lustre/obdfilter/filter_io.c

   1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
   2  * vim:expandtab:shiftwidth=8:tabstop=8:
   3  *
   4  *  linux/fs/obdfilter/filter_io.c
   5  *
   6  *  Copyright (c) 2001-2003 Cluster File Systems, Inc.
   7  *   Author: Peter Braam <braam@clusterfs.com>
   8  *   Author: Andreas Dilger <adilger@clusterfs.com>
   9  *   Author: Phil Schwan <phil@clusterfs.com>
  10  *
  11  *   This file is part of Lustre, http://www.lustre.org.
  12  *
  13  *   Lustre is free software; you can redistribute it and/or
  14  *   modify it under the terms of version 2 of the GNU General Public
  15  *   License as published by the Free Software Foundation.
  16  *
  17  *   Lustre is distributed in the hope that it will be useful,
  18  *   but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  20  *   GNU General Public License for more details.
  21  *
  22  *   You should have received a copy of the GNU General Public License
  23  *   along with Lustre; if not, write to the Free Software
  24  *   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  25  */
  26
  27 #define DEBUG_SUBSYSTEM S_FILTER
  28
  29 #include <linux/config.h>
  30 #include <linux/module.h>
  31 #include <linux/pagemap.h> // XXX kill me soon
  32 #include <linux/version.h>
  33
  34 #include <linux/obd_class.h>
  35 #include <linux/lustre_fsfilt.h>
  36 #include "filter_internal.h"
  37
  38 static int filter_start_page_read(struct inode *inode, struct niobuf_local *lnb)
  39 {
  40         struct address_space *mapping = inode->i_mapping;
  41         struct page *page;
  42         unsigned long index = lnb->offset >> PAGE_SHIFT;
  43         int rc;
  44
  45         page = grab_cache_page(mapping, index); /* locked page */
  46         if (page == NULL)
  47                 return lnb->rc = -ENOMEM;
  48
  49         LASSERT(page->mapping == mapping);
  50
  51         lnb->page = page;
  52
  53         if (inode->i_size < lnb->offset + lnb->len - 1)
  54                 lnb->rc = inode->i_size - lnb->offset;
  55         else
  56                 lnb->rc = lnb->len;
  57
  58         if (PageUptodate(page)) {
  59                 unlock_page(page);
  60                 return 0;
  61         }
  62
  63         rc = mapping->a_ops->readpage(NULL, page);
  64         if (rc < 0) {
  65                 CERROR("page index %lu, rc = %d\n", index, rc);
  66                 lnb->page = NULL;
  67                 page_cache_release(page);
  68                 return lnb->rc = rc;
  69         }
  70
  71         return 0;
  72 }
  73
  74 static int filter_finish_page_read(struct niobuf_local *lnb)
  75 {
  76         if (lnb->page == NULL)
  77                 return 0;
  78
  79         if (PageUptodate(lnb->page))
  80                 return 0;
  81
  82         wait_on_page(lnb->page);
  83         if (!PageUptodate(lnb->page)) {
  84                 CERROR("page index %lu/offset "LPX64" not uptodate\n",
  85                        lnb->page->index, lnb->offset);
  86                 GOTO(err_page, lnb->rc = -EIO);
  87         }
  88         if (PageError(lnb->page)) {
  89                 CERROR("page index %lu/offset "LPX64" has error\n",
  90                        lnb->page->index, lnb->offset);
  91                 GOTO(err_page, lnb->rc = -EIO);
  92         }
  93
  94         return 0;
  95
  96 err_page:
  97         page_cache_release(lnb->page);
  98         lnb->page = NULL;
  99         return lnb->rc;
 100 }
 101
 102 /* Grab the dirty and seen grant announcements from the incoming obdo.
 103  * We will later calculate the clients new grant and return it.
 104  * Caller must hold osfs lock */
 105 static void filter_grant_incoming(struct obd_export *exp, struct obdo *oa)
 106 {
 107         struct filter_export_data *fed;
 108         struct obd_device *obd = exp->exp_obd;
 109         ENTRY;
 110
 111         LASSERT_SPIN_LOCKED(&obd->obd_osfs_lock);
 112
 113         if ((oa->o_valid & (OBD_MD_FLBLOCKS|OBD_MD_FLGRANT)) !=
 114                                         (OBD_MD_FLBLOCKS|OBD_MD_FLGRANT)) {
 115                 oa->o_valid &= ~OBD_MD_FLGRANT;
 116                 EXIT;
 117                 return;
 118         }
 119
 120         fed = &exp->exp_filter_data;
 121
 122         /* Add some margin, since there is a small race if other RPCs arrive
 123          * out-or-order and have already consumed some grant.  We want to
 124          * leave this here in case there is a large error in accounting. */
 125         CDEBUG(oa->o_grant > fed->fed_grant + FILTER_GRANT_CHUNK ?
 126                D_WARNING : D_CACHE,
 127                "%s: cli %s/%p reports grant: "LPU64" dropped: %u, local: %lu\n",
 128                obd->obd_name, exp->exp_client_uuid.uuid, exp, oa->o_grant,
 129                oa->o_dropped, fed->fed_grant);
 130
 131         /* Update our accounting now so that statfs takes it into account.
