err.no Git - linux-2.6/blob - fs/jffs2/gc.c

   1 /*
   2  * JFFS2 -- Journalling Flash File System, Version 2.
   3  *
   4  * Copyright © 2001-2007 Red Hat, Inc.
   5  *
   6  * Created by David Woodhouse <dwmw2@infradead.org>
   7  *
   8  * For licensing information, see the file 'LICENCE' in this directory.
   9  *
  10  */
  11
  12 #include <linux/kernel.h>
  13 #include <linux/mtd/mtd.h>
  14 #include <linux/slab.h>
  15 #include <linux/pagemap.h>
  16 #include <linux/crc32.h>
  17 #include <linux/compiler.h>
  18 #include <linux/stat.h>
  19 #include "nodelist.h"
  20 #include "compr.h"
  21
  22 static int jffs2_garbage_collect_pristine(struct jffs2_sb_info *c,
  23                                           struct jffs2_inode_cache *ic,
  24                                           struct jffs2_raw_node_ref *raw);
  25 static int jffs2_garbage_collect_metadata(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
  26                                         struct jffs2_inode_info *f, struct jffs2_full_dnode *fd);
  27 static int jffs2_garbage_collect_dirent(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
  28                                         struct jffs2_inode_info *f, struct jffs2_full_dirent *fd);
  29 static int jffs2_garbage_collect_deletion_dirent(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
  30                                         struct jffs2_inode_info *f, struct jffs2_full_dirent *fd);
  31 static int jffs2_garbage_collect_hole(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
  32                                       struct jffs2_inode_info *f, struct jffs2_full_dnode *fn,
  33                                       uint32_t start, uint32_t end);
  34 static int jffs2_garbage_collect_dnode(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
  35                                        struct jffs2_inode_info *f, struct jffs2_full_dnode *fn,
  36                                        uint32_t start, uint32_t end);
  37 static int jffs2_garbage_collect_live(struct jffs2_sb_info *c,  struct jffs2_eraseblock *jeb,
  38                                struct jffs2_raw_node_ref *raw, struct jffs2_inode_info *f);
  39
  40 /* Called with erase_completion_lock held */
  41 static struct jffs2_eraseblock *jffs2_find_gc_block(struct jffs2_sb_info *c)
  42 {
  43         struct jffs2_eraseblock *ret;
  44         struct list_head *nextlist = NULL;
  45         int n = jiffies % 128;
  46
  47         /* Pick an eraseblock to garbage collect next. This is where we'll
  48            put the clever wear-levelling algorithms. Eventually.  */
  49         /* We possibly want to favour the dirtier blocks more when the
  50            number of free blocks is low. */
  51 again:
  52         if (!list_empty(&c->bad_used_list) && c->nr_free_blocks > c->resv_blocks_gcbad) {
  53                 D1(printk(KERN_DEBUG "Picking block from bad_used_list to GC next\n"));
  54                 nextlist = &c->bad_used_list;
  55         } else if (n < 50 && !list_empty(&c->erasable_list)) {
  56                 /* Note that most of them will have gone directly to be erased.
  57                    So don't favour the erasable_list _too_ much. */
  58                 D1(printk(KERN_DEBUG "Picking block from erasable_list to GC next\n"));
  59                 nextlist = &c->erasable_list;
  60         } else if (n < 110 && !list_empty(&c->very_dirty_list)) {
  61                 /* Most of the time, pick one off the very_dirty list */
  62                 D1(printk(KERN_DEBUG "Picking block from very_dirty_list to GC next\n"));
  63                 nextlist = &c->very_dirty_list;
  64         } else if (n < 126 && !list_empty(&c->dirty_list)) {
  65                 D1(printk(KERN_DEBUG "Picking block from dirty_list to GC next\n"));
  66                 nextlist = &c->dirty_list;
  67         } else if (!list_empty(&c->clean_list)) {
  68                 D1(printk(KERN_DEBUG "Picking block from clean_list to GC next\n"));
  69                 nextlist = &c->clean_list;
  70         } else if (!list_empty(&c->dirty_list)) {
  71                 D1(printk(KERN_DEBUG "Picking block from dirty_list to GC next (clean_list was empty)\n"));
  72
  73                 nextlist = &c->dirty_list;
  74         } else if (!list_empty(&c->very_dirty_list)) {
  75                 D1(printk(KERN_DEBUG "Picking block from very_dirty_list to GC next (clean_list and dirty_list were empty)\n"));
  76                 nextlist = &c->very_dirty_list;
  77         } else if (!list_empty(&c->erasable_list)) {
  78                 D1(printk(KERN_DEBUG "Picking block from erasable_list to GC next (clean_list and {very_,}dirty_list were empty)\n"));
  79
  80                 nextlist = &c->erasable_list;
  81         } else if (!list_empty(&c->erasable_pending_wbuf_list)) {
  82                 /* There are blocks are wating for the wbuf sync */
  83                 D1(printk(KERN_DEBUG "Synching wbuf in order to reuse erasable_pending_wbuf_list blocks\n"));
  84                 spin_unlock(&c->erase_completion_lock);
  85                 jffs2_flush_wbuf_pad(c);
  86                 spin_lock(&c->erase_completion_lock);
  87                 goto again;
  88         } else {
  89                 /* Eep. All were empty */
  90                 D1(printk(KERN_NOTICE "jffs2: No clean, dirty _or_ erasable blocks to GC from! Where are they all?\n"));
  91                 return NULL;
  92         }
  93
  94         ret = list_entry(nextlist->next, struct jffs2_eraseblock, list);
  95         list_del(&ret->list);
  96         c->gcblock = ret;
  97         ret->gc_node = ret->first_node;
  98         if (!ret->gc_node) {
  99                 printk(KERN_WARNING "Eep. ret->gc_node for block at 0x%08x is NULL\n", ret->offset);
 100                 BUG();
 101         }
 102
 103         /* Have we accidentally picked a clean block with wasted space ? */
 104         if (ret->wasted_size) {
 105                 D1(printk(KERN_DEBUG "Converting wasted_size %08x to dirty_size\n", ret->wasted_size));
 106                 ret->dirty_size += ret->wasted_size;
 107                 c->wasted_size -= ret->wasted_size;
 108                 c->dirty_size += ret->wasted_size;
 109                 ret->wasted_size = 0;
 110         }
 111
 112         return ret;
 113 }
 114
 115 /* jffs2_garbage_collect_pass
 116  * Make a single attempt to progress GC. Move one node, and possibly
 117  * start erasing one eraseblock.
