NFSv4: Ensure the lockowners are labelled using the fl_owner and/or fl_pid
[linux-2.6/kvm.git] / fs / jffs2 / erase.c
blob6286ad9b00f756c48b85283054ae078c910a1d23
1 /*
2 * JFFS2 -- Journalling Flash File System, Version 2.
4 * Copyright © 2001-2007 Red Hat, Inc.
6 * Created by David Woodhouse <dwmw2@infradead.org>
8 * For licensing information, see the file 'LICENCE' in this directory.
12 #include <linux/kernel.h>
13 #include <linux/slab.h>
14 #include <linux/mtd/mtd.h>
15 #include <linux/compiler.h>
16 #include <linux/crc32.h>
17 #include <linux/sched.h>
18 #include <linux/pagemap.h>
19 #include "nodelist.h"
21 struct erase_priv_struct {
22 struct jffs2_eraseblock *jeb;
23 struct jffs2_sb_info *c;
26 #ifndef __ECOS
27 static void jffs2_erase_callback(struct erase_info *);
28 #endif
29 static void jffs2_erase_failed(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb, uint32_t bad_offset);
30 static void jffs2_erase_succeeded(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb);
31 static void jffs2_mark_erased_block(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb);
33 static void jffs2_erase_block(struct jffs2_sb_info *c,
34 struct jffs2_eraseblock *jeb)
36 int ret;
37 uint32_t bad_offset;
38 #ifdef __ECOS
39 ret = jffs2_flash_erase(c, jeb);
40 if (!ret) {
41 jffs2_erase_succeeded(c, jeb);
42 return;
44 bad_offset = jeb->offset;
45 #else /* Linux */
46 struct erase_info *instr;
48 D1(printk(KERN_DEBUG "jffs2_erase_block(): erase block %#08x (range %#08x-%#08x)\n",
49 jeb->offset, jeb->offset, jeb->offset + c->sector_size));
50 instr = kmalloc(sizeof(struct erase_info) + sizeof(struct erase_priv_struct), GFP_KERNEL);
51 if (!instr) {
52 printk(KERN_WARNING "kmalloc for struct erase_info in jffs2_erase_block failed. Refiling block for later\n");
53 mutex_lock(&c->erase_free_sem);
54 spin_lock(&c->erase_completion_lock);
55 list_move(&jeb->list, &c->erase_pending_list);
56 c->erasing_size -= c->sector_size;
57 c->dirty_size += c->sector_size;
58 jeb->dirty_size = c->sector_size;
59 spin_unlock(&c->erase_completion_lock);
60 mutex_unlock(&c->erase_free_sem);
61 return;
64 memset(instr, 0, sizeof(*instr));
66 instr->mtd = c->mtd;
67 instr->addr = jeb->offset;
68 instr->len = c->sector_size;
69 instr->callback = jffs2_erase_callback;
70 instr->priv = (unsigned long)(&instr[1]);
71 instr->fail_addr = MTD_FAIL_ADDR_UNKNOWN;
73 ((struct erase_priv_struct *)instr->priv)->jeb = jeb;
74 ((struct erase_priv_struct *)instr->priv)->c = c;
76 ret = c->mtd->erase(c->mtd, instr);
77 if (!ret)
78 return;
80 bad_offset = instr->fail_addr;
81 kfree(instr);
82 #endif /* __ECOS */
84 if (ret == -ENOMEM || ret == -EAGAIN) {
85 /* Erase failed immediately. Refile it on the list */
86 D1(printk(KERN_DEBUG "Erase at 0x%08x failed: %d. Refiling on erase_pending_list\n", jeb->offset, ret));
87 mutex_lock(&c->erase_free_sem);
88 spin_lock(&c->erase_completion_lock);
89 list_move(&jeb->list, &c->erase_pending_list);
90 c->erasing_size -= c->sector_size;
91 c->dirty_size += c->sector_size;
92 jeb->dirty_size = c->sector_size;
