xref: /openbmc/linux/fs/ceph/addr.c (revision 7587eb18)
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 	/* dirty the head */
91 	spin_lock(&ci->i_ceph_lock);
92 	BUG_ON(ci->i_wr_ref == 0); // caller should hold Fw reference
93 	if (__ceph_have_pending_cap_snap(ci)) {
94 		struct ceph_cap_snap *capsnap =
95 				list_last_entry(&ci->i_cap_snaps,
96 						struct ceph_cap_snap,
97 						ci_item);
98 		snapc = ceph_get_snap_context(capsnap->context);
99 		capsnap->dirty_pages++;
100 	} else {
101 		BUG_ON(!ci->i_head_snapc);
102 		snapc = ceph_get_snap_context(ci->i_head_snapc);
103 		++ci->i_wrbuffer_ref_head;
104 	}
105 	if (ci->i_wrbuffer_ref == 0)
106 		ihold(inode);
107 	++ci->i_wrbuffer_ref;
108 	dout("%p set_page_dirty %p idx %lu head %d/%d -> %d/%d "
109 	     "snapc %p seq %lld (%d snaps)\n",
110 	     mapping->host, page, page->index,
111 	     ci->i_wrbuffer_ref-1, ci->i_wrbuffer_ref_head-1,
112 	     ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
113 	     snapc, snapc->seq, snapc->num_snaps);
114 	spin_unlock(&ci->i_ceph_lock);
115 
116 	/*
117 	 * Reference snap context in page->private.  Also set
118 	 * PagePrivate so that we get invalidatepage callback.
119 	 */
120 	BUG_ON(PagePrivate(page));
121 	page->private = (unsigned long)snapc;
122 	SetPagePrivate(page);
123 
124 	ret = __set_page_dirty_nobuffers(page);
125 	WARN_ON(!PageLocked(page));
126 	WARN_ON(!page->mapping);
127 
128 	return ret;
129 }
130 
131 /*
132  * If we are truncating the full page (i.e. offset == 0), adjust the
133  * dirty page counters appropriately.  Only called if there is private
134  * data on the page.
135  */
136 static void ceph_invalidatepage(struct page *page, unsigned int offset,
137 				unsigned int length)
138 {
139 	struct inode *inode;
140 	struct ceph_inode_info *ci;
141 	struct ceph_snap_context *snapc = page_snap_context(page);
142 
143 	inode = page->mapping->host;
144 	ci = ceph_inode(inode);
145 
146 	if (offset != 0 || length != PAGE_SIZE) {
147 		dout("%p invalidatepage %p idx %lu partial dirty page %u~%u\n",
148 		     inode, page, page->index, offset, length);
149 		return;
150 	}
151 
152 	ceph_invalidate_fscache_page(inode, page);
153 
154 	if (!PagePrivate(page))
155 		return;
156 
157 	/*
158 	 * We can get non-dirty pages here due to races between
159 	 * set_page_dirty and truncate_complete_page; just spit out a
160 	 * warning, in case we end up with accounting problems later.
161 	 */
162 	if (!PageDirty(page))
163 		pr_err("%p invalidatepage %p page not dirty\n", inode, page);
164 
165 	ClearPageChecked(page);
166 
167 	dout("%p invalidatepage %p idx %lu full dirty page\n",
168 	     inode, page, page->index);
169 
170 	ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
171 	ceph_put_snap_context(snapc);
172 	page->private = 0;
173 	ClearPagePrivate(page);
174 }
175 
176 static int ceph_releasepage(struct page *page, gfp_t g)
177 {
178 	dout("%p releasepage %p idx %lu\n", page->mapping->host,
179 	     page, page->index);
180 	WARN_ON(PageDirty(page));
181 
182 	/* Can we release the page from the cache? */
183 	if (!ceph_release_fscache_page(page, g))
184 		return 0;
185 
186 	return !PagePrivate(page);
187 }
188 
189 /*
190  * read a single page, without unlocking it.