 132          * Note that fed_dirty is only approximate and can become incorrect
 133          * if RPCs arrive out-of-order.  No important calculations depend
 134          * on fed_dirty however. */
 135         obd->u.filter.fo_tot_dirty += oa->o_dirty - fed->fed_dirty;
 136         if (fed->fed_grant < oa->o_dropped) {
 137                 CERROR("%s: cli %s/%p reports %u dropped > fed_grant %lu\n",
 138                        obd->obd_name, exp->exp_client_uuid.uuid, exp,
 139                        oa->o_dropped, fed->fed_grant);
 140                 oa->o_dropped = 0;
 141         }
 142         if (obd->u.filter.fo_tot_granted < oa->o_dropped) {
 143                 CERROR("%s: cli %s/%p reports %u dropped > tot_grant "LPU64"\n",
 144                        obd->obd_name, exp->exp_client_uuid.uuid, exp,
 145                        oa->o_dropped, obd->u.filter.fo_tot_granted);
 146                 oa->o_dropped = 0;
 147         }
 148         obd->u.filter.fo_tot_granted -= oa->o_dropped;
 149         fed->fed_grant -= oa->o_dropped;
 150         fed->fed_dirty = oa->o_dirty;
 151         EXIT;
 152 }
 153
 154 #define GRANT_FOR_LLOG(obd) 16
 155
 156 /* Figure out how much space is available between what we've granted
 157  * and what remains in the filesystem.  Compensate for ext3 indirect
 158  * block overhead when computing how much free space is left ungranted.
 159  *
 160  * Caller must hold obd_osfs_lock. */
 161 obd_size filter_grant_space_left(struct obd_export *exp)
 162 {
 163         struct obd_device *obd = exp->exp_obd;
 164         int blockbits = obd->u.filter.fo_sb->s_blocksize_bits;
 165         obd_size tot_granted = obd->u.filter.fo_tot_granted, avail, left = 0;
 166         int rc, statfs_done = 0;
 167
 168         LASSERT_SPIN_LOCKED(&obd->obd_osfs_lock);
 169
 170         if (time_before(obd->obd_osfs_age, jiffies - HZ)) {
 171 restat:
 172                 rc = fsfilt_statfs(obd, obd->u.filter.fo_sb, jiffies + 1);
 173                 if (rc) /* N.B. statfs can't really fail */
 174                         RETURN(0);
 175                 statfs_done = 1;
 176         }
 177
 178         avail = obd->obd_osfs.os_bavail;
 179         left = avail - (avail >> (blockbits - 3)); /* (d)indirect */
 180         if (left > GRANT_FOR_LLOG(obd)) {
 181                 left = (left - GRANT_FOR_LLOG(obd)) << blockbits;
 182         } else {
 183                 left = 0 /* << blockbits */;
 184         }
 185
 186         if (!statfs_done && left < 32 * FILTER_GRANT_CHUNK + tot_granted) {
 187                 CDEBUG(D_CACHE, "fs has no space left and statfs too old\n");
 188                 goto restat;
 189         }
 190
 191         if (left >= tot_granted) {
 192                 left -= tot_granted;
 193         } else {
 194                 static unsigned long next;
 195                 if (left < tot_granted - obd->u.filter.fo_tot_pending &&
 196                     time_after(jiffies, next)) {
 197                         spin_unlock(&obd->obd_osfs_lock);
 198                         CERROR("%s: cli %s/%p grant "LPU64" > available "
 199                                LPU64" and pending "LPU64"\n", obd->obd_name,
 200                                exp->exp_client_uuid.uuid, exp, tot_granted,
 201                                left, obd->u.filter.fo_tot_pending);
 202                         if (next == 0)
 203                                 portals_debug_dumplog();
 204                         next = jiffies + 20 * HZ;
 205                         spin_lock(&obd->obd_osfs_lock);
 206                 }
 207                 left = 0;
 208         }
 209
 210         CDEBUG(D_CACHE, "%s: cli %s/%p free: "LPU64" avail: "LPU64" grant "LPU64
 211                " left: "LPU64" pending: "LPU64"\n", obd->obd_name,
 212                exp->exp_client_uuid.uuid, exp,
 213                obd->obd_osfs.os_bfree << blockbits, avail << blockbits,
 214                tot_granted, left, obd->u.filter.fo_tot_pending);
 215
 216         return left;
 217 }
 218
 219 /* Calculate how much grant space to allocate to this client, based on how
 220  * much space is currently free and how much of that is already granted.