93 spin_unlock(&c->erase_completion_lock);
94 mutex_unlock(&c->erase_free_sem);
95 return;
98 if (ret == -EROFS)
99 printk(KERN_WARNING "Erase at 0x%08x failed immediately: -EROFS. Is the sector locked?\n", jeb->offset);
100 else
101 printk(KERN_WARNING "Erase at 0x%08x failed immediately: errno %d\n", jeb->offset, ret);
103 jffs2_erase_failed(c, jeb, bad_offset);
106 int jffs2_erase_pending_blocks(struct jffs2_sb_info *c, int count)
108 struct jffs2_eraseblock *jeb;
109 int work_done = 0;
111 mutex_lock(&c->erase_free_sem);
113 spin_lock(&c->erase_completion_lock);
115 while (!list_empty(&c->erase_complete_list) ||
116 !list_empty(&c->erase_pending_list)) {
118 if (!list_empty(&c->erase_complete_list)) {
119 jeb = list_entry(c->erase_complete_list.next, struct jffs2_eraseblock, list);
120 list_move(&jeb->list, &c->erase_checking_list);
121 spin_unlock(&c->erase_completion_lock);
122 mutex_unlock(&c->erase_free_sem);
123 jffs2_mark_erased_block(c, jeb);
125 work_done++;
126 if (!--count) {
127 D1(printk(KERN_DEBUG "Count reached. jffs2_erase_pending_blocks leaving\n"));
128 goto done;
131 } else if (!list_empty(&c->erase_pending_list)) {
132 jeb = list_entry(c->erase_pending_list.next, struct jffs2_eraseblock, list);
133 D1(printk(KERN_DEBUG "Starting erase of pending block 0x%08x\n", jeb->offset));
134 list_del(&jeb->list);
135 c->erasing_size += c->sector_size;
136 c->wasted_size -= jeb->wasted_size;
137 c->free_size -= jeb->free_size;
138 c->used_size -= jeb->used_size;
139 c->dirty_size -= jeb->dirty_size;
140 jeb->wasted_size = jeb->used_size = jeb->dirty_size = jeb->free_size = 0;
141 jffs2_free_jeb_node_refs(c, jeb);
142 list_add(&jeb->list, &c->erasing_list);
143 spin_unlock(&c->erase_completion_lock);
144 mutex_unlock(&c->erase_free_sem);
146 jffs2_erase_block(c, jeb);
148 } else {
149 BUG();
152 /* Be nice */
153 yield();
154 mutex_lock(&c->erase_free_sem);
155 spin_lock(&c->erase_completion_lock);
158 spin_unlock(&c->erase_completion_lock);
159 mutex_unlock(&c->erase_free_sem);
160 done:
161 D1(printk(KERN_DEBUG "jffs2_erase_pending_blocks completed\n"));
162 return work_done;
165 static void jffs2_erase_succeeded(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb)
167 D1(printk(KERN_DEBUG "Erase completed successfully at 0x%08x\n", jeb->offset));
168 mutex_lock(&c->erase_free_sem);
169 spin_lock(&c->erase_completion_lock);
170 list_move_tail(&jeb->list, &c->erase_complete_list);
171 /* Wake the GC thread to mark them clean */
172 jffs2_garbage_collect_trigger(c);
173 spin_unlock(&c->erase_completion_lock);
174 mutex_unlock(&c->erase_free_sem);
175 wake_up(&c->erase_wait);
178 static void jffs2_erase_failed(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb, uint32_t bad_offset)
180 /* For NAND, if the failure did not occur at the device level for a
181 specific physical page, don't bother updating the bad block table. */
182 if (jffs2_cleanmarker_oob(c) && (bad_offset != (uint32_t)MTD_FAIL_ADDR_UNKNOWN)) {
183 /* We had a device-level failure to erase. Let's see if we've
