fs/ceph/addr.c

   1 #include <linux/ceph/ceph_debug.h>
   2
   3 #include <linux/backing-dev.h>
   4 #include <linux/fs.h>
   5 #include <linux/mm.h>
   6 #include <linux/pagemap.h>
   7 #include <linux/writeback.h>    /* generic_writepages */
   8 #include <linux/slab.h>
   9 #include <linux/pagevec.h>
  10 #include <linux/task_io_accounting_ops.h>
  11
  12 #include "super.h"
  13 #include "mds_client.h"
  14 #include <linux/ceph/osd_client.h>
  15
  16 /*
  17  * Ceph address space ops.
  18  *
  19  * There are a few funny things going on here.
  20  *
  21  * The page->private field is used to reference a struct
  22  * ceph_snap_context for _every_ dirty page.  This indicates which
  23  * snapshot the page was logically dirtied in, and thus which snap
  24  * context needs to be associated with the osd write during writeback.
  25  *
  26  * Similarly, struct ceph_inode_info maintains a set of counters to
  27  * count dirty pages on the inode.  In the absence of snapshots,
  28  * i_wrbuffer_ref == i_wrbuffer_ref_head == the dirty page count.
  29  *
  30  * When a snapshot is taken (that is, when the client receives
  31  * notification that a snapshot was taken), each inode with caps and
  32  * with dirty pages (dirty pages implies there is a cap) gets a new
  33  * ceph_cap_snap in the i_cap_snaps list (which is sorted in ascending
  34  * order, new snaps go to the tail).  The i_wrbuffer_ref_head count is
  35  * moved to capsnap->dirty. (Unless a sync write is currently in
  36  * progress.  In that case, the capsnap is said to be "pending", new
  37  * writes cannot start, and the capsnap isn't "finalized" until the
  38  * write completes (or fails) and a final size/mtime for the inode for
  39  * that snap can be settled upon.)  i_wrbuffer_ref_head is reset to 0.
  40  *
  41  * On writeback, we must submit writes to the osd IN SNAP ORDER.  So,
  42  * we look for the first capsnap in i_cap_snaps and write out pages in
  43  * that snap context _only_.  Then we move on to the next capsnap,
  44  * eventually reaching the "live" or "head" context (i.e., pages that
  45  * are not yet snapped) and are writing the most recently dirtied
  46  * pages.
  47  *
  48  * Invalidate and so forth must take care to ensure the dirty page
  49  * accounting is preserved.
  50  */
  51
  52 #define CONGESTION_ON_THRESH(congestion_kb) (congestion_kb >> (PAGE_SHIFT-10))
  53 #define CONGESTION_OFF_THRESH(congestion_kb)                            \
  54         (CONGESTION_ON_THRESH(congestion_kb) -                          \
  55          (CONGESTION_ON_THRESH(congestion_kb) >> 2))
  56
  57 static inline struct ceph_snap_context *page_snap_context(struct page *page)
  58 {
  59         if (PagePrivate(page))
  60                 return (void *)page->private;
  61         return NULL;
  62 }
  63
  64 /*
  65  * Dirty a page.  Optimistically adjust accounting, on the assumption
  66  * that we won't race with invalidate.  If we do, readjust.
  67  */
  68 static int ceph_set_page_dirty(struct page *page)
  69 {
  70         struct address_space *mapping = page->mapping;
  71         struct inode *inode;
  72         struct ceph_inode_info *ci;
  73         int undo = 0;
  74         struct ceph_snap_context *snapc;
  75
  76         if (unlikely(!mapping))
  77                 return !TestSetPageDirty(page);
  78
  79         if (TestSetPageDirty(page)) {
  80                 dout("%p set_page_dirty %p idx %lu -- already dirty\n",
  81                      mapping->host, page, page->index);
  82                 return 0;
  83         }
  84
  85         inode = mapping->host;
  86         ci = ceph_inode(inode);
  87
  88         /*
  89          * Note that we're grabbing a snapc ref here without holding
  90          * any locks!
  91          */
  92         snapc = ceph_get_snap_context(ci->i_snap_realm->cached_context);
  93
  94         /* dirty the head */
  95         spin_lock(&ci->i_ceph_lock);
  96         if (ci->i_head_snapc == NULL)
  97                 ci->i_head_snapc = ceph_get_snap_context(snapc);
  98         ++ci->i_wrbuffer_ref_head;
  99         if (ci->i_wrbuffer_ref == 0)
 100                 ihold(inode);
 101         ++ci->i_wrbuffer_ref;
 102         dout("%p set_page_dirty %p idx %lu head %d/%d -> %d/%d "
 103              "snapc %p seq %lld (%d snaps)\n",
 104              mapping->host, page, page->index,
 105              ci->i_wrbuffer_ref-1, ci->i_wrbuffer_ref_head-1,
 106              ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
 107              snapc, snapc->seq, snapc->num_snaps);
 108         spin_unlock(&ci->i_ceph_lock);
 109
 110         /* now adjust page */
 111         spin_lock_irq(&mapping->tree_lock);
 112         if (page->mapping) {    /* Race with truncate? */
 113                 WARN_ON_ONCE(!PageUptodate(page));
 114                 account_page_dirtied(page, page->mapping);
 115                 radix_tree_tag_set(&mapping->page_tree,
 116                                 page_index(page), PAGECACHE_TAG_DIRTY);
 117
 118                 /*
 119                  * Reference snap context in page->private.  Also set
 120                  * PagePrivate so that we get invalidatepage callback.
 121                  */
 122                 page->private = (unsigned long)snapc;
 123                 SetPagePrivate(page);
 124         } else {
 125                 dout("ANON set_page_dirty %p (raced truncate?)\n", page);
 126                 undo = 1;
 127         }
 128
 129         spin_unlock_irq(&mapping->tree_lock);
 130
 131         if (undo)
 132                 /* whoops, we failed to dirty the page */
 133                 ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
 134
 135         __mark_inode_dirty(mapping->host, I_DIRTY_PAGES);
 136
 137         BUG_ON(!PageDirty(page));
 138         return 1;
 139 }
 140
 141 /*
 142  * If we are truncating the full page (i.e. offset == 0), adjust the
 143  * dirty page counters appropriately.  Only called if there is private
 144  * data on the page.
