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 "cache.h"
  15 #include <linux/ceph/osd_client.h>
  16
  17 /*
  18  * Ceph address space ops.
  19  *
  20  * There are a few funny things going on here.
  21  *
  22  * The page->private field is used to reference a struct
  23  * ceph_snap_context for _every_ dirty page.  This indicates which
  24  * snapshot the page was logically dirtied in, and thus which snap
  25  * context needs to be associated with the osd write during writeback.
  26  *
  27  * Similarly, struct ceph_inode_info maintains a set of counters to
  28  * count dirty pages on the inode.  In the absence of snapshots,
  29  * i_wrbuffer_ref == i_wrbuffer_ref_head == the dirty page count.
  30  *
  31  * When a snapshot is taken (that is, when the client receives
  32  * notification that a snapshot was taken), each inode with caps and
  33  * with dirty pages (dirty pages implies there is a cap) gets a new
  34  * ceph_cap_snap in the i_cap_snaps list (which is sorted in ascending
  35  * order, new snaps go to the tail).  The i_wrbuffer_ref_head count is
  36  * moved to capsnap->dirty. (Unless a sync write is currently in
  37  * progress.  In that case, the capsnap is said to be "pending", new
  38  * writes cannot start, and the capsnap isn't "finalized" until the
  39  * write completes (or fails) and a final size/mtime for the inode for
  40  * that snap can be settled upon.)  i_wrbuffer_ref_head is reset to 0.
  41  *
  42  * On writeback, we must submit writes to the osd IN SNAP ORDER.  So,
  43  * we look for the first capsnap in i_cap_snaps and write out pages in
  44  * that snap context _only_.  Then we move on to the next capsnap,
  45  * eventually reaching the "live" or "head" context (i.e., pages that
  46  * are not yet snapped) and are writing the most recently dirtied
  47  * pages.
  48  *
  49  * Invalidate and so forth must take care to ensure the dirty page
  50  * accounting is preserved.
  51  */
  52
  53 #define CONGESTION_ON_THRESH(congestion_kb) (congestion_kb >> (PAGE_SHIFT-10))
  54 #define CONGESTION_OFF_THRESH(congestion_kb)                            \
  55         (CONGESTION_ON_THRESH(congestion_kb) -                          \
  56          (CONGESTION_ON_THRESH(congestion_kb) >> 2))
  57
  58 static inline struct ceph_snap_context *page_snap_context(struct page *page)
  59 {
  60         if (PagePrivate(page))
  61                 return (void *)page->private;
  62         return NULL;
  63 }
  64
  65 /*
  66  * Dirty a page.  Optimistically adjust accounting, on the assumption
  67  * that we won't race with invalidate.  If we do, readjust.
  68  */
  69 static int ceph_set_page_dirty(struct page *page)
  70 {
  71         struct address_space *mapping = page->mapping;
  72         struct inode *inode;
  73         struct ceph_inode_info *ci;
  74         struct ceph_snap_context *snapc;
  75         int ret;
  76
  77         if (unlikely(!mapping))
  78                 return !TestSetPageDirty(page);
  79
  80         if (PageDirty(page)) {
  81                 dout("%p set_page_dirty %p idx %lu -- already dirty\n",
  82                      mapping->host, page, page->index);
  83                 BUG_ON(!PagePrivate(page));
  84                 return 0;
  85         }
  86
  87         inode = mapping->host;
  88         ci = ceph_inode(inode);
  89
  90         /*
  91          * Note that we're grabbing a snapc ref here without holding
  92          * any locks!
  93          */
  94         snapc = ceph_get_snap_context(ci->i_snap_realm->cached_context);
  95
  96         /* dirty the head */
  97         spin_lock(&ci->i_ceph_lock);
  98         if (ci->i_head_snapc == NULL)
  99                 ci->i_head_snapc = ceph_get_snap_context(snapc);
 100         ++ci->i_wrbuffer_ref_head;
 101         if (ci->i_wrbuffer_ref == 0)
 102                 ihold(inode);
 103         ++ci->i_wrbuffer_ref;
 104         dout("%p set_page_dirty %p idx %lu head %d/%d -> %d/%d "
 105              "snapc %p seq %lld (%d snaps)\n",
 106              mapping->host, page, page->index,
 107              ci->i_wrbuffer_ref-1, ci->i_wrbuffer_ref_head-1,
 108              ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
 109              snapc, snapc->seq, snapc->num_snaps);
 110         spin_unlock(&ci->i_ceph_lock);
 111
 112         /*
 113          * Reference snap context in page->private.  Also set
 114          * PagePrivate so that we get invalidatepage callback.
 115          */
 116         BUG_ON(PagePrivate(page));
 117         page->private = (unsigned long)snapc;
 118         SetPagePrivate(page);
 119
 120         ret = __set_page_dirty_nobuffers(page);
 121         WARN_ON(!PageLocked(page));
 122         WARN_ON(!page->mapping);
 123
 124         return ret;
 125 }
 126
 127 /*
 128  * If we are truncating the full page (i.e. offset == 0), adjust the
 129  * dirty page counters appropriately.  Only called if there is private
 130  * data on the page.
 131  */
 132 static void ceph_invalidatepage(struct page *page, unsigned int offset,
 133                                 unsigned int length)
 134 {
 135         struct inode *inode;
 136         struct ceph_inode_info *ci;
 137         struct ceph_snap_context *snapc = page_snap_context(page);
 138
 139         inode = page->mapping->host;
 140         ci = ceph_inode(inode);
 141
 142         if (offset != 0 || length != PAGE_CACHE_SIZE) {
 143                 dout("%p invalidatepage %p idx %lu partial dirty page %u~%u\n",
 144                      inode, page, page->index, offset, length);
 145                 return;
 146         }
 147
 148         ceph_invalidate_fscache_page(inode, page);
 149
 150         if (!PagePrivate(page))
 151                 return;
 152
 153         /*
 154          * We can get non-dirty pages here due to races between
 155          * set_page_dirty and truncate_complete_page; just spit out a
 156          * warning, in case we end up with accounting problems later.
