fs/nfs/objlayout/objio_osd.c

   1 /*
   2  *  pNFS Objects layout implementation over open-osd initiator library
   3  *
   4  *  Copyright (C) 2009 Panasas Inc. [year of first publication]
   5  *  All rights reserved.
   6  *
   7  *  Benny Halevy <bhalevy@panasas.com>
   8  *  Boaz Harrosh <ooo@electrozaur.com>
   9  *
  10  *  This program is free software; you can redistribute it and/or modify
  11  *  it under the terms of the GNU General Public License version 2
  12  *  See the file COPYING included with this distribution for more details.
  13  *
  14  *  Redistribution and use in source and binary forms, with or without
  15  *  modification, are permitted provided that the following conditions
  16  *  are met:
  17  *
  18  *  1. Redistributions of source code must retain the above copyright
  19  *     notice, this list of conditions and the following disclaimer.
  20  *  2. Redistributions in binary form must reproduce the above copyright
  21  *     notice, this list of conditions and the following disclaimer in the
  22  *     documentation and/or other materials provided with the distribution.
  23  *  3. Neither the name of the Panasas company nor the names of its
  24  *     contributors may be used to endorse or promote products derived
  25  *     from this software without specific prior written permission.
  26  *
  27  *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  28  *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  29  *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  30  *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  31  *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  32  *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  33  *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  34  *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  35  *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  36  *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  37  *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  38  */
  39
  40 #include <linux/module.h>
  41 #include <scsi/osd_ore.h>
  42
  43 #include "objlayout.h"
  44 #include "../internal.h"
  45
  46 #define NFSDBG_FACILITY         NFSDBG_PNFS_LD
  47
  48 struct objio_dev_ent {
  49         struct nfs4_deviceid_node id_node;
  50         struct ore_dev od;
  51 };
  52
  53 static void
  54 objio_free_deviceid_node(struct nfs4_deviceid_node *d)
  55 {
  56         struct objio_dev_ent *de = container_of(d, struct objio_dev_ent, id_node);
  57
  58         dprintk("%s: free od=%p\n", __func__, de->od.od);
  59         osduld_put_device(de->od.od);
  60         kfree(de);
  61 }
  62
  63 static struct objio_dev_ent *_dev_list_find(const struct nfs_server *nfss,
  64         const struct nfs4_deviceid *d_id)
  65 {
  66         struct nfs4_deviceid_node *d;
  67         struct objio_dev_ent *de;
  68
  69         d = nfs4_find_get_deviceid(nfss->pnfs_curr_ld, nfss->nfs_client, d_id);
  70         if (!d)
  71                 return NULL;
  72
  73         de = container_of(d, struct objio_dev_ent, id_node);
  74         return de;
  75 }
  76
  77 static struct objio_dev_ent *
  78 _dev_list_add(const struct nfs_server *nfss,
  79         const struct nfs4_deviceid *d_id, struct osd_dev *od,
  80         gfp_t gfp_flags)
  81 {
  82         struct nfs4_deviceid_node *d;
  83         struct objio_dev_ent *de = kzalloc(sizeof(*de), gfp_flags);
  84         struct objio_dev_ent *n;
  85
  86         if (!de) {
  87                 dprintk("%s: -ENOMEM od=%p\n", __func__, od);
  88                 return NULL;
  89         }
  90
  91         dprintk("%s: Adding od=%p\n", __func__, od);
  92         nfs4_init_deviceid_node(&de->id_node,
  93                                 nfss->pnfs_curr_ld,
  94                                 nfss->nfs_client,
  95                                 d_id);
  96         de->od.od = od;
  97
  98         d = nfs4_insert_deviceid_node(&de->id_node);
  99         n = container_of(d, struct objio_dev_ent, id_node);
 100         if (n != de) {
 101                 dprintk("%s: Race with other n->od=%p\n", __func__, n->od.od);
 102                 objio_free_deviceid_node(&de->id_node);
 103                 de = n;
 104         }
 105
 106         return de;
 107 }
 108
 109 struct objio_segment {
 110         struct pnfs_layout_segment lseg;
 111
 112         struct ore_layout layout;
 113         struct ore_components oc;
 114 };
 115
 116 static inline struct objio_segment *
 117 OBJIO_LSEG(struct pnfs_layout_segment *lseg)
 118 {
 119         return container_of(lseg, struct objio_segment, lseg);
