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