 157          */
 158         if (!PageDirty(page))
 159                 pr_err("%p invalidatepage %p page not dirty\n", inode, page);
 160
 161         ClearPageChecked(page);
 162
 163         dout("%p invalidatepage %p idx %lu full dirty page\n",
 164              inode, page, page->index);
 165
 166         ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
 167         ceph_put_snap_context(snapc);
 168         page->private = 0;
 169         ClearPagePrivate(page);
 170 }
 171
 172 static int ceph_releasepage(struct page *page, gfp_t g)
 173 {
 174         struct inode *inode = page->mapping ? page->mapping->host : NULL;
 175         dout("%p releasepage %p idx %lu\n", inode, page, page->index);
 176         WARN_ON(PageDirty(page));
 177
 178         /* Can we release the page from the cache? */
 179         if (!ceph_release_fscache_page(page, g))
 180                 return 0;
 181
 182         return !PagePrivate(page);
 183 }
 184
 185 /*
 186  * read a single page, without unlocking it.
 187  */
 188 static int readpage_nounlock(struct file *filp, struct page *page)
 189 {
 190         struct inode *inode = file_inode(filp);
 191         struct ceph_inode_info *ci = ceph_inode(inode);
 192         struct ceph_osd_client *osdc =
 193                 &ceph_inode_to_client(inode)->client->osdc;
 194         int err = 0;
 195         u64 len = PAGE_CACHE_SIZE;
 196
 197         err = ceph_readpage_from_fscache(inode, page);
 198
 199         if (err == 0)
 200                 goto out;
 201
 202         dout("readpage inode %p file %p page %p index %lu\n",
 203              inode, filp, page, page->index);
 204         err = ceph_osdc_readpages(osdc, ceph_vino(inode), &ci->i_layout,
 205                                   (u64) page_offset(page), &len,
 206                                   ci->i_truncate_seq, ci->i_truncate_size,
 207                                   &page, 1, 0);
 208         if (err == -ENOENT)
 209                 err = 0;
 210         if (err < 0) {
 211                 SetPageError(page);
 212                 goto out;
 213         } else {
 214                 if (err < PAGE_CACHE_SIZE) {
 215                 /* zero fill remainder of page */
 216                         zero_user_segment(page, err, PAGE_CACHE_SIZE);
 217                 } else {
 218                         flush_dcache_page(page);
 219                 }
 220         }
 221         SetPageUptodate(page);
 222
 223         if (err >= 0)
 224                 ceph_readpage_to_fscache(inode, page);
 225
 226 out:
 227         return err < 0 ? err : 0;
 228 }
 229
 230 static int ceph_readpage(struct file *filp, struct page *page)
 231 {
 232         int r = readpage_nounlock(filp, page);
 233         unlock_page(page);
 234         return r;
 235 }
 236
 237 /*
 238  * Finish an async read(ahead) op.
 239  */
 240 static void finish_read(struct ceph_osd_request *req, struct ceph_msg *msg)
 241 {
 242         struct inode *inode = req->r_inode;
 243         struct ceph_osd_data *osd_data;
 244         int rc = req->r_result;
 245         int bytes = le32_to_cpu(msg->hdr.data_len);
 246         int num_pages;
 247         int i;
 248
 249         dout("finish_read %p req %p rc %d bytes %d\n", inode, req, rc, bytes);
 250
 251         /* unlock all pages, zeroing any data we didn't read */
 252         osd_data = osd_req_op_extent_osd_data(req, 0);
 253         BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
 254         num_pages = calc_pages_for((u64)osd_data->alignment,
 255                                         (u64)osd_data->length);
 256         for (i = 0; i < num_pages; i++) {
 257                 struct page *page = osd_data->pages[i];
 258
 259                 if (bytes < (int)PAGE_CACHE_SIZE) {
 260                         /* zero (remainder of) page */
 261                         int s = bytes < 0 ? 0 : bytes;
 262                         zero_user_segment(page, s, PAGE_CACHE_SIZE);
 263                 }
 264                 dout("finish_read %p uptodate %p idx %lu\n", inode, page,
 265                      page->index);
 266                 flush_dcache_page(page);
 267                 SetPageUptodate(page);
 268                 ceph_readpage_to_fscache(inode, page);
 269                 unlock_page(page);
 270                 page_cache_release(page);
 271                 bytes -= PAGE_CACHE_SIZE;
 272         }
 273         kfree(osd_data->pages);
 274 }
 275
 276 static void ceph_unlock_page_vector(struct page **pages, int num_pages)
 277 {
 278         int i;
 279
 280         for (i = 0; i < num_pages; i++)
 281                 unlock_page(pages[i]);
 282 }
 283
 284 /*
 285  * start an async read(ahead) operation.  return nr_pages we submitted
 286  * a read for on success, or negative error code.
