nodemgmt.c 26.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3
/*
 * JFFS2 -- Journalling Flash File System, Version 2.
 *
4
 * Copyright © 2001-2007 Red Hat, Inc.
L
Linus Torvalds 已提交
5 6 7 8 9 10 11 12 13 14 15 16 17
 *
 * Created by David Woodhouse <dwmw2@infradead.org>
 *
 * 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"
18
#include "debug.h"
L
Linus Torvalds 已提交
19 20 21 22 23 24 25 26 27

/**
 *	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
28 29
 *	and puts 'len' into the appropriate place, or returns -ENOSPC or other 
 *	error if appropriate. Doesn't return len since that's 
L
Linus Torvalds 已提交
30 31 32 33 34 35 36 37 38
 *
 *	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.
 */

39
static int jffs2_do_reserve_space(struct jffs2_sb_info *c,  uint32_t minsize,
40
				  uint32_t *len, uint32_t sumsize);
L
Linus Torvalds 已提交
41

42
int jffs2_reserve_space(struct jffs2_sb_info *c, uint32_t minsize,
43
			uint32_t *len, int prio, uint32_t sumsize)
L
Linus Torvalds 已提交
44 45 46 47 48 49 50
{
	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));
51
	mutex_lock(&c->alloc_sem);
L
Linus Torvalds 已提交
52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76

	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) {
			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) {
77
					D1(printk(KERN_NOTICE "jffs2_reserve_space(): Low on dirty space to GC, but it's a deletion. Allowing...\n"));
L
Linus Torvalds 已提交
78 79 80 81 82 83
					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);
84
				mutex_unlock(&c->alloc_sem);
L
Linus Torvalds 已提交
85 86
				return -ENOSPC;
			}
87

L
Linus Torvalds 已提交
88 89 90 91
			/* 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.
92
			 + Return -ENOSPC, if the maximum possibly available space is less or equal than
L
Linus Torvalds 已提交
93 94 95 96 97 98 99
			 * 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) {
100
					D1(printk(KERN_NOTICE "jffs2_reserve_space(): Low on possibly available space, but it's a deletion. Allowing...\n"));
L
Linus Torvalds 已提交
101 102 103 104 105 106
					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);
107
				mutex_unlock(&c->alloc_sem);
L
Linus Torvalds 已提交
108 109 110
				return -ENOSPC;
			}

111
			mutex_unlock(&c->alloc_sem);
L
Linus Torvalds 已提交
112 113 114 115 116

			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);
117

L
Linus Torvalds 已提交
118 119 120 121 122 123 124 125 126
			ret = jffs2_garbage_collect_pass(c);
			if (ret)
				return ret;

			cond_resched();

			if (signal_pending(current))
				return -EINTR;

127
			mutex_lock(&c->alloc_sem);
L
Linus Torvalds 已提交
128 129 130
			spin_lock(&c->erase_completion_lock);
		}

131
		ret = jffs2_do_reserve_space(c, minsize, len, sumsize);
L
Linus Torvalds 已提交
132 133 134 135 136
		if (ret) {
			D1(printk(KERN_DEBUG "jffs2_reserve_space: ret is %d\n", ret));
		}
	}
	spin_unlock(&c->erase_completion_lock);
137
	if (!ret)
138
		ret = jffs2_prealloc_raw_node_refs(c, c->nextblock, 1);
L
Linus Torvalds 已提交
139
	if (ret)
140
		mutex_unlock(&c->alloc_sem);
L
Linus Torvalds 已提交
141 142 143
	return ret;
}

144 145
int jffs2_reserve_space_gc(struct jffs2_sb_info *c, uint32_t minsize,
			   uint32_t *len, uint32_t sumsize)
L
Linus Torvalds 已提交
146 147 148 149 150 151 152 153
{
	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) {
154
		ret = jffs2_do_reserve_space(c, minsize, len, sumsize);
L
Linus Torvalds 已提交
155
		if (ret) {
D
David Woodhouse 已提交
156
			D1(printk(KERN_DEBUG "jffs2_reserve_space_gc: looping, ret is %d\n", ret));
L
Linus Torvalds 已提交
157 158 159
		}
	}
	spin_unlock(&c->erase_completion_lock);
160
	if (!ret)
161
		ret = jffs2_prealloc_raw_node_refs(c, c->nextblock, 1);
162