 118  */
 119 int jffs2_garbage_collect_pass(struct jffs2_sb_info *c)
 120 {
 121         struct jffs2_inode_info *f;
 122         struct jffs2_inode_cache *ic;
 123         struct jffs2_eraseblock *jeb;
 124         struct jffs2_raw_node_ref *raw;
 125         uint32_t gcblock_dirty;
 126         int ret = 0, inum, nlink;
 127         int xattr = 0;
 128
 129         if (down_interruptible(&c->alloc_sem))
 130                 return -EINTR;
 131
 132         for (;;) {
 133                 spin_lock(&c->erase_completion_lock);
 134                 if (!c->unchecked_size)
 135                         break;
 136
 137                 /* We can't start doing GC yet. We haven't finished checking
 138                    the node CRCs etc. Do it now. */
 139
 140                 /* checked_ino is protected by the alloc_sem */
 141                 if (c->checked_ino > c->highest_ino && xattr) {
 142                         printk(KERN_CRIT "Checked all inodes but still 0x%x bytes of unchecked space?\n",
 143                                c->unchecked_size);
 144                         jffs2_dbg_dump_block_lists_nolock(c);
 145                         spin_unlock(&c->erase_completion_lock);
 146                         up(&c->alloc_sem);
 147                         return -ENOSPC;
 148                 }
 149
 150                 spin_unlock(&c->erase_completion_lock);
 151
 152                 if (!xattr)
 153                         xattr = jffs2_verify_xattr(c);
 154
 155                 spin_lock(&c->inocache_lock);
 156
 157                 ic = jffs2_get_ino_cache(c, c->checked_ino++);
 158
 159                 if (!ic) {
 160                         spin_unlock(&c->inocache_lock);
 161                         continue;
 162                 }
 163
 164                 if (!ic->nlink) {
 165                         D1(printk(KERN_DEBUG "Skipping check of ino #%d with nlink zero\n",
 166                                   ic->ino));
 167                         spin_unlock(&c->inocache_lock);
 168                         jffs2_xattr_delete_inode(c, ic);
 169                         continue;
 170                 }
 171                 switch(ic->state) {
 172                 case INO_STATE_CHECKEDABSENT:
 173                 case INO_STATE_PRESENT:
 174                         D1(printk(KERN_DEBUG "Skipping ino #%u already checked\n", ic->ino));
 175                         spin_unlock(&c->inocache_lock);
 176                         continue;
 177
 178                 case INO_STATE_GC:
 179                 case INO_STATE_CHECKING:
 180                         printk(KERN_WARNING "Inode #%u is in state %d during CRC check phase!\n", ic->ino, ic->state);
 181                         spin_unlock(&c->inocache_lock);
 182                         BUG();
 183
 184                 case INO_STATE_READING:
 185                         /* We need to wait for it to finish, lest we move on
 186                            and trigger the BUG() above while we haven't yet
 187                            finished checking all its nodes */
 188                         D1(printk(KERN_DEBUG "Waiting for ino #%u to finish reading\n", ic->ino));
 189                         /* We need to come back again for the _same_ inode. We've
 190                          made no progress in this case, but that should be OK */
 191                         c->checked_ino--;
 192
 193                         up(&c->alloc_sem);
 194                         sleep_on_spinunlock(&c->inocache_wq, &c->inocache_lock);
 195                         return 0;
 196
 197                 default:
 198                         BUG();
 199
 200                 case INO_STATE_UNCHECKED:
 201                         ;
 202                 }
 203                 ic->state = INO_STATE_CHECKING;
 204                 spin_unlock(&c->inocache_lock);
 205
 206                 D1(printk(KERN_DEBUG "jffs2_garbage_collect_pass() triggering inode scan of ino#%u\n", ic->ino));
 207
 208                 ret = jffs2_do_crccheck_inode(c, ic);
 209                 if (ret)
 210                         printk(KERN_WARNING "Returned error for crccheck of ino #%u. Expect badness...\n", ic->ino);
 211
 212                 jffs2_set_inocache_state(c, ic, INO_STATE_CHECKEDABSENT);
 213                 up(&c->alloc_sem);
 214                 return ret;
 215         }
 216
 217         /* First, work out which block we're garbage-collecting */
 218         jeb = c->gcblock;
 219
 220         if (!jeb)
 221                 jeb = jffs2_find_gc_block(c);
 222
 223         if (!jeb) {
 224                 D1 (printk(KERN_NOTICE "jffs2: Couldn't find erase block to garbage collect!\n"));
 225                 spin_unlock(&c->erase_completion_lock);
 226                 up(&c->alloc_sem);
 227                 return -EIO;
 228         }
 229
 230         D1(printk(KERN_DEBUG "GC from block %08x, used_size %08x, dirty_size %08x, free_size %08x\n", jeb->offset, jeb->used_size, jeb->dirty_size, jeb->free_size));
 231         D1(if (c->nextblock)
 232            printk(KERN_DEBUG "Nextblock at  %08x, used_size %08x, dirty_size %08x, wasted_size %08x, free_size %08x\n", c->nextblock->offset, c->nextblock->used_size, c->nextblock->dirty_size, c->nextblock->wasted_size, c->nextblock->free_size));
 233
 234         if (!jeb->used_size) {
 235                 up(&c->alloc_sem);
 236                 goto eraseit;
 237         }
 238
 239         raw = jeb->gc_node;
 240         gcblock_dirty = jeb->dirty_size;
 241
 242         while(ref_obsolete(raw)) {
 243                 D1(printk(KERN_DEBUG "Node at 0x%08x is obsolete... skipping\n", ref_offset(raw)));
 244                 raw = ref_next(raw);
 245                 if (unlikely(!raw)) {
 246                         printk(KERN_WARNING "eep. End of raw list while still supposedly nodes to GC\n");
 247                         printk(KERN_WARNING "erase block at 0x%08x. free_size 0x%08x, dirty_size 0x%08x, used_size 0x%08x\n",
 248                                jeb->offset, jeb->free_size, jeb->dirty_size, jeb->used_size);
 249                         jeb->gc_node = raw;
 250                         spin_unlock(&c->erase_completion_lock);
 251                         up(&c->alloc_sem);
 252                         BUG();
 253                 }
 254         }
 255         jeb->gc_node = raw;
 256
 257         D1(printk(KERN_DEBUG "Going to garbage collect node at 0x%08x\n", ref_offset(raw)));
 258
 259         if (!raw->next_in_ino) {
 260                 /* Inode-less node. Clean marker, snapshot or something like that */
 261                 spin_unlock(&c->erase_completion_lock);
 262                 if (ref_flags(raw) == REF_PRISTINE) {
 263                         /* It's an unknown node with JFFS2_FEATURE_RWCOMPAT_COPY */
 264                         jffs2_garbage_collect_pristine(c, NULL, raw);
 265                 } else {
 266                         /* Just mark it obsolete */
 267                         jffs2_mark_node_obsolete(c, raw);
 268                 }
 269                 up(&c->alloc_sem);
 270                 goto eraseit_lock;
 271         }
 272
 273         ic = jffs2_raw_ref_to_ic(raw);
 274
 275 #ifdef CONFIG_JFFS2_FS_XATTR
 276         /* When 'ic' refers xattr_datum/xattr_ref, this node is GCed as xattr.
 277          * We can decide whether this node is inode or xattr by ic->class.     */
 278         if (ic->class == RAWNODE_CLASS_XATTR_DATUM
 279             || ic->class == RAWNODE_CLASS_XATTR_REF) {
 280                 spin_unlock(&c->erase_completion_lock);
 281
 282                 if (ic->class == RAWNODE_CLASS_XATTR_DATUM) {
 283                         ret = jffs2_garbage_collect_xattr_datum(c, (struct jffs2_xattr_datum *)ic, raw);
 284                 } else {
 285                         ret = jffs2_garbage_collect_xattr_ref(c, (struct jffs2_xattr_ref *)ic, raw);
 286                 }
 287                 goto test_gcnode;
 288         }
 289 #endif
 290
 291         /* We need to hold the inocache. Either the erase_completion_lock or
 292            the inocache_lock are sufficient; we trade down since the inocache_lock
 293            causes less contention. */
 294         spin_lock(&c->inocache_lock);
 295
 296         spin_unlock(&c->erase_completion_lock);
 297
 298         D1(printk(KERN_DEBUG "jffs2_garbage_collect_pass collecting from block @0x%08x. Node @0x%08x(%d), ino #%u\n", jeb->offset, ref_offset(raw), ref_flags(raw), ic->ino));
 299
 300         /* Three possibilities:
 301            1. Inode is already in-core. We must iget it and do proper
 302               updating to its fragtree, etc.
 303            2. Inode is not in-core, node is REF_PRISTINE. We lock the
 304               inocache to prevent a read_inode(), copy the node intact.
 305            3. Inode is not in-core, node is not pristine. We must iget()
 306               and take the slow path.
 307         */
 308
 309         switch(ic->state) {
 310         case INO_STATE_CHECKEDABSENT:
 311                 /* It's been checked, but it's not currently in-core.
 312                    We can just copy any pristine nodes, but have
 313                    to prevent anyone else from doing read_inode() while
 314                    we're at it, so we set the state accordingly */
 315                 if (ref_flags(raw) == REF_PRISTINE)
 316                         ic->state = INO_STATE_GC;
 317                 else {
 318                         D1(printk(KERN_DEBUG "Ino #%u is absent but node not REF_PRISTINE. Reading.\n",
 319                                   ic->ino));
 320                 }
 321                 break;
 322
 323         case INO_STATE_PRESENT:
 324                 /* It's in-core. GC must iget() it. */
 325                 break;
 326
 327         case INO_STATE_UNCHECKED:
 328         case INO_STATE_CHECKING:
 329         case INO_STATE_GC:
 330                 /* Should never happen. We should have finished checking
 331                    by the time we actually start doing any GC, and since
 332                    we're holding the alloc_sem, no other garbage collection
 333                    can happen.