 287  */
 288 static int start_read(struct inode *inode, struct list_head *page_list, int max)
 289 {
 290         struct ceph_osd_client *osdc =
 291                 &ceph_inode_to_client(inode)->client->osdc;
 292         struct ceph_inode_info *ci = ceph_inode(inode);
 293         struct page *page = list_entry(page_list->prev, struct page, lru);
 294         struct ceph_vino vino;
 295         struct ceph_osd_request *req;
 296         u64 off;
 297         u64 len;
 298         int i;
 299         struct page **pages;
 300         pgoff_t next_index;
 301         int nr_pages = 0;
 302         int ret;
 303
 304         off = (u64) page_offset(page);
 305
 306         /* count pages */
 307         next_index = page->index;
 308         list_for_each_entry_reverse(page, page_list, lru) {
 309                 if (page->index != next_index)
 310                         break;
 311                 nr_pages++;
 312                 next_index++;
 313                 if (max && nr_pages == max)
 314                         break;
 315         }
 316         len = nr_pages << PAGE_CACHE_SHIFT;
 317         dout("start_read %p nr_pages %d is %lld~%lld\n", inode, nr_pages,
 318              off, len);
 319         vino = ceph_vino(inode);
 320         req = ceph_osdc_new_request(osdc, &ci->i_layout, vino, off, &len,
 321                                     1, CEPH_OSD_OP_READ,
 322                                     CEPH_OSD_FLAG_READ, NULL,
 323                                     ci->i_truncate_seq, ci->i_truncate_size,
 324                                     false);
 325         if (IS_ERR(req))
 326                 return PTR_ERR(req);
 327
 328         /* build page vector */
 329         nr_pages = calc_pages_for(0, len);
 330         pages = kmalloc(sizeof(*pages) * nr_pages, GFP_NOFS);
 331         ret = -ENOMEM;
 332         if (!pages)
 333                 goto out;
 334         for (i = 0; i < nr_pages; ++i) {
 335                 page = list_entry(page_list->prev, struct page, lru);
 336                 BUG_ON(PageLocked(page));
 337                 list_del(&page->lru);
 338
 339                 dout("start_read %p adding %p idx %lu\n", inode, page,
 340                      page->index);
 341                 if (add_to_page_cache_lru(page, &inode->i_data, page->index,
 342                                           GFP_NOFS)) {
 343                         ceph_fscache_uncache_page(inode, page);
 344                         page_cache_release(page);
 345                         dout("start_read %p add_to_page_cache failed %p\n",
 346                              inode, page);
 347                         nr_pages = i;
 348                         goto out_pages;
 349                 }
 350                 pages[i] = page;
 351         }
 352         osd_req_op_extent_osd_data_pages(req, 0, pages, len, 0, false, false);
 353         req->r_callback = finish_read;
 354         req->r_inode = inode;
 355
 356         ceph_osdc_build_request(req, off, NULL, vino.snap, NULL);
 357
 358         dout("start_read %p starting %p %lld~%lld\n", inode, req, off, len);
 359         ret = ceph_osdc_start_request(osdc, req, false);
 360         if (ret < 0)
 361                 goto out_pages;
 362         ceph_osdc_put_request(req);
 363         return nr_pages;
 364
 365 out_pages:
 366         ceph_unlock_page_vector(pages, nr_pages);
 367         ceph_release_page_vector(pages, nr_pages);
 368 out:
 369         ceph_osdc_put_request(req);
 370         return ret;
 371 }
 372
 373
 374 /*
 375  * Read multiple pages.  Leave pages we don't read + unlock in page_list;
 376  * the caller (VM) cleans them up.
 377  */
 378 static int ceph_readpages(struct file *file, struct address_space *mapping,
 379                           struct list_head *page_list, unsigned nr_pages)
 380 {
 381         struct inode *inode = file_inode(file);
 382         struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 383         int rc = 0;
 384         int max = 0;
 385
 386         rc = ceph_readpages_from_fscache(mapping->host, mapping, page_list,
 387                                          &nr_pages);
 388
 389         if (rc == 0)
 390                 goto out;
 391
 392         if (fsc->mount_options->rsize >= PAGE_CACHE_SIZE)
 393                 max = (fsc->mount_options->rsize + PAGE_CACHE_SIZE - 1)
 394                         >> PAGE_SHIFT;
 395
 396         dout("readpages %p file %p nr_pages %d max %d\n", inode,
 397                 file, nr_pages,
 398              max);
 399         while (!list_empty(page_list)) {
 400                 rc = start_read(inode, page_list, max);
 401                 if (rc < 0)
 402                         goto out;
 403                 BUG_ON(rc == 0);
 404         }
 405 out:
 406         ceph_fscache_readpages_cancel(inode, page_list);
 407
 408         dout("readpages %p file %p ret %d\n", inode, file, rc);
 409         return rc;
 410 }
 411
 412 /*
 413  * Get ref for the oldest snapc for an inode with dirty data... that is, the
 414  * only snap context we are allowed to write back.
 415  */
 416 static struct ceph_snap_context *get_oldest_context(struct inode *inode,
 417                                                     u64 *snap_size)
 418 {
 419         struct ceph_inode_info *ci = ceph_inode(inode);
 420         struct ceph_snap_context *snapc = NULL;
 421         struct ceph_cap_snap *capsnap = NULL;
 422
 423         spin_lock(&ci->i_ceph_lock);
 424         list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
 425                 dout(" cap_snap %p snapc %p has %d dirty pages\n", capsnap,
 426                      capsnap->context, capsnap->dirty_pages);
 427                 if (capsnap->dirty_pages) {
 428                         snapc = ceph_get_snap_context(capsnap->context);
 429                         if (snap_size)
 430                                 *snap_size = capsnap->size;
 431                         break;
 432                 }
 433         }
 434         if (!snapc && ci->i_wrbuffer_ref_head) {
 435                 snapc = ceph_get_snap_context(ci->i_head_snapc);
 436                 dout(" head snapc %p has %d dirty pages\n",
 437                      snapc, ci->i_wrbuffer_ref_head);
 438         }
 439         spin_unlock(&ci->i_ceph_lock);
 440         return snapc;
 441 }
 442
 443 /*
 444  * Write a single page, but leave the page locked.
 445  *
 446  * If we get a write error, set the page error bit, but still adjust the
 447  * dirty page accounting (i.e., page is no longer dirty).