 221  *
 222  * Caller must hold obd_osfs_lock. */
 223 long filter_grant(struct obd_export *exp, obd_size current_grant,
 224                   obd_size want, obd_size fs_space_left)
 225 {
 226         struct obd_device *obd = exp->exp_obd;
 227         struct filter_export_data *fed = &exp->exp_filter_data;
 228         int blockbits = obd->u.filter.fo_sb->s_blocksize_bits;
 229         __u64 grant = 0;
 230
 231         LASSERT_SPIN_LOCKED(&obd->obd_osfs_lock);
 232
 233         /* Grant some fraction of the client's requested grant space so that
 234          * they are not always waiting for write credits (not all of it to
 235          * avoid overgranting in face of multiple RPCs in flight).  This
 236          * essentially will be able to control the OSC_MAX_RIF for a client.
 237          *
 238          * If we do have a large disparity between what the client thinks it
 239          * has and what we think it has, don't grant very much and let the
 240          * client consume its grant first.  Either it just has lots of RPCs
 241          * in flight, or it was evicted and its grants will soon be used up. */
 242         if (current_grant < want &&
 243             current_grant < fed->fed_grant + FILTER_GRANT_CHUNK) {
 244                 grant = min((want >> blockbits) / 2,
 245                             (fs_space_left >> blockbits) / 8);
 246                 grant <<= blockbits;
 247
 248                 if (grant) {
 249                         if (grant > FILTER_GRANT_CHUNK)
 250                                 grant = FILTER_GRANT_CHUNK;
 251
 252                         obd->u.filter.fo_tot_granted += grant;
 253                         fed->fed_grant += grant;
 254                 }
 255         }
 256
 257         CDEBUG(D_CACHE,"%s: cli %s/%p wants: "LPU64" granting: "LPU64"\n",
 258                obd->obd_name, exp->exp_client_uuid.uuid, exp, want, grant);
 259         CDEBUG(D_CACHE,
 260                "%s: cli %s/%p tot cached:"LPU64" granted:"LPU64
 261                " num_exports: %d\n", obd->obd_name, exp->exp_client_uuid.uuid,
 262                exp, obd->u.filter.fo_tot_dirty,
 263                obd->u.filter.fo_tot_granted, obd->obd_num_exports);
 264
 265         return grant;
 266 }
 267
 268 static int filter_preprw_read(int cmd, struct obd_export *exp, struct obdo *oa,
 269                               int objcount, struct obd_ioobj *obj,
 270                               int niocount, struct niobuf_remote *nb,
 271                               struct niobuf_local *res,
 272                               struct obd_trans_info *oti)
 273 {
 274         struct obd_device *obd = exp->exp_obd;
 275         struct obd_run_ctxt saved;
 276         struct obd_ioobj *o;
 277         struct niobuf_remote *rnb;
 278         struct niobuf_local *lnb = NULL;
 279         struct fsfilt_objinfo *fso;
 280         struct dentry *dentry;
 281         struct inode *inode;
 282         int rc = 0, i, j, tot_bytes = 0, cleanup_phase = 0;
 283         unsigned long now = jiffies;
 284         ENTRY;
 285
 286         /* We are currently not supporting multi-obj BRW_READ RPCS at all.
 287          * When we do this function's dentry cleanup will need to be fixed */
 288         LASSERT(objcount == 1);
 289         LASSERT(obj->ioo_bufcnt > 0);
 290
 291         if (oa && oa->o_valid & OBD_MD_FLGRANT) {
 292                 spin_lock(&obd->obd_osfs_lock);
 293                 filter_grant_incoming(exp, oa);
 294
 295 #if 0
 296                 /* Reads do not increase grants */
 297                 oa->o_grant = filter_grant(exp, oa->o_grant, oa->o_undirty,
 298                                            filter_grant_space_left(exp));
 299 #else
 300                 oa->o_grant = 0;
 301 #endif
 302                 spin_unlock(&obd->obd_osfs_lock);
 303         }
 304
 305         OBD_ALLOC(fso, objcount * sizeof(*fso));
 306         if (fso == NULL)
 307                 RETURN(-ENOMEM);
 308
 309         memset(res, 0, niocount * sizeof(*res));
 310
 311         push_ctxt(&saved, &exp->exp_obd->obd_ctxt, NULL);
 312         for (i = 0, o = obj; i < objcount; i++, o++) {
 313                 LASSERT(o->ioo_bufcnt);
 314
 315                 dentry = filter_oa2dentry(obd, oa);
 316                 if (IS_ERR(dentry))
 317                         GOTO(cleanup, rc = PTR_ERR(dentry));
 318
 319                 if (dentry->d_inode == NULL) {
 320                         CERROR("trying to BRW to non-existent file "LPU64"\n",
 321                                o->ioo_id);
 322                         f_dput(dentry);
 323                         GOTO(cleanup, rc = -ENOENT);
 324                 }
 325
 326                 fso[i].fso_dentry = dentry;
 327                 fso[i].fso_bufcnt = o->ioo_bufcnt;
 328         }
 329
 330         if (time_after(jiffies, now + 15 * HZ))
 331                 CERROR("slow preprw_read setup %lus\n", (jiffies - now) / HZ);
 332         else
 333                 CDEBUG(D_INFO, "preprw_read setup: %lu jiffies\n",
 334                        (jiffies - now));
 335
 336         for (i = 0, o = obj, rnb = nb, lnb = res; i < objcount; i++, o++) {
 337                 dentry = fso[i].fso_dentry;
 338                 inode = dentry->d_inode;
 339
 340                 for (j = 0; j < o->ioo_bufcnt; j++, rnb++, lnb++) {
 341                         lnb->dentry = dentry;
 342                         lnb->offset = rnb->offset;
 343                         lnb->len    = rnb->len;
 344                         lnb->flags  = rnb->flags;
 345
 346                         if (inode->i_size <= rnb->offset) {
 347                                 /* If there's no more data, abort early.