191  */
192 static int readpage_nounlock(struct file *filp, struct page *page)
193 {
194 	struct inode *inode = file_inode(filp);
195 	struct ceph_inode_info *ci = ceph_inode(inode);
196 	struct ceph_osd_client *osdc =
197 		&ceph_inode_to_client(inode)->client->osdc;
198 	int err = 0;
199 	u64 off = page_offset(page);
200 	u64 len = PAGE_SIZE;
201 
202 	if (off >= i_size_read(inode)) {
203 		zero_user_segment(page, 0, PAGE_SIZE);
204 		SetPageUptodate(page);
205 		return 0;
206 	}
207 
208 	if (ci->i_inline_version != CEPH_INLINE_NONE) {
209 		/*
210 		 * Uptodate inline data should have been added
211 		 * into page cache while getting Fcr caps.
212 		 */
213 		if (off == 0)
214 			return -EINVAL;
215 		zero_user_segment(page, 0, PAGE_SIZE);
216 		SetPageUptodate(page);
217 		return 0;
218 	}
219 
220 	err = ceph_readpage_from_fscache(inode, page);
221 	if (err == 0)
222 		goto out;
223 
224 	dout("readpage inode %p file %p page %p index %lu\n",
225 	     inode, filp, page, page->index);
226 	err = ceph_osdc_readpages(osdc, ceph_vino(inode), &ci->i_layout,
227 				  off, &len,
228 				  ci->i_truncate_seq, ci->i_truncate_size,
229 				  &page, 1, 0);
230 	if (err == -ENOENT)
231 		err = 0;
232 	if (err < 0) {
233 		SetPageError(page);
234 		ceph_fscache_readpage_cancel(inode, page);
235 		goto out;
236 	}
237 	if (err < PAGE_SIZE)
238 		/* zero fill remainder of page */
239 		zero_user_segment(page, err, PAGE_SIZE);
240 	else
241 		flush_dcache_page(page);
242 
243 	SetPageUptodate(page);
244 	ceph_readpage_to_fscache(inode, page);
245 
246 out:
247 	return err < 0 ? err : 0;
248 }
249 
250 static int ceph_readpage(struct file *filp, struct page *page)
251 {
252 	int r = readpage_nounlock(filp, page);
253 	unlock_page(page);
254 	return r;
255 }
256 
257 /*
258  * Finish an async read(ahead) op.
259  */
260 static void finish_read(struct ceph_osd_request *req)
261 {
262 	struct inode *inode = req->r_inode;
263 	struct ceph_osd_data *osd_data;
264 	int rc = req->r_result <= 0 ? req->r_result : 0;
265 	int bytes = req->r_result >= 0 ? req->r_result : 0;
266 	int num_pages;
267 	int i;
268 
269 	dout("finish_read %p req %p rc %d bytes %d\n", inode, req, rc, bytes);
270 
271 	/* unlock all pages, zeroing any data we didn't read */
272 	osd_data = osd_req_op_extent_osd_data(req, 0);
273 	BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
274 	num_pages = calc_pages_for((u64)osd_data->alignment,
275 					(u64)osd_data->length);
276 	for (i = 0; i < num_pages; i++) {
277 		struct page *page = osd_data->pages[i];
278 
279 		if (rc < 0 && rc != -ENOENT) {
280 			ceph_fscache_readpage_cancel(inode, page);
281 			goto unlock;
282 		}
283 		if (bytes < (int)PAGE_SIZE) {
284 			/* zero (remainder of) page */
285 			int s = bytes < 0 ? 0 : bytes;
286 			zero_user_segment(page, s, PAGE_SIZE);
287 		}
288  		dout("finish_read %p uptodate %p idx %lu\n", inode, page,
289 		     page->index);
290 		flush_dcache_page(page);
291 		SetPageUptodate(page);
292 		ceph_readpage_to_fscache(inode, page);
293 unlock:
294 		unlock_page(page);
295 		put_page(page);
296 		bytes -= PAGE_SIZE;
297 	}
298 	kfree(osd_data->pages);
299 }
300 
301 static void ceph_unlock_page_vector(struct page **pages, int num_pages)
302 {
303 	int i;
304 
305 	for (i = 0; i < num_pages; i++)
306 		unlock_page(pages[i]);
307 }
308 
309 /*
310  * start an async read(ahead) operation.  return nr_pages we submitted
311  * a read for on success, or negative error code.
