SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/sunrpc/cache.c

   1 /*
   2  * net/sunrpc/cache.c
   3  *
   4  * Generic code for various authentication-related caches
   5  * used by sunrpc clients and servers.
   6  *
   7  * Copyright (C) 2002 Neil Brown <neilb@cse.unsw.edu.au>
   8  *
   9  * Released under terms in GPL version 2.  See COPYING.
  10  *
  11  */
  12
  13 #include <linux/types.h>
  14 #include <linux/fs.h>
  15 #include <linux/file.h>
  16 #include <linux/slab.h>
  17 #include <linux/signal.h>
  18 #include <linux/sched.h>
  19 #include <linux/kmod.h>
  20 #include <linux/list.h>
  21 #include <linux/module.h>
  22 #include <linux/ctype.h>
  23 #include <asm/uaccess.h>
  24 #include <linux/poll.h>
  25 #include <linux/seq_file.h>
  26 #include <linux/proc_fs.h>
  27 #include <linux/net.h>
  28 #include <linux/workqueue.h>
  29 #include <linux/mutex.h>
  30 #include <linux/pagemap.h>
  31 #include <asm/ioctls.h>
  32 #include <linux/sunrpc/types.h>
  33 #include <linux/sunrpc/cache.h>
  34 #include <linux/sunrpc/stats.h>
  35 #include <linux/sunrpc/rpc_pipe_fs.h>
  36
  37 #define  RPCDBG_FACILITY RPCDBG_CACHE
  38
  39 static int cache_defer_req(struct cache_req *req, struct cache_head *item);
  40 static void cache_revisit_request(struct cache_head *item);
  41
  42 static void cache_init(struct cache_head *h)
  43 {
  44         time_t now = get_seconds();
  45         h->next = NULL;
  46         h->flags = 0;
  47         kref_init(&h->ref);
  48         h->expiry_time = now + CACHE_NEW_EXPIRY;
  49         h->last_refresh = now;
  50 }
  51
  52 struct cache_head *sunrpc_cache_lookup(struct cache_detail *detail,
  53                                        struct cache_head *key, int hash)
  54 {
  55         struct cache_head **head,  **hp;
  56         struct cache_head *new = NULL;
  57
  58         head = &detail->hash_table[hash];
  59
  60         read_lock(&detail->hash_lock);
  61
  62         for (hp=head; *hp != NULL ; hp = &(*hp)->next) {
  63                 struct cache_head *tmp = *hp;
  64                 if (detail->match(tmp, key)) {
  65                         cache_get(tmp);
  66                         read_unlock(&detail->hash_lock);
  67                         return tmp;
  68                 }
  69         }
  70         read_unlock(&detail->hash_lock);
  71         /* Didn't find anything, insert an empty entry */
  72
  73         new = detail->alloc();
  74         if (!new)
  75                 return NULL;
  76         /* must fully initialise 'new', else
  77          * we might get lose if we need to
  78          * cache_put it soon.
  79          */
  80         cache_init(new);
  81         detail->init(new, key);
  82
  83         write_lock(&detail->hash_lock);
  84
  85         /* check if entry appeared while we slept */
  86         for (hp=head; *hp != NULL ; hp = &(*hp)->next) {
  87                 struct cache_head *tmp = *hp;
  88                 if (detail->match(tmp, key)) {
  89                         cache_get(tmp);
  90                         write_unlock(&detail->hash_lock);
  91                         cache_put(new, detail);
  92                         return tmp;
  93                 }
  94         }
  95         new->next = *head;
  96         *head = new;
  97         detail->entries++;
  98         cache_get(new);
  99         write_unlock(&detail->hash_lock);
 100
 101         return new;
 102 }
 103 EXPORT_SYMBOL_GPL(sunrpc_cache_lookup);
 104
 105
 106 static void cache_dequeue(struct cache_detail *detail, struct cache_head *ch);
 107
 108 static int cache_fresh_locked(struct cache_head *head, time_t expiry)
 109 {
 110         head->expiry_time = expiry;
 111         head->last_refresh = get_seconds();
 112         return !test_and_set_bit(CACHE_VALID, &head->flags);
 113 }
 114
 115 static void cache_fresh_unlocked(struct cache_head *head,
 116                         struct cache_detail *detail, int new)
 117 {
 118         if (new)
 119                 cache_revisit_request(head);
 120         if (test_and_clear_bit(CACHE_PENDING, &head->flags)) {
 121                 cache_revisit_request(head);
 122                 cache_dequeue(detail, head);
 123         }
 124 }
 125
 126 struct cache_head *sunrpc_cache_update(struct cache_detail *detail,
 127                                        struct cache_head *new, struct cache_head *old, int hash)
 128 {
 129         /* The 'old' entry is to be replaced by 'new'.
 130          * If 'old' is not VALID, we update it directly,
 131          * otherwise we need to replace it
 132          */
 133         struct cache_head **head;
 134         struct cache_head *tmp;
 135         int is_new;
 136
 137         if (!test_bit(CACHE_VALID, &old->flags)) {
 138                 write_lock(&detail->hash_lock);
 139                 if (!test_bit(CACHE_VALID, &old->flags)) {
 140                         if (test_bit(CACHE_NEGATIVE, &new->flags))
 141                                 set_bit(CACHE_NEGATIVE, &old->flags);
 142                         else
 143                                 detail->update(old, new);
 144                         is_new = cache_fresh_locked(old, new->expiry_time);
 145                         write_unlock(&detail->hash_lock);
 146                         cache_fresh_unlocked(old, detail, is_new);
 147                         return old;
 148                 }
 149                 write_unlock(&detail->hash_lock);
 150         }
 151         /* We need to insert a new entry */
 152         tmp = detail->alloc();
 153         if (!tmp) {
 154                 cache_put(old, detail);
 155                 return NULL;
 156         }
 157         cache_init(tmp);
 158         detail->init(tmp, old);
 159         head = &detail->hash_table[hash];
 160
 161         write_lock(&detail->hash_lock);
 162         if (test_bit(CACHE_NEGATIVE, &new->flags))
 163                 set_bit(CACHE_NEGATIVE, &tmp->flags);
 164         else
 165                 detail->update(tmp, new);
 166         tmp->next = *head;
 167         *head = tmp;
 168         detail->entries++;
 169         cache_get(tmp);
 170         is_new = cache_fresh_locked(tmp, new->expiry_time);
 171         cache_fresh_locked(old, 0);
 172         write_unlock(&detail->hash_lock);
 173         cache_fresh_unlocked(tmp, detail, is_new);
 174         cache_fresh_unlocked(old, detail, 0);
 175         cache_put(old, detail);
 176         return tmp;
 177 }
 178 EXPORT_SYMBOL_GPL(sunrpc_cache_update);
 179
 180 static int cache_make_upcall(struct cache_detail *cd, struct cache_head *h)
 181 {
 182         if (!cd->cache_upcall)
 183                 return -EINVAL;
 184         return cd->cache_upcall(cd, h);
 185 }
 186
 187 static inline int cache_is_valid(struct cache_detail *detail, struct cache_head *h)
 188 {
 189         if (!test_bit(CACHE_VALID, &h->flags) ||
 190             h->expiry_time < get_seconds())
 191                 return -EAGAIN;
 192         else if (detail->flush_time > h->last_refresh)
 193                 return -EAGAIN;
 194         else {
 195                 /* entry is valid */
 196                 if (test_bit(CACHE_NEGATIVE, &h->flags))
 197                         return -ENOENT;
 198                 else
 199                         return 0;
 200         }
 201 }
 202
 203 /*
 204  * This is the generic cache management routine for all
 205  * the authentication caches.