 348                                  * lnb->page == NULL and lnb->rc == 0, so it's
 349                                  * easy to detect later. */
 350                                 break;
 351                         } else {
 352                                 rc = filter_start_page_read(inode, lnb);
 353                         }
 354
 355                         if (rc) {
 356                                 CDEBUG(rc == -ENOSPC ? D_INODE : D_ERROR,
 357                                        "page err %u@"LPU64" %u/%u %p: rc %d\n",
 358                                        lnb->len, lnb->offset, j, o->ioo_bufcnt,
 359                                        dentry, rc);
 360                                 cleanup_phase = 1;
 361                                 GOTO(cleanup, rc);
 362                         }
 363
 364                         tot_bytes += lnb->rc;
 365                         if (lnb->rc < lnb->len) {
 366                                 /* short read, be sure to wait on it */
 367                                 lnb++;
 368                                 break;
 369                         }
 370                 }
 371         }
 372
 373         if (time_after(jiffies, now + 15 * HZ))
 374                 CERROR("slow start_page_read %lus\n", (jiffies - now) / HZ);
 375         else
 376                 CDEBUG(D_INFO, "start_page_read: %lu jiffies\n",
 377                        (jiffies - now));
 378
 379         lprocfs_counter_add(obd->obd_stats, LPROC_FILTER_READ_BYTES, tot_bytes);
 380         while (lnb-- > res) {
 381                 rc = filter_finish_page_read(lnb);
 382                 if (rc) {
 383                         CERROR("error page %u@"LPU64" %u %p: rc %d\n", lnb->len,
 384                                lnb->offset, (int)(lnb - res), lnb->dentry, rc);
 385                         cleanup_phase = 1;
 386                         GOTO(cleanup, rc);
 387                 }
 388         }
 389
 390         if (time_after(jiffies, now + 15 * HZ))
 391                 CERROR("slow finish_page_read %lus\n", (jiffies - now) / HZ);
 392         else
 393                 CDEBUG(D_INFO, "finish_page_read: %lu jiffies\n",
 394                        (jiffies - now));
 395
 396         filter_tally_read(&exp->exp_obd->u.filter, res, niocount);
 397
 398         EXIT;
 399
 400  cleanup:
 401         switch (cleanup_phase) {
 402         case 1:
 403                 for (lnb = res; lnb < (res + niocount); lnb++) {
 404                         if (lnb->page)
 405                                 page_cache_release(lnb->page);
 406                 }
 407                 if (res->dentry != NULL)
 408                         f_dput(res->dentry);
 409                 else
 410                         CERROR("NULL dentry in cleanup -- tell CFS\n");
 411         case 0:
 412                 OBD_FREE(fso, objcount * sizeof(*fso));
 413                 pop_ctxt(&saved, &exp->exp_obd->obd_ctxt, NULL);
 414         }
 415         return rc;
 416 }
 417
 418 /* When clients have dirtied as much space as they've been granted they
 419  * fall through to sync writes.  These sync writes haven't been expressed
 420  * in grants and need to error with ENOSPC when there isn't room in the
 421  * filesystem for them after grants are taken into account.  However,
 422  * writeback of the dirty data that was already granted space can write
 423  * right on through.