 334                 */
 335                 printk(KERN_CRIT "Inode #%u already in state %d in jffs2_garbage_collect_pass()!\n",
 336                        ic->ino, ic->state);
 337                 up(&c->alloc_sem);
 338                 spin_unlock(&c->inocache_lock);
 339                 BUG();
 340
 341         case INO_STATE_READING:
 342                 /* Someone's currently trying to read it. We must wait for
 343                    them to finish and then go through the full iget() route
 344                    to do the GC. However, sometimes read_inode() needs to get
 345                    the alloc_sem() (for marking nodes invalid) so we must
 346                    drop the alloc_sem before sleeping. */
 347
 348                 up(&c->alloc_sem);
 349                 D1(printk(KERN_DEBUG "jffs2_garbage_collect_pass() waiting for ino #%u in state %d\n",
 350                           ic->ino, ic->state));
 351                 sleep_on_spinunlock(&c->inocache_wq, &c->inocache_lock);
 352                 /* And because we dropped the alloc_sem we must start again from the
 353                    beginning. Ponder chance of livelock here -- we're returning success
 354                    without actually making any progress.
 355
 356                    Q: What are the chances that the inode is back in INO_STATE_READING
 357                    again by the time we next enter this function? And that this happens
 358                    enough times to cause a real delay?
 359
 360                    A: Small enough that I don't care :)
 361                 */
 362                 return 0;
 363         }
 364
 365         /* OK. Now if the inode is in state INO_STATE_GC, we are going to copy the
 366            node intact, and we don't have to muck about with the fragtree etc.
 367            because we know it's not in-core. If it _was_ in-core, we go through
 368            all the iget() crap anyway */
 369
 370         if (ic->state == INO_STATE_GC) {
 371                 spin_unlock(&c->inocache_lock);
 372
 373                 ret = jffs2_garbage_collect_pristine(c, ic, raw);
 374
 375                 spin_lock(&c->inocache_lock);
 376                 ic->state = INO_STATE_CHECKEDABSENT;
 377                 wake_up(&c->inocache_wq);
 378
 379                 if (ret != -EBADFD) {
 380                         spin_unlock(&c->inocache_lock);
 381                         goto test_gcnode;
 382                 }
 383
 384                 /* Fall through if it wanted us to, with inocache_lock held */
 385         }
 386
 387         /* Prevent the fairly unlikely race where the gcblock is
 388            entirely obsoleted by the final close of a file which had
 389            the only valid nodes in the block, followed by erasure,
 390            followed by freeing of the ic because the erased block(s)
 391            held _all_ the nodes of that inode.... never been seen but
 392            it's vaguely possible. */
 393
 394         inum = ic->ino;
 395         nlink = ic->nlink;
 396         spin_unlock(&c->inocache_lock);
 397
 398         f = jffs2_gc_fetch_inode(c, inum, nlink);
 399         if (IS_ERR(f)) {
 400                 ret = PTR_ERR(f);
 401                 goto release_sem;
 402         }
 403         if (!f) {
 404                 ret = 0;
 405                 goto release_sem;
 406         }
 407
 408         ret = jffs2_garbage_collect_live(c, jeb, raw, f);
 409
 410         jffs2_gc_release_inode(c, f);
 411
 412  test_gcnode:
 413         if (jeb->dirty_size == gcblock_dirty && !ref_obsolete(jeb->gc_node)) {
 414                 /* Eep. This really should never happen. GC is broken */
 415                 printk(KERN_ERR "Error garbage collecting node at %08x!\n", ref_offset(jeb->gc_node));
 416                 ret = -ENOSPC;
 417         }
 418  release_sem:
 419         up(&c->alloc_sem);
 420
 421  eraseit_lock:
 422         /* If we've finished this block, start it erasing */
 423         spin_lock(&c->erase_completion_lock);
 424
 425  eraseit:
 426         if (c->gcblock && !c->gcblock->used_size) {
 427                 D1(printk(KERN_DEBUG "Block at 0x%08x completely obsoleted by GC. Moving to erase_pending_list\n", c->gcblock->offset));
 428                 /* We're GC'ing an empty block? */
 429                 list_add_tail(&c->gcblock->list, &c->erase_pending_list);
 430                 c->gcblock = NULL;
 431                 c->nr_erasing_blocks++;
 432                 jffs2_erase_pending_trigger(c);
 433         }
 434         spin_unlock(&c->erase_completion_lock);
 435
 436         return ret;
 437 }
 438
 439 static int jffs2_garbage_collect_live(struct jffs2_sb_info *c,  struct jffs2_eraseblock *jeb,
 440                                       struct jffs2_raw_node_ref *raw, struct jffs2_inode_info *f)
 441 {
 442         struct jffs2_node_frag *frag;
 443         struct jffs2_full_dnode *fn = NULL;
 444         struct jffs2_full_dirent *fd;
 445         uint32_t start = 0, end = 0, nrfrags = 0;
 446         int ret = 0;
 447
 448         down(&f->sem);
 449
 450         /* Now we have the lock for this inode. Check that it's still the one at the head
 451            of the list. */
 452
 453         spin_lock(&c->erase_completion_lock);
 454
 455         if (c->gcblock != jeb) {
 456                 spin_unlock(&c->erase_completion_lock);
 457                 D1(printk(KERN_DEBUG "GC block is no longer gcblock. Restart\n"));
 458                 goto upnout;
 459         }
 460         if (ref_obsolete(raw)) {
 461                 spin_unlock(&c->erase_completion_lock);
 462                 D1(printk(KERN_DEBUG "node to be GC'd was obsoleted in the meantime.\n"));
 463                 /* They'll call again */
 464                 goto upnout;
 465         }
 466         spin_unlock(&c->erase_completion_lock);
 467
 468         /* OK. Looks safe. And nobody can get us now because we have the semaphore. Move the block */
 469         if (f->metadata && f->metadata->raw == raw) {
 470                 fn = f->metadata;
 471                 ret = jffs2_garbage_collect_metadata(c, jeb, f, fn);
 472                 goto upnout;
 473         }
 474
 475         /* FIXME. Read node and do lookup? */
 476         for (frag = frag_first(&f->fragtree); frag; frag = frag_next(frag)) {
 477                 if (frag->node && frag->node->raw == raw) {
 478                         fn = frag->node;
 479                         end = frag->ofs + frag->size;
 480                         if (!nrfrags++)
 481                                 start = frag->ofs;
 482                         if (nrfrags == frag->node->frags)
 483                                 break; /* We've found them all */
 484                 }
 485         }
 486         if (fn) {
 487                 if (ref_flags(raw) == REF_PRISTINE) {
 488                         ret = jffs2_garbage_collect_pristine(c, f->inocache, raw);
 489                         if (!ret) {
 490                                 /* Urgh. Return it sensibly. */
 491                                 frag->node->raw = f->inocache->nodes;
 492                         }
 493                         if (ret != -EBADFD)
 494                                 goto upnout;
 495                 }
 496                 /* We found a datanode. Do the GC */
 497                 if((start >> PAGE_CACHE_SHIFT) < ((end-1) >> PAGE_CACHE_SHIFT)) {
 498                         /* It crosses a page boundary. Therefore, it must be a hole. */
 499                         ret = jffs2_garbage_collect_hole(c, jeb, f, fn, start, end);
 500                 } else {
 501                         /* It could still be a hole. But we GC the page this way anyway */
 502                         ret = jffs2_garbage_collect_dnode(c, jeb, f, fn, start, end);
 503                 }
 504                 goto upnout;
 505         }
 506
 507         /* Wasn't a dnode. Try dirent */