L
Linus Torvalds 已提交
163 164 165
	return ret;
}

166 167 168 169

/* 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)
L
Linus Torvalds 已提交
170
{
171

172 173 174 175 176
	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;
	}
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
	/* 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);
		}
192
	} else {
193 194 195 196 197 198 199 200 201 202 203 204 205
		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;
206

207 208 209 210 211 212 213 214 215
	/* 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);
A
Akinobu Mita 已提交
216
			list_move_tail(&ejeb->list, &c->erase_pending_list);
217 218 219 220 221 222 223 224 225 226
			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 */
L
Linus Torvalds 已提交
227 228 229
			spin_unlock(&c->erase_completion_lock);
			jffs2_flush_wbuf_pad(c);
			spin_lock(&c->erase_completion_lock);
230 231
			/* Have another go. It'll be on the erasable_list now */
			return -EAGAIN;
L
Linus Torvalds 已提交
232
		}
233 234 235 236

		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. */
237 238
			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",
239 240
				   list_empty(&c->erasing_list)?"yes":"no", list_empty(&c->erase_pending_list)?"yes":"no");
			return -ENOSPC;
L
Linus Torvalds 已提交
241
		}
242 243 244 245 246 247 248 249 250 251

		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;
L
Linus Torvalds 已提交
252
	}
253 254 255 256 257

	next = c->free_list.next;
	list_del(next);
	c->nextblock = list_entry(next, struct jffs2_eraseblock, list);
	c->nr_free_blocks--;
258

259 260 261 262 263 264 265 266
	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 */
267 268
static int jffs2_do_reserve_space(struct jffs2_sb_info *c, uint32_t minsize,
				  uint32_t *len, uint32_t sumsize)
269 270
{
	struct jffs2_eraseblock *jeb = c->nextblock;
271
	uint32_t reserved_size;				/* for summary information at the end of the jeb */
272 273 274 275 276 277 278 279 280 281
	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);
282
			dbg_summary("minsize=%d , jeb->free=%d ,"
283 284 285 286 287 288 289 290 291 292 293 294 295 296
						"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;
L
Linus Torvalds 已提交
297 298
			}

299
			/* Writing out the collected summary information */
300
			dbg_summary("generating summary for 0x%08x.\n", jeb->offset);
301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
			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;
316 317
			/* keep always valid value in reserved_size */
			reserved_size = PAD(sumsize + c->summary->sum_size + JFFS2_SUMMARY_FRAME_SIZE);
318 319 320
		}
	} else {
		if (jeb && minsize > jeb->free_size) {
321 322
			uint32_t waste;

323 324 325 326
			/* 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)) {
L
Linus Torvalds 已提交
327
				spin_unlock(&c->erase_completion_lock);
328
				D1(printk(KERN_DEBUG "jffs2_do_reserve_space: Flushing write buffer\n"));
L
Linus Torvalds 已提交
329 330
				jffs2_flush_wbuf_pad(c);
				spin_lock(&c->erase_completion_lock);
331 332
				jeb = c->nextblock;
				goto restart;
L
Linus Torvalds 已提交
333 334
			}

335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354
			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;
L
Linus Torvalds 已提交
355

356 357
			jffs2_close_nextblock(c, jeb);
			jeb = NULL;
L
Linus Torvalds 已提交
358
		}
359 360 361 362 363 364 365
	}

	if (!jeb) {

		ret = jffs2_find_nextblock(c);
		if (ret)
			return ret;
L
Linus Torvalds 已提交
366