 424  *
 425  * Caller must hold obd_osfs_lock. */
 426 static int filter_grant_check(struct obd_export *exp, int objcount,
 427                               struct fsfilt_objinfo *fso, int niocount,
 428                               struct niobuf_remote *rnb,
 429                               struct niobuf_local *lnb, obd_size *left,
 430                               struct inode *inode)
 431 {
 432         struct filter_export_data *fed = &exp->exp_filter_data;
 433         int blocksize = exp->exp_obd->u.filter.fo_sb->s_blocksize;
 434         unsigned long used = 0, ungranted = 0, using;
 435         int i, rc = -ENOSPC, obj, n = 0, mask = D_CACHE;
 436
 437         LASSERT_SPIN_LOCKED(&exp->exp_obd->obd_osfs_lock);
 438
 439         for (obj = 0; obj < objcount; obj++) {
 440                 for (i = 0; i < fso[obj].fso_bufcnt; i++, n++) {
 441                         int tmp, bytes;
 442
 443                         /* FIXME: this is calculated with PAGE_SIZE on client */
 444                         bytes = rnb[n].len;
 445                         bytes += rnb[n].offset & (blocksize - 1);
 446                         tmp = (rnb[n].offset + rnb[n].len) & (blocksize - 1);
 447                         if (tmp)
 448                                 bytes += blocksize - tmp;
 449
 450                         if (rnb[n].flags & OBD_BRW_FROM_GRANT) {
 451                                 if (fed->fed_grant < used + bytes) {
 452                                         CDEBUG(D_CACHE,
 453                                                "%s: cli %s/%p claims %ld+%d "
 454                                                "GRANT, real grant %lu idx %d\n",
 455                                                exp->exp_obd->obd_name,
 456                                                exp->exp_client_uuid.uuid, exp,
 457                                                used, bytes, fed->fed_grant, n);
 458                                         mask = D_ERROR;
 459                                 } else {
 460                                         used += bytes;
 461                                         rnb[n].flags |= OBD_BRW_GRANTED;
 462                                         lnb[n].lnb_grant_used = bytes;
 463                                         CDEBUG(0, "idx %d used=%lu\n", n, used);
 464                                         rc = 0;
 465                                         continue;
 466                                 }
 467                         }
 468                         if (*left > ungranted) {
 469                                 /* if enough space, pretend it was granted */
 470                                 ungranted += bytes;
 471                                 rnb[n].flags |= OBD_BRW_GRANTED;
 472                                 CDEBUG(0, "idx %d ungranted=%lu\n",n,ungranted);
 473                                 rc = 0;
 474                                 continue;
 475                         }
 476
 477                         /* We can't check for already-mapped blocks here, as
 478                          * it requires dropping the osfs lock to do the bmap.
 479                          * Instead, we return ENOSPC and in that case we need
 480                          * to go through and verify if all of the blocks not
 481                          * marked BRW_GRANTED are already mapped and we can
 482                          * ignore this error. */
 483                         lnb[n].rc = -ENOSPC;
 484                         rnb[n].flags &= OBD_BRW_GRANTED;
 485                         CDEBUG(D_CACHE,"%s: cli %s/%p idx %d no space for %d\n",
 486                                exp->exp_obd->obd_name,
 487                                exp->exp_client_uuid.uuid, exp, n, bytes);
 488                 }
 489         }
 490
 491         /* Now substract what client have used already.  We don't subtract
 492          * this from the tot_granted yet, so that other client's can't grab
 493          * that space before we have actually allocated our blocks.  That
 494          * happens in filter_grant_commit() after the writes are done. */
 495         *left -= ungranted;
 496         fed->fed_grant -= used;
 497         fed->fed_pending += used;
 498         exp->exp_obd->u.filter.fo_tot_pending += used;
 499
 500         CDEBUG(mask,
 501                "%s: cli %s/%p used: %lu ungranted: %lu grant: %lu dirty: %lu\n",
 502                exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp, used,
 503                ungranted, fed->fed_grant, fed->fed_dirty);
 504
 505         /* Rough calc in case we don't refresh cached statfs data */
 506         using = (used + ungranted + 1 ) >>
 507                 exp->exp_obd->u.filter.fo_sb->s_blocksize_bits;
 508         if (exp->exp_obd->obd_osfs.os_bavail > using)
 509                 exp->exp_obd->obd_osfs.os_bavail -= using;
 510         else
 511                 exp->exp_obd->obd_osfs.os_bavail = 0;
 512
 513         if (fed->fed_dirty < used) {
 514                 CERROR("%s: cli %s/%p claims used %lu > fed_dirty %lu\n",
 515                        exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
 516                        used, fed->fed_dirty);
 517                 used = fed->fed_dirty;
 518         }
 519         exp->exp_obd->u.filter.fo_tot_dirty -= used;
 520         fed->fed_dirty -= used;
 521
 522         return rc;
 523 }
 524
 525 static int filter_start_page_write(struct inode *inode,
 526                                    struct niobuf_local *lnb)
 527 {
 528         struct page *page = alloc_pages(GFP_HIGHUSER, 0);
 529         if (page == NULL) {
 530                 CERROR("no memory for a temp page\n");
 531                 RETURN(lnb->rc = -ENOMEM);
 532         }
 533         POISON_PAGE(page, 0xf1);
 534         if (lnb->len != PAGE_SIZE) {
 535                 memset(kmap(page) + lnb->len, 0, PAGE_SIZE - lnb->len);
 536                 kunmap(page);
 537         }
 538         page->index = lnb->offset >> PAGE_SHIFT;
 539         lnb->page = page;
 540
 541         return 0;
 542 }
 543
 544 /* If we ever start to support multi-object BRW RPCs, we will need to get locks
 545  * on mulitple inodes.  That isn't all, because there still exists the
 546  * possibility of a truncate starting a new transaction while holding the ext3
 547  * rwsem = write while some writes (which have started their transactions here)
 548  * blocking on the ext3 rwsem = read => lock inversion.