 145  */
 146 static void ceph_invalidatepage(struct page *page, unsigned long offset)
 147 {
 148         struct inode *inode;
 149         struct ceph_inode_info *ci;
 150         struct ceph_snap_context *snapc = page_snap_context(page);
 151
 152         BUG_ON(!PageLocked(page));
 153         BUG_ON(!PagePrivate(page));
 154         BUG_ON(!page->mapping);
 155
 156         inode = page->mapping->host;
 157
 158         /*
 159          * We can get non-dirty pages here due to races between
 160          * set_page_dirty and truncate_complete_page; just spit out a
 161          * warning, in case we end up with accounting problems later.
 162          */
 163         if (!PageDirty(page))
 164                 pr_err("%p invalidatepage %p page not dirty\n", inode, page);
 165
 166         if (offset == 0)
 167                 ClearPageChecked(page);
 168
 169         ci = ceph_inode(inode);
 170         if (offset == 0) {
 171                 dout("%p invalidatepage %p idx %lu full dirty page %lu\n",
 172                      inode, page, page->index, offset);
 173                 ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
 174                 ceph_put_snap_context(snapc);
 175                 page->private = 0;
 176                 ClearPagePrivate(page);
 177         } else {
 178                 dout("%p invalidatepage %p idx %lu partial dirty page\n",
 179                      inode, page, page->index);
 180         }
 181 }
 182
 183 /* just a sanity check */
 184 static int ceph_releasepage(struct page *page, gfp_t g)
 185 {
 186         struct inode *inode = page->mapping ? page->mapping->host : NULL;
 187         dout("%p releasepage %p idx %lu\n", inode, page, page->index);
 188         WARN_ON(PageDirty(page));
 189         WARN_ON(PagePrivate(page));
 190         return 0;
 191 }
 192
 193 /*
 194  * read a single page, without unlocking it.
 195  */
 196 static int readpage_nounlock(struct file *filp, struct page *page)
 197 {
 198         struct inode *inode = file_inode(filp);
 199         struct ceph_inode_info *ci = ceph_inode(inode);
 200         struct ceph_osd_client *osdc =
 201                 &ceph_inode_to_client(inode)->client->osdc;
 202         int err = 0;
 203         u64 len = PAGE_CACHE_SIZE;
 204
 205         dout("readpage inode %p file %p page %p index %lu\n",
 206              inode, filp, page, page->index);
 207         err = ceph_osdc_readpages(osdc, ceph_vino(inode), &ci->i_layout,
 208                                   (u64) page_offset(page), &len,
 209                                   ci->i_truncate_seq, ci->i_truncate_size,
 210                                   &page, 1, 0);
 211         if (err == -ENOENT)
 212                 err = 0;
 213         if (err < 0) {
 214                 SetPageError(page);
 215                 goto out;
 216         } else if (err < PAGE_CACHE_SIZE) {
 217                 /* zero fill remainder of page */
 218                 zero_user_segment(page, err, PAGE_CACHE_SIZE);
 219         }
 220         SetPageUptodate(page);
 221
 222 out:
 223         return err < 0 ? err : 0;
 224 }
 225
 226 static int ceph_readpage(struct file *filp, struct page *page)
 227 {
 228         int r = readpage_nounlock(filp, page);
 229         unlock_page(page);
 230         return r;
 231 }
 232
 233 /*
 234  * Finish an async read(ahead) op.
 235  */
 236 static void finish_read(struct ceph_osd_request *req, struct ceph_msg *msg)
 237 {
 238         struct inode *inode = req->r_inode;
 239         int rc = req->r_result;
 240         int bytes = le32_to_cpu(msg->hdr.data_len);
 241         int i;
 242
 243         dout("finish_read %p req %p rc %d bytes %d\n", inode, req, rc, bytes);
 244
 245         /* unlock all pages, zeroing any data we didn't read */
 246         for (i = 0; i < req->r_data.num_pages; i++, bytes -= PAGE_CACHE_SIZE) {
 247                 struct page *page = req->r_data.pages[i];
 248
 249                 if (bytes < (int)PAGE_CACHE_SIZE) {
 250                         /* zero (remainder of) page */
 251                         int s = bytes < 0 ? 0 : bytes;
 252                         zero_user_segment(page, s, PAGE_CACHE_SIZE);
 253                 }
 254                 dout("finish_read %p uptodate %p idx %lu\n", inode, page,
 255                      page->index);
 256                 flush_dcache_page(page);
 257                 SetPageUptodate(page);
 258                 unlock_page(page);
 259                 page_cache_release(page);
 260         }
 261         kfree(req->r_data.pages);
 262 }
 263
 264 static void ceph_unlock_page_vector(struct page **pages, int num_pages)
 265 {
 266         int i;
 267
 268         for (i = 0; i < num_pages; i++)
 269                 unlock_page(pages[i]);
 270 }
 271
 272 /*
 273  * start an async read(ahead) operation.  return nr_pages we submitted
 274  * a read for on success, or negative error code.