 508         for (fd = f->dents; fd; fd=fd->next) {
 509                 if (fd->raw == raw)
 510                         break;
 511         }
 512
 513         if (fd && fd->ino) {
 514                 ret = jffs2_garbage_collect_dirent(c, jeb, f, fd);
 515         } else if (fd) {
 516                 ret = jffs2_garbage_collect_deletion_dirent(c, jeb, f, fd);
 517         } else {
 518                 printk(KERN_WARNING "Raw node at 0x%08x wasn't in node lists for ino #%u\n",
 519                        ref_offset(raw), f->inocache->ino);
 520                 if (ref_obsolete(raw)) {
 521                         printk(KERN_WARNING "But it's obsolete so we don't mind too much\n");
 522                 } else {
 523                         jffs2_dbg_dump_node(c, ref_offset(raw));
 524                         BUG();
 525                 }
 526         }
 527  upnout:
 528         up(&f->sem);
 529
 530         return ret;
 531 }
 532
 533 static int jffs2_garbage_collect_pristine(struct jffs2_sb_info *c,
 534                                           struct jffs2_inode_cache *ic,
 535                                           struct jffs2_raw_node_ref *raw)
 536 {
 537         union jffs2_node_union *node;
 538         size_t retlen;
 539         int ret;
 540         uint32_t phys_ofs, alloclen;
 541         uint32_t crc, rawlen;
 542         int retried = 0;
 543
 544         D1(printk(KERN_DEBUG "Going to GC REF_PRISTINE node at 0x%08x\n", ref_offset(raw)));
 545
 546         alloclen = rawlen = ref_totlen(c, c->gcblock, raw);
 547
 548         /* Ask for a small amount of space (or the totlen if smaller) because we
 549            don't want to force wastage of the end of a block if splitting would
 550            work. */
 551         if (ic && alloclen > sizeof(struct jffs2_raw_inode) + JFFS2_MIN_DATA_LEN)
 552                 alloclen = sizeof(struct jffs2_raw_inode) + JFFS2_MIN_DATA_LEN;
 553
 554         ret = jffs2_reserve_space_gc(c, alloclen, &alloclen, rawlen);
 555         /* 'rawlen' is not the exact summary size; it is only an upper estimation */
 556
 557         if (ret)
 558                 return ret;
 559
 560         if (alloclen < rawlen) {
 561                 /* Doesn't fit untouched. We'll go the old route and split it */
 562                 return -EBADFD;
 563         }
 564
 565         node = kmalloc(rawlen, GFP_KERNEL);
 566         if (!node)
 567                 return -ENOMEM;
 568
 569         ret = jffs2_flash_read(c, ref_offset(raw), rawlen, &retlen, (char *)node);
 570         if (!ret && retlen != rawlen)
 571                 ret = -EIO;
 572         if (ret)
 573                 goto out_node;
 574
 575         crc = crc32(0, node, sizeof(struct jffs2_unknown_node)-4);
 576         if (je32_to_cpu(node->u.hdr_crc) != crc) {
 577                 printk(KERN_WARNING "Header CRC failed on REF_PRISTINE node at 0x%08x: Read 0x%08x, calculated 0x%08x\n",
 578                        ref_offset(raw), je32_to_cpu(node->u.hdr_crc), crc);
 579                 goto bail;
 580         }
 581
 582         switch(je16_to_cpu(node->u.nodetype)) {
 583         case JFFS2_NODETYPE_INODE:
 584                 crc = crc32(0, node, sizeof(node->i)-8);
 585                 if (je32_to_cpu(node->i.node_crc) != crc) {
 586                         printk(KERN_WARNING "Node CRC failed on REF_PRISTINE data node at 0x%08x: Read 0x%08x, calculated 0x%08x\n",
 587                                ref_offset(raw), je32_to_cpu(node->i.node_crc), crc);
 588                         goto bail;
 589                 }
 590
 591                 if (je32_to_cpu(node->i.dsize)) {
 592                         crc = crc32(0, node->i.data, je32_to_cpu(node->i.csize));
 593                         if (je32_to_cpu(node->i.data_crc) != crc) {
 594                                 printk(KERN_WARNING "Data CRC failed on REF_PRISTINE data node at 0x%08x: Read 0x%08x, calculated 0x%08x\n",
 595                                        ref_offset(raw), je32_to_cpu(node->i.data_crc), crc);
 596                                 goto bail;
 597                         }
 598                 }
 599                 break;
 600
 601         case JFFS2_NODETYPE_DIRENT:
 602                 crc = crc32(0, node, sizeof(node->d)-8);
 603                 if (je32_to_cpu(node->d.node_crc) != crc) {
 604                         printk(KERN_WARNING "Node CRC failed on REF_PRISTINE dirent node at 0x%08x: Read 0x%08x, calculated 0x%08x\n",
 605                                ref_offset(raw), je32_to_cpu(node->d.node_crc), crc);
 606                         goto bail;
 607                 }
 608
 609                 if (strnlen(node->d.name, node->d.nsize) != node->d.nsize) {
 610                         printk(KERN_WARNING "Name in dirent node at 0x%08x contains zeroes\n", ref_offset(raw));
 611                         goto bail;
 612                 }
 613
 614                 if (node->d.nsize) {
 615                         crc = crc32(0, node->d.name, node->d.nsize);
 616                         if (je32_to_cpu(node->d.name_crc) != crc) {
 617                                 printk(KERN_WARNING "Name CRC failed on REF_PRISTINE dirent node at 0x%08x: Read 0x%08x, calculated 0x%08x\n",
 618                                        ref_offset(raw), je32_to_cpu(node->d.name_crc), crc);
 619                                 goto bail;
 620                         }
 621                 }
 622                 break;
 623         default:
 624                 /* If it's inode-less, we don't _know_ what it is. Just copy it intact */
 625                 if (ic) {
 626                         printk(KERN_WARNING "Unknown node type for REF_PRISTINE node at 0x%08x: 0x%04x\n",
 627                                ref_offset(raw), je16_to_cpu(node->u.nodetype));
 628                         goto bail;
 629                 }
 630         }
 631
 632         /* OK, all the CRCs are good; this node can just be copied as-is. */
 633  retry:
 634         phys_ofs = write_ofs(c);
 635
 636         ret = jffs2_flash_write(c, phys_ofs, rawlen, &retlen, (char *)node);
 637
 638         if (ret || (retlen != rawlen)) {
 639                 printk(KERN_NOTICE "Write of %d bytes at 0x%08x failed. returned %d, retlen %zd\n",
 640                        rawlen, phys_ofs, ret, retlen);
 641                 if (retlen) {
 642                         jffs2_add_physical_node_ref(c, phys_ofs | REF_OBSOLETE, rawlen, NULL);
 643                 } else {
 644                         printk(KERN_NOTICE "Not marking the space at 0x%08x as dirty because the flash driver returned retlen zero\n", phys_ofs);
 645                 }
 646                 if (!retried) {
 647                         /* Try to reallocate space and retry */
 648                         uint32_t dummy;
 649                         struct jffs2_eraseblock *jeb = &c->blocks[phys_ofs / c->sector_size];
 650
 651                         retried = 1;
 652
 653                         D1(printk(KERN_DEBUG "Retrying failed write of REF_PRISTINE node.\n"));
 654
 655                         jffs2_dbg_acct_sanity_check(c,jeb);
 656                         jffs2_dbg_acct_paranoia_check(c, jeb);
 657
 658                         ret = jffs2_reserve_space_gc(c, rawlen, &dummy, rawlen);
 659                                                 /* this is not the exact summary size of it,
 660                                                         it is only an upper estimation */
 661
 662                         if (!ret) {
 663                                 D1(printk(KERN_DEBUG "Allocated space at 0x%08x to retry failed write.\n", phys_ofs));
 664
 665                                 jffs2_dbg_acct_sanity_check(c,jeb);
 666                                 jffs2_dbg_acct_paranoia_check(c, jeb);
 667
 668                                 goto retry;
 669                         }