 448  */
 449 static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
 450 {
 451         struct inode *inode;
 452         struct ceph_inode_info *ci;
 453         struct ceph_fs_client *fsc;
 454         struct ceph_osd_client *osdc;
 455         struct ceph_snap_context *snapc, *oldest;
 456         loff_t page_off = page_offset(page);
 457         long writeback_stat;
 458         u64 truncate_size, snap_size = 0;
 459         u32 truncate_seq;
 460         int err = 0, len = PAGE_CACHE_SIZE;
 461
 462         dout("writepage %p idx %lu\n", page, page->index);
 463
 464         if (!page->mapping || !page->mapping->host) {
 465                 dout("writepage %p - no mapping\n", page);
 466                 return -EFAULT;
 467         }
 468         inode = page->mapping->host;
 469         ci = ceph_inode(inode);
 470         fsc = ceph_inode_to_client(inode);
 471         osdc = &fsc->client->osdc;
 472
 473         /* verify this is a writeable snap context */
 474         snapc = page_snap_context(page);
 475         if (snapc == NULL) {
 476                 dout("writepage %p page %p not dirty?\n", inode, page);
 477                 goto out;
 478         }
 479         oldest = get_oldest_context(inode, &snap_size);
 480         if (snapc->seq > oldest->seq) {
 481                 dout("writepage %p page %p snapc %p not writeable - noop\n",
 482                      inode, page, snapc);
 483                 /* we should only noop if called by kswapd */
 484                 WARN_ON((current->flags & PF_MEMALLOC) == 0);
 485                 ceph_put_snap_context(oldest);
 486                 goto out;
 487         }
 488         ceph_put_snap_context(oldest);
 489
 490         spin_lock(&ci->i_ceph_lock);
 491         truncate_seq = ci->i_truncate_seq;
 492         truncate_size = ci->i_truncate_size;
 493         if (!snap_size)
 494                 snap_size = i_size_read(inode);
 495         spin_unlock(&ci->i_ceph_lock);
 496
 497         /* is this a partial page at end of file? */
 498         if (page_off >= snap_size) {
 499                 dout("%p page eof %llu\n", page, snap_size);
 500                 goto out;
 501         }
 502         if (snap_size < page_off + len)
 503                 len = snap_size - page_off;
 504
 505         dout("writepage %p page %p index %lu on %llu~%u snapc %p\n",
 506              inode, page, page->index, page_off, len, snapc);
 507
 508         writeback_stat = atomic_long_inc_return(&fsc->writeback_count);
 509         if (writeback_stat >
 510             CONGESTION_ON_THRESH(fsc->mount_options->congestion_kb))
 511                 set_bdi_congested(&fsc->backing_dev_info, BLK_RW_ASYNC);
 512
 513         ceph_readpage_to_fscache(inode, page);
 514
 515         set_page_writeback(page);
 516         err = ceph_osdc_writepages(osdc, ceph_vino(inode),
 517                                    &ci->i_layout, snapc,
 518                                    page_off, len,
 519                                    truncate_seq, truncate_size,
 520                                    &inode->i_mtime, &page, 1);
 521         if (err < 0) {
 522                 dout("writepage setting page/mapping error %d %p\n", err, page);
 523                 SetPageError(page);
 524                 mapping_set_error(&inode->i_data, err);
 525                 if (wbc)
 526                         wbc->pages_skipped++;
 527         } else {
 528                 dout("writepage cleaned page %p\n", page);
 529                 err = 0;  /* vfs expects us to return 0 */
 530         }
 531         page->private = 0;
 532         ClearPagePrivate(page);
 533         end_page_writeback(page);
 534         ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
 535         ceph_put_snap_context(snapc);  /* page's reference */
 536 out:
 537         return err;
 538 }
 539
 540 static int ceph_writepage(struct page *page, struct writeback_control *wbc)
 541 {
 542         int err;
 543         struct inode *inode = page->mapping->host;
 544         BUG_ON(!inode);
 545         ihold(inode);
 546         err = writepage_nounlock(page, wbc);
 547         unlock_page(page);
 548         iput(inode);
 549         return err;
 550 }
 551
 552
 553 /*
 554  * lame release_pages helper.  release_pages() isn't exported to
 555  * modules.
 556  */
 557 static void ceph_release_pages(struct page **pages, int num)
 558 {
 559         struct pagevec pvec;
 560         int i;
 561
 562         pagevec_init(&pvec, 0);
 563         for (i = 0; i < num; i++) {
 564                 if (pagevec_add(&pvec, pages[i]) == 0)
 565                         pagevec_release(&pvec);
 566         }
 567         pagevec_release(&pvec);
 568 }
 569
 570 /*
 571  * async writeback completion handler.
 572  *
 573  * If we get an error, set the mapping error bit, but not the individual
 574  * page error bits.
 575  */
 576 static void writepages_finish(struct ceph_osd_request *req,
 577                               struct ceph_msg *msg)
 578 {
 579         struct inode *inode = req->r_inode;
 580         struct ceph_inode_info *ci = ceph_inode(inode);
 581         struct ceph_osd_data *osd_data;
 582         unsigned wrote;
 583         struct page *page;
 584         int num_pages;
 585         int i;
 586         struct ceph_snap_context *snapc = req->r_snapc;
 587         struct address_space *mapping = inode->i_mapping;
 588         int rc = req->r_result;
 589         u64 bytes = req->r_ops[0].extent.length;
 590         struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 591         long writeback_stat;
 592         unsigned issued = ceph_caps_issued(ci);
 593
 594         osd_data = osd_req_op_extent_osd_data(req, 0);
 595         BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
 596         num_pages = calc_pages_for((u64)osd_data->alignment,
 597                                         (u64)osd_data->length);
 598         if (rc >= 0) {
 599                 /*
 600                  * Assume we wrote the pages we originally sent.  The
 601                  * osd might reply with fewer pages if our writeback
 602                  * raced with a truncation and was adjusted at the osd,
 603                  * so don't believe the reply.