 206  * It checks the currency of a cache item and will (later)
 207  * initiate an upcall to fill it if needed.
 208  *
 209  *
 210  * Returns 0 if the cache_head can be used, or cache_puts it and returns
 211  * -EAGAIN if upcall is pending and request has been queued
 212  * -ETIMEDOUT if upcall failed or request could not be queue or
 213  *           upcall completed but item is still invalid (implying that
 214  *           the cache item has been replaced with a newer one).
 215  * -ENOENT if cache entry was negative
 216  */
 217 int cache_check(struct cache_detail *detail,
 218                     struct cache_head *h, struct cache_req *rqstp)
 219 {
 220         int rv;
 221         long refresh_age, age;
 222
 223         /* First decide return status as best we can */
 224         rv = cache_is_valid(detail, h);
 225
 226         /* now see if we want to start an upcall */
 227         refresh_age = (h->expiry_time - h->last_refresh);
 228         age = get_seconds() - h->last_refresh;
 229
 230         if (rqstp == NULL) {
 231                 if (rv == -EAGAIN)
 232                         rv = -ENOENT;
 233         } else if (rv == -EAGAIN || age > refresh_age/2) {
 234                 dprintk("RPC:       Want update, refage=%ld, age=%ld\n",
 235                                 refresh_age, age);
 236                 if (!test_and_set_bit(CACHE_PENDING, &h->flags)) {
 237                         switch (cache_make_upcall(detail, h)) {
 238                         case -EINVAL:
 239                                 clear_bit(CACHE_PENDING, &h->flags);
 240                                 cache_revisit_request(h);
 241                                 if (rv == -EAGAIN) {
 242                                         set_bit(CACHE_NEGATIVE, &h->flags);
 243                                         cache_fresh_unlocked(h, detail,
 244                                              cache_fresh_locked(h, get_seconds()+CACHE_NEW_EXPIRY));
 245                                         rv = -ENOENT;
 246                                 }
 247                                 break;
 248
 249                         case -EAGAIN:
 250                                 clear_bit(CACHE_PENDING, &h->flags);
 251                                 cache_revisit_request(h);
 252                                 break;
 253                         }
 254                 }
 255         }
 256
 257         if (rv == -EAGAIN) {
 258                 if (cache_defer_req(rqstp, h) < 0) {
 259                         /* Request is not deferred */
 260                         rv = cache_is_valid(detail, h);
 261                         if (rv == -EAGAIN)
 262                                 rv = -ETIMEDOUT;
 263                 }
 264         }
 265         if (rv)
 266                 cache_put(h, detail);
 267         return rv;
 268 }
 269 EXPORT_SYMBOL_GPL(cache_check);
 270
 271 /*
 272  * caches need to be periodically cleaned.
 273  * For this we maintain a list of cache_detail and
 274  * a current pointer into that list and into the table
 275  * for that entry.
 276  *
 277  * Each time clean_cache is called it finds the next non-empty entry
 278  * in the current table and walks the list in that entry
 279  * looking for entries that can be removed.
 280  *
 281  * An entry gets removed if:
 282  * - The expiry is before current time
 283  * - The last_refresh time is before the flush_time for that cache
 284  *
 285  * later we might drop old entries with non-NEVER expiry if that table
 286  * is getting 'full' for some definition of 'full'
 287  *
 288  * The question of "how often to scan a table" is an interesting one
 289  * and is answered in part by the use of the "nextcheck" field in the
 290  * cache_detail.
 291  * When a scan of a table begins, the nextcheck field is set to a time
 292  * that is well into the future.
 293  * While scanning, if an expiry time is found that is earlier than the
 294  * current nextcheck time, nextcheck is set to that expiry time.
 295  * If the flush_time is ever set to a time earlier than the nextcheck
 296  * time, the nextcheck time is then set to that flush_time.
 297  *
 298  * A table is then only scanned if the current time is at least
 299  * the nextcheck time.
 300  *
 301  */
 302
 303 static LIST_HEAD(cache_list);
 304 static DEFINE_SPINLOCK(cache_list_lock);
 305 static struct cache_detail *current_detail;
 306 static int current_index;
 307
 308 static void do_cache_clean(struct work_struct *work);
 309 static DECLARE_DELAYED_WORK(cache_cleaner, do_cache_clean);
 310
 311 static void sunrpc_init_cache_detail(struct cache_detail *cd)
 312 {
 313         rwlock_init(&cd->hash_lock);
 314         INIT_LIST_HEAD(&cd->queue);
 315         spin_lock(&cache_list_lock);
 316         cd->nextcheck = 0;
 317         cd->entries = 0;
 318         atomic_set(&cd->readers, 0);
 319         cd->last_close = 0;
 320         cd->last_warn = -1;
 321         list_add(&cd->others, &cache_list);
 322         spin_unlock(&cache_list_lock);
 323
 324         /* start the cleaning process */
 325         schedule_delayed_work(&cache_cleaner, 0);
 326 }
 327
 328 static void sunrpc_destroy_cache_detail(struct cache_detail *cd)
 329 {
 330         cache_purge(cd);
 331         spin_lock(&cache_list_lock);
 332         write_lock(&cd->hash_lock);
 333         if (cd->entries || atomic_read(&cd->inuse)) {
 334                 write_unlock(&cd->hash_lock);
 335                 spin_unlock(&cache_list_lock);
 336                 goto out;
 337         }
 338         if (current_detail == cd)
 339                 current_detail = NULL;
 340         list_del_init(&cd->others);
 341         write_unlock(&cd->hash_lock);
 342         spin_unlock(&cache_list_lock);
 343         if (list_empty(&cache_list)) {
 344                 /* module must be being unloaded so its safe to kill the worker */
 345                 cancel_delayed_work_sync(&cache_cleaner);
 346         }
 347         return;
 348 out:
 349         printk(KERN_ERR "nfsd: failed to unregister %s cache\n", cd->name);
 350 }
 351
 352 /* clean cache tries to find something to clean
 353  * and cleans it.
 354  * It returns 1 if it cleaned something,
 355  *            0 if it didn't find anything this time
 356  *           -1 if it fell off the end of the list.