367
		jeb = c->nextblock;
L
Linus Torvalds 已提交
368 369 370 371 372 373 374 375

		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 */
376
	*len = jeb->free_size - reserved_size;
L
Linus Torvalds 已提交
377 378 379

	if (c->cleanmarker_size && jeb->used_size == c->cleanmarker_size &&
	    !jeb->first_node->next_in_ino) {
380
		/* Only node in it beforehand was a CLEANMARKER node (we think).
L
Linus Torvalds 已提交
381
		   So mark it obsolete now that there's going to be another node
382
		   in the block. This will reduce used_size to zero but We've
L
Linus Torvalds 已提交
383 384 385 386 387 388 389 390
		   already set c->nextblock so that jffs2_mark_node_obsolete()
		   won't try to refile it to the dirty_list.
		*/
		spin_unlock(&c->erase_completion_lock);
		jffs2_mark_node_obsolete(c, jeb->first_node);
		spin_lock(&c->erase_completion_lock);
	}

391 392
	D1(printk(KERN_DEBUG "jffs2_do_reserve_space(): Giving 0x%x bytes at 0x%x\n",
		  *len, jeb->offset + (c->sector_size - jeb->free_size)));
L
Linus Torvalds 已提交
393 394 395 396 397 398 399 400 401
	return 0;
}

/**
 *	jffs2_add_physical_node_ref - add a physical node reference to the list
 *	@c: superblock info
 *	@new: new node reference to add
 *	@len: length of this physical node
 *
402
 *	Should only be used to report nodes for which space has been allocated
L
Linus Torvalds 已提交
403 404 405 406
 *	by jffs2_reserve_space.
 *
 *	Must be called with the alloc_sem held.
 */
407

408 409 410
struct jffs2_raw_node_ref *jffs2_add_physical_node_ref(struct jffs2_sb_info *c,
						       uint32_t ofs, uint32_t len,
						       struct jffs2_inode_cache *ic)
L
Linus Torvalds 已提交
411 412
{
	struct jffs2_eraseblock *jeb;
413
	struct jffs2_raw_node_ref *new;
L
Linus Torvalds 已提交
414

415
	jeb = &c->blocks[ofs / c->sector_size];
L
Linus Torvalds 已提交
416

417 418
	D1(printk(KERN_DEBUG "jffs2_add_physical_node_ref(): Node at 0x%x(%d), size 0x%x\n",
		  ofs & ~3, ofs & 3, len));
L
Linus Torvalds 已提交
419
#if 1
420 421 422 423 424
	/* Allow non-obsolete nodes only to be added at the end of c->nextblock, 
	   if c->nextblock is set. Note that wbuf.c will file obsolete nodes
	   even after refiling c->nextblock */
	if ((c->nextblock || ((ofs & 3) != REF_OBSOLETE))
	    && (jeb != c->nextblock || (ofs & ~3) != jeb->offset + (c->sector_size - jeb->free_size))) {
425 426 427 428 429 430
		printk(KERN_WARNING "argh. node added in wrong place at 0x%08x(%d)\n", ofs & ~3, ofs & 3);
		if (c->nextblock)
			printk(KERN_WARNING "nextblock 0x%08x", c->nextblock->offset);
		else
			printk(KERN_WARNING "No nextblock");
		printk(", expected at %08x\n", jeb->offset + (c->sector_size - jeb->free_size));
431
		return ERR_PTR(-EINVAL);
L
Linus Torvalds 已提交
432 433 434 435
	}
#endif
	spin_lock(&c->erase_completion_lock);

436
	new = jffs2_link_node_ref(c, jeb, ofs, len, ic);
L
Linus Torvalds 已提交
437

E
Estelle Hammache 已提交
438
	if (!jeb->free_size && !jeb->dirty_size && !ISDIRTY(jeb->wasted_size)) {
L
Linus Torvalds 已提交
439 440 441 442 443 444 445 446 447 448 449 450 451
		/* If it lives on the dirty_list, jffs2_reserve_space will put it there */
		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));
		if (jffs2_wbuf_dirty(c)) {
			/* Flush the last write in the block if it's outstanding */
			spin_unlock(&c->erase_completion_lock);
			jffs2_flush_wbuf_pad(c);
			spin_lock(&c->erase_completion_lock);
		}

		list_add_tail(&jeb->list, &c->clean_list);
		c->nextblock = NULL;
	}
452 453
	jffs2_dbg_acct_sanity_check_nolock(c,jeb);
	jffs2_dbg_acct_paranoia_check_nolock(c, jeb);
L
Linus Torvalds 已提交
454 455 456

	spin_unlock(&c->erase_completion_lock);