 604                  */
 605                 wrote = num_pages;
 606         } else {
 607                 wrote = 0;
 608                 mapping_set_error(mapping, rc);
 609         }
 610         dout("writepages_finish %p rc %d bytes %llu wrote %d (pages)\n",
 611              inode, rc, bytes, wrote);
 612
 613         /* clean all pages */
 614         for (i = 0; i < num_pages; i++) {
 615                 page = osd_data->pages[i];
 616                 BUG_ON(!page);
 617                 WARN_ON(!PageUptodate(page));
 618
 619                 writeback_stat =
 620                         atomic_long_dec_return(&fsc->writeback_count);
 621                 if (writeback_stat <
 622                     CONGESTION_OFF_THRESH(fsc->mount_options->congestion_kb))
 623                         clear_bdi_congested(&fsc->backing_dev_info,
 624                                             BLK_RW_ASYNC);
 625
 626                 ceph_put_snap_context(page_snap_context(page));
 627                 page->private = 0;
 628                 ClearPagePrivate(page);
 629                 dout("unlocking %d %p\n", i, page);
 630                 end_page_writeback(page);
 631
 632                 /*
 633                  * We lost the cache cap, need to truncate the page before
 634                  * it is unlocked, otherwise we'd truncate it later in the
 635                  * page truncation thread, possibly losing some data that
 636                  * raced its way in
 637                  */
 638                 if ((issued & (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) == 0)
 639                         generic_error_remove_page(inode->i_mapping, page);
 640
 641                 unlock_page(page);
 642         }
 643         dout("%p wrote+cleaned %d pages\n", inode, wrote);
 644         ceph_put_wrbuffer_cap_refs(ci, num_pages, snapc);
 645
 646         ceph_release_pages(osd_data->pages, num_pages);
 647         if (osd_data->pages_from_pool)
 648                 mempool_free(osd_data->pages,
 649                              ceph_sb_to_client(inode->i_sb)->wb_pagevec_pool);
 650         else
 651                 kfree(osd_data->pages);
 652         ceph_osdc_put_request(req);
 653 }
 654
 655 /*
 656  * initiate async writeback
 657  */
 658 static int ceph_writepages_start(struct address_space *mapping,
 659                                  struct writeback_control *wbc)
 660 {
 661         struct inode *inode = mapping->host;
 662         struct ceph_inode_info *ci = ceph_inode(inode);
 663         struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
 664         struct ceph_vino vino = ceph_vino(inode);
 665         pgoff_t index, start, end;
 666         int range_whole = 0;
 667         int should_loop = 1;
 668         pgoff_t max_pages = 0, max_pages_ever = 0;
 669         struct ceph_snap_context *snapc = NULL, *last_snapc = NULL, *pgsnapc;
 670         struct pagevec pvec;
 671         int done = 0;
 672         int rc = 0;
 673         unsigned wsize = 1 << inode->i_blkbits;
 674         struct ceph_osd_request *req = NULL;
 675         int do_sync;
 676         u64 truncate_size, snap_size;
 677         u32 truncate_seq;
 678
 679         /*
 680          * Include a 'sync' in the OSD request if this is a data
 681          * integrity write (e.g., O_SYNC write or fsync()), or if our
 682          * cap is being revoked.
 683          */
 684         if ((wbc->sync_mode == WB_SYNC_ALL) ||
 685                 ceph_caps_revoking(ci, CEPH_CAP_FILE_BUFFER))
 686                 do_sync = 1;
 687         dout("writepages_start %p dosync=%d (mode=%s)\n",
 688              inode, do_sync,
 689              wbc->sync_mode == WB_SYNC_NONE ? "NONE" :
 690              (wbc->sync_mode == WB_SYNC_ALL ? "ALL" : "HOLD"));
 691
 692         if (fsc->mount_state == CEPH_MOUNT_SHUTDOWN) {
 693                 pr_warning("writepage_start %p on forced umount\n", inode);
 694                 return -EIO; /* we're in a forced umount, don't write! */
 695         }
 696         if (fsc->mount_options->wsize && fsc->mount_options->wsize < wsize)
 697                 wsize = fsc->mount_options->wsize;
 698         if (wsize < PAGE_CACHE_SIZE)
 699                 wsize = PAGE_CACHE_SIZE;
 700         max_pages_ever = wsize >> PAGE_CACHE_SHIFT;
 701
 702         pagevec_init(&pvec, 0);
 703
 704         /* where to start/end? */
 705         if (wbc->range_cyclic) {
 706                 start = mapping->writeback_index; /* Start from prev offset */
 707                 end = -1;
 708                 dout(" cyclic, start at %lu\n", start);
 709         } else {
 710                 start = wbc->range_start >> PAGE_CACHE_SHIFT;
 711                 end = wbc->range_end >> PAGE_CACHE_SHIFT;
 712                 if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
 713                         range_whole = 1;
 714                 should_loop = 0;
 715                 dout(" not cyclic, %lu to %lu\n", start, end);
 716         }
 717         index = start;
 718
 719 retry:
 720         /* find oldest snap context with dirty data */
 721         ceph_put_snap_context(snapc);
 722         snap_size = 0;
 723         snapc = get_oldest_context(inode, &snap_size);
 724         if (!snapc) {
 725                 /* hmm, why does writepages get called when there
 726                    is no dirty data? */
 727                 dout(" no snap context with dirty data?\n");
 728                 goto out;
 729         }
 730         if (snap_size == 0)
 731                 snap_size = i_size_read(inode);
 732         dout(" oldest snapc is %p seq %lld (%d snaps)\n",
 733              snapc, snapc->seq, snapc->num_snaps);
 734
 735         spin_lock(&ci->i_ceph_lock);
 736         truncate_seq = ci->i_truncate_seq;
 737         truncate_size = ci->i_truncate_size;
 738         if (!snap_size)
 739                 snap_size = i_size_read(inode);
 740         spin_unlock(&ci->i_ceph_lock);
 741
 742         if (last_snapc && snapc != last_snapc) {
 743                 /* if we switched to a newer snapc, restart our scan at the
 744                  * start of the original file range. */
 745                 dout("  snapc differs from last pass, restarting at %lu\n",
 746                      index);
 747                 index = start;
 748         }
 749         last_snapc = snapc;
 750
 751         while (!done && index <= end) {
 752                 int num_ops = do_sync ? 2 : 1;
 753                 unsigned i;
 754                 int first;
 755                 pgoff_t next;
 756                 int pvec_pages, locked_pages;
 757                 struct page **pages = NULL;
 758                 mempool_t *pool = NULL; /* Becomes non-null if mempool used */
 759                 struct page *page;
 760                 int want;
 761                 u64 offset, len;
 762                 long writeback_stat;
 763
 764                 next = 0;
 765                 locked_pages = 0;
 766                 max_pages = max_pages_ever;
 767
 768 get_more_pages:
 769                 first = -1;
 770                 want = min(end - index,
 771                            min((pgoff_t)PAGEVEC_SIZE,
 772                                max_pages - (pgoff_t)locked_pages) - 1)
 773                         + 1;
 774                 pvec_pages = pagevec_lookup_tag(&pvec, mapping, &index,
 775                                                 PAGECACHE_TAG_DIRTY,
 776                                                 want);
 777                 dout("pagevec_lookup_tag got %d\n", pvec_pages);
 778                 if (!pvec_pages && !locked_pages)
 779                         break;
 780                 for (i = 0; i < pvec_pages && locked_pages < max_pages; i++) {
 781                         page = pvec.pages[i];
 782                         dout("? %p idx %lu\n", page, page->index);
 783                         if (locked_pages == 0)
 784                                 lock_page(page);  /* first page */
 785                         else if (!trylock_page(page))
 786                                 break;
 787
 788                         /* only dirty pages, or our accounting breaks */
 789                         if (unlikely(!PageDirty(page)) ||
 790                             unlikely(page->mapping != mapping)) {
 791                                 dout("!dirty or !mapping %p\n", page);
 792                                 unlock_page(page);
 793                                 break;
 794                         }
 795                         if (!wbc->range_cyclic && page->index > end) {
 796                                 dout("end of range %p\n", page);
 797                                 done = 1;
 798                                 unlock_page(page);
 799                                 break;
 800                         }