 357  */
 358 static int cache_clean(void)
 359 {
 360         int rv = 0;
 361         struct list_head *next;
 362
 363         spin_lock(&cache_list_lock);
 364
 365         /* find a suitable table if we don't already have one */
 366         while (current_detail == NULL ||
 367             current_index >= current_detail->hash_size) {
 368                 if (current_detail)
 369                         next = current_detail->others.next;
 370                 else
 371                         next = cache_list.next;
 372                 if (next == &cache_list) {
 373                         current_detail = NULL;
 374                         spin_unlock(&cache_list_lock);
 375                         return -1;
 376                 }
 377                 current_detail = list_entry(next, struct cache_detail, others);
 378                 if (current_detail->nextcheck > get_seconds())
 379                         current_index = current_detail->hash_size;
 380                 else {
 381                         current_index = 0;
 382                         current_detail->nextcheck = get_seconds()+30*60;
 383                 }
 384         }
 385
 386         /* find a non-empty bucket in the table */
 387         while (current_detail &&
 388                current_index < current_detail->hash_size &&
 389                current_detail->hash_table[current_index] == NULL)
 390                 current_index++;
 391
 392         /* find a cleanable entry in the bucket and clean it, or set to next bucket */
 393
 394         if (current_detail && current_index < current_detail->hash_size) {
 395                 struct cache_head *ch, **cp;
 396                 struct cache_detail *d;
 397
 398                 write_lock(&current_detail->hash_lock);
 399
 400                 /* Ok, now to clean this strand */
 401
 402                 cp = & current_detail->hash_table[current_index];
 403                 ch = *cp;
 404                 for (; ch; cp= & ch->next, ch= *cp) {
 405                         if (current_detail->nextcheck > ch->expiry_time)
 406                                 current_detail->nextcheck = ch->expiry_time+1;
 407                         if (ch->expiry_time >= get_seconds()
 408                             && ch->last_refresh >= current_detail->flush_time
 409                                 )
 410                                 continue;
 411                         if (test_and_clear_bit(CACHE_PENDING, &ch->flags))
 412                                 cache_dequeue(current_detail, ch);
 413
 414                         if (atomic_read(&ch->ref.refcount) == 1)
 415                                 break;
 416                 }
 417                 if (ch) {
 418                         *cp = ch->next;
 419                         ch->next = NULL;
 420                         current_detail->entries--;
 421                         rv = 1;
 422                 }
 423                 write_unlock(&current_detail->hash_lock);
 424                 d = current_detail;
 425                 if (!ch)
 426                         current_index ++;
 427                 spin_unlock(&cache_list_lock);
 428                 if (ch) {
 429                         cache_revisit_request(ch);
 430                         cache_put(ch, d);
 431                 }
 432         } else
 433                 spin_unlock(&cache_list_lock);
 434
 435         return rv;
 436 }
 437
 438 /*
 439  * We want to regularly clean the cache, so we need to schedule some work ...
 440  */
 441 static void do_cache_clean(struct work_struct *work)
 442 {
 443         int delay = 5;
 444         if (cache_clean() == -1)
 445                 delay = round_jiffies_relative(30*HZ);
 446
 447         if (list_empty(&cache_list))
 448                 delay = 0;
 449
 450         if (delay)
 451                 schedule_delayed_work(&cache_cleaner, delay);
 452 }
 453
 454
 455 /*
 456  * Clean all caches promptly.  This just calls cache_clean
 457  * repeatedly until we are sure that every cache has had a chance to
 458  * be fully cleaned
 459  */
 460 void cache_flush(void)
 461 {
 462         while (cache_clean() != -1)
 463                 cond_resched();
 464         while (cache_clean() != -1)
 465                 cond_resched();
 466 }
 467 EXPORT_SYMBOL_GPL(cache_flush);
 468
 469 void cache_purge(struct cache_detail *detail)
 470 {
 471         detail->flush_time = LONG_MAX;
 472         detail->nextcheck = get_seconds();
 473         cache_flush();
 474         detail->flush_time = 1;
 475 }
 476 EXPORT_SYMBOL_GPL(cache_purge);
 477
 478
 479 /*
 480  * Deferral and Revisiting of Requests.
 481  *
 482  * If a cache lookup finds a pending entry, we
 483  * need to defer the request and revisit it later.
 484  * All deferred requests are stored in a hash table,
 485  * indexed by "struct cache_head *".
 486  * As it may be wasteful to store a whole request
 487  * structure, we allow the request to provide a
 488  * deferred form, which must contain a
 489  * 'struct cache_deferred_req'
 490  * This cache_deferred_req contains a method to allow
 491  * it to be revisited when cache info is available
 492  */
 493
 494 #define DFR_HASHSIZE    (PAGE_SIZE/sizeof(struct list_head))
 495 #define DFR_HASH(item)  ((((long)item)>>4 ^ (((long)item)>>13)) % DFR_HASHSIZE)
 496
 497 #define DFR_MAX 300     /* ??? */
 498
 499 static DEFINE_SPINLOCK(cache_defer_lock);
 500 static LIST_HEAD(cache_defer_list);
 501 static struct list_head cache_defer_hash[DFR_HASHSIZE];
 502 static int cache_defer_cnt;
 503
 504 static int cache_defer_req(struct cache_req *req, struct cache_head *item)
 505 {
 506         struct cache_deferred_req *dreq;
 507         int hash = DFR_HASH(item);
 508
 509         if (cache_defer_cnt >= DFR_MAX) {
 510                 /* too much in the cache, randomly drop this one,
 511                  * or continue and drop the oldest below
 512                  */
 513                 if (net_random()&1)
 514                         return -ENOMEM;
 515         }
 516         dreq = req->defer(req);
 517         if (dreq == NULL)
 518                 return -ENOMEM;
 519
 520         dreq->item = item;
 521
 522         spin_lock(&cache_defer_lock);
 523
 524         list_add(&dreq->recent, &cache_defer_list);
 525
 526         if (cache_defer_hash[hash].next == NULL)
 527                 INIT_LIST_HEAD(&cache_defer_hash[hash]);
 528         list_add(&dreq->hash, &cache_defer_hash[hash]);
 529
 530         /* it is in, now maybe clean up */
 531         dreq = NULL;
 532         if (++cache_defer_cnt > DFR_MAX) {
 533                 dreq = list_entry(cache_defer_list.prev,
 534                                   struct cache_deferred_req, recent);
 535                 list_del(&dreq->recent);
 536                 list_del(&dreq->hash);
 537                 cache_defer_cnt--;
 538         }
 539         spin_unlock(&cache_defer_lock);
 540
 541         if (dreq) {
 542                 /* there was one too many */
 543                 dreq->revisit(dreq, 1);
 544         }
 545         if (!test_bit(CACHE_PENDING, &item->flags)) {
 546                 /* must have just been validated... */
 547                 cache_revisit_request(item);
 548                 return -EAGAIN;
 549         }
 550         return 0;
 551 }
 552
 553 static void cache_revisit_request(struct cache_head *item)
 554 {
 555         struct cache_deferred_req *dreq;
 556         struct list_head pending;
 557
 558         struct list_head *lp;
 559         int hash = DFR_HASH(item);
 560
 561         INIT_LIST_HEAD(&pending);
 562         spin_lock(&cache_defer_lock);
 563
 564         lp = cache_defer_hash[hash].next;
 565         if (lp) {
 566                 while (lp != &cache_defer_hash[hash]) {
 567                         dreq = list_entry(lp, struct cache_deferred_req, hash);
 568                         lp = lp->next;
 569                         if (dreq->item == item) {
 570                                 list_del(&dreq->hash);
 571                                 list_move(&dreq->recent, &pending);
 572                                 cache_defer_cnt--;
 573                         }
 574                 }
 575         }
 576         spin_unlock(&cache_defer_lock);
 577
 578         while (!list_empty(&pending)) {
 579                 dreq = list_entry(pending.next, struct cache_deferred_req, recent);
 580                 list_del_init(&dreq->recent);
 581                 dreq->revisit(dreq, 0);
 582         }
 583 }
 584
 585 void cache_clean_deferred(void *owner)
 586 {
 587         struct cache_deferred_req *dreq, *tmp;
 588         struct list_head pending;
 589
 590
 591         INIT_LIST_HEAD(&pending);
 592         spin_lock(&cache_defer_lock);
 593
 594         list_for_each_entry_safe(dreq, tmp, &cache_defer_list, recent) {
 595                 if (dreq->owner == owner) {
 596                         list_del(&dreq->hash);
 597                         list_move(&dreq->recent, &pending);
 598                         cache_defer_cnt--;
 599                 }
 600         }
 601         spin_unlock(&cache_defer_lock);
 602
 603         while (!list_empty(&pending)) {
 604                 dreq = list_entry(pending.next, struct cache_deferred_req, recent);
 605                 list_del_init(&dreq->recent);
 606                 dreq->revisit(dreq, 1);
 607         }
 608 }
 609
 610 /*
 611  * communicate with user-space
 612  *
 613  * We have a magic /proc file - /proc/sunrpc/<cachename>/channel.