 120 }
 121
 122 struct objio_state {
 123         /* Generic layer */
 124         struct objlayout_io_res oir;
 125
 126         bool sync;
 127         /*FIXME: Support for extra_bytes at ore_get_rw_state() */
 128         struct ore_io_state *ios;
 129 };
 130
 131 /* Send and wait for a get_device_info of devices in the layout,
 132    then look them up with the osd_initiator library */
 133 static int objio_devices_lookup(struct pnfs_layout_hdr *pnfslay,
 134         struct objio_segment *objio_seg, unsigned c, struct nfs4_deviceid *d_id,
 135         gfp_t gfp_flags)
 136 {
 137         struct pnfs_osd_deviceaddr *deviceaddr;
 138         struct objio_dev_ent *ode;
 139         struct osd_dev *od;
 140         struct osd_dev_info odi;
 141         bool retry_flag = true;
 142         int err;
 143
 144         ode = _dev_list_find(NFS_SERVER(pnfslay->plh_inode), d_id);
 145         if (ode) {
 146                 objio_seg->oc.ods[c] = &ode->od; /* must use container_of */
 147                 return 0;
 148         }
 149
 150         err = objlayout_get_deviceinfo(pnfslay, d_id, &deviceaddr, gfp_flags);
 151         if (unlikely(err)) {
 152                 dprintk("%s: objlayout_get_deviceinfo dev(%llx:%llx) =>%d\n",
 153                         __func__, _DEVID_LO(d_id), _DEVID_HI(d_id), err);
 154                 return err;
 155         }
 156
 157         odi.systemid_len = deviceaddr->oda_systemid.len;
 158         if (odi.systemid_len > sizeof(odi.systemid)) {
 159                 dprintk("%s: odi.systemid_len > sizeof(systemid=%zd)\n",
 160                         __func__, sizeof(odi.systemid));
 161                 err = -EINVAL;
 162                 goto out;
 163         } else if (odi.systemid_len)
 164                 memcpy(odi.systemid, deviceaddr->oda_systemid.data,
 165                        odi.systemid_len);
 166         odi.osdname_len  = deviceaddr->oda_osdname.len;
 167         odi.osdname      = (u8 *)deviceaddr->oda_osdname.data;
 168
 169         if (!odi.osdname_len && !odi.systemid_len) {
 170                 dprintk("%s: !odi.osdname_len && !odi.systemid_len\n",
 171                         __func__);
 172                 err = -ENODEV;
 173                 goto out;
 174         }
 175
 176 retry_lookup:
 177         od = osduld_info_lookup(&odi);
 178         if (unlikely(IS_ERR(od))) {
 179                 err = PTR_ERR(od);
 180                 dprintk("%s: osduld_info_lookup => %d\n", __func__, err);
 181                 if (err == -ENODEV && retry_flag) {
 182                         err = objlayout_autologin(deviceaddr);
 183                         if (likely(!err)) {
 184                                 retry_flag = false;
 185                                 goto retry_lookup;
 186                         }
 187                 }
 188                 goto out;
 189         }
 190
 191         ode = _dev_list_add(NFS_SERVER(pnfslay->plh_inode), d_id, od,
 192                             gfp_flags);
 193         objio_seg->oc.ods[c] = &ode->od; /* must use container_of */
 194         dprintk("Adding new dev_id(%llx:%llx)\n",
 195                 _DEVID_LO(d_id), _DEVID_HI(d_id));
 196 out:
 197         objlayout_put_deviceinfo(deviceaddr);
 198         return err;
 199 }
 200
 201 static void copy_single_comp(struct ore_components *oc, unsigned c,
 202                              struct pnfs_osd_object_cred *src_comp)
 203 {
 204         struct ore_comp *ocomp = &oc->comps[c];
 205
 206         WARN_ON(src_comp->oc_cap_key.cred_len > 0); /* libosd is NO_SEC only */
 207         WARN_ON(src_comp->oc_cap.cred_len > sizeof(ocomp->cred));
 208
 209         ocomp->obj.partition = src_comp->oc_object_id.oid_partition_id;
 210         ocomp->obj.id = src_comp->oc_object_id.oid_object_id;
 211
 212         memcpy(ocomp->cred, src_comp->oc_cap.cred, sizeof(ocomp->cred));
 213 }
 214
 215 static int __alloc_objio_seg(unsigned numdevs, gfp_t gfp_flags,
 216                        struct objio_segment **pseg)
 217 {
 218 /*      This is the in memory structure of the objio_segment
 219  *
 220  *      struct __alloc_objio_segment {
 221  *              struct objio_segment olseg;
 222  *              struct ore_dev *ods[numdevs];
 223  *              struct ore_comp comps[numdevs];
 224  *      } *aolseg;
 225  *      NOTE: The code as above compiles and runs perfectly. It is elegant,
 226  *      type safe and compact. At some Past time Linus has decided he does not
 227  *      like variable length arrays, For the sake of this principal we uglify
 228  *      the code as below.