312  */
313 static int start_read(struct inode *inode, struct list_head *page_list, int max)
314 {
315 	struct ceph_osd_client *osdc =
316 		&ceph_inode_to_client(inode)->client->osdc;
317 	struct ceph_inode_info *ci = ceph_inode(inode);
318 	struct page *page = list_entry(page_list->prev, struct page, lru);
319 	struct ceph_vino vino;
320 	struct ceph_osd_request *req;
321 	u64 off;
322 	u64 len;
323 	int i;
324 	struct page **pages;
325 	pgoff_t next_index;
326 	int nr_pages = 0;
327 	int ret;
328 
329 	off = (u64) page_offset(page);
330 
331 	/* count pages */
332 	next_index = page->index;
333 	list_for_each_entry_reverse(page, page_list, lru) {
334 		if (page->index != next_index)
335 			break;
336 		nr_pages++;
337 		next_index++;
338 		if (max && nr_pages == max)
339 			break;
340 	}
341 	len = nr_pages << PAGE_SHIFT;
342 	dout("start_read %p nr_pages %d is %lld~%lld\n", inode, nr_pages,
343 	     off, len);
344 	vino = ceph_vino(inode);
345 	req = ceph_osdc_new_request(osdc, &ci->i_layout, vino, off, &len,
346 				    0, 1, CEPH_OSD_OP_READ,
347 				    CEPH_OSD_FLAG_READ, NULL,
348 				    ci->i_truncate_seq, ci->i_truncate_size,
349 				    false);
350 	if (IS_ERR(req))
351 		return PTR_ERR(req);
352 
353 	/* build page vector */
354 	nr_pages = calc_pages_for(0, len);
355 	pages = kmalloc(sizeof(*pages) * nr_pages, GFP_KERNEL);
356 	ret = -ENOMEM;
357 	if (!pages)
358 		goto out;
359 	for (i = 0; i < nr_pages; ++i) {
360 		page = list_entry(page_list->prev, struct page, lru);
361 		BUG_ON(PageLocked(page));
362 		list_del(&page->lru);
363 
364  		dout("start_read %p adding %p idx %lu\n", inode, page,
365 		     page->index);
366 		if (add_to_page_cache_lru(page, &inode->i_data, page->index,
367 					  GFP_KERNEL)) {
368 			ceph_fscache_uncache_page(inode, page);
369 			put_page(page);
370 			dout("start_read %p add_to_page_cache failed %p\n",
371 			     inode, page);
372 			nr_pages = i;
373 			goto out_pages;
374 		}
375 		pages[i] = page;
376 	}
377 	osd_req_op_extent_osd_data_pages(req, 0, pages, len, 0, false, false);
378 	req->r_callback = finish_read;
379 	req->r_inode = inode;
380 
381 	dout("start_read %p starting %p %lld~%lld\n", inode, req, off, len);
382 	ret = ceph_osdc_start_request(osdc, req, false);
383 	if (ret < 0)
384 		goto out_pages;
385 	ceph_osdc_put_request(req);
386 	return nr_pages;
387 
388 out_pages:
389 	ceph_unlock_page_vector(pages, nr_pages);
390 	ceph_release_page_vector(pages, nr_pages);
391 out:
392 	ceph_osdc_put_request(req);
393 	return ret;
394 }
395 
396 
397 /*
398  * Read multiple pages.  Leave pages we don't read + unlock in page_list;
399  * the caller (VM) cleans them up.