457
	return new;
L
Linus Torvalds 已提交
458 459 460 461 462 463 464
}


void jffs2_complete_reservation(struct jffs2_sb_info *c)
{
	D1(printk(KERN_DEBUG "jffs2_complete_reservation()\n"));
	jffs2_garbage_collect_trigger(c);
465
	mutex_unlock(&c->alloc_sem);
L
Linus Torvalds 已提交
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488
}

static inline int on_list(struct list_head *obj, struct list_head *head)
{
	struct list_head *this;

	list_for_each(this, head) {
		if (this == obj) {
			D1(printk("%p is on list at %p\n", obj, head));
			return 1;

		}
	}
	return 0;
}

void jffs2_mark_node_obsolete(struct jffs2_sb_info *c, struct jffs2_raw_node_ref *ref)
{
	struct jffs2_eraseblock *jeb;
	int blocknr;
	struct jffs2_unknown_node n;
	int ret, addedsize;
	size_t retlen;
489
	uint32_t freed_len;
L
Linus Torvalds 已提交
490

491
	if(unlikely(!ref)) {
L
Linus Torvalds 已提交
492 493 494 495 496 497 498 499 500 501 502 503 504 505 506
		printk(KERN_NOTICE "EEEEEK. jffs2_mark_node_obsolete called with NULL node\n");
		return;
	}
	if (ref_obsolete(ref)) {
		D1(printk(KERN_DEBUG "jffs2_mark_node_obsolete called with already obsolete node at 0x%08x\n", ref_offset(ref)));
		return;
	}
	blocknr = ref->flash_offset / c->sector_size;
	if (blocknr >= c->nr_blocks) {
		printk(KERN_NOTICE "raw node at 0x%08x is off the end of device!\n", ref->flash_offset);
		BUG();
	}
	jeb = &c->blocks[blocknr];

	if (jffs2_can_mark_obsolete(c) && !jffs2_is_readonly(c) &&
507
	    !(c->flags & (JFFS2_SB_FLAG_SCANNING | JFFS2_SB_FLAG_BUILDING))) {
508 509
		/* Hm. This may confuse static lock analysis. If any of the above
		   three conditions is false, we're going to return from this
L
Linus Torvalds 已提交
510 511 512 513
		   function without actually obliterating any nodes or freeing
		   any jffs2_raw_node_refs. So we don't need to stop erases from
		   happening, or protect against people holding an obsolete
		   jffs2_raw_node_ref without the erase_completion_lock. */
514
		mutex_lock(&c->erase_free_sem);
L
Linus Torvalds 已提交
515 516 517 518
	}

	spin_lock(&c->erase_completion_lock);

519 520
	freed_len = ref_totlen(c, jeb, ref);

L
Linus Torvalds 已提交
521
	if (ref_flags(ref) == REF_UNCHECKED) {
522
		D1(if (unlikely(jeb->unchecked_size < freed_len)) {
L
Linus Torvalds 已提交
523
			printk(KERN_NOTICE "raw unchecked node of size 0x%08x freed from erase block %d at 0x%08x, but unchecked_size was already 0x%08x\n",
524
			       freed_len, blocknr, ref->flash_offset, jeb->used_size);
L
Linus Torvalds 已提交
525 526
			BUG();
		})
527 528 529
		D1(printk(KERN_DEBUG "Obsoleting previously unchecked node at 0x%08x of len %x: ", ref_offset(ref), freed_len));
		jeb->unchecked_size -= freed_len;
		c->unchecked_size -= freed_len;
L
Linus Torvalds 已提交
530
	} else {
531
		D1(if (unlikely(jeb->used_size < freed_len)) {
L
Linus Torvalds 已提交
532
			printk(KERN_NOTICE "raw node of size 0x%08x freed from erase block %d at 0x%08x, but used_size was already 0x%08x\n",
533
			       freed_len, blocknr, ref->flash_offset, jeb->used_size);
L
Linus Torvalds 已提交
534 535
			BUG();
		})
536 537 538
		D1(printk(KERN_DEBUG "Obsoleting node at 0x%08x of len %#x: ", ref_offset(ref), freed_len));
		jeb->used_size -= freed_len;
		c->used_size -= freed_len;
L
Linus Torvalds 已提交
539 540 541
	}