 229  */
 230         struct objio_segment *lseg;
 231         size_t lseg_size = sizeof(*lseg) +
 232                         numdevs * sizeof(lseg->oc.ods[0]) +
 233                         numdevs * sizeof(*lseg->oc.comps);
 234
 235         lseg = kzalloc(lseg_size, gfp_flags);
 236         if (unlikely(!lseg)) {
 237                 dprintk("%s: Failed allocation numdevs=%d size=%zd\n", __func__,
 238                         numdevs, lseg_size);
 239                 return -ENOMEM;
 240         }
 241
 242         lseg->oc.numdevs = numdevs;
 243         lseg->oc.single_comp = EC_MULTPLE_COMPS;
 244         lseg->oc.ods = (void *)(lseg + 1);
 245         lseg->oc.comps = (void *)(lseg->oc.ods + numdevs);
 246
 247         *pseg = lseg;
 248         return 0;
 249 }
 250
 251 int objio_alloc_lseg(struct pnfs_layout_segment **outp,
 252         struct pnfs_layout_hdr *pnfslay,
 253         struct pnfs_layout_range *range,
 254         struct xdr_stream *xdr,
 255         gfp_t gfp_flags)
 256 {
 257         struct objio_segment *objio_seg;
 258         struct pnfs_osd_xdr_decode_layout_iter iter;
 259         struct pnfs_osd_layout layout;
 260         struct pnfs_osd_object_cred src_comp;
 261         unsigned cur_comp;
 262         int err;
 263
 264         err = pnfs_osd_xdr_decode_layout_map(&layout, &iter, xdr);
 265         if (unlikely(err))
 266                 return err;
 267
 268         err = __alloc_objio_seg(layout.olo_num_comps, gfp_flags, &objio_seg);
 269         if (unlikely(err))
 270                 return err;
 271
 272         objio_seg->layout.stripe_unit = layout.olo_map.odm_stripe_unit;
 273         objio_seg->layout.group_width = layout.olo_map.odm_group_width;
 274         objio_seg->layout.group_depth = layout.olo_map.odm_group_depth;
 275         objio_seg->layout.mirrors_p1 = layout.olo_map.odm_mirror_cnt + 1;
 276         objio_seg->layout.raid_algorithm = layout.olo_map.odm_raid_algorithm;
 277
 278         err = ore_verify_layout(layout.olo_map.odm_num_comps,
 279                                           &objio_seg->layout);
 280         if (unlikely(err))
 281                 goto err;
 282
 283         objio_seg->oc.first_dev = layout.olo_comps_index;
 284         cur_comp = 0;
 285         while (pnfs_osd_xdr_decode_layout_comp(&src_comp, &iter, xdr, &err)) {
 286                 copy_single_comp(&objio_seg->oc, cur_comp, &src_comp);
 287                 err = objio_devices_lookup(pnfslay, objio_seg, cur_comp,
 288                                            &src_comp.oc_object_id.oid_device_id,
 289                                            gfp_flags);
 290                 if (err)
 291                         goto err;
 292                 ++cur_comp;
 293         }
 294         /* pnfs_osd_xdr_decode_layout_comp returns false on error */
 295         if (unlikely(err))
 296                 goto err;
 297
 298         *outp = &objio_seg->lseg;
 299         return 0;
 300
 301 err:
 302         kfree(objio_seg);
 303         dprintk("%s: Error: return %d\n", __func__, err);
 304         *outp = NULL;
 305         return err;
 306 }
 307
 308 void objio_free_lseg(struct pnfs_layout_segment *lseg)
 309 {
 310         int i;
 311         struct objio_segment *objio_seg = OBJIO_LSEG(lseg);
 312
 313         for (i = 0; i < objio_seg->oc.numdevs; i++) {
 314                 struct ore_dev *od = objio_seg->oc.ods[i];
 315                 struct objio_dev_ent *ode;
 316
 317                 if (!od)
 318                         break;