400  */
401 static int ceph_readpages(struct file *file, struct address_space *mapping,
402 			  struct list_head *page_list, unsigned nr_pages)
403 {
404 	struct inode *inode = file_inode(file);
405 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
406 	int rc = 0;
407 	int max = 0;
408 
409 	if (ceph_inode(inode)->i_inline_version != CEPH_INLINE_NONE)
410 		return -EINVAL;
411 
412 	rc = ceph_readpages_from_fscache(mapping->host, mapping, page_list,
413 					 &nr_pages);
414 
415 	if (rc == 0)
416 		goto out;
417 
418 	if (fsc->mount_options->rsize >= PAGE_SIZE)
419 		max = (fsc->mount_options->rsize + PAGE_SIZE - 1)
420 			>> PAGE_SHIFT;
421 
422 	dout("readpages %p file %p nr_pages %d max %d\n", inode,
423 		file, nr_pages,
424 	     max);
425 	while (!list_empty(page_list)) {
426 		rc = start_read(inode, page_list, max);
427 		if (rc < 0)
428 			goto out;
429 		BUG_ON(rc == 0);
430 	}
431 out:
432 	ceph_fscache_readpages_cancel(inode, page_list);
433 
434 	dout("readpages %p file %p ret %d\n", inode, file, rc);
435 	return rc;
436 }
437 
438 /*
439  * Get ref for the oldest snapc for an inode with dirty data... that is, the
440  * only snap context we are allowed to write back.
441  */
442 static struct ceph_snap_context *get_oldest_context(struct inode *inode,
443 						    loff_t *snap_size)
444 {
445 	struct ceph_inode_info *ci = ceph_inode(inode);
446 	struct ceph_snap_context *snapc = NULL;
447 	struct ceph_cap_snap *capsnap = NULL;
448 
449 	spin_lock(&ci->i_ceph_lock);
450 	list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
451 		dout(" cap_snap %p snapc %p has %d dirty pages\n", capsnap,
452 		     capsnap->context, capsnap->dirty_pages);
453 		if (capsnap->dirty_pages) {
454 			snapc = ceph_get_snap_context(capsnap->context);
455 			if (snap_size)
456 				*snap_size = capsnap->size;
457 			break;
458 		}
459 	}
460 	if (!snapc && ci->i_wrbuffer_ref_head) {
461 		snapc = ceph_get_snap_context(ci->i_head_snapc);
462 		dout(" head snapc %p has %d dirty pages\n",
463 		     snapc, ci->i_wrbuffer_ref_head);
464 	}
465 	spin_unlock(&ci->i_ceph_lock);
466 	return snapc;
467 }
468 
469 /*
470  * Write a single page, but leave the page locked.
471  *
472  * If we get a write error, set the page error bit, but still adjust the
473  * dirty page accounting (i.e., page is no longer dirty).