184 failed too many times. */
185 if (!jffs2_write_nand_badblock(c, jeb, bad_offset)) {
186 /* We'd like to give this block another try. */
187 mutex_lock(&c->erase_free_sem);
188 spin_lock(&c->erase_completion_lock);
189 list_move(&jeb->list, &c->erase_pending_list);
190 c->erasing_size -= c->sector_size;
191 c->dirty_size += c->sector_size;
192 jeb->dirty_size = c->sector_size;
193 spin_unlock(&c->erase_completion_lock);
194 mutex_unlock(&c->erase_free_sem);
195 return;
199 mutex_lock(&c->erase_free_sem);
200 spin_lock(&c->erase_completion_lock);
201 c->erasing_size -= c->sector_size;
202 c->bad_size += c->sector_size;
203 list_move(&jeb->list, &c->bad_list);
204 c->nr_erasing_blocks--;
205 spin_unlock(&c->erase_completion_lock);
206 mutex_unlock(&c->erase_free_sem);
207 wake_up(&c->erase_wait);
210 #ifndef __ECOS
211 static void jffs2_erase_callback(struct erase_info *instr)
213 struct erase_priv_struct *priv = (void *)instr->priv;
215 if(instr->state != MTD_ERASE_DONE) {
216 printk(KERN_WARNING "Erase at 0x%08llx finished, but state != MTD_ERASE_DONE. State is 0x%x instead.\n",
217 (unsigned long long)instr->addr, instr->state);
218 jffs2_erase_failed(priv->c, priv->jeb, instr->fail_addr);
219 } else {
220 jffs2_erase_succeeded(priv->c, priv->jeb);
222 kfree(instr);
224 #endif /* !__ECOS */
226 /* Hmmm. Maybe we should accept the extra space it takes and make
227 this a standard doubly-linked list? */
228 static inline void jffs2_remove_node_refs_from_ino_list(struct jffs2_sb_info *c,
229 struct jffs2_raw_node_ref *ref, struct jffs2_eraseblock *jeb)
231 struct jffs2_inode_cache *ic = NULL;
232 struct jffs2_raw_node_ref **prev;
234 prev = &ref->next_in_ino;
236 /* Walk the inode's list once, removing any nodes from this eraseblock */
237 while (1) {
238 if (!(*prev)->next_in_ino) {
239 /* We're looking at the jffs2_inode_cache, which is
240 at the end of the linked list. Stash it and continue
241 from the beginning of the list */
242 ic = (struct jffs2_inode_cache *)(*prev);
243 prev = &ic->nodes;
244 continue;
247 if (SECTOR_ADDR((*prev)->flash_offset) == jeb->offset) {
248 /* It's in the block we're erasing */
249 struct jffs2_raw_node_ref *this;
251 this = *prev;
252 *prev = this->next_in_ino;
253 this->next_in_ino = NULL;
255 if (this == ref)
256 break;
258 continue;
260 /* Not to be deleted. Skip */
261 prev = &((*prev)->next_in_ino);
264 /* PARANOIA */
265 if (!ic) {
266 JFFS2_WARNING("inode_cache/xattr_datum/xattr_ref"
267 " not found in remove_node_refs()!!\n");
268 return;
271 D1(printk(KERN_DEBUG "Removed nodes in range 0x%08x-0x%08x from ino #%u\n",
272 jeb->offset, jeb->offset + c->sector_size, ic->ino));
274 D2({
275 int i=0;
276 struct jffs2_raw_node_ref *this;
277 printk(KERN_DEBUG "After remove_node_refs_from_ino_list: \n");
279 this = ic->nodes;
281 printk(KERN_DEBUG);
282 while(this) {
283 printk(KERN_CONT "0x%08x(%d)->",
284 ref_offset(this), ref_flags(this));
285 if (++i == 5) {
286 printk(KERN_DEBUG);
287 i=0;
289 this = this->next_in_ino;
291 printk(KERN_CONT "\n");
294 switch (ic->class) {