 801                         if (next && (page->index != next)) {
 802                                 dout("not consecutive %p\n", page);
 803                                 unlock_page(page);
 804                                 break;
 805                         }
 806                         if (wbc->sync_mode != WB_SYNC_NONE) {
 807                                 dout("waiting on writeback %p\n", page);
 808                                 wait_on_page_writeback(page);
 809                         }
 810                         if (page_offset(page) >= snap_size) {
 811                                 dout("%p page eof %llu\n", page, snap_size);
 812                                 done = 1;
 813                                 unlock_page(page);
 814                                 break;
 815                         }
 816                         if (PageWriteback(page)) {
 817                                 dout("%p under writeback\n", page);
 818                                 unlock_page(page);
 819                                 break;
 820                         }
 821
 822                         /* only if matching snap context */
 823                         pgsnapc = page_snap_context(page);
 824                         if (pgsnapc->seq > snapc->seq) {
 825                                 dout("page snapc %p %lld > oldest %p %lld\n",
 826                                      pgsnapc, pgsnapc->seq, snapc, snapc->seq);
 827                                 unlock_page(page);
 828                                 if (!locked_pages)
 829                                         continue; /* keep looking for snap */
 830                                 break;
 831                         }
 832
 833                         if (!clear_page_dirty_for_io(page)) {
 834                                 dout("%p !clear_page_dirty_for_io\n", page);
 835                                 unlock_page(page);
 836                                 break;
 837                         }
 838
 839                         /*
 840                          * We have something to write.  If this is
 841                          * the first locked page this time through,
 842                          * allocate an osd request and a page array
 843                          * that it will use.
 844                          */
 845                         if (locked_pages == 0) {
 846                                 BUG_ON(pages);
 847                                 /* prepare async write request */
 848                                 offset = (u64)page_offset(page);
 849                                 len = wsize;
 850                                 req = ceph_osdc_new_request(&fsc->client->osdc,
 851                                                         &ci->i_layout, vino,
 852                                                         offset, &len, num_ops,
 853                                                         CEPH_OSD_OP_WRITE,
 854                                                         CEPH_OSD_FLAG_WRITE |
 855                                                         CEPH_OSD_FLAG_ONDISK,
 856                                                         snapc, truncate_seq,
 857                                                         truncate_size, true);
 858                                 if (IS_ERR(req)) {
 859                                         rc = PTR_ERR(req);
 860                                         unlock_page(page);
 861                                         break;
 862                                 }
 863
 864                                 req->r_callback = writepages_finish;
 865                                 req->r_inode = inode;
 866
 867                                 max_pages = calc_pages_for(0, (u64)len);
 868                                 pages = kmalloc(max_pages * sizeof (*pages),
 869                                                 GFP_NOFS);
 870                                 if (!pages) {
 871                                         pool = fsc->wb_pagevec_pool;
 872                                         pages = mempool_alloc(pool, GFP_NOFS);
 873                                         BUG_ON(!pages);
 874                                 }
 875                         }
 876
 877                         /* note position of first page in pvec */
 878                         if (first < 0)
 879                                 first = i;
 880                         dout("%p will write page %p idx %lu\n",
 881                              inode, page, page->index);
 882
 883                         writeback_stat =
 884                                atomic_long_inc_return(&fsc->writeback_count);
 885                         if (writeback_stat > CONGESTION_ON_THRESH(
 886                                     fsc->mount_options->congestion_kb)) {
 887                                 set_bdi_congested(&fsc->backing_dev_info,
 888                                                   BLK_RW_ASYNC);
 889                         }
 890
 891                         set_page_writeback(page);
 892                         pages[locked_pages] = page;
 893                         locked_pages++;
 894                         next = page->index + 1;
 895                 }
 896
 897                 /* did we get anything? */
 898                 if (!locked_pages)
 899                         goto release_pvec_pages;
 900                 if (i) {
 901                         int j;
 902                         BUG_ON(!locked_pages || first < 0);
 903
 904                         if (pvec_pages && i == pvec_pages &&
 905                             locked_pages < max_pages) {
 906                                 dout("reached end pvec, trying for more\n");
 907                                 pagevec_reinit(&pvec);
 908                                 goto get_more_pages;
 909                         }
 910
 911                         /* shift unused pages over in the pvec...  we
 912                          * will need to release them below. */
 913                         for (j = i; j < pvec_pages; j++) {
 914                                 dout(" pvec leftover page %p\n",
 915                                      pvec.pages[j]);
 916                                 pvec.pages[j-i+first] = pvec.pages[j];
 917                         }
 918                         pvec.nr -= i-first;
 919                 }
 920
 921                 /* Format the osd request message and submit the write */
 922
 923                 offset = page_offset(pages[0]);
 924                 len = min(snap_size - offset,
 925                           (u64)locked_pages << PAGE_CACHE_SHIFT);
 926                 dout("writepages got %d pages at %llu~%llu\n",
 927                      locked_pages, offset, len);
 928
 929                 osd_req_op_extent_osd_data_pages(req, 0, pages, len, 0,
 930                                                         !!pool, false);
 931
 932                 pages = NULL;   /* request message now owns the pages array */
 933                 pool = NULL;
 934
 935                 /* Update the write op length in case we changed it */
 936
 937                 osd_req_op_extent_update(req, 0, len);
 938
 939                 vino = ceph_vino(inode);
 940                 ceph_osdc_build_request(req, offset, snapc, vino.snap,
 941                                         &inode->i_mtime);
 942
 943                 rc = ceph_osdc_start_request(&fsc->client->osdc, req, true);
 944                 BUG_ON(rc);
 945                 req = NULL;
 946
 947                 /* continue? */
 948                 index = next;
 949                 wbc->nr_to_write -= locked_pages;
 950                 if (wbc->nr_to_write <= 0)
 951                         done = 1;
 952
 953 release_pvec_pages:
 954                 dout("pagevec_release on %d pages (%p)\n", (int)pvec.nr,
 955                      pvec.nr ? pvec.pages[0] : NULL);
 956                 pagevec_release(&pvec);
 957
 958                 if (locked_pages && !done)
 959                         goto retry;
 960         }
 961
 962         if (should_loop && !done) {
 963                 /* more to do; loop back to beginning of file */
 964                 dout("writepages looping back to beginning of file\n");
 965                 should_loop = 0;
 966                 index = 0;
 967                 goto retry;
 968         }
 969
 970         if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
 971                 mapping->writeback_index = index;
 972
 973 out:
 974         if (req)
 975                 ceph_osdc_put_request(req);
 976         ceph_put_snap_context(snapc);
 977         dout("writepages done, rc = %d\n", rc);
 978         return rc;
 979 }
 980
 981
 982
 983 /*
 984  * See if a given @snapc is either writeable, or already written.