 614  * On read, you get a full request, or block.
 615  * On write, an update request is processed.
 616  * Poll works if anything to read, and always allows write.
 617  *
 618  * Implemented by linked list of requests.  Each open file has
 619  * a ->private that also exists in this list.  New requests are added
 620  * to the end and may wakeup and preceding readers.
 621  * New readers are added to the head.  If, on read, an item is found with
 622  * CACHE_UPCALLING clear, we free it from the list.
 623  *
 624  */
 625
 626 static DEFINE_SPINLOCK(queue_lock);
 627 static DEFINE_MUTEX(queue_io_mutex);
 628
 629 struct cache_queue {
 630         struct list_head        list;
 631         int                     reader; /* if 0, then request */
 632 };
 633 struct cache_request {
 634         struct cache_queue      q;
 635         struct cache_head       *item;
 636         char                    * buf;
 637         int                     len;
 638         int                     readers;
 639 };
 640 struct cache_reader {
 641         struct cache_queue      q;
 642         int                     offset; /* if non-0, we have a refcnt on next request */
 643 };
 644
 645 static ssize_t cache_read(struct file *filp, char __user *buf, size_t count,
 646                           loff_t *ppos, struct cache_detail *cd)
 647 {
 648         struct cache_reader *rp = filp->private_data;
 649         struct cache_request *rq;
 650         struct inode *inode = filp->f_path.dentry->d_inode;
 651         int err;
 652
 653         if (count == 0)
 654                 return 0;
 655
 656         mutex_lock(&inode->i_mutex); /* protect against multiple concurrent
 657                               * readers on this file */
 658  again:
 659         spin_lock(&queue_lock);
 660         /* need to find next request */
 661         while (rp->q.list.next != &cd->queue &&
 662                list_entry(rp->q.list.next, struct cache_queue, list)
 663                ->reader) {
 664                 struct list_head *next = rp->q.list.next;
 665                 list_move(&rp->q.list, next);
 666         }
 667         if (rp->q.list.next == &cd->queue) {
 668                 spin_unlock(&queue_lock);
 669                 mutex_unlock(&inode->i_mutex);
 670                 BUG_ON(rp->offset);
 671                 return 0;
 672         }
 673         rq = container_of(rp->q.list.next, struct cache_request, q.list);
 674         BUG_ON(rq->q.reader);
 675         if (rp->offset == 0)
 676                 rq->readers++;
 677         spin_unlock(&queue_lock);
 678
 679         if (rp->offset == 0 && !test_bit(CACHE_PENDING, &rq->item->flags)) {
 680                 err = -EAGAIN;
 681                 spin_lock(&queue_lock);
 682                 list_move(&rp->q.list, &rq->q.list);
 683                 spin_unlock(&queue_lock);
 684         } else {
 685                 if (rp->offset + count > rq->len)
 686                         count = rq->len - rp->offset;
 687                 err = -EFAULT;
 688                 if (copy_to_user(buf, rq->buf + rp->offset, count))
 689                         goto out;
 690                 rp->offset += count;
 691                 if (rp->offset >= rq->len) {
 692                         rp->offset = 0;
 693                         spin_lock(&queue_lock);
 694                         list_move(&rp->q.list, &rq->q.list);
 695                         spin_unlock(&queue_lock);
 696                 }
 697                 err = 0;
 698         }
 699  out:
 700         if (rp->offset == 0) {
 701                 /* need to release rq */
 702                 spin_lock(&queue_lock);
 703                 rq->readers--;
 704                 if (rq->readers == 0 &&
 705                     !test_bit(CACHE_PENDING, &rq->item->flags)) {
 706                         list_del(&rq->q.list);
 707                         spin_unlock(&queue_lock);
 708                         cache_put(rq->item, cd);
 709                         kfree(rq->buf);
 710                         kfree(rq);
 711                 } else
 712                         spin_unlock(&queue_lock);
 713         }
 714         if (err == -EAGAIN)
 715                 goto again;
 716         mutex_unlock(&inode->i_mutex);
 717         return err ? err :  count;
 718 }
 719
 720 static ssize_t cache_do_downcall(char *kaddr, const char __user *buf,
 721                                  size_t count, struct cache_detail *cd)
 722 {
 723         ssize_t ret;
 724
 725         if (copy_from_user(kaddr, buf, count))
 726                 return -EFAULT;
 727         kaddr[count] = '\0';
 728         ret = cd->cache_parse(cd, kaddr, count);
 729         if (!ret)
 730                 ret = count;
 731         return ret;
 732 }
 733
 734 static ssize_t cache_slow_downcall(const char __user *buf,
 735                                    size_t count, struct cache_detail *cd)
 736 {
 737         static char write_buf[8192]; /* protected by queue_io_mutex */
 738         ssize_t ret = -EINVAL;
 739
 740         if (count >= sizeof(write_buf))
 741                 goto out;
 742         mutex_lock(&queue_io_mutex);
 743         ret = cache_do_downcall(write_buf, buf, count, cd);
 744         mutex_unlock(&queue_io_mutex);
 745 out:
 746         return ret;
 747 }
 748
 749 static ssize_t cache_downcall(struct address_space *mapping,
 750                               const char __user *buf,
 751                               size_t count, struct cache_detail *cd)
 752 {
 753         struct page *page;
 754         char *kaddr;
 755         ssize_t ret = -ENOMEM;
 756
 757         if (count >= PAGE_CACHE_SIZE)
 758                 goto out_slow;
 759
 760         page = find_or_create_page(mapping, 0, GFP_KERNEL);
 761         if (!page)
 762                 goto out_slow;
 763
 764         kaddr = kmap(page);
 765         ret = cache_do_downcall(kaddr, buf, count, cd);
 766         kunmap(page);
 767         unlock_page(page);
 768         page_cache_release(page);
 769         return ret;
 770 out_slow:
 771         return cache_slow_downcall(buf, count, cd);
 772 }
 773
 774 static ssize_t cache_write(struct file *filp, const char __user *buf,
 775                            size_t count, loff_t *ppos,
 776                            struct cache_detail *cd)
 777 {
 778         struct address_space *mapping = filp->f_mapping;
 779         struct inode *inode = filp->f_path.dentry->d_inode;
 780         ssize_t ret = -EINVAL;
 781
 782         if (!cd->cache_parse)
 783                 goto out;
 784
 785         mutex_lock(&inode->i_mutex);
 786         ret = cache_downcall(mapping, buf, count, cd);
 787         mutex_unlock(&inode->i_mutex);
 788 out:
 789         return ret;
 790 }
 791
 792 static DECLARE_WAIT_QUEUE_HEAD(queue_wait);
 793
 794 static unsigned int cache_poll(struct file *filp, poll_table *wait,
 795                                struct cache_detail *cd)
 796 {
 797         unsigned int mask;
 798         struct cache_reader *rp = filp->private_data;
 799         struct cache_queue *cq;
 800
 801         poll_wait(filp, &queue_wait, wait);
 802
 803         /* alway allow write */
 804         mask = POLL_OUT | POLLWRNORM;
 805
 806         if (!rp)
 807                 return mask;
 808
 809         spin_lock(&queue_lock);
 810