 670                         D1(printk(KERN_DEBUG "Failed to allocate space to retry failed write: %d!\n", ret));
 671                 }
 672
 673                 if (!ret)
 674                         ret = -EIO;
 675                 goto out_node;
 676         }
 677         jffs2_add_physical_node_ref(c, phys_ofs | REF_PRISTINE, rawlen, ic);
 678
 679         jffs2_mark_node_obsolete(c, raw);
 680         D1(printk(KERN_DEBUG "WHEEE! GC REF_PRISTINE node at 0x%08x succeeded\n", ref_offset(raw)));
 681
 682  out_node:
 683         kfree(node);
 684         return ret;
 685  bail:
 686         ret = -EBADFD;
 687         goto out_node;
 688 }
 689
 690 static int jffs2_garbage_collect_metadata(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
 691                                         struct jffs2_inode_info *f, struct jffs2_full_dnode *fn)
 692 {
 693         struct jffs2_full_dnode *new_fn;
 694         struct jffs2_raw_inode ri;
 695         struct jffs2_node_frag *last_frag;
 696         union jffs2_device_node dev;
 697         char *mdata = NULL, mdatalen = 0;
 698         uint32_t alloclen, ilen;
 699         int ret;
 700
 701         if (S_ISBLK(JFFS2_F_I_MODE(f)) ||
 702             S_ISCHR(JFFS2_F_I_MODE(f)) ) {
 703                 /* For these, we don't actually need to read the old node */
 704                 mdatalen = jffs2_encode_dev(&dev, JFFS2_F_I_RDEV(f));
 705                 mdata = (char *)&dev;
 706                 D1(printk(KERN_DEBUG "jffs2_garbage_collect_metadata(): Writing %d bytes of kdev_t\n", mdatalen));
 707         } else if (S_ISLNK(JFFS2_F_I_MODE(f))) {
 708                 mdatalen = fn->size;
 709                 mdata = kmalloc(fn->size, GFP_KERNEL);
 710                 if (!mdata) {
 711                         printk(KERN_WARNING "kmalloc of mdata failed in jffs2_garbage_collect_metadata()\n");
 712                         return -ENOMEM;
 713                 }
 714                 ret = jffs2_read_dnode(c, f, fn, mdata, 0, mdatalen);
 715                 if (ret) {
 716                         printk(KERN_WARNING "read of old metadata failed in jffs2_garbage_collect_metadata(): %d\n", ret);
 717                         kfree(mdata);
 718                         return ret;
 719                 }
 720                 D1(printk(KERN_DEBUG "jffs2_garbage_collect_metadata(): Writing %d bites of symlink target\n", mdatalen));
 721
 722         }
 723
 724         ret = jffs2_reserve_space_gc(c, sizeof(ri) + mdatalen, &alloclen,
 725                                 JFFS2_SUMMARY_INODE_SIZE);
 726         if (ret) {
 727                 printk(KERN_WARNING "jffs2_reserve_space_gc of %zd bytes for garbage_collect_metadata failed: %d\n",
 728                        sizeof(ri)+ mdatalen, ret);
 729                 goto out;
 730         }
 731
 732         last_frag = frag_last(&f->fragtree);
 733         if (last_frag)
 734                 /* Fetch the inode length from the fragtree rather then
 735                  * from i_size since i_size may have not been updated yet */
 736                 ilen = last_frag->ofs + last_frag->size;
 737         else
 738                 ilen = JFFS2_F_I_SIZE(f);
 739
 740         memset(&ri, 0, sizeof(ri));
 741         ri.magic = cpu_to_je16(JFFS2_MAGIC_BITMASK);
 742         ri.nodetype = cpu_to_je16(JFFS2_NODETYPE_INODE);
 743         ri.totlen = cpu_to_je32(sizeof(ri) + mdatalen);
 744         ri.hdr_crc = cpu_to_je32(crc32(0, &ri, sizeof(struct jffs2_unknown_node)-4));
 745
 746         ri.ino = cpu_to_je32(f->inocache->ino);
 747         ri.version = cpu_to_je32(++f->highest_version);
 748         ri.mode = cpu_to_jemode(JFFS2_F_I_MODE(f));
 749         ri.uid = cpu_to_je16(JFFS2_F_I_UID(f));
 750         ri.gid = cpu_to_je16(JFFS2_F_I_GID(f));
 751         ri.isize = cpu_to_je32(ilen);
 752         ri.atime = cpu_to_je32(JFFS2_F_I_ATIME(f));
 753         ri.ctime = cpu_to_je32(JFFS2_F_I_CTIME(f));
 754         ri.mtime = cpu_to_je32(JFFS2_F_I_MTIME(f));
 755         ri.offset = cpu_to_je32(0);
 756         ri.csize = cpu_to_je32(mdatalen);
 757         ri.dsize = cpu_to_je32(mdatalen);
 758         ri.compr = JFFS2_COMPR_NONE;
 759         ri.node_crc = cpu_to_je32(crc32(0, &ri, sizeof(ri)-8));
 760         ri.data_crc = cpu_to_je32(crc32(0, mdata, mdatalen));
 761
 762         new_fn = jffs2_write_dnode(c, f, &ri, mdata, mdatalen, ALLOC_GC);
 763
 764         if (IS_ERR(new_fn)) {
 765                 printk(KERN_WARNING "Error writing new dnode: %ld\n", PTR_ERR(new_fn));
 766                 ret = PTR_ERR(new_fn);
 767                 goto out;
 768         }
 769         jffs2_mark_node_obsolete(c, fn->raw);
 770         jffs2_free_full_dnode(fn);
 771         f->metadata = new_fn;
 772  out:
 773         if (S_ISLNK(JFFS2_F_I_MODE(f)))
 774                 kfree(mdata);
 775         return ret;
 776 }
 777
 778 static int jffs2_garbage_collect_dirent(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
 779                                         struct jffs2_inode_info *f, struct jffs2_full_dirent *fd)
 780 {
 781         struct jffs2_full_dirent *new_fd;
 782         struct jffs2_raw_dirent rd;
 783         uint32_t alloclen;
 784         int ret;
 785
 786         rd.magic = cpu_to_je16(JFFS2_MAGIC_BITMASK);
 787         rd.nodetype = cpu_to_je16(JFFS2_NODETYPE_DIRENT);
 788         rd.nsize = strlen(fd->name);
 789         rd.totlen = cpu_to_je32(sizeof(rd) + rd.nsize);
 790         rd.hdr_crc = cpu_to_je32(crc32(0, &rd, sizeof(struct jffs2_unknown_node)-4));
 791
 792         rd.pino = cpu_to_je32(f->inocache->ino);
 793         rd.version = cpu_to_je32(++f->highest_version);
 794         rd.ino = cpu_to_je32(fd->ino);
 795         /* If the times on this inode were set by explicit utime() they can be different,
 796            so refrain from splatting them. */
 797         if (JFFS2_F_I_MTIME(f) == JFFS2_F_I_CTIME(f))
 798                 rd.mctime = cpu_to_je32(JFFS2_F_I_MTIME(f));
 799         else
 800                 rd.mctime = cpu_to_je32(0);
 801         rd.type = fd->type;
 802         rd.node_crc = cpu_to_je32(crc32(0, &rd, sizeof(rd)-8));
 803         rd.name_crc = cpu_to_je32(crc32(0, fd->name, rd.nsize));
 804
 805         ret = jffs2_reserve_space_gc(c, sizeof(rd)+rd.nsize, &alloclen,
 806                                 JFFS2_SUMMARY_DIRENT_SIZE(rd.nsize));
 807         if (ret) {
 808                 printk(KERN_WARNING "jffs2_reserve_space_gc of %zd bytes for garbage_collect_dirent failed: %d\n",
 809                        sizeof(rd)+rd.nsize, ret);
 810                 return ret;
 811         }
 812         new_fd = jffs2_write_dirent(c, f, &rd, fd->name, rd.nsize, ALLOC_GC);
 813
 814         if (IS_ERR(new_fd)) {
 815                 printk(KERN_WARNING "jffs2_write_dirent in garbage_collect_dirent failed: %ld\n", PTR_ERR(new_fd));
 816                 return PTR_ERR(new_fd);
 817         }
 818         jffs2_add_fd_to_list(c, new_fd, &f->dents);
 819         return 0;
 820 }
 821
 822 static int jffs2_garbage_collect_deletion_dirent(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
 823                                         struct jffs2_inode_info *f, struct jffs2_full_dirent *fd)
 824 {
 825         struct jffs2_full_dirent **fdp = &f->dents;
 826         int found = 0;
 827
 828         /* On a medium where we can't actually mark nodes obsolete
 829            pernamently, such as NAND flash, we need to work out
 830            whether this deletion dirent is still needed to actively
 831            delete a 'real' dirent with the same name that's still
 832            somewhere else on the flash. */
 833         if (!jffs2_can_mark_obsolete(c)) {
 834                 struct jffs2_raw_dirent *rd;
 835                 struct jffs2_raw_node_ref *raw;
 836                 int ret;
 837                 size_t retlen;
 838                 int name_len = strlen(fd->name);
 839                 uint32_t name_crc = crc32(0, fd->name, name_len);