295 #ifdef CONFIG_JFFS2_FS_XATTR
296 case RAWNODE_CLASS_XATTR_DATUM:
297 jffs2_release_xattr_datum(c, (struct jffs2_xattr_datum *)ic);
298 break;
299 case RAWNODE_CLASS_XATTR_REF:
300 jffs2_release_xattr_ref(c, (struct jffs2_xattr_ref *)ic);
301 break;
302 #endif
303 default:
304 if (ic->nodes == (void *)ic && ic->pino_nlink == 0)
305 jffs2_del_ino_cache(c, ic);
309 void jffs2_free_jeb_node_refs(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb)
311 struct jffs2_raw_node_ref *block, *ref;
312 D1(printk(KERN_DEBUG "Freeing all node refs for eraseblock offset 0x%08x\n", jeb->offset));
314 block = ref = jeb->first_node;
316 while (ref) {
317 if (ref->flash_offset == REF_LINK_NODE) {
318 ref = ref->next_in_ino;
319 jffs2_free_refblock(block);
320 block = ref;
321 continue;
323 if (ref->flash_offset != REF_EMPTY_NODE && ref->next_in_ino)
324 jffs2_remove_node_refs_from_ino_list(c, ref, jeb);
325 /* else it was a non-inode node or already removed, so don't bother */
327 ref++;
329 jeb->first_node = jeb->last_node = NULL;
332 static int jffs2_block_check_erase(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb, uint32_t *bad_offset)
334 void *ebuf;
335 uint32_t ofs;
336 size_t retlen;
337 int ret = -EIO;
339 if (c->mtd->point) {
340 unsigned long *wordebuf;
342 ret = c->mtd->point(c->mtd, jeb->offset, c->sector_size,
343 &retlen, &ebuf, NULL);
344 if (ret) {
345 D1(printk(KERN_DEBUG "MTD point failed %d\n", ret));
346 goto do_flash_read;
348 if (retlen < c->sector_size) {
349 /* Don't muck about if it won't let us point to the whole erase sector */
350 D1(printk(KERN_DEBUG "MTD point returned len too short: 0x%zx\n", retlen));
351 c->mtd->unpoint(c->mtd, jeb->offset, retlen);
352 goto do_flash_read;
354 wordebuf = ebuf-sizeof(*wordebuf);
355 retlen /= sizeof(*wordebuf);
356 do {
357 if (*++wordebuf != ~0)
358 break;
359 } while(--retlen);
360 c->mtd->unpoint(c->mtd, jeb->offset, c->sector_size);
361 if (retlen) {
362 printk(KERN_WARNING "Newly-erased block contained word 0x%lx at offset 0x%08tx\n",
363 *wordebuf, jeb->offset + c->sector_size-retlen*sizeof(*wordebuf));
364 return -EIO;
366 return 0;
368 do_flash_read:
369 ebuf = kmalloc(PAGE_SIZE, GFP_KERNEL);
370 if (!ebuf) {
371 printk(KERN_WARNING "Failed to allocate page buffer for verifying erase at 0x%08x. Refiling\n", jeb->offset);
372 return -EAGAIN;
375 D1(printk(KERN_DEBUG "Verifying erase at 0x%08x\n", jeb->offset));
377 for (ofs = jeb->offset; ofs < jeb->offset + c->sector_size; ) {
378 uint32_t readlen = min((uint32_t)PAGE_SIZE, jeb->offset + c->sector_size - ofs);
379 int i;
381 *bad_offset = ofs;
383 ret = c->mtd->read(c->mtd, ofs, readlen, &retlen, ebuf);
384 if (ret) {
385 printk(KERN_WARNING "Read of newly-erased block at 0x%08x failed: %d. Putting on bad_list\n", ofs, ret);
386 ret = -EIO;
387 goto fail;
389 if (retlen != readlen) {
390 printk(KERN_WARNING "Short read from newly-erased block at 0x%08x. Wanted %d, got %zd\n", ofs, readlen, retlen);
391 ret = -EIO;
392 goto fail;
394 for (i=0; i<readlen; i += sizeof(unsigned long)) {