	// Take care, that wasted size is taken into concern
542
	if ((jeb->dirty_size || ISDIRTY(jeb->wasted_size + freed_len)) && jeb != c->nextblock) {
543
		D1(printk("Dirtying\n"));
544 545 546
		addedsize = freed_len;
		jeb->dirty_size += freed_len;
		c->dirty_size += freed_len;
L
Linus Torvalds 已提交
547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564

		/* Convert wasted space to dirty, if not a bad block */
		if (jeb->wasted_size) {
			if (on_list(&jeb->list, &c->bad_used_list)) {
				D1(printk(KERN_DEBUG "Leaving block at %08x on the bad_used_list\n",
					  jeb->offset));
				addedsize = 0; /* To fool the refiling code later */
			} else {
				D1(printk(KERN_DEBUG "Converting %d bytes of wasted space to dirty in block at %08x\n",
					  jeb->wasted_size, jeb->offset));
				addedsize += jeb->wasted_size;
				jeb->dirty_size += jeb->wasted_size;
				c->dirty_size += jeb->wasted_size;
				c->wasted_size -= jeb->wasted_size;
				jeb->wasted_size = 0;
			}
		}
	} else {
565
		D1(printk("Wasting\n"));
L
Linus Torvalds 已提交
566
		addedsize = 0;
567 568
		jeb->wasted_size += freed_len;
		c->wasted_size += freed_len;
L
Linus Torvalds 已提交
569 570
	}
	ref->flash_offset = ref_offset(ref) | REF_OBSOLETE;
571

572 573
	jffs2_dbg_acct_sanity_check_nolock(c, jeb);
	jffs2_dbg_acct_paranoia_check_nolock(c, jeb);
L
Linus Torvalds 已提交
574

575 576
	if (c->flags & JFFS2_SB_FLAG_SCANNING) {
		/* Flash scanning is in progress. Don't muck about with the block
L
Linus Torvalds 已提交
577
		   lists because they're not ready yet, and don't actually
578
		   obliterate nodes that look obsolete. If they weren't
L
Linus Torvalds 已提交
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611
		   marked obsolete on the flash at the time they _became_
		   obsolete, there was probably a reason for that. */
		spin_unlock(&c->erase_completion_lock);
		/* We didn't lock the erase_free_sem */
		return;
	}