 985  */
 986 static int context_is_writeable_or_written(struct inode *inode,
 987                                            struct ceph_snap_context *snapc)
 988 {
 989         struct ceph_snap_context *oldest = get_oldest_context(inode, NULL);
 990         int ret = !oldest || snapc->seq <= oldest->seq;
 991
 992         ceph_put_snap_context(oldest);
 993         return ret;
 994 }
 995
 996 /*
 997  * We are only allowed to write into/dirty the page if the page is
 998  * clean, or already dirty within the same snap context.
 999  *
1000  * called with page locked.
1001  * return success with page locked,
1002  * or any failure (incl -EAGAIN) with page unlocked.
1003  */
1004 static int ceph_update_writeable_page(struct file *file,
1005                             loff_t pos, unsigned len,
1006                             struct page *page)
1007 {
1008         struct inode *inode = file_inode(file);
1009         struct ceph_inode_info *ci = ceph_inode(inode);
1010         struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
1011         loff_t page_off = pos & PAGE_CACHE_MASK;
1012         int pos_in_page = pos & ~PAGE_CACHE_MASK;
1013         int end_in_page = pos_in_page + len;
1014         loff_t i_size;
1015         int r;
1016         struct ceph_snap_context *snapc, *oldest;
1017
1018 retry_locked:
1019         /* writepages currently holds page lock, but if we change that later, */
1020         wait_on_page_writeback(page);
1021
1022         /* check snap context */
1023         BUG_ON(!ci->i_snap_realm);
1024         down_read(&mdsc->snap_rwsem);
1025         BUG_ON(!ci->i_snap_realm->cached_context);
1026         snapc = page_snap_context(page);
1027         if (snapc && snapc != ci->i_head_snapc) {
1028                 /*
1029                  * this page is already dirty in another (older) snap
1030                  * context!  is it writeable now?
1031                  */
1032                 oldest = get_oldest_context(inode, NULL);
1033                 up_read(&mdsc->snap_rwsem);
1034
1035                 if (snapc->seq > oldest->seq) {
1036                         ceph_put_snap_context(oldest);
1037                         dout(" page %p snapc %p not current or oldest\n",
1038                              page, snapc);
1039                         /*
1040                          * queue for writeback, and wait for snapc to
1041                          * be writeable or written
1042                          */
1043                         snapc = ceph_get_snap_context(snapc);
1044                         unlock_page(page);
1045                         ceph_queue_writeback(inode);
1046                         r = wait_event_interruptible(ci->i_cap_wq,
1047                                context_is_writeable_or_written(inode, snapc));
1048                         ceph_put_snap_context(snapc);
1049                         if (r == -ERESTARTSYS)
1050                                 return r;
1051                         return -EAGAIN;
1052                 }
1053                 ceph_put_snap_context(oldest);
1054
1055                 /* yay, writeable, do it now (without dropping page lock) */
1056                 dout(" page %p snapc %p not current, but oldest\n",
1057                      page, snapc);
1058                 if (!clear_page_dirty_for_io(page))
1059                         goto retry_locked;
1060                 r = writepage_nounlock(page, NULL);
1061                 if (r < 0)
1062                         goto fail_nosnap;
1063                 goto retry_locked;
1064         }
1065
1066         if (PageUptodate(page)) {
1067                 dout(" page %p already uptodate\n", page);
1068                 return 0;
1069         }
1070
1071         /* full page? */
1072         if (pos_in_page == 0 && len == PAGE_CACHE_SIZE)
1073                 return 0;
1074
1075         /* past end of file? */
1076         i_size = inode->i_size;   /* caller holds i_mutex */
1077
1078         if (i_size + len > inode->i_sb->s_maxbytes) {
1079                 /* file is too big */
1080                 r = -EINVAL;
1081                 goto fail;
1082         }
1083
1084         if (page_off >= i_size ||
1085             (pos_in_page == 0 && (pos+len) >= i_size &&
1086              end_in_page - pos_in_page != PAGE_CACHE_SIZE)) {
1087                 dout(" zeroing %p 0 - %d and %d - %d\n",
1088                      page, pos_in_page, end_in_page, (int)PAGE_CACHE_SIZE);
1089                 zero_user_segments(page,
1090                                    0, pos_in_page,
1091                                    end_in_page, PAGE_CACHE_SIZE);
1092                 return 0;
1093         }
1094
1095         /* we need to read it. */
1096         up_read(&mdsc->snap_rwsem);
1097         r = readpage_nounlock(file, page);
1098         if (r < 0)
1099                 goto fail_nosnap;
1100         goto retry_locked;
1101
1102 fail:
1103         up_read(&mdsc->snap_rwsem);
1104 fail_nosnap:
1105         unlock_page(page);
1106         return r;
1107 }
1108
1109 /*
1110  * We are only allowed to write into/dirty the page if the page is
1111  * clean, or already dirty within the same snap context.