474  */
475 static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
476 {
477 	struct inode *inode;
478 	struct ceph_inode_info *ci;
479 	struct ceph_fs_client *fsc;
480 	struct ceph_osd_client *osdc;
481 	struct ceph_snap_context *snapc, *oldest;
482 	loff_t page_off = page_offset(page);
483 	loff_t snap_size = -1;
484 	long writeback_stat;
485 	u64 truncate_size;
486 	u32 truncate_seq;
487 	int err = 0, len = PAGE_SIZE;
488 
489 	dout("writepage %p idx %lu\n", page, page->index);
490 
491 	if (!page->mapping || !page->mapping->host) {
492 		dout("writepage %p - no mapping\n", page);
493 		return -EFAULT;
494 	}
495 	inode = page->mapping->host;
496 	ci = ceph_inode(inode);
497 	fsc = ceph_inode_to_client(inode);
498 	osdc = &fsc->client->osdc;
499 
500 	/* verify this is a writeable snap context */
501 	snapc = page_snap_context(page);
502 	if (snapc == NULL) {
503 		dout("writepage %p page %p not dirty?\n", inode, page);
504 		goto out;
505 	}
506 	oldest = get_oldest_context(inode, &snap_size);
507 	if (snapc->seq > oldest->seq) {
508 		dout("writepage %p page %p snapc %p not writeable - noop\n",
509 		     inode, page, snapc);
510 		/* we should only noop if called by kswapd */
511 		WARN_ON((current->flags & PF_MEMALLOC) == 0);
512 		ceph_put_snap_context(oldest);
513 		goto out;
514 	}
515 	ceph_put_snap_context(oldest);
516 
517 	spin_lock(&ci->i_ceph_lock);
518 	truncate_seq = ci->i_truncate_seq;
519 	truncate_size = ci->i_truncate_size;
520 	if (snap_size == -1)
521 		snap_size = i_size_read(inode);
522 	spin_unlock(&ci->i_ceph_lock);
523 
524 	/* is this a partial page at end of file? */
525 	if (page_off >= snap_size) {
526 		dout("%p page eof %llu\n", page, snap_size);
527 		goto out;
528 	}
529 	if (snap_size < page_off + len)
530 		len = snap_size - page_off;
531 
532 	dout("writepage %p page %p index %lu on %llu~%u snapc %p\n",
533 	     inode, page, page->index, page_off, len, snapc);
534 
535 	writeback_stat = atomic_long_inc_return(&fsc->writeback_count);
536 	if (writeback_stat >
537 	    CONGESTION_ON_THRESH(fsc->mount_options->congestion_kb))
538 		set_bdi_congested(&fsc->backing_dev_info, BLK_RW_ASYNC);
539 
540 	set_page_writeback(page);
541 	err = ceph_osdc_writepages(osdc, ceph_vino(inode),
542 				   &ci->i_layout, snapc,
543 				   page_off, len,
544 				   truncate_seq, truncate_size,
545 				   &inode->i_mtime, &page, 1);
546 	if (err < 0) {
547 		struct writeback_control tmp_wbc;
548 		if (!wbc)
549 			wbc = &tmp_wbc;
550 		if (err == -ERESTARTSYS) {
551 			/* killed by SIGKILL */
552 			dout("writepage interrupted page %p\n", page);
553 			redirty_page_for_writepage(wbc, page);
554 			end_page_writeback(page);
555 			goto out;
556 		}
557 		dout("writepage setting page/mapping error %d %p\n",
558 		     err, page);
559 		SetPageError(page);
560 		mapping_set_error(&inode->i_data, err);
561 		wbc->pages_skipped++;
562 	} else {
563 		dout("writepage cleaned page %p\n", page);
564 		err = 0;  /* vfs expects us to return 0 */
565 	}
566 	page->private = 0;
567 	ClearPagePrivate(page);
568 	end_page_writeback(page);
569 	ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
570 	ceph_put_snap_context(snapc);  /* page's reference */
571 out:
572 	return err;
573 }
574 
575 static int ceph_writepage(struct page *page, struct writeback_control *wbc)
576 {
577 	int err;
578 	struct inode *inode = page->mapping->host;
579 	BUG_ON(!inode);
580 	ihold(inode);
581 	err = writepage_nounlock(page, wbc);
582 	if (err == -ERESTARTSYS) {
583 		/* direct memory reclaimer was killed by SIGKILL. return 0
584 		 * to prevent caller from setting mapping/page error */
585 		err = 0;
586 	}
587 	unlock_page(page);
588 	iput(inode);
589 	return err;
590 }
591 
592 /*
593  * lame release_pages helper.  release_pages() isn't exported to
594  * modules.
595  */
596 static void ceph_release_pages(struct page **pages, int num)
597 {
598 	struct pagevec pvec;
599 	int i;
600 
601 	pagevec_init(&pvec, 0);
602 	for (i = 0; i < num; i++) {
603 		if (pagevec_add(&pvec, pages[i]) == 0)
604 			pagevec_release(&pvec);
605 	}
606 	pagevec_release(&pvec);
607 }
608 
609 /*
610  * async writeback completion handler.