	if (jeb == c->nextblock) {
		D2(printk(KERN_DEBUG "Not moving nextblock 0x%08x to dirty/erase_pending list\n", jeb->offset));
	} else if (!jeb->used_size && !jeb->unchecked_size) {
		if (jeb == c->gcblock) {
			D1(printk(KERN_DEBUG "gcblock at 0x%08x completely dirtied. Clearing gcblock...\n", jeb->offset));
			c->gcblock = NULL;
		} else {
			D1(printk(KERN_DEBUG "Eraseblock at 0x%08x completely dirtied. Removing from (dirty?) list...\n", jeb->offset));
			list_del(&jeb->list);
		}
		if (jffs2_wbuf_dirty(c)) {
			D1(printk(KERN_DEBUG "...and adding to erasable_pending_wbuf_list\n"));
			list_add_tail(&jeb->list, &c->erasable_pending_wbuf_list);
		} else {
			if (jiffies & 127) {
				/* Most of the time, we just erase it immediately. Otherwise we
				   spend ages scanning it on mount, etc. */
				D1(printk(KERN_DEBUG "...and adding to erase_pending_list\n"));
				list_add_tail(&jeb->list, &c->erase_pending_list);
				c->nr_erasing_blocks++;
				jffs2_erase_pending_trigger(c);
			} else {
				/* Sometimes, however, we leave it elsewhere so it doesn't get
				   immediately reused, and we spread the load a bit. */
				D1(printk(KERN_DEBUG "...and adding to erasable_list\n"));
				list_add_tail(&jeb->list, &c->erasable_list);
612
			}
L
Linus Torvalds 已提交
613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629
		}
		D1(printk(KERN_DEBUG "Done OK\n"));
	} else if (jeb == c->gcblock) {
		D2(printk(KERN_DEBUG "Not moving gcblock 0x%08x to dirty_list\n", jeb->offset));
	} else if (ISDIRTY(jeb->dirty_size) && !ISDIRTY(jeb->dirty_size - addedsize)) {
		D1(printk(KERN_DEBUG "Eraseblock at 0x%08x is freshly dirtied. Removing from clean list...\n", jeb->offset));
		list_del(&jeb->list);
		D1(printk(KERN_DEBUG "...and adding to dirty_list\n"));
		list_add_tail(&jeb->list, &c->dirty_list);
	} else if (VERYDIRTY(c, jeb->dirty_size) &&
		   !VERYDIRTY(c, jeb->dirty_size - addedsize)) {
		D1(printk(KERN_DEBUG "Eraseblock at 0x%08x is now very dirty. Removing from dirty list...\n", jeb->offset));
		list_del(&jeb->list);
		D1(printk(KERN_DEBUG "...and adding to very_dirty_list\n"));
		list_add_tail(&jeb->list, &c->very_dirty_list);
	} else {
		D1(printk(KERN_DEBUG "Eraseblock at 0x%08x not moved anywhere. (free 0x%08x, dirty 0x%08x, used 0x%08x)\n",
630 631
			  jeb->offset, jeb->free_size, jeb->dirty_size, jeb->used_size));
	}
L
Linus Torvalds 已提交
632 633 634

	spin_unlock(&c->erase_completion_lock);

635 636
	if (!jffs2_can_mark_obsolete(c) || jffs2_is_readonly(c) ||
		(c->flags & JFFS2_SB_FLAG_BUILDING)) {
L
Linus Torvalds 已提交
637 638 639 640 641 642 643
		/* We didn't lock the erase_free_sem */
		return;
	}

	/* The erase_free_sem is locked, and has been since before we marked the node obsolete
	   and potentially put its eraseblock onto the erase_pending_list. Thus, we know that
	   the block hasn't _already_ been erased, and that 'ref' itself hasn't been freed yet
644
	   by jffs2_free_jeb_node_refs() in erase.c. Which is nice. */
L
Linus Torvalds 已提交
645 646 647 648 649 650 651 652 653 654 655

	D1(printk(KERN_DEBUG "obliterating obsoleted node at 0x%08x\n", ref_offset(ref)));
	ret = jffs2_flash_read(c, ref_offset(ref), sizeof(n), &retlen, (char *)&n);
	if (ret) {
		printk(KERN_WARNING "Read error reading from obsoleted node at 0x%08x: %d\n", ref_offset(ref), ret);
		goto out_erase_sem;
	}
	if (retlen != sizeof(n)) {
		printk(KERN_WARNING "Short read from obsoleted node at 0x%08x: %zd\n", ref_offset(ref), retlen);
		goto out_erase_sem;
	}
656 657
	if (PAD(je32_to_cpu(n.totlen)) != PAD(freed_len)) {
		printk(KERN_WARNING "Node totlen on flash (0x%08x) != totlen from node ref (0x%08x)\n", je32_to_cpu(n.totlen), freed_len);
L
Linus Torvalds 已提交
658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677
		goto out_erase_sem;
	}
	if (!(je16_to_cpu(n.nodetype) & JFFS2_NODE_ACCURATE)) {
		D1(printk(KERN_DEBUG "Node at 0x%08x was already marked obsolete (nodetype 0x%04x)\n", ref_offset(ref), je16_to_cpu(n.nodetype)));
		goto out_erase_sem;
	}
	/* XXX FIXME: This is ugly now */
	n.nodetype = cpu_to_je16(je16_to_cpu(n.nodetype) & ~JFFS2_NODE_ACCURATE);
	ret = jffs2_flash_write(c, ref_offset(ref), sizeof(n), &retlen, (char *)&n);
	if (ret) {
		printk(KERN_WARNING "Write error in obliterating obsoleted node at 0x%08x: %d\n", ref_offset(ref), ret);
		goto out_erase_sem;
	}
	if (retlen != sizeof(n)) {
		printk(KERN_WARNING "Short write in obliterating obsoleted node at 0x%08x: %zd\n", ref_offset(ref), retlen);
		goto out_erase_sem;
	}