 811         for (cq= &rp->q; &cq->list != &cd->queue;
 812              cq = list_entry(cq->list.next, struct cache_queue, list))
 813                 if (!cq->reader) {
 814                         mask |= POLLIN | POLLRDNORM;
 815                         break;
 816                 }
 817         spin_unlock(&queue_lock);
 818         return mask;
 819 }
 820
 821 static int cache_ioctl(struct inode *ino, struct file *filp,
 822                        unsigned int cmd, unsigned long arg,
 823                        struct cache_detail *cd)
 824 {
 825         int len = 0;
 826         struct cache_reader *rp = filp->private_data;
 827         struct cache_queue *cq;
 828
 829         if (cmd != FIONREAD || !rp)
 830                 return -EINVAL;
 831
 832         spin_lock(&queue_lock);
 833
 834         /* only find the length remaining in current request,
 835          * or the length of the next request
 836          */
 837         for (cq= &rp->q; &cq->list != &cd->queue;
 838              cq = list_entry(cq->list.next, struct cache_queue, list))
 839                 if (!cq->reader) {
 840                         struct cache_request *cr =
 841                                 container_of(cq, struct cache_request, q);
 842                         len = cr->len - rp->offset;
 843                         break;
 844                 }
 845         spin_unlock(&queue_lock);
 846
 847         return put_user(len, (int __user *)arg);
 848 }
 849
 850 static int cache_open(struct inode *inode, struct file *filp,
 851                       struct cache_detail *cd)
 852 {
 853         struct cache_reader *rp = NULL;
 854
 855         if (!cd || !try_module_get(cd->owner))
 856                 return -EACCES;
 857         nonseekable_open(inode, filp);
 858         if (filp->f_mode & FMODE_READ) {
 859                 rp = kmalloc(sizeof(*rp), GFP_KERNEL);
 860                 if (!rp)
 861                         return -ENOMEM;
 862                 rp->offset = 0;
 863                 rp->q.reader = 1;
 864                 atomic_inc(&cd->readers);
 865                 spin_lock(&queue_lock);
 866                 list_add(&rp->q.list, &cd->queue);
 867                 spin_unlock(&queue_lock);
 868         }
 869         filp->private_data = rp;
 870         return 0;
 871 }
 872
 873 static int cache_release(struct inode *inode, struct file *filp,
 874                          struct cache_detail *cd)
 875 {
 876         struct cache_reader *rp = filp->private_data;
 877
 878         if (rp) {
 879                 spin_lock(&queue_lock);
 880                 if (rp->offset) {
 881                         struct cache_queue *cq;
 882                         for (cq= &rp->q; &cq->list != &cd->queue;
 883                              cq = list_entry(cq->list.next, struct cache_queue, list))
 884                                 if (!cq->reader) {
 885                                         container_of(cq, struct cache_request, q)
 886                                                 ->readers--;
 887                                         break;
 888                                 }
 889                         rp->offset = 0;
 890                 }
 891                 list_del(&rp->q.list);
 892                 spin_unlock(&queue_lock);
 893
 894                 filp->private_data = NULL;
 895                 kfree(rp);
 896
 897                 cd->last_close = get_seconds();
 898                 atomic_dec(&cd->readers);
 899         }
 900         module_put(cd->owner);
 901         return 0;
 902 }
 903
 904
 905
 906 static void cache_dequeue(struct cache_detail *detail, struct cache_head *ch)
 907 {
 908         struct cache_queue *cq;
 909         spin_lock(&queue_lock);
 910         list_for_each_entry(cq, &detail->queue, list)
 911                 if (!cq->reader) {
 912                         struct cache_request *cr = container_of(cq, struct cache_request, q);
 913                         if (cr->item != ch)
 914                                 continue;
 915                         if (cr->readers != 0)
 916                                 continue;
 917                         list_del(&cr->q.list);
 918                         spin_unlock(&queue_lock);
 919                         cache_put(cr->item, detail);
 920                         kfree(cr->buf);
 921                         kfree(cr);
 922                         return;
 923                 }
 924         spin_unlock(&queue_lock);
 925 }
 926
 927 /*
 928  * Support routines for text-based upcalls.
 929  * Fields are separated by spaces.
 930  * Fields are either mangled to quote space tab newline slosh with slosh
 931  * or a hexified with a leading \x
 932  * Record is terminated with newline.
 933  *
 934  */
 935
 936 void qword_add(char **bpp, int *lp, char *str)
 937 {
 938         char *bp = *bpp;
 939         int len = *lp;
 940         char c;
 941
 942         if (len < 0) return;
 943
 944         while ((c=*str++) && len)
 945                 switch(c) {
 946                 case ' ':
 947                 case '\t':
 948                 case '\n':
 949                 case '\\':
 950                         if (len >= 4) {
 951                                 *bp++ = '\\';
 952                                 *bp++ = '0' + ((c & 0300)>>6);
 953                                 *bp++ = '0' + ((c & 0070)>>3);
 954                                 *bp++ = '0' + ((c & 0007)>>0);
 955                         }
 956                         len -= 4;
 957                         break;
 958                 default:
 959                         *bp++ = c;
 960                         len--;
 961                 }
 962         if (c || len <1) len = -1;
 963         else {
 964                 *bp++ = ' ';
 965                 len--;
 966         }
 967         *bpp = bp;
 968         *lp = len;
 969 }
 970 EXPORT_SYMBOL_GPL(qword_add);
 971
 972 void qword_addhex(char **bpp, int *lp, char *buf, int blen)
 973 {
 974         char *bp = *bpp;
 975         int len = *lp;
 976
 977         if (len < 0) return;
 978
 979         if (len > 2) {
 980                 *bp++ = '\\';
 981                 *bp++ = 'x';
 982                 len -= 2;
 983                 while (blen && len >= 2) {
 984                         unsigned char c = *buf++;
 985                         *bp++ = '0' + ((c&0xf0)>>4) + (c>=0xa0)*('a'-'9'-1);
 986                         *bp++ = '0' + (c&0x0f) + ((c&0x0f)>=0x0a)*('a'-'9'-1);
 987                         len -= 2;
 988                         blen--;
 989                 }
 990         }
 991         if (blen || len<1) len = -1;
 992         else {
 993                 *bp++ = ' ';
 994                 len--;
 995         }
 996         *bpp = bp;
 997         *lp = len;
 998 }
 999 EXPORT_SYMBOL_GPL(qword_addhex);
1000
1001 static void warn_no_listener(struct cache_detail *detail)
1002 {
1003         if (detail->last_warn != detail->last_close) {
1004                 detail->last_warn = detail->last_close;
1005                 if (detail->warn_no_listener)
1006                         detail->warn_no_listener(detail, detail->last_close != 0);
1007         }
1008 }
1009
1010 /*
1011  * register an upcall request to user-space and queue it up for read() by the
1012  * upcall daemon.
1013  *
1014  * Each request is at most one page long.
