIN;
		}

		if (!c->nr_erasing_blocks) {
			/* Ouch. We're in GC, or we wouldn't have got here.
			   And there's no space left. At all. */
			printk(KERN_CRIT "Argh. No free space left for GC. nr_erasing_blocks is %d. nr_free_blocks is %d. (erasableempty: %s, erasingempty: %s, erasependingempty: %s)\n",
				   c->nr_erasing_blocks, c->nr_free_blocks, list_empty(&c->erasable_list)?"yes":"no",
				   list_empty(&c->erasing_list)?"yes":"no", list_empty(&c->erase_pending_list)?"yes":"no");
			return -ENOSPC;
		}

		spin_unlock(&c->erase_completion_lock);
		/* Don't wait for it; just erase one right now */
		jffs2_erase_pending_blocks(c, 1);
		spin_lock(&c->erase_completion_lock);

		/* An erase may have failed, decreasing the
		   amount of free space available. So we must
		   restart from the beginning */
		return -EAGAIN;
	}

	next = c->free_list.next;
	list_del(next);
	c->nextblock = list_entry(next, struct jffs2_eraseblock, list);
	c->nr_free_blocks--;

	jffs2_sum_reset_collected(c->summary); /* reset collected summary */

	D1(printk(KERN_DEBUG "jffs2_find_nextblock(): new nextblock = 0x%08x\n", c->nextblock->offset));

	return 0;
}

/* Called with alloc sem _and_ erase_completion_lock */
static int jffs2_do_reserve_space(struct jffs2_sb_info *c, uint32_t minsize,
				  uint32_t *len, uint32_t sumsize)
{
	struct jffs2_eraseblock *jeb = c->nextblock;
	uint32_t reserved_size;				/* for summary information at the end of the jeb */
	int ret;

 restart:
	reserved_size = 0;

	if (jffs2_sum_active() && (sumsize != JFFS2_SUMMARY_NOSUM_SIZE)) {
							/* NOSUM_SIZE means not to generate summary */

		if (jeb) {
			reserved_size = PAD(sumsize + c->summary->sum_size + JFFS2_SUMMARY_FRAME_SIZE);
			dbg_summary("minsize=%d , jeb->free=%d ,"
						"summary->size=%d , sumsize=%d\n",
						minsize, jeb->free_size,
						c->summary->sum_size, sumsize);
		}

		/* Is there enough space for writing out the current node, or we have to
		   write out summary information now, close this jeb and select new nextblock? */
		if (jeb && (PAD(minsize) + PAD(c->summary->sum_size + sumsize +
					JFFS2_SUMMARY_FRAME_SIZE) > jeb->free_size)) {

			/* Has summary been disabled for this jeb? */
			if (jffs2_sum_is_disabled(c->summary)) {
				sumsize = JFFS2_SUMMARY_NOSUM_SIZE;
				goto restart;
			}

			/* Writing out the collected summary information */
			dbg_summary("generating summary for 0x%08x.\n", jeb->offset);
			ret = jffs2_sum_write_sumnode(c);

			if (ret)
				return ret;

			if (jffs2_sum_is_disabled(c->summary)) {
				/* jffs2_write_sumnode() couldn't write out the summary information
				   diabling summary for this jeb and free the collected information
				 */
				sumsize = JFFS2_SUMMARY_NOSUM_SIZE;
				goto restart;
			}

			jffs2_close_nextblock(c, jeb);
			jeb = NULL;
			/* keep always valid value in reserved_size */
			reserved_size = PAD(sumsize + c->summary->sum_size + JFFS2_SUMMARY_FRAME_SIZE);
		}
	} else {
		if (jeb && minsize > jeb->free_size) {
			uint32_t waste;

			/* Skip the end of this block and file it as having some dirty space */
			/* If there's a pending write to it, flush now */

			if (jffs2_wbuf_dirty(c)) {
				spin_unlock(&c->erase_completion_lock);
				D1(printk(KERN_DEBUG "jffs2_do_reserve_space: Flushing write buffer\n"));
				jffs2_flush_wbuf_pad(c);
				spin_lock(&c->erase_completion_lock);
				jeb = c->nextblock;
				goto restart;
			}

			spin_unlock(&c->erase_completion_lock);

			ret = jffs2_prealloc_raw_node_refs(c, jeb, 1);
			if (ret)
				return ret;
			/* Just lock it again and continue. Nothing much can change because
			   we hold c->alloc_sem anyway. In fact, it's not entirely clear why
			   we hold c->erase_completion_lock in the majority of this function...
			   but that's a question for another (more caffeine-rich) day. */
			spin_lock(&c->erase_completion_lock);

			waste = jeb->free_size;
			jffs2_link_node_ref(c, jeb,
					    (jeb->offset + c->sector_size - waste) | REF_OBSOLETE,
					    waste, NULL);
			/* FIXME: that made it count as dirty. Convert to wasted */
			jeb->dirty_size -= waste;
			c->dirty_size -= waste;
			jeb->wasted_size += waste;
			c->wasted_size += waste;

			jffs2_close_nextblock(c, jeb);
			jeb = NULL;
		}
	}

	if (!jeb) {

		ret = jffs2_find_nextblock(c);
		if (ret)
			return ret;

		jeb = c->nextblock;

		if (jeb->free_size != c->sector_size - c->cleanmarker_size) {
			printk(KERN_WARNING "Eep. Block 0x%08x taken from free_list had free_size of 0x%08x!!\n", jeb->offset, jeb->free_size);
			goto restart;
		}
	}
	/* OK, jeb (==c->nextblock) is now pointing at a block which definitely has
	   enough space */
	*len = jeb->free_size - reserved_size;

	if (c->cleanmarker_size && jeb->used_size == c->cleanmarker_size &&
	    !jeb->first_node->next_in_ino) {
		/* Only node in it beforehand was a CLEANMARKER node (we think).
		   So mark it obsolete now that there's going to be another node