	/* Nodes which have been marked obsolete no longer need to be
	   associated with any inode. Remove them from the per-inode list.
678 679

	   Note we can't do this for NAND at the moment because we need
L
Linus Torvalds 已提交
680 681
	   obsolete dirent nodes to stay on the lists, because of the
	   horridness in jffs2_garbage_collect_deletion_dirent(). Also
682
	   because we delete the inocache, and on NAND we need that to
L
Linus Torvalds 已提交
683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698
	   stay around until all the nodes are actually erased, in order
	   to stop us from giving the same inode number to another newly
	   created inode. */
	if (ref->next_in_ino) {
		struct jffs2_inode_cache *ic;
		struct jffs2_raw_node_ref **p;

		spin_lock(&c->erase_completion_lock);

		ic = jffs2_raw_ref_to_ic(ref);
		for (p = &ic->nodes; (*p) != ref; p = &((*p)->next_in_ino))
			;

		*p = ref->next_in_ino;
		ref->next_in_ino = NULL;

699 700 701 702 703 704 705 706 707 708 709 710 711 712
		switch (ic->class) {
#ifdef CONFIG_JFFS2_FS_XATTR
			case RAWNODE_CLASS_XATTR_DATUM:
				jffs2_release_xattr_datum(c, (struct jffs2_xattr_datum *)ic);
				break;
			case RAWNODE_CLASS_XATTR_REF:
				jffs2_release_xattr_ref(c, (struct jffs2_xattr_ref *)ic);
				break;
#endif
			default:
				if (ic->nodes == (void *)ic && ic->nlink == 0)
					jffs2_del_ino_cache(c, ic);
				break;
		}
L
Linus Torvalds 已提交
713 714 715 716
		spin_unlock(&c->erase_completion_lock);
	}

 out_erase_sem:
717
	mutex_unlock(&c->erase_free_sem);
L
Linus Torvalds 已提交
718 719 720 721 722 723
}

int jffs2_thread_should_wake(struct jffs2_sb_info *c)
{
	int ret = 0;
	uint32_t dirty;
724 725
	int nr_very_dirty = 0;
	struct jffs2_eraseblock *jeb;
L
Linus Torvalds 已提交
726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742

	if (c->unchecked_size) {
		D1(printk(KERN_DEBUG "jffs2_thread_should_wake(): unchecked_size %d, checked_ino #%d\n",
			  c->unchecked_size, c->checked_ino));
		return 1;
	}

	/* 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.
	 */
	dirty = c->dirty_size + c->erasing_size - c->nr_erasing_blocks * c->sector_size;

743 744
	if (c->nr_free_blocks + c->nr_erasing_blocks < c->resv_blocks_gctrigger &&
			(dirty > c->nospc_dirty_size))
L
Linus Torvalds 已提交
745 746
		ret = 1;

747 748 749 750
	list_for_each_entry(jeb, &c->very_dirty_list, list) {
		nr_very_dirty++;
		if (nr_very_dirty == c->vdirty_blocks_gctrigger) {
			ret = 1;
751 752 753
			/* In debug mode, actually go through and count them all */
			D1(continue);
			break;
754 755 756 757 758
		}
	}

	D1(printk(KERN_DEBUG "jffs2_thread_should_wake(): nr_free_blocks %d, nr_erasing_blocks %d, dirty_size 0x%x, vdirty_blocks %d: %s\n",
		  c->nr_free_blocks, c->nr_erasing_blocks, c->dirty_size, nr_very_dirty, ret?"yes":"no"));
L
Linus Torvalds 已提交
759 760 761

	return ret;
}