 275  */
 276 static int start_read(struct inode *inode, struct list_head *page_list, int max)
 277 {
 278         struct ceph_osd_client *osdc =
 279                 &ceph_inode_to_client(inode)->client->osdc;
 280         struct ceph_inode_info *ci = ceph_inode(inode);
 281         struct page *page = list_entry(page_list->prev, struct page, lru);
 282         struct ceph_osd_request *req;
 283         u64 off;
 284         u64 len;
 285         int i;
 286         struct page **pages;
 287         pgoff_t next_index;
 288         int nr_pages = 0;
 289         int ret;
 290
 291         off = (u64) page_offset(page);
 292
 293         /* count pages */
 294         next_index = page->index;
 295         list_for_each_entry_reverse(page, page_list, lru) {
 296                 if (page->index != next_index)
 297                         break;
 298                 nr_pages++;
 299                 next_index++;
 300                 if (max && nr_pages == max)
 301                         break;
 302         }
 303         len = nr_pages << PAGE_CACHE_SHIFT;
 304         dout("start_read %p nr_pages %d is %lld~%lld\n", inode, nr_pages,
 305              off, len);
 306
 307         req = ceph_osdc_new_request(osdc, &ci->i_layout, ceph_vino(inode),
 308                                     off, &len,
 309                                     CEPH_OSD_OP_READ, CEPH_OSD_FLAG_READ,
 310                                     NULL, 0,
 311                                     ci->i_truncate_seq, ci->i_truncate_size,
 312                                     NULL, false);
 313         if (IS_ERR(req))
 314                 return PTR_ERR(req);
 315
 316         /* build page vector */
 317         nr_pages = calc_pages_for(0, len);
 318         pages = kmalloc(sizeof(*pages) * nr_pages, GFP_NOFS);
 319         ret = -ENOMEM;
 320         if (!pages)
 321                 goto out;
 322         for (i = 0; i < nr_pages; ++i) {
 323                 page = list_entry(page_list->prev, struct page, lru);
 324                 BUG_ON(PageLocked(page));
 325                 list_del(&page->lru);
 326
 327                 dout("start_read %p adding %p idx %lu\n", inode, page,
 328                      page->index);
 329                 if (add_to_page_cache_lru(page, &inode->i_data, page->index,
 330                                           GFP_NOFS)) {
 331                         page_cache_release(page);
 332                         dout("start_read %p add_to_page_cache failed %p\n",
 333                              inode, page);
 334                         nr_pages = i;
 335                         goto out_pages;
 336                 }
 337                 pages[i] = page;
 338         }
 339         req->r_data.pages = pages;
 340         req->r_data.num_pages = nr_pages;
 341         req->r_data.alignment = 0;
 342         req->r_callback = finish_read;
 343         req->r_inode = inode;
 344
 345         dout("start_read %p starting %p %lld~%lld\n", inode, req, off, len);
 346         ret = ceph_osdc_start_request(osdc, req, false);
 347         if (ret < 0)
 348                 goto out_pages;
 349         ceph_osdc_put_request(req);
 350         return nr_pages;
 351
 352 out_pages:
 353         ceph_unlock_page_vector(pages, nr_pages);
 354         ceph_release_page_vector(pages, nr_pages);
 355 out:
 356         ceph_osdc_put_request(req);
 357         return ret;
 358 }
 359
 360
 361 /*
 362  * Read multiple pages.  Leave pages we don't read + unlock in page_list;
 363  * the caller (VM) cleans them up.
 364  */
 365 static int ceph_readpages(struct file *file, struct address_space *mapping,
 366                           struct list_head *page_list, unsigned nr_pages)
 367 {
 368         struct inode *inode = file_inode(file);
 369         struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 370         int rc = 0;
 371         int max = 0;
 372
 373         if (fsc->mount_options->rsize >= PAGE_CACHE_SIZE)
 374                 max = (fsc->mount_options->rsize + PAGE_CACHE_SIZE - 1)
 375                         >> PAGE_SHIFT;
 376
 377         dout("readpages %p file %p nr_pages %d max %d\n", inode,
 378                 file, nr_pages,
 379              max);
 380         while (!list_empty(page_list)) {
 381                 rc = start_read(inode, page_list, max);
 382                 if (rc < 0)
 383                         goto out;
 384                 BUG_ON(rc == 0);
 385         }
 386 out:
 387         dout("readpages %p file %p ret %d\n", inode, file, rc);
 388         return rc;
 389 }
 390
 391 /*
 392  * Get ref for the oldest snapc for an inode with dirty data... that is, the
 393  * only snap context we are allowed to write back.
 394  */
 395 static struct ceph_snap_context *get_oldest_context(struct inode *inode,
 396                                                     u64 *snap_size)
 397 {
 398         struct ceph_inode_info *ci = ceph_inode(inode);
 399         struct ceph_snap_context *snapc = NULL;
 400         struct ceph_cap_snap *capsnap = NULL;
 401
 402         spin_lock(&ci->i_ceph_lock);
 403         list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
 404                 dout(" cap_snap %p snapc %p has %d dirty pages\n", capsnap,
 405                      capsnap->context, capsnap->dirty_pages);
 406                 if (capsnap->dirty_pages) {
 407                         snapc = ceph_get_snap_context(capsnap->context);
 408                         if (snap_size)
 409                                 *snap_size = capsnap->size;
 410                         break;
 411                 }
 412         }
 413         if (!snapc && ci->i_wrbuffer_ref_head) {
 414                 snapc = ceph_get_snap_context(ci->i_head_snapc);
 415                 dout(" head snapc %p has %d dirty pages\n",
 416                      snapc, ci->i_wrbuffer_ref_head);
 417         }
 418         spin_unlock(&ci->i_ceph_lock);
 419         return snapc;
 420 }
 421
 422 /*
 423  * Write a single page, but leave the page locked.
 424  *
 425  * If we get a write error, set the page error bit, but still adjust the
 426  * dirty page accounting (i.e., page is no longer dirty).