1015  */
1016 int sunrpc_cache_pipe_upcall(struct cache_detail *detail, struct cache_head *h,
1017                 void (*cache_request)(struct cache_detail *,
1018                                       struct cache_head *,
1019                                       char **,
1020                                       int *))
1021 {
1022
1023         char *buf;
1024         struct cache_request *crq;
1025         char *bp;
1026         int len;
1027
1028         if (atomic_read(&detail->readers) == 0 &&
1029             detail->last_close < get_seconds() - 30) {
1030                         warn_no_listener(detail);
1031                         return -EINVAL;
1032         }
1033
1034         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
1035         if (!buf)
1036                 return -EAGAIN;
1037
1038         crq = kmalloc(sizeof (*crq), GFP_KERNEL);
1039         if (!crq) {
1040                 kfree(buf);
1041                 return -EAGAIN;
1042         }
1043
1044         bp = buf; len = PAGE_SIZE;
1045
1046         cache_request(detail, h, &bp, &len);
1047
1048         if (len < 0) {
1049                 kfree(buf);
1050                 kfree(crq);
1051                 return -EAGAIN;
1052         }
1053         crq->q.reader = 0;
1054         crq->item = cache_get(h);
1055         crq->buf = buf;
1056         crq->len = PAGE_SIZE - len;
1057         crq->readers = 0;
1058         spin_lock(&queue_lock);
1059         list_add_tail(&crq->q.list, &detail->queue);
1060         spin_unlock(&queue_lock);
1061         wake_up(&queue_wait);
1062         return 0;
1063 }
1064 EXPORT_SYMBOL_GPL(sunrpc_cache_pipe_upcall);
1065
1066 /*
1067  * parse a message from user-space and pass it
1068  * to an appropriate cache
1069  * Messages are, like requests, separated into fields by
1070  * spaces and dequotes as \xHEXSTRING or embedded \nnn octal
1071  *
1072  * Message is
1073  *   reply cachename expiry key ... content....
1074  *
1075  * key and content are both parsed by cache
1076  */
1077
1078 #define isodigit(c) (isdigit(c) && c <= '7')
1079 int qword_get(char **bpp, char *dest, int bufsize)
1080 {
1081         /* return bytes copied, or -1 on error */
1082         char *bp = *bpp;
1083         int len = 0;
1084
1085         while (*bp == ' ') bp++;
1086
1087         if (bp[0] == '\\' && bp[1] == 'x') {
1088                 /* HEX STRING */
1089                 bp += 2;
1090                 while (isxdigit(bp[0]) && isxdigit(bp[1]) && len < bufsize) {
1091                         int byte = isdigit(*bp) ? *bp-'0' : toupper(*bp)-'A'+10;
1092                         bp++;
1093                         byte <<= 4;
1094                         byte |= isdigit(*bp) ? *bp-'0' : toupper(*bp)-'A'+10;
1095                         *dest++ = byte;
1096                         bp++;
1097                         len++;
1098                 }
1099         } else {
1100                 /* text with \nnn octal quoting */
1101                 while (*bp != ' ' && *bp != '\n' && *bp && len < bufsize-1) {
1102                         if (*bp == '\\' &&
1103                             isodigit(bp[1]) && (bp[1] <= '3') &&
1104                             isodigit(bp[2]) &&
1105                             isodigit(bp[3])) {
1106                                 int byte = (*++bp -'0');
1107                                 bp++;
1108                                 byte = (byte << 3) | (*bp++ - '0');
1109                                 byte = (byte << 3) | (*bp++ - '0');
1110                                 *dest++ = byte;
1111                                 len++;
1112                         } else {
1113                                 *dest++ = *bp++;
1114                                 len++;
1115                         }
1116                 }
1117         }
1118
1119         if (*bp != ' ' && *bp != '\n' && *bp != '\0')
1120                 return -1;
1121         while (*bp == ' ') bp++;
1122         *bpp = bp;
1123         *dest = '\0';
1124         return len;
1125 }
1126 EXPORT_SYMBOL_GPL(qword_get);
1127
1128
1129 /*
1130  * support /proc/sunrpc/cache/$CACHENAME/content
1131  * as a seqfile.
1132  * We call ->cache_show passing NULL for the item to
1133  * get a header, then pass each real item in the cache
1134  */
1135
1136 struct handle {
1137         struct cache_detail *cd;
1138 };
1139
1140 static void *c_start(struct seq_file *m, loff_t *pos)
1141         __acquires(cd->hash_lock)
1142 {
1143         loff_t n = *pos;
1144         unsigned hash, entry;
1145         struct cache_head *ch;
1146         struct cache_detail *cd = ((struct handle*)m->private)->cd;
1147
1148
1149         read_lock(&cd->hash_lock);
1150         if (!n--)
1151                 return SEQ_START_TOKEN;
1152         hash = n >> 32;
1153         entry = n & ((1LL<<32) - 1);
1154
1155         for (ch=cd->hash_table[hash]; ch; ch=ch->next)
1156                 if (!entry--)
1157                         return ch;
1158         n &= ~((1LL<<32) - 1);
1159         do {
1160                 hash++;
1161                 n += 1LL<<32;
1162         } while(hash < cd->hash_size &&
1163                 cd->hash_table[hash]==NULL);
1164         if (hash >= cd->hash_size)
1165                 return NULL;
1166         *pos = n+1;
1167         return cd->hash_table[hash];
1168 }
1169
1170 static void *c_next(struct seq_file *m, void *p, loff_t *pos)
1171 {
1172         struct cache_head *ch = p;
1173         int hash = (*pos >> 32);
1174         struct cache_detail *cd = ((struct handle*)m->private)->cd;
1175
1176         if (p == SEQ_START_TOKEN)
1177                 hash = 0;
1178         else if (ch->next == NULL) {
1179                 hash++;
1180                 *pos += 1LL<<32;
1181         } else {
1182                 ++*pos;
1183                 return ch->next;
1184         }
1185         *pos &= ~((1LL<<32) - 1);
1186         while (hash < cd->hash_size &&
1187                cd->hash_table[hash] == NULL) {
1188                 hash++;
1189                 *pos += 1LL<<32;
1190         }
1191         if (hash >= cd->hash_size)
1192                 return NULL;
1193         ++*pos;
1194         return cd->hash_table[hash];
1195 }
1196
1197 static void c_stop(struct seq_file *m, void *p)
1198         __releases(cd->hash_lock)
1199 {
1200         struct cache_detail *cd = ((struct handle*)m->private)->cd;
1201         read_unlock(&cd->hash_lock);
1202 }
1203
1204 static int c_show(struct seq_file *m, void *p)
1205 {
1206         struct cache_head *cp = p;
1207         struct cache_detail *cd = ((struct handle*)m->private)->cd;
1208
1209         if (p == SEQ_START_TOKEN)
1210                 return cd->cache_show(m, cd, NULL);
1211
1212         ifdebug(CACHE)
1213                 seq_printf(m, "# expiry=%ld refcnt=%d flags=%lx\n",
1214                            cp->expiry_time, atomic_read(&cp->ref.refcount), cp->flags);
1215         cache_get(cp);
1216         if (cache_check(cd, cp, NULL))
1217                 /* cache_check does a cache_put on failure */
1218                 seq_printf(m, "# ");
1219         else
1220                 cache_put(cp, cd);
1221
1222         return cd->cache_show(m, cd, cp);
1223 }
1224
1225 static const struct seq_operations cache_content_op = {
1226         .start  = c_start,
1227         .next   = c_next,
1228         .stop   = c_stop,
1229         .show   = c_show,
1230 };
1231
1232 static int content_open(struct inode *inode, struct file *file,