 840                 uint32_t rawlen = ref_totlen(c, jeb, fd->raw);
 841
 842                 rd = kmalloc(rawlen, GFP_KERNEL);
 843                 if (!rd)
 844                         return -ENOMEM;
 845
 846                 /* Prevent the erase code from nicking the obsolete node refs while
 847                    we're looking at them. I really don't like this extra lock but
 848                    can't see any alternative. Suggestions on a postcard to... */
 849                 down(&c->erase_free_sem);
 850
 851                 for (raw = f->inocache->nodes; raw != (void *)f->inocache; raw = raw->next_in_ino) {
 852
 853                         cond_resched();
 854
 855                         /* We only care about obsolete ones */
 856                         if (!(ref_obsolete(raw)))
 857                                 continue;
 858
 859                         /* Any dirent with the same name is going to have the same length... */
 860                         if (ref_totlen(c, NULL, raw) != rawlen)
 861                                 continue;
 862
 863                         /* Doesn't matter if there's one in the same erase block. We're going to
 864                            delete it too at the same time. */
 865                         if (SECTOR_ADDR(raw->flash_offset) == SECTOR_ADDR(fd->raw->flash_offset))
 866                                 continue;
 867
 868                         D1(printk(KERN_DEBUG "Check potential deletion dirent at %08x\n", ref_offset(raw)));
 869
 870                         /* This is an obsolete node belonging to the same directory, and it's of the right
 871                            length. We need to take a closer look...*/
 872                         ret = jffs2_flash_read(c, ref_offset(raw), rawlen, &retlen, (char *)rd);
 873                         if (ret) {
 874                                 printk(KERN_WARNING "jffs2_g_c_deletion_dirent(): Read error (%d) reading obsolete node at %08x\n", ret, ref_offset(raw));
 875                                 /* If we can't read it, we don't need to continue to obsolete it. Continue */
 876                                 continue;
 877                         }
 878                         if (retlen != rawlen) {
 879                                 printk(KERN_WARNING "jffs2_g_c_deletion_dirent(): Short read (%zd not %u) reading header from obsolete node at %08x\n",
 880                                        retlen, rawlen, ref_offset(raw));
 881                                 continue;
 882                         }
 883
 884                         if (je16_to_cpu(rd->nodetype) != JFFS2_NODETYPE_DIRENT)
 885                                 continue;
 886
 887                         /* If the name CRC doesn't match, skip */
 888                         if (je32_to_cpu(rd->name_crc) != name_crc)
 889                                 continue;
 890
 891                         /* If the name length doesn't match, or it's another deletion dirent, skip */
 892                         if (rd->nsize != name_len || !je32_to_cpu(rd->ino))
 893                                 continue;
 894
 895                         /* OK, check the actual name now */
 896                         if (memcmp(rd->name, fd->name, name_len))
 897                                 continue;
 898
 899                         /* OK. The name really does match. There really is still an older node on
 900                            the flash which our deletion dirent obsoletes. So we have to write out
 901                            a new deletion dirent to replace it */
 902                         up(&c->erase_free_sem);
 903
 904                         D1(printk(KERN_DEBUG "Deletion dirent at %08x still obsoletes real dirent \"%s\" at %08x for ino #%u\n",
 905                                   ref_offset(fd->raw), fd->name, ref_offset(raw), je32_to_cpu(rd->ino)));
 906                         kfree(rd);
 907
 908                         return jffs2_garbage_collect_dirent(c, jeb, f, fd);
 909                 }
 910
 911                 up(&c->erase_free_sem);
 912                 kfree(rd);
 913         }
 914
 915         /* FIXME: If we're deleting a dirent which contains the current mtime and ctime,
 916            we should update the metadata node with those times accordingly */
 917
 918         /* No need for it any more. Just mark it obsolete and remove it from the list */
 919         while (*fdp) {
 920                 if ((*fdp) == fd) {
 921                         found = 1;
 922                         *fdp = fd->next;
 923                         break;
 924                 }
 925                 fdp = &(*fdp)->next;
 926         }
 927         if (!found) {
 928                 printk(KERN_WARNING "Deletion dirent \"%s\" not found in list for ino #%u\n", fd->name, f->inocache->ino);
 929         }
 930         jffs2_mark_node_obsolete(c, fd->raw);
 931         jffs2_free_full_dirent(fd);
 932         return 0;
 933 }
 934
 935 static int jffs2_garbage_collect_hole(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
 936                                       struct jffs2_inode_info *f, struct jffs2_full_dnode *fn,
 937                                       uint32_t start, uint32_t end)
 938 {
 939         struct jffs2_raw_inode ri;
 940         struct jffs2_node_frag *frag;
 941         struct jffs2_full_dnode *new_fn;
 942         uint32_t alloclen, ilen;
 943         int ret;
 944
 945         D1(printk(KERN_DEBUG "Writing replacement hole node for ino #%u from offset 0x%x to 0x%x\n",
 946                   f->inocache->ino, start, end));
 947
 948         memset(&ri, 0, sizeof(ri));
 949
 950         if(fn->frags > 1) {
 951                 size_t readlen;
 952                 uint32_t crc;
 953                 /* It's partially obsoleted by a later write. So we have to
 954                    write it out again with the _same_ version as before */
 955                 ret = jffs2_flash_read(c, ref_offset(fn->raw), sizeof(ri), &readlen, (char *)&ri);
 956                 if (readlen != sizeof(ri) || ret) {
 957                         printk(KERN_WARNING "Node read failed in jffs2_garbage_collect_hole. Ret %d, retlen %zd. Data will be lost by writing new hole node\n", ret, readlen);
 958                         goto fill;
 959                 }
 960                 if (je16_to_cpu(ri.nodetype) != JFFS2_NODETYPE_INODE) {
 961                         printk(KERN_WARNING "jffs2_garbage_collect_hole: Node at 0x%08x had node type 0x%04x instead of JFFS2_NODETYPE_INODE(0x%04x)\n",
 962                                ref_offset(fn->raw),
 963                                je16_to_cpu(ri.nodetype), JFFS2_NODETYPE_INODE);
 964                         return -EIO;
 965                 }
 966                 if (je32_to_cpu(ri.totlen) != sizeof(ri)) {
 967                         printk(KERN_WARNING "jffs2_garbage_collect_hole: Node at 0x%08x had totlen 0x%x instead of expected 0x%zx\n",
 968                                ref_offset(fn->raw),
 969                                je32_to_cpu(ri.totlen), sizeof(ri));
 970                         return -EIO;
 971                 }
 972                 crc = crc32(0, &ri, sizeof(ri)-8);
 973                 if (crc != je32_to_cpu(ri.node_crc)) {
 974                         printk(KERN_WARNING "jffs2_garbage_collect_hole: Node at 0x%08x had CRC 0x%08x which doesn't match calculated CRC 0x%08x\n",
 975                                ref_offset(fn->raw),
 976                                je32_to_cpu(ri.node_crc), crc);
 977                         /* FIXME: We could possibly deal with this by writing new holes for each frag */
 978                         printk(KERN_WARNING "Data in the range 0x%08x to 0x%08x of inode #%u will be lost\n",
 979                                start, end, f->inocache->ino);
 980                         goto fill;
 981                 }