 319                 ode = container_of(od, typeof(*ode), od);
 320                 nfs4_put_deviceid_node(&ode->id_node);
 321         }
 322         kfree(objio_seg);
 323 }
 324
 325 static int
 326 objio_alloc_io_state(struct pnfs_layout_hdr *pnfs_layout_type, bool is_reading,
 327         struct pnfs_layout_segment *lseg, struct page **pages, unsigned pgbase,
 328         loff_t offset, size_t count, void *rpcdata, gfp_t gfp_flags,
 329         struct objio_state **outp)
 330 {
 331         struct objio_segment *objio_seg = OBJIO_LSEG(lseg);
 332         struct ore_io_state *ios;
 333         int ret;
 334         struct __alloc_objio_state {
 335                 struct objio_state objios;
 336                 struct pnfs_osd_ioerr ioerrs[objio_seg->oc.numdevs];
 337         } *aos;
 338
 339         aos = kzalloc(sizeof(*aos), gfp_flags);
 340         if (unlikely(!aos))
 341                 return -ENOMEM;
 342
 343         objlayout_init_ioerrs(&aos->objios.oir, objio_seg->oc.numdevs,
 344                         aos->ioerrs, rpcdata, pnfs_layout_type);
 345
 346         ret = ore_get_rw_state(&objio_seg->layout, &objio_seg->oc, is_reading,
 347                                offset, count, &ios);
 348         if (unlikely(ret)) {
 349                 kfree(aos);
 350                 return ret;
 351         }
 352
 353         ios->pages = pages;
 354         ios->pgbase = pgbase;
 355         ios->private = aos;
 356         BUG_ON(ios->nr_pages > (pgbase + count + PAGE_SIZE - 1) >> PAGE_SHIFT);
 357
 358         aos->objios.sync = 0;
 359         aos->objios.ios = ios;
 360         *outp = &aos->objios;
 361         return 0;
 362 }
 363
 364 void objio_free_result(struct objlayout_io_res *oir)
 365 {
 366         struct objio_state *objios = container_of(oir, struct objio_state, oir);
 367
 368         ore_put_io_state(objios->ios);
 369         kfree(objios);
 370 }
 371
 372 static enum pnfs_osd_errno osd_pri_2_pnfs_err(enum osd_err_priority oep)
 373 {
 374         switch (oep) {
 375         case OSD_ERR_PRI_NO_ERROR:
 376                 return (enum pnfs_osd_errno)0;
 377
 378         case OSD_ERR_PRI_CLEAR_PAGES:
 379                 BUG_ON(1);
 380                 return 0;
 381
 382         case OSD_ERR_PRI_RESOURCE:
 383                 return PNFS_OSD_ERR_RESOURCE;
 384         case OSD_ERR_PRI_BAD_CRED:
 385                 return PNFS_OSD_ERR_BAD_CRED;
 386         case OSD_ERR_PRI_NO_ACCESS:
 387                 return PNFS_OSD_ERR_NO_ACCESS;
 388         case OSD_ERR_PRI_UNREACHABLE:
 389                 return PNFS_OSD_ERR_UNREACHABLE;
 390         case OSD_ERR_PRI_NOT_FOUND:
 391                 return PNFS_OSD_ERR_NOT_FOUND;
 392         case OSD_ERR_PRI_NO_SPACE:
 393                 return PNFS_OSD_ERR_NO_SPACE;
 394         default:
 395                 WARN_ON(1);
 396                 /* fallthrough */
 397         case OSD_ERR_PRI_EIO:
 398                 return PNFS_OSD_ERR_EIO;
 399         }
 400 }
 401
 402 static void __on_dev_error(struct ore_io_state *ios,
 403         struct ore_dev *od, unsigned dev_index, enum osd_err_priority oep,
 404         u64 dev_offset, u64  dev_len)
 405 {
 406         struct objio_state *objios = ios->private;
 407         struct pnfs_osd_objid pooid;
 408         struct objio_dev_ent *ode = container_of(od, typeof(*ode), od);
 409         /* FIXME: what to do with more-then-one-group layouts. We need to
 410          * translate from ore_io_state index to oc->comps index
 411          */
 412         unsigned comp = dev_index;
 413