1233                         struct cache_detail *cd)
1234 {
1235         struct handle *han;
1236
1237         if (!cd || !try_module_get(cd->owner))
1238                 return -EACCES;
1239         han = __seq_open_private(file, &cache_content_op, sizeof(*han));
1240         if (han == NULL)
1241                 return -ENOMEM;
1242
1243         han->cd = cd;
1244         return 0;
1245 }
1246
1247 static int content_release(struct inode *inode, struct file *file,
1248                 struct cache_detail *cd)
1249 {
1250         int ret = seq_release_private(inode, file);
1251         module_put(cd->owner);
1252         return ret;
1253 }
1254
1255 static int open_flush(struct inode *inode, struct file *file,
1256                         struct cache_detail *cd)
1257 {
1258         if (!cd || !try_module_get(cd->owner))
1259                 return -EACCES;
1260         return nonseekable_open(inode, file);
1261 }
1262
1263 static int release_flush(struct inode *inode, struct file *file,
1264                         struct cache_detail *cd)
1265 {
1266         module_put(cd->owner);
1267         return 0;
1268 }
1269
1270 static ssize_t read_flush(struct file *file, char __user *buf,
1271                           size_t count, loff_t *ppos,
1272                           struct cache_detail *cd)
1273 {
1274         char tbuf[20];
1275         unsigned long p = *ppos;
1276         size_t len;
1277
1278         sprintf(tbuf, "%lu\n", cd->flush_time);
1279         len = strlen(tbuf);
1280         if (p >= len)
1281                 return 0;
1282         len -= p;
1283         if (len > count)
1284                 len = count;
1285         if (copy_to_user(buf, (void*)(tbuf+p), len))
1286                 return -EFAULT;
1287         *ppos += len;
1288         return len;
1289 }
1290
1291 static ssize_t write_flush(struct file *file, const char __user *buf,
1292                            size_t count, loff_t *ppos,
1293                            struct cache_detail *cd)
1294 {
1295         char tbuf[20];
1296         char *ep;
1297         long flushtime;
1298         if (*ppos || count > sizeof(tbuf)-1)
1299                 return -EINVAL;
1300         if (copy_from_user(tbuf, buf, count))
1301                 return -EFAULT;
1302         tbuf[count] = 0;
1303         flushtime = simple_strtoul(tbuf, &ep, 0);
1304         if (*ep && *ep != '\n')
1305                 return -EINVAL;
1306
1307         cd->flush_time = flushtime;
1308         cd->nextcheck = get_seconds();
1309         cache_flush();
1310
1311         *ppos += count;
1312         return count;
1313 }
1314
1315 static ssize_t cache_read_procfs(struct file *filp, char __user *buf,
1316                                  size_t count, loff_t *ppos)
1317 {
1318         struct cache_detail *cd = PDE(filp->f_path.dentry->d_inode)->data;
1319
1320         return cache_read(filp, buf, count, ppos, cd);
1321 }
1322
1323 static ssize_t cache_write_procfs(struct file *filp, const char __user *buf,
1324                                   size_t count, loff_t *ppos)
1325 {
1326         struct cache_detail *cd = PDE(filp->f_path.dentry->d_inode)->data;
1327
1328         return cache_write(filp, buf, count, ppos, cd);
1329 }
1330
1331 static unsigned int cache_poll_procfs(struct file *filp, poll_table *wait)
1332 {
1333         struct cache_detail *cd = PDE(filp->f_path.dentry->d_inode)->data;
1334
1335         return cache_poll(filp, wait, cd);
1336 }
1337
1338 static int cache_ioctl_procfs(struct inode *inode, struct file *filp,
1339                               unsigned int cmd, unsigned long arg)
1340 {
1341         struct cache_detail *cd = PDE(inode)->data;
1342
1343         return cache_ioctl(inode, filp, cmd, arg, cd);
1344 }
1345
1346 static int cache_open_procfs(struct inode *inode, struct file *filp)
1347 {
1348         struct cache_detail *cd = PDE(inode)->data;
1349
1350         return cache_open(inode, filp, cd);
1351 }
1352
1353 static int cache_release_procfs(struct inode *inode, struct file *filp)
1354 {
1355         struct cache_detail *cd = PDE(inode)->data;
1356
1357         return cache_release(inode, filp, cd);
1358 }
1359
1360 static const struct file_operations cache_file_operations_procfs = {
1361         .owner          = THIS_MODULE,
1362         .llseek         = no_llseek,
1363         .read           = cache_read_procfs,
1364         .write          = cache_write_procfs,
1365         .poll           = cache_poll_procfs,
1366         .ioctl          = cache_ioctl_procfs, /* for FIONREAD */
1367         .open           = cache_open_procfs,
1368         .release        = cache_release_procfs,
1369 };
1370
1371 static int content_open_procfs(struct inode *inode, struct file *filp)
1372 {
1373         struct cache_detail *cd = PDE(inode)->data;
1374
1375         return content_open(inode, filp, cd);
1376 }
1377
1378 static int content_release_procfs(struct inode *inode, struct file *filp)
1379 {
1380         struct cache_detail *cd = PDE(inode)->data;
1381
1382         return content_release(inode, filp, cd);
1383 }
1384
1385 static const struct file_operations content_file_operations_procfs = {
1386         .open           = content_open_procfs,
1387         .read           = seq_read,
1388         .llseek         = seq_lseek,
1389         .release        = content_release_procfs,
1390 };
1391
1392 static int open_flush_procfs(struct inode *inode, struct file *filp)
1393 {
1394         struct cache_detail *cd = PDE(inode)->data;
1395
1396         return open_flush(inode, filp, cd);
1397 }
1398
1399 static int release_flush_procfs(struct inode *inode, struct file *filp)
1400 {
1401         struct cache_detail *cd = PDE(inode)->data;
1402
1403         return release_flush(inode, filp, cd);
1404 }
1405
1406 static ssize_t read_flush_procfs(struct file *filp, char __user *buf,
1407                             size_t count, loff_t *ppos)
1408 {
1409         struct cache_detail *cd = PDE(filp->f_path.dentry->d_inode)->data;
1410
1411         return read_flush(filp, buf, count, ppos, cd);
1412 }
1413
1414 static ssize_t write_flush_procfs(struct file *filp,
1415                                   const char __user *buf,
1416                                   size_t count, loff_t *ppos)
1417 {
1418         struct cache_detail *cd = PDE(filp->f_path.dentry->d_inode)->data;
1419
1420         return write_flush(filp, buf, count, ppos, cd);
1421 }
1422
1423 static const struct file_operations cache_flush_operations_procfs = {
1424         .open           = open_flush_procfs,
1425         .read           = read_flush_procfs,
1426         .write          = write_flush_procfs,
1427         .release        = release_flush_procfs,
1428 };
1429
1430 static void remove_cache_proc_entries(struct cache_detail *cd)
1431 {
1432         if (cd->u.procfs.proc_ent == NULL)
1433                 return;
1434         if (cd->u.procfs.flush_ent)
1435                 remove_proc_entry("flush", cd->u.procfs.proc_ent);
1436         if (cd->u.procfs.channel_ent)
1437                 remove_proc_entry("channel", cd->u.procfs.proc_ent);
1438         if (cd->u.procfs.content_ent)
1439                 remove_proc_entry("content", cd->u.procfs.proc_ent);
1440         cd->u.procfs.proc_ent = NULL;
1441         remove_proc_entry(cd->name, proc_net_rpc);
1442 }
1443
1444 #ifdef CONFIG_PROC_FS
1445 static int create_cache_proc_entries(struct cache_detail *cd)
1446 {