611  *
612  * If we get an error, set the mapping error bit, but not the individual
613  * page error bits.
614  */
615 static void writepages_finish(struct ceph_osd_request *req)
616 {
617 	struct inode *inode = req->r_inode;
618 	struct ceph_inode_info *ci = ceph_inode(inode);
619 	struct ceph_osd_data *osd_data;
620 	struct page *page;
621 	int num_pages, total_pages = 0;
622 	int i, j;
623 	int rc = req->r_result;
624 	struct ceph_snap_context *snapc = req->r_snapc;
625 	struct address_space *mapping = inode->i_mapping;
626 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
627 	bool remove_page;
628 
629 	dout("writepages_finish %p rc %d\n", inode, rc);
630 	if (rc < 0)
631 		mapping_set_error(mapping, rc);
632 
633 	/*
634 	 * We lost the cache cap, need to truncate the page before
635 	 * it is unlocked, otherwise we'd truncate it later in the
636 	 * page truncation thread, possibly losing some data that
637 	 * raced its way in
638 	 */
639 	remove_page = !(ceph_caps_issued(ci) &
640 			(CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO));
641 
642 	/* clean all pages */
643 	for (i = 0; i < req->r_num_ops; i++) {
644 		if (req->r_ops[i].op != CEPH_OSD_OP_WRITE)
645 			break;
646 
647 		osd_data = osd_req_op_extent_osd_data(req, i);
648 		BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
649 		num_pages = calc_pages_for((u64)osd_data->alignment,
650 					   (u64)osd_data->length);
651 		total_pages += num_pages;
652 		for (j = 0; j < num_pages; j++) {
653 			page = osd_data->pages[j];
654 			BUG_ON(!page);
655 			WARN_ON(!PageUptodate(page));
656 
657 			if (atomic_long_dec_return(&fsc->writeback_count) <
658 			     CONGESTION_OFF_THRESH(
659 					fsc->mount_options->congestion_kb))
660 				clear_bdi_congested(&fsc->backing_dev_info,
661 						    BLK_RW_ASYNC);
662 
663 			if (rc < 0)
664 				SetPageError(page);
665 
666 			ceph_put_snap_context(page_snap_context(page));
667 			page->private = 0;
668 			ClearPagePrivate(page);
669 			dout("unlocking %p\n", page);
670 			end_page_writeback(page);
671 
672 			if (remove_page)
673 				generic_error_remove_page(inode->i_mapping,
674 							  page);
675 
676 			unlock_page(page);
677 		}
678 		dout("writepages_finish %p wrote %llu bytes cleaned %d pages\n",
679 		     inode, osd_data->length, rc >= 0 ? num_pages : 0);
680 
681 		ceph_release_pages(osd_data->pages, num_pages);
682 	}
683 
684 	ceph_put_wrbuffer_cap_refs(ci, total_pages, snapc);
685 
686 	osd_data = osd_req_op_extent_osd_data(req, 0);
687 	if (osd_data->pages_from_pool)
688 		mempool_free(osd_data->pages,
689 			     ceph_sb_to_client(inode->i_sb)->wb_pagevec_pool);
690 	else
691 		kfree(osd_data->pages);
692 	ceph_osdc_put_request(req);
693 }
694 
695 /*
696  * initiate async writeback
697  */
698 static int ceph_writepages_start(struct address_space *mapping,
699 				 struct writeback_control *wbc)
700 {
701 	struct inode *inode = mapping->host;
702 	struct ceph_inode_info *ci = ceph_inode(inode);
703 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
704 	struct ceph_vino vino = ceph_vino(inode);
705 	pgoff_t index, start, end;
706 	int range_whole = 0;
707 	int should_loop = 1;
708 	pgoff_t max_pages = 0, max_pages_ever = 0;
709 	struct ceph_snap_context *snapc = NULL, *last_snapc = NULL, *pgsnapc;
710 	struct pagevec pvec;
711 	int done = 0;
712 	int rc = 0;
713 	unsigned wsize = 1 << inode->i_blkbits;
714 	struct ceph_osd_request *req = NULL;
715 	int do_sync = 0;
716 	loff_t snap_size, i_size;
717 	u64 truncate_size;
718 	u32 truncate_seq;
719 
720 	/*
721 	 * Include a 'sync' in the OSD request if this is a data
722 	 * integrity write (e.g., O_SYNC write or fsync()), or if our
723 	 * cap is being revoked.