 549  *
 550  * The handling gets very ugly when dealing with locked pages.  It may be easier
 551  * to just get rid of the locked page code (which has problems of its own) and
 552  * either discover we do not need it anymore (i.e. it was a symptom of another
 553  * bug) or ensure we get the page locks in an appropriate order. */
 554 static int filter_preprw_write(int cmd, struct obd_export *exp, struct obdo *oa,
 555                                int objcount, struct obd_ioobj *obj,
 556                                int niocount, struct niobuf_remote *nb,
 557                                struct niobuf_local *res,
 558                                struct obd_trans_info *oti)
 559 {
 560         struct obd_run_ctxt saved;
 561         struct niobuf_remote *rnb;
 562         struct niobuf_local *lnb;
 563         struct fsfilt_objinfo fso;
 564         struct dentry *dentry;
 565         obd_size left;
 566         unsigned long now = jiffies;
 567         int rc = 0, i, tot_bytes = 0, cleanup_phase = 1;
 568         ENTRY;
 569         LASSERT(objcount == 1);
 570         LASSERT(obj->ioo_bufcnt > 0);
 571
 572         memset(res, 0, niocount * sizeof(*res));
 573
 574         push_ctxt(&saved, &exp->exp_obd->obd_ctxt, NULL);
 575         dentry = filter_fid2dentry(exp->exp_obd, NULL, obj->ioo_gr,
 576                                    obj->ioo_id);
 577         if (IS_ERR(dentry))
 578                 GOTO(cleanup, rc = PTR_ERR(dentry));
 579
 580         if (dentry->d_inode == NULL) {
 581                 CERROR("trying to BRW to non-existent file "LPU64"\n",
 582                        obj->ioo_id);
 583                 f_dput(dentry);
 584                 GOTO(cleanup, rc = -ENOENT);
 585         }
 586
 587         fso.fso_dentry = dentry;
 588         fso.fso_bufcnt = obj->ioo_bufcnt;
 589
 590         if (time_after(jiffies, now + 15 * HZ))
 591                 CERROR("slow preprw_write setup %lus\n", (jiffies - now) / HZ);
 592         else
 593                 CDEBUG(D_INFO, "preprw_write setup: %lu jiffies\n",
 594                        (jiffies - now));
 595
 596         spin_lock(&exp->exp_obd->obd_osfs_lock);
 597         if (oa)
 598                 filter_grant_incoming(exp, oa);
 599         cleanup_phase = 0;
 600
 601         left = filter_grant_space_left(exp);
 602
 603         rc = filter_grant_check(exp, objcount, &fso, niocount, nb, res,
 604                                 &left, dentry->d_inode);
 605         if (oa && oa->o_valid & OBD_MD_FLGRANT)
 606                 oa->o_grant = filter_grant(exp,oa->o_grant,oa->o_undirty,left);
 607
 608         spin_unlock(&exp->exp_obd->obd_osfs_lock);
 609
 610         if (rc) {
 611                 f_dput(dentry);
 612                 GOTO(cleanup, rc);
 613         }
 614
 615         for (i = 0, rnb = nb, lnb = res; i < obj->ioo_bufcnt;
 616              i++, lnb++, rnb++) {
 617                 /* We still set up for ungranted pages so that granted pages
 618                  * can be written to disk as they were promised, and portals
 619                  * needs to keep the pages all aligned properly. */
 620                 lnb->dentry = dentry;
 621                 lnb->offset = rnb->offset;
 622                 lnb->len    = rnb->len;
 623                 lnb->flags  = rnb->flags;
 624
 625                 rc = filter_start_page_write(dentry->d_inode, lnb);
 626                 if (rc) {
 627                         CERROR("page err %u@"LPU64" %u/%u %p: rc %d\n",
 628                                lnb->len, lnb->offset,
 629                                i, obj->ioo_bufcnt, dentry, rc);
 630                         while (lnb-- > res)
 631                                 __free_pages(lnb->page, 0);
 632                         f_dput(dentry);
 633                         GOTO(cleanup, rc);
 634                 }
 635                 if (lnb->rc == 0)
 636                         tot_bytes += lnb->len;
 637         }
 638
 639         if (time_after(jiffies, now + 15 * HZ))
 640                 CERROR("slow start_page_write %lus\n", (jiffies - now) / HZ);
 641         else
 642                 CDEBUG(D_INFO, "start_page_write: %lu jiffies\n",
 643                        (jiffies - now));
 644
 645         lprocfs_counter_add(exp->exp_obd->obd_stats, LPROC_FILTER_WRITE_BYTES,
 646                             tot_bytes);