1447         struct proc_dir_entry *p;
1448
1449         cd->u.procfs.proc_ent = proc_mkdir(cd->name, proc_net_rpc);
1450         if (cd->u.procfs.proc_ent == NULL)
1451                 goto out_nomem;
1452         cd->u.procfs.channel_ent = NULL;
1453         cd->u.procfs.content_ent = NULL;
1454
1455         p = proc_create_data("flush", S_IFREG|S_IRUSR|S_IWUSR,
1456                              cd->u.procfs.proc_ent,
1457                              &cache_flush_operations_procfs, cd);
1458         cd->u.procfs.flush_ent = p;
1459         if (p == NULL)
1460                 goto out_nomem;
1461
1462         if (cd->cache_upcall || cd->cache_parse) {
1463                 p = proc_create_data("channel", S_IFREG|S_IRUSR|S_IWUSR,
1464                                      cd->u.procfs.proc_ent,
1465                                      &cache_file_operations_procfs, cd);
1466                 cd->u.procfs.channel_ent = p;
1467                 if (p == NULL)
1468                         goto out_nomem;
1469         }
1470         if (cd->cache_show) {
1471                 p = proc_create_data("content", S_IFREG|S_IRUSR|S_IWUSR,
1472                                 cd->u.procfs.proc_ent,
1473                                 &content_file_operations_procfs, cd);
1474                 cd->u.procfs.content_ent = p;
1475                 if (p == NULL)
1476                         goto out_nomem;
1477         }
1478         return 0;
1479 out_nomem:
1480         remove_cache_proc_entries(cd);
1481         return -ENOMEM;
1482 }
1483 #else /* CONFIG_PROC_FS */
1484 static int create_cache_proc_entries(struct cache_detail *cd)
1485 {
1486         return 0;
1487 }
1488 #endif
1489
1490 int cache_register(struct cache_detail *cd)
1491 {
1492         int ret;
1493
1494         sunrpc_init_cache_detail(cd);
1495         ret = create_cache_proc_entries(cd);
1496         if (ret)
1497                 sunrpc_destroy_cache_detail(cd);
1498         return ret;
1499 }
1500 EXPORT_SYMBOL_GPL(cache_register);
1501
1502 void cache_unregister(struct cache_detail *cd)
1503 {
1504         remove_cache_proc_entries(cd);
1505         sunrpc_destroy_cache_detail(cd);
1506 }
1507 EXPORT_SYMBOL_GPL(cache_unregister);
1508
1509 static ssize_t cache_read_pipefs(struct file *filp, char __user *buf,
1510                                  size_t count, loff_t *ppos)
1511 {
1512         struct cache_detail *cd = RPC_I(filp->f_path.dentry->d_inode)->private;
1513
1514         return cache_read(filp, buf, count, ppos, cd);
1515 }
1516
1517 static ssize_t cache_write_pipefs(struct file *filp, const char __user *buf,
1518                                   size_t count, loff_t *ppos)
1519 {
1520         struct cache_detail *cd = RPC_I(filp->f_path.dentry->d_inode)->private;
1521
1522         return cache_write(filp, buf, count, ppos, cd);
1523 }
1524
1525 static unsigned int cache_poll_pipefs(struct file *filp, poll_table *wait)
1526 {
1527         struct cache_detail *cd = RPC_I(filp->f_path.dentry->d_inode)->private;
1528
1529         return cache_poll(filp, wait, cd);
1530 }
1531
1532 static int cache_ioctl_pipefs(struct inode *inode, struct file *filp,
1533                               unsigned int cmd, unsigned long arg)
1534 {
1535         struct cache_detail *cd = RPC_I(inode)->private;
1536
1537         return cache_ioctl(inode, filp, cmd, arg, cd);
1538 }
1539
1540 static int cache_open_pipefs(struct inode *inode, struct file *filp)
1541 {
1542         struct cache_detail *cd = RPC_I(inode)->private;
1543
1544         return cache_open(inode, filp, cd);
1545 }
1546
1547 static int cache_release_pipefs(struct inode *inode, struct file *filp)
1548 {
1549         struct cache_detail *cd = RPC_I(inode)->private;
1550
1551         return cache_release(inode, filp, cd);
1552 }
1553
1554 const struct file_operations cache_file_operations_pipefs = {
1555         .owner          = THIS_MODULE,
1556         .llseek         = no_llseek,
1557         .read           = cache_read_pipefs,
1558         .write          = cache_write_pipefs,
1559         .poll           = cache_poll_pipefs,
1560         .ioctl          = cache_ioctl_pipefs, /* for FIONREAD */
1561         .open           = cache_open_pipefs,
1562         .release        = cache_release_pipefs,
1563 };
1564
1565 static int content_open_pipefs(struct inode *inode, struct file *filp)
1566 {
1567         struct cache_detail *cd = RPC_I(inode)->private;
1568
1569         return content_open(inode, filp, cd);
1570 }
1571
1572 static int content_release_pipefs(struct inode *inode, struct file *filp)
1573 {
1574         struct cache_detail *cd = RPC_I(inode)->private;
1575
1576         return content_release(inode, filp, cd);
1577 }
1578
1579 const struct file_operations content_file_operations_pipefs = {
1580         .open           = content_open_pipefs,
1581         .read           = seq_read,
1582         .llseek         = seq_lseek,
1583         .release        = content_release_pipefs,
1584 };
1585
1586 static int open_flush_pipefs(struct inode *inode, struct file *filp)
1587 {
1588         struct cache_detail *cd = RPC_I(inode)->private;
1589
1590         return open_flush(inode, filp, cd);
1591 }
1592
1593 static int release_flush_pipefs(struct inode *inode, struct file *filp)
1594 {
1595         struct cache_detail *cd = RPC_I(inode)->private;
1596
1597         return release_flush(inode, filp, cd);
1598 }
1599
1600 static ssize_t read_flush_pipefs(struct file *filp, char __user *buf,
1601                             size_t count, loff_t *ppos)
1602 {
1603         struct cache_detail *cd = RPC_I(filp->f_path.dentry->d_inode)->private;
1604
1605         return read_flush(filp, buf, count, ppos, cd);
1606 }
1607
1608 static ssize_t write_flush_pipefs(struct file *filp,
1609                                   const char __user *buf,
1610                                   size_t count, loff_t *ppos)
1611 {
1612         struct cache_detail *cd = RPC_I(filp->f_path.dentry->d_inode)->private;
1613
1614         return write_flush(filp, buf, count, ppos, cd);
1615 }
1616
1617 const struct file_operations cache_flush_operations_pipefs = {
1618         .open           = open_flush_pipefs,
1619         .read           = read_flush_pipefs,
1620         .write          = write_flush_pipefs,
1621         .release        = release_flush_pipefs,
1622 };
1623
1624 int sunrpc_cache_register_pipefs(struct dentry *parent,
1625                                  const char *name, mode_t umode,
1626                                  struct cache_detail *cd)
1627 {
1628         struct qstr q;
1629         struct dentry *dir;
1630         int ret = 0;
1631
1632         sunrpc_init_cache_detail(cd);
1633         q.name = name;
1634         q.len = strlen(name);
1635         q.hash = full_name_hash(q.name, q.len);
1636         dir = rpc_create_cache_dir(parent, &q, umode, cd);
1637         if (!IS_ERR(dir))
1638                 cd->u.pipefs.dir = dir;
1639         else {
1640                 sunrpc_destroy_cache_detail(cd);
1641                 ret = PTR_ERR(dir);
1642         }
1643         return ret;
1644 }
1645 EXPORT_SYMBOL_GPL(sunrpc_cache_register_pipefs);
1646
1647 void sunrpc_cache_unregister_pipefs(struct cache_detail *cd)
1648 {
1649         rpc_remove_cache_dir(cd->u.pipefs.dir);
1650         cd->u.pipefs.dir = NULL;
1651         sunrpc_destroy_cache_detail(cd);
1652 }
1653 EXPORT_SYMBOL_GPL(sunrpc_cache_unregister_pipefs);
1654