 427  */
 428 static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 429 {
 430         struct inode *inode;
 431         struct ceph_inode_info *ci;
 432         struct ceph_fs_client *fsc;
 433         struct ceph_osd_client *osdc;
 434         loff_t page_off = page_offset(page);
 435         int len = PAGE_CACHE_SIZE;
 436         loff_t i_size;
 437         int err = 0;
 438         struct ceph_snap_context *snapc, *oldest;
 439         u64 snap_size = 0;
 440         long writeback_stat;
 441
 442         dout("writepage %p idx %lu\n", page, page->index);
 443
 444         if (!page->mapping || !page->mapping->host) {
 445                 dout("writepage %p - no mapping\n", page);
 446                 return -EFAULT;
 447         }
 448         inode = page->mapping->host;
 449         ci = ceph_inode(inode);
 450         fsc = ceph_inode_to_client(inode);
 451         osdc = &fsc->client->osdc;
 452
 453         /* verify this is a writeable snap context */
 454         snapc = page_snap_context(page);
 455         if (snapc == NULL) {
 456                 dout("writepage %p page %p not dirty?\n", inode, page);
 457                 goto out;
 458         }
 459         oldest = get_oldest_context(inode, &snap_size);
 460         if (snapc->seq > oldest->seq) {
 461                 dout("writepage %p page %p snapc %p not writeable - noop\n",
 462                      inode, page, snapc);
 463                 /* we should only noop if called by kswapd */
 464                 WARN_ON((current->flags & PF_MEMALLOC) == 0);
 465                 ceph_put_snap_context(oldest);
 466                 goto out;
 467         }
 468         ceph_put_snap_context(oldest);
 469
 470         /* is this a partial page at end of file? */
 471         if (snap_size)
 472                 i_size = snap_size;
 473         else
 474                 i_size = i_size_read(inode);
 475         if (i_size < page_off + len)
 476                 len = i_size - page_off;
 477
 478         dout("writepage %p page %p index %lu on %llu~%u snapc %p\n",
 479              inode, page, page->index, page_off, len, snapc);
 480
 481         writeback_stat = atomic_long_inc_return(&fsc->writeback_count);
 482         if (writeback_stat >
 483             CONGESTION_ON_THRESH(fsc->mount_options->congestion_kb))
 484                 set_bdi_congested(&fsc->backing_dev_info, BLK_RW_ASYNC);
 485
 486         set_page_writeback(page);
 487         err = ceph_osdc_writepages(osdc, ceph_vino(inode),
 488                                    &ci->i_layout, snapc,
 489                                    page_off, len,
 490                                    ci->i_truncate_seq, ci->i_truncate_size,
 491                                    &inode->i_mtime, &page, 1);
 492         if (err < 0) {
 493                 dout("writepage setting page/mapping error %d %p\n", err, page);
 494                 SetPageError(page);
 495                 mapping_set_error(&inode->i_data, err);
 496                 if (wbc)
 497                         wbc->pages_skipped++;
 498         } else {
 499                 dout("writepage cleaned page %p\n", page);
 500                 err = 0;  /* vfs expects us to return 0 */
 501         }
 502         page->private = 0;
 503         ClearPagePrivate(page);
 504         end_page_writeback(page);
 505         ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
 506         ceph_put_snap_context(snapc);  /* page's reference */
 507 out:
 508         return err;
 509 }
 510
 511 static int ceph_writepage(struct page *page, struct writeback_control *wbc)
 512 {
 513         int err;
 514         struct inode *inode = page->mapping->host;
 515         BUG_ON(!inode);
 516         ihold(inode);
 517         err = writepage_nounlock(page, wbc);
 518         unlock_page(page);
 519         iput(inode);
 520         return err;
 521 }
 522
 523
 524 /*
 525  * lame release_pages helper.  release_pages() isn't exported to
 526  * modules.
 527  */
 528 static void ceph_release_pages(struct page **pages, int num)
 529 {
 530         struct pagevec pvec;
 531         int i;
 532
 533         pagevec_init(&pvec, 0);
 534         for (i = 0; i < num; i++) {
 535                 if (pagevec_add(&pvec, pages[i]) == 0)
 536                         pagevec_release(&pvec);
 537         }
 538         pagevec_release(&pvec);
 539 }
 540
 541
 542 /*
 543  * async writeback completion handler.
 544  *
 545  * If we get an error, set the mapping error bit, but not the individual
 546  * page error bits.
 547  */
 548 static void writepages_finish(struct ceph_osd_request *req,
 549                               struct ceph_msg *msg)
 550 {
 551         struct inode *inode = req->r_inode;
 552         struct ceph_inode_info *ci = ceph_inode(inode);
 553         unsigned wrote;
 554         struct page *page;
 555         int i;
 556         struct ceph_snap_context *snapc = req->r_snapc;
 557         struct address_space *mapping = inode->i_mapping;
 558         int rc = req->r_result;
 559         u64 bytes = le64_to_cpu(req->r_request_ops[0].extent.length);
 560         struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 561         long writeback_stat;
 562         unsigned issued = ceph_caps_issued(ci);
 563
 564         if (rc >= 0) {
 565                 /*
 566                  * Assume we wrote the pages we originally sent.  The
 567                  * osd might reply with fewer pages if our writeback
 568                  * raced with a truncation and was adjusted at the osd,
 569                  * so don't believe the reply.
 570                  */
 571                 wrote = req->r_data.num_pages;
 572         } else {
 573                 wrote = 0;
 574                 mapping_set_error(mapping, rc);
 575         }
 576         dout("writepages_finish %p rc %d bytes %llu wrote %d (pages)\n",
 577              inode, rc, bytes, wrote);
 578
 579         /* clean all pages */
 580         for (i = 0; i < req->r_data.num_pages; i++) {
 581                 page = req->r_data.pages[i];
 582                 BUG_ON(!page);
 583                 WARN_ON(!PageUptodate(page));
 584
 585                 writeback_stat =
 586                         atomic_long_dec_return(&fsc->writeback_count);
 587                 if (writeback_stat <
 588                     CONGESTION_OFF_THRESH(fsc->mount_options->congestion_kb))
 589                         clear_bdi_congested(&fsc->backing_dev_info,
 590                                             BLK_RW_ASYNC);
 591
 592                 ceph_put_snap_context(page_snap_context(page));
 593                 page->private = 0;
 594                 ClearPagePrivate(page);
 595                 dout("unlocking %d %p\n", i, page);
 596                 end_page_writeback(page);
 597
 598                 /*
 599                  * We lost the cache cap, need to truncate the page before
 600                  * it is unlocked, otherwise we'd truncate it later in the
 601                  * page truncation thread, possibly losing some data that
 602                  * raced its way in
 603                  */
 604                 if ((issued & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0)
 605                         generic_error_remove_page(inode->i_mapping, page);
 606
 607                 unlock_page(page);
 608         }
 609         dout("%p wrote+cleaned %d pages\n", inode, wrote);
 610         ceph_put_wrbuffer_cap_refs(ci, req->r_data.num_pages, snapc);
 611
 612         ceph_release_pages(req->r_data.pages, req->r_data.num_pages);
 613         if (req->r_data.pages_from_pool)
 614                 mempool_free(req->r_data.pages,
 615                              ceph_sb_to_client(inode->i_sb)->wb_pagevec_pool);
 616         else
 617                 kfree(req->r_data.pages);
 618         ceph_osdc_put_request(req);
 619 }
 620
 621 /*
 622  * allocate a page vec, either directly, or if necessary, via a the
 623  * mempool.  we avoid the mempool if we can because req->r_data.num_pages
 624  * may be less than the maximum write size.