 647         EXIT;
 648 cleanup:
 649         switch(cleanup_phase) {
 650         case 1:
 651                 spin_lock(&exp->exp_obd->obd_osfs_lock);
 652                 if (oa)
 653                         filter_grant_incoming(exp, oa);
 654                 spin_unlock(&exp->exp_obd->obd_osfs_lock);
 655         default: ;
 656         }
 657         pop_ctxt(&saved, &exp->exp_obd->obd_ctxt, NULL);
 658         return rc;
 659 }
 660
 661 int filter_preprw(int cmd, struct obd_export *exp, struct obdo *oa,
 662                   int objcount, struct obd_ioobj *obj, int niocount,
 663                   struct niobuf_remote *nb, struct niobuf_local *res,
 664                   struct obd_trans_info *oti)
 665 {
 666         if (cmd == OBD_BRW_WRITE)
 667                 return filter_preprw_write(cmd, exp, oa, objcount, obj,
 668                                            niocount, nb, res, oti);
 669
 670         if (cmd == OBD_BRW_READ)
 671                 return filter_preprw_read(cmd, exp, oa, objcount, obj,
 672                                           niocount, nb, res, oti);
 673
 674         LBUG();
 675         return -EPROTO;
 676 }
 677
 678 static int filter_commitrw_read(struct obd_export *exp, struct obdo *oa,
 679                                 int objcount, struct obd_ioobj *obj,
 680                                 int niocount, struct niobuf_local *res,
 681                                 struct obd_trans_info *oti, int rc)
 682 {
 683         struct obd_ioobj *o;
 684         struct niobuf_local *lnb;
 685         int i, j, drop = 0;
 686         ENTRY;
 687
 688         if (res->dentry != NULL)
 689                 drop = (res->dentry->d_inode->i_size >
 690                         exp->exp_obd->u.filter.fo_readcache_max_filesize);
 691
 692         for (i = 0, o = obj, lnb = res; i < objcount; i++, o++) {
 693                 for (j = 0 ; j < o->ioo_bufcnt ; j++, lnb++) {
 694                         if (lnb->page == NULL)
 695                                 continue;
 696                         /* drop from cache like truncate_list_pages() */
 697                         if (drop && !TryLockPage(lnb->page)) {
 698                                 if (lnb->page->mapping)
 699                                         ll_truncate_complete_page(lnb->page);
 700                                 unlock_page(lnb->page);
 701                         }
 702                         page_cache_release(lnb->page);
 703                 }
 704         }
 705
 706         if (res->dentry != NULL)
 707                 f_dput(res->dentry);
 708         RETURN(rc);
 709 }
 710
 711 void flip_into_page_cache(struct inode *inode, struct page *new_page)
 712 {
 713         struct page *old_page;
 714         int rc;
 715
 716         do {
 717                 /* the dlm is protecting us from read/write concurrency, so we
 718                  * expect this find_lock_page to return quickly.  even if we
 719                  * race with another writer it won't be doing much work with
 720                  * the page locked.  we do this 'cause t_c_p expects a
 721                  * locked page, and it wants to grab the pagecache lock
 722                  * as well. */
 723                 old_page = find_lock_page(inode->i_mapping, new_page->index);
 724                 if (old_page) {
 725                         ll_truncate_complete_page(old_page);
 726                         unlock_page(old_page);
 727                         page_cache_release(old_page);
 728                 }
 729
 730 #if 0 /* this should be a /proc tunable someday */
 731                 /* racing o_directs (no locking ioctl) could race adding
 732                  * their pages, so we repeat the page invalidation unless
 733                  * we successfully added our new page */
 734                 rc = add_to_page_cache_unique(new_page, inode->i_mapping,
 735                                               new_page->index,
 736                                               page_hash(inode->i_mapping,
 737                                                         new_page->index));
 738                 if (rc == 0) {
 739                         /* add_to_page_cache clears uptodate|dirty and locks
 740                          * the page */
 741                         SetPageUptodate(new_page);