 982                 if (ri.compr != JFFS2_COMPR_ZERO) {
 983                         printk(KERN_WARNING "jffs2_garbage_collect_hole: Node 0x%08x wasn't a hole node!\n", ref_offset(fn->raw));
 984                         printk(KERN_WARNING "Data in the range 0x%08x to 0x%08x of inode #%u will be lost\n",
 985                                start, end, f->inocache->ino);
 986                         goto fill;
 987                 }
 988         } else {
 989         fill:
 990                 ri.magic = cpu_to_je16(JFFS2_MAGIC_BITMASK);
 991                 ri.nodetype = cpu_to_je16(JFFS2_NODETYPE_INODE);
 992                 ri.totlen = cpu_to_je32(sizeof(ri));
 993                 ri.hdr_crc = cpu_to_je32(crc32(0, &ri, sizeof(struct jffs2_unknown_node)-4));
 994
 995                 ri.ino = cpu_to_je32(f->inocache->ino);
 996                 ri.version = cpu_to_je32(++f->highest_version);
 997                 ri.offset = cpu_to_je32(start);
 998                 ri.dsize = cpu_to_je32(end - start);
 999                 ri.csize = cpu_to_je32(0);
1000                 ri.compr = JFFS2_COMPR_ZERO;
1001         }
1002
1003         frag = frag_last(&f->fragtree);
1004         if (frag)
1005                 /* Fetch the inode length from the fragtree rather then
1006                  * from i_size since i_size may have not been updated yet */
1007                 ilen = frag->ofs + frag->size;
1008         else
1009                 ilen = JFFS2_F_I_SIZE(f);
1010
1011         ri.mode = cpu_to_jemode(JFFS2_F_I_MODE(f));
1012         ri.uid = cpu_to_je16(JFFS2_F_I_UID(f));
1013         ri.gid = cpu_to_je16(JFFS2_F_I_GID(f));
1014         ri.isize = cpu_to_je32(ilen);
1015         ri.atime = cpu_to_je32(JFFS2_F_I_ATIME(f));
1016         ri.ctime = cpu_to_je32(JFFS2_F_I_CTIME(f));
1017         ri.mtime = cpu_to_je32(JFFS2_F_I_MTIME(f));
1018         ri.data_crc = cpu_to_je32(0);
1019         ri.node_crc = cpu_to_je32(crc32(0, &ri, sizeof(ri)-8));
1020
1021         ret = jffs2_reserve_space_gc(c, sizeof(ri), &alloclen,
1022                                      JFFS2_SUMMARY_INODE_SIZE);
1023         if (ret) {
1024                 printk(KERN_WARNING "jffs2_reserve_space_gc of %zd bytes for garbage_collect_hole failed: %d\n",
1025                        sizeof(ri), ret);
1026                 return ret;
1027         }
1028         new_fn = jffs2_write_dnode(c, f, &ri, NULL, 0, ALLOC_GC);
1029
1030         if (IS_ERR(new_fn)) {
1031                 printk(KERN_WARNING "Error writing new hole node: %ld\n", PTR_ERR(new_fn));
1032                 return PTR_ERR(new_fn);
1033         }
1034         if (je32_to_cpu(ri.version) == f->highest_version) {
1035                 jffs2_add_full_dnode_to_inode(c, f, new_fn);
1036                 if (f->metadata) {
1037                         jffs2_mark_node_obsolete(c, f->metadata->raw);
1038                         jffs2_free_full_dnode(f->metadata);
1039                         f->metadata = NULL;
1040                 }
1041                 return 0;
1042         }
1043
1044         /*
1045          * We should only get here in the case where the node we are
1046          * replacing had more than one frag, so we kept the same version
1047          * number as before. (Except in case of error -- see 'goto fill;'
1048          * above.)
1049          */
1050         D1(if(unlikely(fn->frags <= 1)) {
1051                 printk(KERN_WARNING "jffs2_garbage_collect_hole: Replacing fn with %d frag(s) but new ver %d != highest_version %d of ino #%d\n",
1052                        fn->frags, je32_to_cpu(ri.version), f->highest_version,
1053                        je32_to_cpu(ri.ino));
1054         });
1055
1056         /* This is a partially-overlapped hole node. Mark it REF_NORMAL not REF_PRISTINE */
1057         mark_ref_normal(new_fn->raw);
1058
1059         for (frag = jffs2_lookup_node_frag(&f->fragtree, fn->ofs);
1060              frag; frag = frag_next(frag)) {
1061                 if (frag->ofs > fn->size + fn->ofs)
1062                         break;
1063                 if (frag->node == fn) {
1064                         frag->node = new_fn;
1065                         new_fn->frags++;
1066                         fn->frags--;
1067                 }
1068         }
1069         if (fn->frags) {
1070                 printk(KERN_WARNING "jffs2_garbage_collect_hole: Old node still has frags!\n");
1071                 BUG();
1072         }
1073         if (!new_fn->frags) {
1074                 printk(KERN_WARNING "jffs2_garbage_collect_hole: New node has no frags!\n");
1075                 BUG();
1076         }
1077
1078         jffs2_mark_node_obsolete(c, fn->raw);
1079         jffs2_free_full_dnode(fn);
1080
1081         return 0;
1082 }
1083
1084 static int jffs2_garbage_collect_dnode(struct jffs2_sb_info *c, struct jffs2_eraseblock *jeb,
1085                                        struct jffs2_inode_info *f, struct jffs2_full_dnode *fn,
1086                                        uint32_t start, uint32_t end)
1087 {
1088         struct jffs2_full_dnode *new_fn;
1089         struct jffs2_raw_inode ri;
1090         uint32_t alloclen, offset, orig_end, orig_start;
1091         int ret = 0;
1092         unsigned char *comprbuf = NULL, *writebuf;
1093         unsigned long pg;
1094         unsigned char *pg_ptr;
1095
1096         memset(&ri, 0, sizeof(ri));
1097
1098         D1(printk(KERN_DEBUG "Writing replacement dnode for ino #%u from offset 0x%x to 0x%x\n",
1099                   f->inocache->ino, start, end));
1100
1101         orig_end = end;
1102         orig_start = start;
1103
1104         if (c->nr_free_blocks + c->nr_erasing_blocks > c->resv_blocks_gcmerge) {
1105                 /* Attempt to do some merging. But only expand to cover logically
1106                    adjacent frags if the block containing them is already considered
1107                    to be dirty. Otherwise we end up with GC just going round in
1108                    circles dirtying the nodes it already wrote out, especially
1109                    on NAND where we have small eraseblocks and hence a much higher
1110                    chance of nodes having to be split to cross boundaries. */
1111
1112                 struct jffs2_node_frag *frag;
1113                 uint32_t min, max;
1114
1115                 min = start & ~(PAGE_CACHE_SIZE-1);
1116                 max = min + PAGE_CACHE_SIZE;
1117
1118                 frag = jffs2_lookup_node_frag(&f->fragtree, start);
1119
1120                 /* BUG_ON(!frag) but that'll happen anyway... */
1121
1122                 BUG_ON(frag->ofs != start);
1123
1124                 /* First grow down... */
1125                 while((frag = frag_prev(frag)) && frag->ofs >= min) {
1126
1127                         /* If the previous frag doesn't even reach the beginning, there's
1128                            excessive fragmentation. Just merge. */
1129                         if (frag->ofs > min) {
1130                                 D1(printk(KERN_DEBUG "Expanding down to cover partial frag (0x%x-0x%x)\n",
1131                                           frag->ofs, frag->ofs+frag->size));
1132                                 start = frag->ofs;
1133                                 continue;
1134                         }
1135                         /* OK. This frag holds the first byte of the page. */
1136                         if (!frag->node || !frag->node->raw) {
1137                                 D1(printk(KERN_DEBUG "First frag in page is hole (0x%x-0x%x). Not expanding down.\n",
1138                                           frag->ofs, frag->ofs+frag->size));
1139                                 break;
1140                         } else {
1141
1142                                 /* OK, it's a frag which extends to the beginning of the page. Does it live
1143                                    in a block which is still considered clean? If so, don't obsolete it.