 625  */
 626 static void alloc_page_vec(struct ceph_fs_client *fsc,
 627                            struct ceph_osd_request *req)
 628 {
 629         req->r_data.pages = kmalloc(sizeof(struct page *) * req->r_data.num_pages,
 630                                GFP_NOFS);
 631         if (!req->r_data.pages) {
 632                 req->r_data.pages = mempool_alloc(fsc->wb_pagevec_pool, GFP_NOFS);
 633                 req->r_data.pages_from_pool = 1;
 634                 WARN_ON(!req->r_data.pages);
 635         }
 636 }
 637
 638 /*
 639  * initiate async writeback
 640  */
 641 static int ceph_writepages_start(struct address_space *mapping,
 642                                  struct writeback_control *wbc)
 643 {
 644         struct inode *inode = mapping->host;
 645         struct ceph_inode_info *ci = ceph_inode(inode);
 646         struct ceph_fs_client *fsc;
 647         pgoff_t index, start, end;
 648         int range_whole = 0;
 649         int should_loop = 1;
 650         pgoff_t max_pages = 0, max_pages_ever = 0;
 651         struct ceph_snap_context *snapc = NULL, *last_snapc = NULL, *pgsnapc;
 652         struct pagevec pvec;
 653         int done = 0;
 654         int rc = 0;
 655         unsigned wsize = 1 << inode->i_blkbits;
 656         struct ceph_osd_request *req = NULL;
 657         int do_sync;
 658         u64 snap_size = 0;
 659
 660         /*
 661          * Include a 'sync' in the OSD request if this is a data
 662          * integrity write (e.g., O_SYNC write or fsync()), or if our
 663          * cap is being revoked.
 664          */
 665         do_sync = wbc->sync_mode == WB_SYNC_ALL;
 666         if (ceph_caps_revoking(ci, CEPH_CAP_FILE_BUFFER))
 667                 do_sync = 1;
 668         dout("writepages_start %p dosync=%d (mode=%s)\n",
 669              inode, do_sync,
 670              wbc->sync_mode == WB_SYNC_NONE ? "NONE" :
 671              (wbc->sync_mode == WB_SYNC_ALL ? "ALL" : "HOLD"));
 672
 673         fsc = ceph_inode_to_client(inode);
 674         if (fsc->mount_state == CEPH_MOUNT_SHUTDOWN) {
 675                 pr_warning("writepage_start %p on forced umount\n", inode);
 676                 return -EIO; /* we're in a forced umount, don't write! */
 677         }
 678         if (fsc->mount_options->wsize && fsc->mount_options->wsize < wsize)
 679                 wsize = fsc->mount_options->wsize;
 680         if (wsize < PAGE_CACHE_SIZE)
 681                 wsize = PAGE_CACHE_SIZE;
 682         max_pages_ever = wsize >> PAGE_CACHE_SHIFT;
 683
 684         pagevec_init(&pvec, 0);
 685
 686         /* where to start/end? */
 687         if (wbc->range_cyclic) {
 688                 start = mapping->writeback_index; /* Start from prev offset */
 689                 end = -1;
 690                 dout(" cyclic, start at %lu\n", start);
 691         } else {
 692                 start = wbc->range_start >> PAGE_CACHE_SHIFT;
 693                 end = wbc->range_end >> PAGE_CACHE_SHIFT;
 694                 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
 695                         range_whole = 1;
 696                 should_loop = 0;
 697                 dout(" not cyclic, %lu to %lu\n", start, end);
 698         }
 699         index = start;
 700
 701 retry:
 702         /* find oldest snap context with dirty data */
 703         ceph_put_snap_context(snapc);
 704         snapc = get_oldest_context(inode, &snap_size);
 705         if (!snapc) {
 706                 /* hmm, why does writepages get called when there
 707                    is no dirty data? */
 708                 dout(" no snap context with dirty data?\n");
 709                 goto out;
 710         }
 711         dout(" oldest snapc is %p seq %lld (%d snaps)\n",
 712              snapc, snapc->seq, snapc->num_snaps);
 713         if (last_snapc && snapc != last_snapc) {
 714                 /* if we switched to a newer snapc, restart our scan at the
 715                  * start of the original file range. */
 716                 dout("  snapc differs from last pass, restarting at %lu\n",
 717                      index);
 718                 index = start;
 719         }
 720         last_snapc = snapc;
 721
 722         while (!done && index <= end) {
 723                 unsigned i;
 724                 int first;
 725                 pgoff_t next;
 726                 int pvec_pages, locked_pages;
 727                 struct page *page;
 728                 int want;
 729                 u64 offset, len;
 730                 long writeback_stat;
 731
 732                 next = 0;
 733                 locked_pages = 0;
 734                 max_pages = max_pages_ever;
 735
 736 get_more_pages:
 737                 first = -1;
 738                 want = min(end - index,
 739                            min((pgoff_t)PAGEVEC_SIZE,
 740                                max_pages - (pgoff_t)locked_pages) - 1)
 741                         + 1;
 742                 pvec_pages = pagevec_lookup_tag(&pvec, mapping, &index,
 743                                                 PAGECACHE_TAG_DIRTY,
 744                                                 want);
 745                 dout("pagevec_lookup_tag got %d\n", pvec_pages);
 746                 if (!pvec_pages && !locked_pages)
 747                         break;
 748                 for (i = 0; i < pvec_pages && locked_pages < max_pages; i++) {
 749                         page = pvec.pages[i];
 750                         dout("? %p idx %lu\n", page, page->index);
 751                         if (locked_pages == 0)
 752                                 lock_page(page);  /* first page */
 753                         else if (!trylock_page(page))
 754                                 break;
 755
 756                         /* only dirty pages, or our accounting breaks */
 757                         if (unlikely(!PageDirty(page)) ||
 758                             unlikely(page->mapping != mapping)) {
 759                                 dout("!dirty or !mapping %p\n", page);
 760                                 unlock_page(page);
 761                                 break;
 762                         }
 763                         if (!wbc->range_cyclic && page->index > end) {
 764                                 dout("end of range %p\n", page);
 765                                 done = 1;
 766                                 unlock_page(page);
 767                                 break;
 768                         }
 769                         if (next && (page->index != next)) {
 770                                 dout("not consecutive %p\n", page);
 771                                 unlock_page(page);
 772                                 break;
 773                         }
 774                         if (wbc->sync_mode != WB_SYNC_NONE) {
 775                                 dout("waiting on writeback %p\n", page);
 776                                 wait_on_page_writeback(page);
 777                         }
 778                         if ((snap_size && page_offset(page) > snap_size) ||
 779                             (!snap_size &&
 780                              page_offset(page) > i_size_read(inode))) {
 781                                 dout("%p page eof %llu\n", page, snap_size ?