 414         pooid.oid_device_id = ode->id_node.deviceid;
 415         pooid.oid_partition_id = ios->oc->comps[comp].obj.partition;
 416         pooid.oid_object_id = ios->oc->comps[comp].obj.id;
 417
 418         objlayout_io_set_result(&objios->oir, comp,
 419                                 &pooid, osd_pri_2_pnfs_err(oep),
 420                                 dev_offset, dev_len, !ios->reading);
 421 }
 422
 423 /*
 424  * read
 425  */
 426 static void _read_done(struct ore_io_state *ios, void *private)
 427 {
 428         struct objio_state *objios = private;
 429         ssize_t status;
 430         int ret = ore_check_io(ios, &__on_dev_error);
 431
 432         /* FIXME: _io_free(ios) can we dealocate the libosd resources; */
 433
 434         if (likely(!ret))
 435                 status = ios->length;
 436         else
 437                 status = ret;
 438
 439         objlayout_read_done(&objios->oir, status, objios->sync);
 440 }
 441
 442 int objio_read_pagelist(struct nfs_pgio_header *hdr)
 443 {
 444         struct objio_state *objios;
 445         int ret;
 446
 447         ret = objio_alloc_io_state(NFS_I(hdr->inode)->layout, true,
 448                         hdr->lseg, hdr->args.pages, hdr->args.pgbase,
 449                         hdr->args.offset, hdr->args.count, hdr,
 450                         GFP_KERNEL, &objios);
 451         if (unlikely(ret))
 452                 return ret;
 453
 454         objios->ios->done = _read_done;
 455         dprintk("%s: offset=0x%llx length=0x%x\n", __func__,
 456                 hdr->args.offset, hdr->args.count);
 457         ret = ore_read(objios->ios);
 458         if (unlikely(ret))
 459                 objio_free_result(&objios->oir);
 460         return ret;
 461 }
 462
 463 /*
 464  * write
 465  */
 466 static void _write_done(struct ore_io_state *ios, void *private)
 467 {
 468         struct objio_state *objios = private;
 469         ssize_t status;
 470         int ret = ore_check_io(ios, &__on_dev_error);
 471
 472         /* FIXME: _io_free(ios) can we dealocate the libosd resources; */
 473
 474         if (likely(!ret)) {
 475                 /* FIXME: should be based on the OSD's persistence model
 476                  * See OSD2r05 Section 4.13 Data persistence model */
 477                 objios->oir.committed = NFS_FILE_SYNC;
 478                 status = ios->length;
 479         } else {
 480                 status = ret;
 481         }
 482
 483         objlayout_write_done(&objios->oir, status, objios->sync);
 484 }
 485
 486 static struct page *__r4w_get_page(void *priv, u64 offset, bool *uptodate)
 487 {
 488         struct objio_state *objios = priv;
 489         struct nfs_pgio_header *hdr = objios->oir.rpcdata;
 490         struct address_space *mapping = hdr->inode->i_mapping;
 491         pgoff_t index = offset / PAGE_SIZE;
 492         struct page *page;
 493         loff_t i_size = i_size_read(hdr->inode);
 494
 495         if (offset >= i_size) {
 496                 *uptodate = true;
 497                 dprintk("%s: g_zero_page index=0x%lx\n", __func__, index);
 498                 return ZERO_PAGE(0);
 499         }
 500
 501         page = find_get_page(mapping, index);
 502         if (!page) {
 503                 page = find_or_create_page(mapping, index, GFP_NOFS);
 504                 if (unlikely(!page)) {
 505                         dprintk("%s: grab_cache_page Failed index=0x%lx\n",
 506                                 __func__, index);
 507                         return NULL;
 508                 }