1144                                    If not, cover it anyway. */
1145
1146                                 struct jffs2_raw_node_ref *raw = frag->node->raw;
1147                                 struct jffs2_eraseblock *jeb;
1148
1149                                 jeb = &c->blocks[raw->flash_offset / c->sector_size];
1150
1151                                 if (jeb == c->gcblock) {
1152                                         D1(printk(KERN_DEBUG "Expanding down to cover frag (0x%x-0x%x) in gcblock at %08x\n",
1153                                                   frag->ofs, frag->ofs+frag->size, ref_offset(raw)));
1154                                         start = frag->ofs;
1155                                         break;
1156                                 }
1157                                 if (!ISDIRTY(jeb->dirty_size + jeb->wasted_size)) {
1158                                         D1(printk(KERN_DEBUG "Not expanding down to cover frag (0x%x-0x%x) in clean block %08x\n",
1159                                                   frag->ofs, frag->ofs+frag->size, jeb->offset));
1160                                         break;
1161                                 }
1162
1163                                 D1(printk(KERN_DEBUG "Expanding down to cover frag (0x%x-0x%x) in dirty block %08x\n",
1164                                                   frag->ofs, frag->ofs+frag->size, jeb->offset));
1165                                 start = frag->ofs;
1166                                 break;
1167                         }
1168                 }
1169
1170                 /* ... then up */
1171
1172                 /* Find last frag which is actually part of the node we're to GC. */
1173                 frag = jffs2_lookup_node_frag(&f->fragtree, end-1);
1174
1175                 while((frag = frag_next(frag)) && frag->ofs+frag->size <= max) {
1176
1177                         /* If the previous frag doesn't even reach the beginning, there's lots
1178                            of fragmentation. Just merge. */
1179                         if (frag->ofs+frag->size < max) {
1180                                 D1(printk(KERN_DEBUG "Expanding up to cover partial frag (0x%x-0x%x)\n",
1181                                           frag->ofs, frag->ofs+frag->size));
1182                                 end = frag->ofs + frag->size;
1183                                 continue;
1184                         }
1185
1186                         if (!frag->node || !frag->node->raw) {
1187                                 D1(printk(KERN_DEBUG "Last frag in page is hole (0x%x-0x%x). Not expanding up.\n",
1188                                           frag->ofs, frag->ofs+frag->size));
1189                                 break;
1190                         } else {
1191
1192                                 /* OK, it's a frag which extends to the beginning of the page. Does it live
1193                                    in a block which is still considered clean? If so, don't obsolete it.
1194                                    If not, cover it anyway. */
1195
1196                                 struct jffs2_raw_node_ref *raw = frag->node->raw;
1197                                 struct jffs2_eraseblock *jeb;
1198
1199                                 jeb = &c->blocks[raw->flash_offset / c->sector_size];
1200
1201                                 if (jeb == c->gcblock) {
1202                                         D1(printk(KERN_DEBUG "Expanding up to cover frag (0x%x-0x%x) in gcblock at %08x\n",
1203                                                   frag->ofs, frag->ofs+frag->size, ref_offset(raw)));
1204                                         end = frag->ofs + frag->size;
1205                                         break;
1206                                 }
1207                                 if (!ISDIRTY(jeb->dirty_size + jeb->wasted_size)) {
1208                                         D1(printk(KERN_DEBUG "Not expanding up to cover frag (0x%x-0x%x) in clean block %08x\n",
1209                                                   frag->ofs, frag->ofs+frag->size, jeb->offset));
1210                                         break;
1211                                 }
1212
1213                                 D1(printk(KERN_DEBUG "Expanding up to cover frag (0x%x-0x%x) in dirty block %08x\n",
1214                                                   frag->ofs, frag->ofs+frag->size, jeb->offset));
1215                                 end = frag->ofs + frag->size;
1216                                 break;
1217                         }
1218                 }
1219                 D1(printk(KERN_DEBUG "Expanded dnode to write from (0x%x-0x%x) to (0x%x-0x%x)\n",
1220                           orig_start, orig_end, start, end));
1221
1222                 D1(BUG_ON(end > frag_last(&f->fragtree)->ofs + frag_last(&f->fragtree)->size));
1223                 BUG_ON(end < orig_end);
1224                 BUG_ON(start > orig_start);
1225         }
1226
1227         /* First, use readpage() to read the appropriate page into the page cache */
1228         /* Q: What happens if we actually try to GC the _same_ page for which commit_write()
1229          *    triggered garbage collection in the first place?
1230          * A: I _think_ it's OK. read_cache_page shouldn't deadlock, we'll write out the
1231          *    page OK. We'll actually write it out again in commit_write, which is a little
1232          *    suboptimal, but at least we're correct.
1233          */
1234         pg_ptr = jffs2_gc_fetch_page(c, f, start, &pg);
1235
1236         if (IS_ERR(pg_ptr)) {
1237                 printk(KERN_WARNING "read_cache_page() returned error: %ld\n", PTR_ERR(pg_ptr));
1238                 return PTR_ERR(pg_ptr);
1239         }
1240
1241         offset = start;
1242         while(offset < orig_end) {
1243                 uint32_t datalen;
1244                 uint32_t cdatalen;
1245                 uint16_t comprtype = JFFS2_COMPR_NONE;
1246
1247                 ret = jffs2_reserve_space_gc(c, sizeof(ri) + JFFS2_MIN_DATA_LEN,
1248                                         &alloclen, JFFS2_SUMMARY_INODE_SIZE);
1249
1250                 if (ret) {
1251                         printk(KERN_WARNING "jffs2_reserve_space_gc of %zd bytes for garbage_collect_dnode failed: %d\n",
1252                                sizeof(ri)+ JFFS2_MIN_DATA_LEN, ret);
1253                         break;
1254                 }
1255                 cdatalen = min_t(uint32_t, alloclen - sizeof(ri), end - offset);
1256                 datalen = end - offset;
1257
1258                 writebuf = pg_ptr + (offset & (PAGE_CACHE_SIZE -1));
1259
1260                 comprtype = jffs2_compress(c, f, writebuf, &comprbuf, &datalen, &cdatalen);
1261
1262                 ri.magic = cpu_to_je16(JFFS2_MAGIC_BITMASK);
1263                 ri.nodetype = cpu_to_je16(JFFS2_NODETYPE_INODE);
1264                 ri.totlen = cpu_to_je32(sizeof(ri) + cdatalen);
1265                 ri.hdr_crc = cpu_to_je32(crc32(0, &ri, sizeof(struct jffs2_unknown_node)-4));
1266
1267                 ri.ino = cpu_to_je32(f->inocache->ino);
1268                 ri.version = cpu_to_je32(++f->highest_version);
1269                 ri.mode = cpu_to_jemode(JFFS2_F_I_MODE(f));
1270                 ri.uid = cpu_to_je16(JFFS2_F_I_UID(f));
1271                 ri.gid = cpu_to_je16(JFFS2_F_I_GID(f));
1272                 ri.isize = cpu_to_je32(JFFS2_F_I_SIZE(f));
1273                 ri.atime = cpu_to_je32(JFFS2_F_I_ATIME(f));
1274                 ri.ctime = cpu_to_je32(JFFS2_F_I_CTIME(f));
1275                 ri.mtime = cpu_to_je32(JFFS2_F_I_MTIME(f));
1276                 ri.offset = cpu_to_je32(offset);
1277                 ri.csize = cpu_to_je32(cdatalen);
1278                 ri.dsize = cpu_to_je32(datalen);
1279                 ri.compr = comprtype & 0xff;
1280                 ri.usercompr = (comprtype >> 8) & 0xff;
1281                 ri.node_crc = cpu_to_je32(crc32(0, &ri, sizeof(ri)-8));
1282                 ri.data_crc = cpu_to_je32(crc32(0, comprbuf, cdatalen));
1283
1284                 new_fn = jffs2_write_dnode(c, f, &ri, comprbuf, cdatalen, ALLOC_GC);
1285
1286                 jffs2_free_comprbuf(comprbuf, writebuf);
1287
1288                 if (IS_ERR(new_fn)) {
1289                         printk(KERN_WARNING "Error writing new dnode: %ld\n", PTR_ERR(new_fn));
1290                         ret = PTR_ERR(new_fn);
1291                         break;
1292                 }
1293                 ret = jffs2_add_full_dnode_to_inode(c, f, new_fn);
1294                 offset += datalen;
1295                 if (f->metadata) {
1296                         jffs2_mark_node_obsolete(c, f->metadata->raw);
1297                         jffs2_free_full_dnode(f->metadata);
1298                         f->metadata = NULL;
1299                 }
1300         }
1301
1302         jffs2_gc_release_page(c, pg_ptr, &pg);
1303         return ret;
1304 }