 782                                      snap_size : i_size_read(inode));
 783                                 done = 1;
 784                                 unlock_page(page);
 785                                 break;
 786                         }
 787                         if (PageWriteback(page)) {
 788                                 dout("%p under writeback\n", page);
 789                                 unlock_page(page);
 790                                 break;
 791                         }
 792
 793                         /* only if matching snap context */
 794                         pgsnapc = page_snap_context(page);
 795                         if (pgsnapc->seq > snapc->seq) {
 796                                 dout("page snapc %p %lld > oldest %p %lld\n",
 797                                      pgsnapc, pgsnapc->seq, snapc, snapc->seq);
 798                                 unlock_page(page);
 799                                 if (!locked_pages)
 800                                         continue; /* keep looking for snap */
 801                                 break;
 802                         }
 803
 804                         if (!clear_page_dirty_for_io(page)) {
 805                                 dout("%p !clear_page_dirty_for_io\n", page);
 806                                 unlock_page(page);
 807                                 break;
 808                         }
 809
 810                         /* ok */
 811                         if (locked_pages == 0) {
 812                                 /* prepare async write request */
 813                                 offset = (u64) page_offset(page);
 814                                 len = wsize;
 815                                 req = ceph_osdc_new_request(&fsc->client->osdc,
 816                                             &ci->i_layout,
 817                                             ceph_vino(inode),
 818                                             offset, &len,
 819                                             CEPH_OSD_OP_WRITE,
 820                                             CEPH_OSD_FLAG_WRITE |
 821                                                     CEPH_OSD_FLAG_ONDISK,
 822                                             snapc, do_sync,
 823                                             ci->i_truncate_seq,
 824                                             ci->i_truncate_size,
 825                                             &inode->i_mtime, true);
 826
 827                                 if (IS_ERR(req)) {
 828                                         rc = PTR_ERR(req);
 829                                         unlock_page(page);
 830                                         break;
 831                                 }
 832
 833                                 req->r_data.num_pages = calc_pages_for(0, len);
 834                                 req->r_data.alignment = 0;
 835                                 max_pages = req->r_data.num_pages;
 836
 837                                 alloc_page_vec(fsc, req);
 838                                 req->r_callback = writepages_finish;
 839                                 req->r_inode = inode;
 840                         }
 841
 842                         /* note position of first page in pvec */
 843                         if (first < 0)
 844                                 first = i;
 845                         dout("%p will write page %p idx %lu\n",
 846                              inode, page, page->index);
 847
 848                         writeback_stat =
 849                                atomic_long_inc_return(&fsc->writeback_count);
 850                         if (writeback_stat > CONGESTION_ON_THRESH(
 851                                     fsc->mount_options->congestion_kb)) {
 852                                 set_bdi_congested(&fsc->backing_dev_info,
 853                                                   BLK_RW_ASYNC);
 854                         }
 855
 856                         set_page_writeback(page);
 857                         req->r_data.pages[locked_pages] = page;
 858                         locked_pages++;
 859                         next = page->index + 1;
 860                 }
 861
 862                 /* did we get anything? */
 863                 if (!locked_pages)
 864                         goto release_pvec_pages;
 865                 if (i) {
 866                         int j;
 867                         BUG_ON(!locked_pages || first < 0);
 868
 869                         if (pvec_pages && i == pvec_pages &&
 870                             locked_pages < max_pages) {
 871                                 dout("reached end pvec, trying for more\n");
 872                                 pagevec_reinit(&pvec);
 873                                 goto get_more_pages;
 874                         }
 875
 876                         /* shift unused pages over in the pvec...  we
 877                          * will need to release them below. */
 878                         for (j = i; j < pvec_pages; j++) {
 879                                 dout(" pvec leftover page %p\n",
 880                                      pvec.pages[j]);
 881                                 pvec.pages[j-i+first] = pvec.pages[j];
 882                         }
 883                         pvec.nr -= i-first;
 884                 }
 885
 886                 /* submit the write */
 887                 offset = req->r_data.pages[0]->index << PAGE_CACHE_SHIFT;
 888                 len = min((snap_size ? snap_size : i_size_read(inode)) - offset,
 889                           (u64)locked_pages << PAGE_CACHE_SHIFT);
 890                 dout("writepages got %d pages at %llu~%llu\n",
 891                      locked_pages, offset, len);
 892
 893                 /* revise final length, page count */
 894                 req->r_data.num_pages = locked_pages;
 895                 req->r_request_ops[0].extent.length = cpu_to_le64(len);
 896                 req->r_request_ops[0].payload_len = cpu_to_le32(len);
 897                 req->r_request->hdr.data_len = cpu_to_le32(len);
 898
 899                 rc = ceph_osdc_start_request(&fsc->client->osdc, req, true);
 900                 BUG_ON(rc);
 901                 req = NULL;
 902
 903                 /* continue? */
 904                 index = next;
 905                 wbc->nr_to_write -= locked_pages;
 906                 if (wbc->nr_to_write <= 0)
 907                         done = 1;
 908
 909 release_pvec_pages:
 910                 dout("pagevec_release on %d pages (%p)\n", (int)pvec.nr,
 911                      pvec.nr ? pvec.pages[0] : NULL);
 912                 pagevec_release(&pvec);
 913
 914                 if (locked_pages && !done)
 915                         goto retry;
 916         }
 917
 918         if (should_loop && !done) {
 919                 /* more to do; loop back to beginning of file */
 920                 dout("writepages looping back to beginning of file\n");
 921                 should_loop = 0;
 922                 index = 0;
 923                 goto retry;
 924         }
 925
 926         if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
 927                 mapping->writeback_index = index;
 928
 929 out:
 930         if (req)
 931                 ceph_osdc_put_request(req);
 932         ceph_put_snap_context(snapc);
 933         dout("writepages done, rc = %d\n", rc);
 934         return rc;
 935 }
 936
 937
 938
 939 /*
 940  * See if a given @snapc is either writeable, or already written.