 742                         unlock_page(new_page);
 743                 }
 744 #else
 745                 rc = 0;
 746 #endif
 747         } while (rc != 0);
 748 }
 749
 750 void filter_grant_commit(struct obd_export *exp, int niocount,
 751                          struct niobuf_local *res)
 752 {
 753         struct filter_obd *filter = &exp->exp_obd->u.filter;
 754         struct niobuf_local *lnb = res;
 755         unsigned long pending = 0;
 756         int i;
 757
 758         spin_lock(&exp->exp_obd->obd_osfs_lock);
 759         for (i = 0, lnb = res; i < niocount; i++, lnb++)
 760                 pending += lnb->lnb_grant_used;
 761
 762         LASSERTF(exp->exp_filter_data.fed_pending >= pending,
 763                  "%s: cli %s/%p fed_pending: %lu grant_used: %lu\n",
 764                  exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
 765                  exp->exp_filter_data.fed_pending, pending);
 766         exp->exp_filter_data.fed_pending -= pending;
 767         LASSERTF(filter->fo_tot_granted >= pending,
 768                  "%s: cli %s/%p tot_granted: "LPU64" grant_used: %lu\n",
 769                  exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
 770                  exp->exp_obd->u.filter.fo_tot_granted, pending);
 771         filter->fo_tot_granted -= pending;
 772         LASSERTF(filter->fo_tot_pending >= pending,
 773                  "%s: cli %s/%p tot_pending: "LPU64" grant_used: %lu\n",
 774                  exp->exp_obd->obd_name, exp->exp_client_uuid.uuid, exp,
 775                  filter->fo_tot_pending, pending);
 776         filter->fo_tot_pending -= pending;
 777
 778         spin_unlock(&exp->exp_obd->obd_osfs_lock);
 779 }
 780
 781 int filter_commitrw(int cmd, struct obd_export *exp, struct obdo *oa,
 782                     int objcount, struct obd_ioobj *obj, int niocount,
 783                     struct niobuf_local *res, struct obd_trans_info *oti,int rc)
 784 {
 785         if (cmd == OBD_BRW_WRITE)
 786                 return filter_commitrw_write(exp, oa, objcount, obj, niocount,
 787                                              res, oti, rc);
 788         if (cmd == OBD_BRW_READ)
 789                 return filter_commitrw_read(exp, oa, objcount, obj, niocount,
 790                                             res, oti, rc);
 791         LBUG();
 792         return -EPROTO;
 793 }
 794
 795 int filter_brw(int cmd, struct obd_export *exp, struct obdo *oa,
 796                struct lov_stripe_md *lsm, obd_count oa_bufs,
 797                struct brw_page *pga, struct obd_trans_info *oti)
 798 {
 799         struct obd_ioobj ioo;
 800         struct niobuf_local *lnb;
 801         struct niobuf_remote *rnb;
 802         obd_count i;
 803         int ret = 0;
 804         ENTRY;
 805
 806         OBD_ALLOC(lnb, oa_bufs * sizeof(struct niobuf_local));
 807         OBD_ALLOC(rnb, oa_bufs * sizeof(struct niobuf_remote));
 808
 809         if (lnb == NULL || rnb == NULL)
 810                 GOTO(out, ret = -ENOMEM);
 811
 812         for (i = 0; i < oa_bufs; i++) {
 813                 rnb[i].offset = pga[i].off;
 814                 rnb[i].len = pga[i].count;
 815         }
 816
 817         obdo_to_ioobj(oa, &ioo);
 818         ioo.ioo_bufcnt = oa_bufs;
 819
 820         ret = filter_preprw(cmd, exp, oa, 1, &ioo, oa_bufs, rnb, lnb, oti);
 821         if (ret != 0)
 822                 GOTO(out, ret);
 823
 824         for (i = 0; i < oa_bufs; i++) {
 825                 void *virt = kmap(pga[i].pg);
 826                 obd_off off = pga[i].off & ~PAGE_MASK;
 827                 void *addr = kmap(lnb[i].page);
 828
 829                 /* 2 kmaps == vanishingly small deadlock opportunity */
 830
 831                 if (cmd & OBD_BRW_WRITE)
 832                         memcpy(addr + off, virt + off, pga[i].count);
 833                 else
 834                         memcpy(virt + off, addr + off, pga[i].count);
 835
 836                 kunmap(lnb[i].page);
 837                 kunmap(pga[i].pg);
 838         }
 839
 840         ret = filter_commitrw(cmd, exp, oa, 1, &ioo, oa_bufs, lnb, oti, ret);
 841
 842 out:
 843         if (lnb)
 844                 OBD_FREE(lnb, oa_bufs * sizeof(struct niobuf_local));
 845         if (rnb)
 846                 OBD_FREE(rnb, oa_bufs * sizeof(struct niobuf_remote));
 847         RETURN(ret);
 848 }