395 /* It's OK. We know it's properly aligned */
396 unsigned long *datum = ebuf + i;
397 if (*datum + 1) {
398 *bad_offset += i;
399 printk(KERN_WARNING "Newly-erased block contained word 0x%lx at offset 0x%08x\n", *datum, *bad_offset);
400 ret = -EIO;
401 goto fail;
404 ofs += readlen;
405 cond_resched();
407 ret = 0;
408 fail:
409 kfree(ebuf);
410 return ret;
413 static void jffs2_mark_erased_block(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb)
415 size_t retlen;
416 int ret;
417 uint32_t uninitialized_var(bad_offset);
419 switch (jffs2_block_check_erase(c, jeb, &bad_offset)) {
420 case -EAGAIN: goto refile;
421 case -EIO: goto filebad;
424 /* Write the erase complete marker */
425 D1(printk(KERN_DEBUG "Writing erased marker to block at 0x%08x\n", jeb->offset));
426 bad_offset = jeb->offset;
428 /* Cleanmarker in oob area or no cleanmarker at all ? */
429 if (jffs2_cleanmarker_oob(c) || c->cleanmarker_size == 0) {
431 if (jffs2_cleanmarker_oob(c)) {
432 if (jffs2_write_nand_cleanmarker(c, jeb))
433 goto filebad;
435 } else {
437 struct kvec vecs[1];
438 struct jffs2_unknown_node marker = {
439 .magic = cpu_to_je16(JFFS2_MAGIC_BITMASK),
440 .nodetype = cpu_to_je16(JFFS2_NODETYPE_CLEANMARKER),
441 .totlen = cpu_to_je32(c->cleanmarker_size)
444 jffs2_prealloc_raw_node_refs(c, jeb, 1);
446 marker.hdr_crc = cpu_to_je32(crc32(0, &marker, sizeof(struct jffs2_unknown_node)-4));
448 vecs[0].iov_base = (unsigned char *) &marker;
449 vecs[0].iov_len = sizeof(marker);
450 ret = jffs2_flash_direct_writev(c, vecs, 1, jeb->offset, &retlen);
452 if (ret || retlen != sizeof(marker)) {
453 if (ret)
454 printk(KERN_WARNING "Write clean marker to block at 0x%08x failed: %d\n",
455 jeb->offset, ret);
456 else
457 printk(KERN_WARNING "Short write to newly-erased block at 0x%08x: Wanted %zd, got %zd\n",
458 jeb->offset, sizeof(marker), retlen);
460 goto filebad;
463 /* Everything else got zeroed before the erase */
464 jeb->free_size = c->sector_size;
466 mutex_lock(&c->erase_free_sem);
467 spin_lock(&c->erase_completion_lock);
469 c->erasing_size -= c->sector_size;
470 c->free_size += c->sector_size;
472 /* Account for cleanmarker now, if it's in-band */
473 if (c->cleanmarker_size && !jffs2_cleanmarker_oob(c))
474 jffs2_link_node_ref(c, jeb, jeb->offset | REF_NORMAL, c->cleanmarker_size, NULL);
476 list_move_tail(&jeb->list, &c->free_list);
477 c->nr_erasing_blocks--;
478 c->nr_free_blocks++;
480 jffs2_dbg_acct_sanity_check_nolock(c, jeb);
481 jffs2_dbg_acct_paranoia_check_nolock(c, jeb);
483 spin_unlock(&c->erase_completion_lock);
484 mutex_unlock(&c->erase_free_sem);
485 wake_up(&c->erase_wait);
486 return;
488 filebad:
489 jffs2_erase_failed(c, jeb, bad_offset);
490 return;
492 refile:
493 /* Stick it back on the list from whence it came and come back later */
494 mutex_lock(&c->erase_free_sem);
495 spin_lock(&c->erase_completion_lock);
496 jffs2_garbage_collect_trigger(c);
497 list_move(&jeb->list, &c->erase_complete_list);
498 spin_unlock(&c->erase_completion_lock);
499 mutex_unlock(&c->erase_free_sem);
500 return;