 941  */
 942 static int context_is_writeable_or_written(struct inode *inode,
 943                                            struct ceph_snap_context *snapc)
 944 {
 945         struct ceph_snap_context *oldest = get_oldest_context(inode, NULL);
 946         int ret = !oldest || snapc->seq <= oldest->seq;
 947
 948         ceph_put_snap_context(oldest);
 949         return ret;
 950 }
 951
 952 /*
 953  * We are only allowed to write into/dirty the page if the page is
 954  * clean, or already dirty within the same snap context.
 955  *
 956  * called with page locked.
 957  * return success with page locked,
 958  * or any failure (incl -EAGAIN) with page unlocked.
 959  */
 960 static int ceph_update_writeable_page(struct file *file,
 961                             loff_t pos, unsigned len,
 962                             struct page *page)
 963 {
 964         struct inode *inode = file_inode(file);
 965         struct ceph_inode_info *ci = ceph_inode(inode);
 966         struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
 967         loff_t page_off = pos & PAGE_CACHE_MASK;
 968         int pos_in_page = pos & ~PAGE_CACHE_MASK;
 969         int end_in_page = pos_in_page + len;
 970         loff_t i_size;
 971         int r;
 972         struct ceph_snap_context *snapc, *oldest;
 973
 974 retry_locked:
 975         /* writepages currently holds page lock, but if we change that later, */
 976         wait_on_page_writeback(page);
 977
 978         /* check snap context */
 979         BUG_ON(!ci->i_snap_realm);
 980         down_read(&mdsc->snap_rwsem);
 981         BUG_ON(!ci->i_snap_realm->cached_context);
 982         snapc = page_snap_context(page);
 983         if (snapc && snapc != ci->i_head_snapc) {
 984                 /*
 985                  * this page is already dirty in another (older) snap
 986                  * context!  is it writeable now?
 987                  */
 988                 oldest = get_oldest_context(inode, NULL);
 989                 up_read(&mdsc->snap_rwsem);
 990
 991                 if (snapc->seq > oldest->seq) {
 992                         ceph_put_snap_context(oldest);
 993                         dout(" page %p snapc %p not current or oldest\n",
 994                              page, snapc);
 995                         /*
 996                          * queue for writeback, and wait for snapc to
 997                          * be writeable or written
 998                          */
 999                         snapc = ceph_get_snap_context(snapc);
1000                         unlock_page(page);
1001                         ceph_queue_writeback(inode);
1002                         r = wait_event_interruptible(ci->i_cap_wq,
1003                                context_is_writeable_or_written(inode, snapc));
1004                         ceph_put_snap_context(snapc);
1005                         if (r == -ERESTARTSYS)
1006                                 return r;
1007                         return -EAGAIN;
1008                 }
1009                 ceph_put_snap_context(oldest);
1010
1011                 /* yay, writeable, do it now (without dropping page lock) */
1012                 dout(" page %p snapc %p not current, but oldest\n",
1013                      page, snapc);
1014                 if (!clear_page_dirty_for_io(page))
1015                         goto retry_locked;
1016                 r = writepage_nounlock(page, NULL);
1017                 if (r < 0)
1018                         goto fail_nosnap;
1019                 goto retry_locked;
1020         }
1021
1022         if (PageUptodate(page)) {
1023                 dout(" page %p already uptodate\n", page);
1024                 return 0;
1025         }
1026
1027         /* full page? */
1028         if (pos_in_page == 0 && len == PAGE_CACHE_SIZE)
1029                 return 0;
1030
1031         /* past end of file? */
1032         i_size = inode->i_size;   /* caller holds i_mutex */
1033
1034         if (i_size + len > inode->i_sb->s_maxbytes) {
1035                 /* file is too big */
1036                 r = -EINVAL;
1037                 goto fail;
1038         }
1039
1040         if (page_off >= i_size ||
1041             (pos_in_page == 0 && (pos+len) >= i_size &&
1042              end_in_page - pos_in_page != PAGE_CACHE_SIZE)) {
1043                 dout(" zeroing %p 0 - %d and %d - %d\n",
1044                      page, pos_in_page, end_in_page, (int)PAGE_CACHE_SIZE);
1045                 zero_user_segments(page,
1046                                    0, pos_in_page,
1047                                    end_in_page, PAGE_CACHE_SIZE);
1048                 return 0;
1049         }
1050
1051         /* we need to read it. */
1052         up_read(&mdsc->snap_rwsem);
1053         r = readpage_nounlock(file, page);
1054         if (r < 0)
1055                 goto fail_nosnap;
1056         goto retry_locked;
1057
1058 fail:
1059         up_read(&mdsc->snap_rwsem);
1060 fail_nosnap:
1061         unlock_page(page);
1062         return r;
1063 }
1064
1065 /*
1066  * We are only allowed to write into/dirty the page if the page is
1067  * clean, or already dirty within the same snap context.