1112  */
1113 static int ceph_write_begin(struct file *file, struct address_space *mapping,
1114                             loff_t pos, unsigned len, unsigned flags,
1115                             struct page **pagep, void **fsdata)
1116 {
1117         struct inode *inode = file_inode(file);
1118         struct page *page;
1119         pgoff_t index = pos >> PAGE_CACHE_SHIFT;
1120         int r;
1121
1122         do {
1123                 /* get a page */
1124                 page = grab_cache_page_write_begin(mapping, index, 0);
1125                 if (!page)
1126                         return -ENOMEM;
1127                 *pagep = page;
1128
1129                 dout("write_begin file %p inode %p page %p %d~%d\n", file,
1130                      inode, page, (int)pos, (int)len);
1131
1132                 r = ceph_update_writeable_page(file, pos, len, page);
1133         } while (r == -EAGAIN);
1134
1135         return r;
1136 }
1137
1138 /*
1139  * we don't do anything in here that simple_write_end doesn't do
1140  * except adjust dirty page accounting and drop read lock on
1141  * mdsc->snap_rwsem.
1142  */
1143 static int ceph_write_end(struct file *file, struct address_space *mapping,
1144                           loff_t pos, unsigned len, unsigned copied,
1145                           struct page *page, void *fsdata)
1146 {
1147         struct inode *inode = file_inode(file);
1148         struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1149         struct ceph_mds_client *mdsc = fsc->mdsc;
1150         unsigned from = pos & (PAGE_CACHE_SIZE - 1);
1151         int check_cap = 0;
1152
1153         dout("write_end file %p inode %p page %p %d~%d (%d)\n", file,
1154              inode, page, (int)pos, (int)copied, (int)len);
1155
1156         /* zero the stale part of the page if we did a short copy */
1157         if (copied < len)
1158                 zero_user_segment(page, from+copied, len);
1159
1160         /* did file size increase? */
1161         /* (no need for i_size_read(); we caller holds i_mutex */
1162         if (pos+copied > inode->i_size)
1163                 check_cap = ceph_inode_set_size(inode, pos+copied);
1164
1165         if (!PageUptodate(page))
1166                 SetPageUptodate(page);
1167
1168         set_page_dirty(page);
1169
1170         unlock_page(page);
1171         up_read(&mdsc->snap_rwsem);
1172         page_cache_release(page);
1173
1174         if (check_cap)
1175                 ceph_check_caps(ceph_inode(inode), CHECK_CAPS_AUTHONLY, NULL);
1176
1177         return copied;
1178 }
1179
1180 /*
1181  * we set .direct_IO to indicate direct io is supported, but since we
1182  * intercept O_DIRECT reads and writes early, this function should
1183  * never get called.
1184  */
1185 static ssize_t ceph_direct_io(int rw, struct kiocb *iocb,
1186                               const struct iovec *iov,
1187                               loff_t pos, unsigned long nr_segs)
1188 {
1189         WARN_ON(1);
1190         return -EINVAL;
1191 }
1192
1193 const struct address_space_operations ceph_aops = {
1194         .readpage = ceph_readpage,
1195         .readpages = ceph_readpages,
1196         .writepage = ceph_writepage,
1197         .writepages = ceph_writepages_start,
1198         .write_begin = ceph_write_begin,
1199         .write_end = ceph_write_end,
1200         .set_page_dirty = ceph_set_page_dirty,
1201         .invalidatepage = ceph_invalidatepage,
1202         .releasepage = ceph_releasepage,
1203         .direct_IO = ceph_direct_io,
1204 };
1205
1206
1207 /*
1208  * vm ops
1209  */
1210
1211 /*
1212  * Reuse write_begin here for simplicity.
1213  */
1214 static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
1215 {
1216         struct inode *inode = file_inode(vma->vm_file);
1217         struct page *page = vmf->page;
1218         struct ceph_mds_client *mdsc = ceph_inode_to_client(inode)->mdsc;
1219         loff_t off = page_offset(page);
1220         loff_t size, len;
1221         int ret;
1222
1223         /* Update time before taking page lock */
1224         file_update_time(vma->vm_file);
1225
1226         size = i_size_read(inode);
1227         if (off + PAGE_CACHE_SIZE <= size)
1228                 len = PAGE_CACHE_SIZE;
1229         else
1230                 len = size & ~PAGE_CACHE_MASK;
1231
1232         dout("page_mkwrite %p %llu~%llu page %p idx %lu\n", inode,
1233              off, len, page, page->index);
1234
1235         lock_page(page);
1236
1237         ret = VM_FAULT_NOPAGE;
1238         if ((off > size) ||
1239             (page->mapping != inode->i_mapping))
1240                 goto out;
1241
1242         ret = ceph_update_writeable_page(vma->vm_file, off, len, page);
1243         if (ret == 0) {
1244                 /* success.  we'll keep the page locked. */
1245                 set_page_dirty(page);
1246                 up_read(&mdsc->snap_rwsem);
1247                 ret = VM_FAULT_LOCKED;
1248         } else {
1249                 if (ret == -ENOMEM)
1250                         ret = VM_FAULT_OOM;
1251                 else
1252                         ret = VM_FAULT_SIGBUS;
1253         }
1254 out:
1255         dout("page_mkwrite %p %llu~%llu = %d\n", inode, off, len, ret);
1256         if (ret != VM_FAULT_LOCKED)
1257                 unlock_page(page);
1258         return ret;
1259 }
1260
1261 static struct vm_operations_struct ceph_vmops = {
1262         .fault          = filemap_fault,
1263         .page_mkwrite   = ceph_page_mkwrite,
1264         .remap_pages    = generic_file_remap_pages,
1265 };
1266
1267 int ceph_mmap(struct file *file, struct vm_area_struct *vma)
1268 {
1269         struct address_space *mapping = file->f_mapping;
1270
1271         if (!mapping->a_ops->readpage)
1272                 return -ENOEXEC;
1273         file_accessed(file);
1274         vma->vm_ops = &ceph_vmops;
1275         return 0;
1276 }