 509                 unlock_page(page);
 510         }
 511         if (PageDirty(page) || PageWriteback(page))
 512                 *uptodate = true;
 513         else
 514                 *uptodate = PageUptodate(page);
 515         dprintk("%s: index=0x%lx uptodate=%d\n", __func__, index, *uptodate);
 516         return page;
 517 }
 518
 519 static void __r4w_put_page(void *priv, struct page *page)
 520 {
 521         dprintk("%s: index=0x%lx\n", __func__,
 522                 (page == ZERO_PAGE(0)) ? -1UL : page->index);
 523         if (ZERO_PAGE(0) != page)
 524                 page_cache_release(page);
 525         return;
 526 }
 527
 528 static const struct _ore_r4w_op _r4w_op = {
 529         .get_page = &__r4w_get_page,
 530         .put_page = &__r4w_put_page,
 531 };
 532
 533 int objio_write_pagelist(struct nfs_pgio_header *hdr, int how)
 534 {
 535         struct objio_state *objios;
 536         int ret;
 537
 538         ret = objio_alloc_io_state(NFS_I(hdr->inode)->layout, false,
 539                         hdr->lseg, hdr->args.pages, hdr->args.pgbase,
 540                         hdr->args.offset, hdr->args.count, hdr, GFP_NOFS,
 541                         &objios);
 542         if (unlikely(ret))
 543                 return ret;
 544
 545         objios->sync = 0 != (how & FLUSH_SYNC);
 546         objios->ios->r4w = &_r4w_op;
 547
 548         if (!objios->sync)
 549                 objios->ios->done = _write_done;
 550
 551         dprintk("%s: offset=0x%llx length=0x%x\n", __func__,
 552                 hdr->args.offset, hdr->args.count);
 553         ret = ore_write(objios->ios);
 554         if (unlikely(ret)) {
 555                 objio_free_result(&objios->oir);
 556                 return ret;
 557         }
 558
 559         if (objios->sync)
 560                 _write_done(objios->ios, objios);
 561
 562         return 0;
 563 }
 564
 565 /*
 566  * Return 0 if @req cannot be coalesced into @pgio, otherwise return the number
 567  * of bytes (maximum @req->wb_bytes) that can be coalesced.
 568  */
 569 static size_t objio_pg_test(struct nfs_pageio_descriptor *pgio,
 570                           struct nfs_page *prev, struct nfs_page *req)
 571 {
 572         unsigned int size;
 573
 574         size = pnfs_generic_pg_test(pgio, prev, req);
 575
 576         if (!size || pgio->pg_count + req->wb_bytes >
 577             (unsigned long)pgio->pg_layout_private)
 578                 return 0;
 579
 580         return min(size, req->wb_bytes);
 581 }
 582
 583 static void objio_init_read(struct nfs_pageio_descriptor *pgio, struct nfs_page *req)
 584 {
 585         pnfs_generic_pg_init_read(pgio, req);
 586         if (unlikely(pgio->pg_lseg == NULL))
 587                 return; /* Not pNFS */
 588
 589         pgio->pg_layout_private = (void *)
 590                                 OBJIO_LSEG(pgio->pg_lseg)->layout.max_io_length;
 591 }
 592
 593 static bool aligned_on_raid_stripe(u64 offset, struct ore_layout *layout,
 594                                    unsigned long *stripe_end)
 595 {
 596         u32 stripe_off;
 597         unsigned stripe_size;
 598
 599         if (layout->raid_algorithm == PNFS_OSD_RAID_0)
 600                 return true;
 601
 602         stripe_size = layout->stripe_unit *
 603                                 (layout->group_width - layout->parity);
 604
 605         div_u64_rem(offset, stripe_size, &stripe_off);
 606         if (!stripe_off)
 607                 return true;
 608
 609         *stripe_end = stripe_size - stripe_off;
 610         return false;