1068  */
1069 static int ceph_write_begin(struct file *file, struct address_space *mapping,
1070                             loff_t pos, unsigned len, unsigned flags,
1071                             struct page **pagep, void **fsdata)
1072 {
1073         struct inode *inode = file_inode(file);
1074         struct page *page;
1075         pgoff_t index = pos >> PAGE_CACHE_SHIFT;
1076         int r;
1077
1078         do {
1079                 /* get a page */
1080                 page = grab_cache_page_write_begin(mapping, index, 0);
1081                 if (!page)
1082                         return -ENOMEM;
1083                 *pagep = page;
1084
1085                 dout("write_begin file %p inode %p page %p %d~%d\n", file,
1086                      inode, page, (int)pos, (int)len);
1087
1088                 r = ceph_update_writeable_page(file, pos, len, page);
1089         } while (r == -EAGAIN);
1090
1091         return r;
1092 }
1093
1094 /*
1095  * we don't do anything in here that simple_write_end doesn't do
1096  * except adjust dirty page accounting and drop read lock on
1097  * mdsc->snap_rwsem.
1098  */
1099 static int ceph_write_end(struct file *file, struct address_space *mapping,
1100                           loff_t pos, unsigned len, unsigned copied,
1101                           struct page *page, void *fsdata)
1102 {
1103         struct inode *inode = file_inode(file);
1104         struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1105         struct ceph_mds_client *mdsc = fsc->mdsc;
1106         unsigned from = pos & (PAGE_CACHE_SIZE - 1);
1107         int check_cap = 0;
1108
1109         dout("write_end file %p inode %p page %p %d~%d (%d)\n", file,
1110              inode, page, (int)pos, (int)copied, (int)len);
1111
1112         /* zero the stale part of the page if we did a short copy */
1113         if (copied < len)
1114                 zero_user_segment(page, from+copied, len);
1115
1116         /* did file size increase? */
1117         /* (no need for i_size_read(); we caller holds i_mutex */
1118         if (pos+copied > inode->i_size)
1119                 check_cap = ceph_inode_set_size(inode, pos+copied);
1120
1121         if (!PageUptodate(page))
1122                 SetPageUptodate(page);
1123
1124         set_page_dirty(page);
1125
1126         unlock_page(page);
1127         up_read(&mdsc->snap_rwsem);
1128         page_cache_release(page);
1129
1130         if (check_cap)
1131                 ceph_check_caps(ceph_inode(inode), CHECK_CAPS_AUTHONLY, NULL);
1132
1133         return copied;
1134 }
1135
1136 /*
1137  * we set .direct_IO to indicate direct io is supported, but since we
1138  * intercept O_DIRECT reads and writes early, this function should
1139  * never get called.
1140  */
1141 static ssize_t ceph_direct_io(int rw, struct kiocb *iocb,
1142                               const struct iovec *iov,
1143                               loff_t pos, unsigned long nr_segs)
1144 {
1145         WARN_ON(1);
1146         return -EINVAL;
1147 }
1148
1149 const struct address_space_operations ceph_aops = {
1150         .readpage = ceph_readpage,
1151         .readpages = ceph_readpages,
1152         .writepage = ceph_writepage,
1153         .writepages = ceph_writepages_start,
1154         .write_begin = ceph_write_begin,
1155         .write_end = ceph_write_end,
1156         .set_page_dirty = ceph_set_page_dirty,
1157         .invalidatepage = ceph_invalidatepage,
1158         .releasepage = ceph_releasepage,
1159         .direct_IO = ceph_direct_io,
1160 };
1161
1162
1163 /*
1164  * vm ops
1165  */
1166
1167 /*
1168  * Reuse write_begin here for simplicity.
1169  */
1170 static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
1171 {
1172         struct inode *inode = file_inode(vma->vm_file);
1173         struct page *page = vmf->page;
1174         struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
1175         loff_t off = page_offset(page);
1176         loff_t size, len;
1177         int ret;
1178
1179         /* Update time before taking page lock */
1180         file_update_time(vma->vm_file);
1181
1182         size = i_size_read(inode);
1183         if (off + PAGE_CACHE_SIZE <= size)
1184                 len = PAGE_CACHE_SIZE;
1185         else
1186                 len = size & ~PAGE_CACHE_MASK;
1187
1188         dout("page_mkwrite %p %llu~%llu page %p idx %lu\n", inode,
1189              off, len, page, page->index);
1190
1191         lock_page(page);
1192
1193         ret = VM_FAULT_NOPAGE;
1194         if ((off > size) ||
1195             (page->mapping != inode->i_mapping))
1196                 goto out;
1197
1198         ret = ceph_update_writeable_page(vma->vm_file, off, len, page);
1199         if (ret == 0) {
1200                 /* success.  we'll keep the page locked. */
1201                 set_page_dirty(page);
1202                 up_read(&mdsc->snap_rwsem);
1203                 ret = VM_FAULT_LOCKED;
1204         } else {
1205                 if (ret == -ENOMEM)
1206                         ret = VM_FAULT_OOM;
1207                 else
1208                         ret = VM_FAULT_SIGBUS;
1209         }
1210 out:
1211         dout("page_mkwrite %p %llu~%llu = %d\n", inode, off, len, ret);
1212         if (ret != VM_FAULT_LOCKED)
1213                 unlock_page(page);
1214         return ret;
1215 }
1216
1217 static struct vm_operations_struct ceph_vmops = {
1218         .fault          = filemap_fault,
1219         .page_mkwrite   = ceph_page_mkwrite,
1220         .remap_pages    = generic_file_remap_pages,
1221 };
1222
1223 int ceph_mmap(struct file *file, struct vm_area_struct *vma)
1224 {
1225         struct address_space *mapping = file->f_mapping;
1226
1227         if (!mapping->a_ops->readpage)
1228                 return -ENOEXEC;
1229         file_accessed(file);
1230         vma->vm_ops = &ceph_vmops;
1231         return 0;
1232 }