 611 }
 612
 613 static void objio_init_write(struct nfs_pageio_descriptor *pgio, struct nfs_page *req)
 614 {
 615         unsigned long stripe_end = 0;
 616         u64 wb_size;
 617
 618         if (pgio->pg_dreq == NULL)
 619                 wb_size = i_size_read(pgio->pg_inode) - req_offset(req);
 620         else
 621                 wb_size = nfs_dreq_bytes_left(pgio->pg_dreq);
 622
 623         pnfs_generic_pg_init_write(pgio, req, wb_size);
 624         if (unlikely(pgio->pg_lseg == NULL))
 625                 return; /* Not pNFS */
 626
 627         if (req->wb_offset ||
 628             !aligned_on_raid_stripe(req->wb_index * PAGE_SIZE,
 629                                &OBJIO_LSEG(pgio->pg_lseg)->layout,
 630                                &stripe_end)) {
 631                 pgio->pg_layout_private = (void *)stripe_end;
 632         } else {
 633                 pgio->pg_layout_private = (void *)
 634                                 OBJIO_LSEG(pgio->pg_lseg)->layout.max_io_length;
 635         }
 636 }
 637
 638 static const struct nfs_pageio_ops objio_pg_read_ops = {
 639         .pg_init = objio_init_read,
 640         .pg_test = objio_pg_test,
 641         .pg_doio = pnfs_generic_pg_readpages,
 642 };
 643
 644 static const struct nfs_pageio_ops objio_pg_write_ops = {
 645         .pg_init = objio_init_write,
 646         .pg_test = objio_pg_test,
 647         .pg_doio = pnfs_generic_pg_writepages,
 648 };
 649
 650 static struct pnfs_layoutdriver_type objlayout_type = {
 651         .id = LAYOUT_OSD2_OBJECTS,
 652         .name = "LAYOUT_OSD2_OBJECTS",
 653         .flags                   = PNFS_LAYOUTRET_ON_SETATTR |
 654                                    PNFS_LAYOUTRET_ON_ERROR,
 655
 656         .owner                   = THIS_MODULE,
 657         .alloc_layout_hdr        = objlayout_alloc_layout_hdr,
 658         .free_layout_hdr         = objlayout_free_layout_hdr,
 659
 660         .alloc_lseg              = objlayout_alloc_lseg,
 661         .free_lseg               = objlayout_free_lseg,
 662
 663         .read_pagelist           = objlayout_read_pagelist,
 664         .write_pagelist          = objlayout_write_pagelist,
 665         .pg_read_ops             = &objio_pg_read_ops,
 666         .pg_write_ops            = &objio_pg_write_ops,
 667
 668         .free_deviceid_node      = objio_free_deviceid_node,
 669
 670         .encode_layoutcommit     = objlayout_encode_layoutcommit,
 671         .encode_layoutreturn     = objlayout_encode_layoutreturn,
 672 };
 673
 674 MODULE_DESCRIPTION("pNFS Layout Driver for OSD2 objects");
 675 MODULE_AUTHOR("Benny Halevy <bhalevy@panasas.com>");
 676 MODULE_LICENSE("GPL");
 677
 678 static int __init
 679 objlayout_init(void)
 680 {
 681         int ret = pnfs_register_layoutdriver(&objlayout_type);
 682
 683         if (ret)
 684                 printk(KERN_INFO
 685                         "NFS: %s: Registering OSD pNFS Layout Driver failed: error=%d\n",
 686                         __func__, ret);
 687         else
 688                 printk(KERN_INFO "NFS: %s: Registered OSD pNFS Layout Driver\n",
 689                         __func__);
 690         return ret;
 691 }
 692
 693 static void __exit
 694 objlayout_exit(void)
 695 {
 696         pnfs_unregister_layoutdriver(&objlayout_type);
 697         printk(KERN_INFO "NFS: %s: Unregistered OSD pNFS Layout Driver\n",
 698                __func__);
 699 }
 700
 701 MODULE_ALIAS("nfs-layouttype4-2");
 702
 703 module_init(objlayout_init);
 704 module_exit(objlayout_exit);