724 	 */
725 	if ((wbc->sync_mode == WB_SYNC_ALL) ||
726 		ceph_caps_revoking(ci, CEPH_CAP_FILE_BUFFER))
727 		do_sync = 1;
728 	dout("writepages_start %p dosync=%d (mode=%s)\n",
729 	     inode, do_sync,
730 	     wbc->sync_mode == WB_SYNC_NONE ? "NONE" :
731 	     (wbc->sync_mode == WB_SYNC_ALL ? "ALL" : "HOLD"));
732 
733 	if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
734 		if (ci->i_wrbuffer_ref > 0) {
735 			pr_warn_ratelimited(
736 				"writepage_start %p %lld forced umount\n",
737 				inode, ceph_ino(inode));
738 		}
739 		mapping_set_error(mapping, -EIO);
740 		return -EIO; /* we're in a forced umount, don't write! */
741 	}
742 	if (fsc->mount_options->wsize && fsc->mount_options->wsize < wsize)
743 		wsize = fsc->mount_options->wsize;
744 	if (wsize < PAGE_SIZE)
745 		wsize = PAGE_SIZE;
746 	max_pages_ever = wsize >> PAGE_SHIFT;
747 
748 	pagevec_init(&pvec, 0);
749 
750 	/* where to start/end? */
751 	if (wbc->range_cyclic) {
752 		start = mapping->writeback_index; /* Start from prev offset */
753 		end = -1;
754 		dout(" cyclic, start at %lu\n", start);
755 	} else {
756 		start = wbc->range_start >> PAGE_SHIFT;
757 		end = wbc->range_end >> PAGE_SHIFT;
758 		if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
759 			range_whole = 1;
760 		should_loop = 0;
761 		dout(" not cyclic, %lu to %lu\n", start, end);
762 	}
763 	index = start;
764 
765 retry:
766 	/* find oldest snap context with dirty data */
767 	ceph_put_snap_context(snapc);
768 	snap_size = -1;
769 	snapc = get_oldest_context(inode, &snap_size);
770 	if (!snapc) {
771 		/* hmm, why does writepages get called when there
772 		   is no dirty data? */
773 		dout(" no snap context with dirty data?\n");
774 		goto out;
775 	}
776 	dout(" oldest snapc is %p seq %lld (%d snaps)\n",
777 	     snapc, snapc->seq, snapc->num_snaps);
778 
779 	spin_lock(&ci->i_ceph_lock);
780 	truncate_seq = ci->i_truncate_seq;
781 	truncate_size = ci->i_truncate_size;
782 	i_size = i_size_read(inode);
783 	spin_unlock(&ci->i_ceph_lock);
784 
785 	if (last_snapc && snapc != last_snapc) {
786 		/* if we switched to a newer snapc, restart our scan at the
787 		 * start of the original file range. */
788 		dout("  snapc differs from last pass, restarting at %lu\n",
789 		     index);
790 		index = start;
791 	}
792 	last_snapc = snapc;
793 
794 	while (!done && index <= end) {
795 		unsigned i;
796 		int first;
797 		pgoff_t strip_unit_end = 0;
798 		int num_ops = 0, op_idx;
799 		int pvec_pages, locked_pages = 0;
800 		struct page **pages = NULL, **data_pages;
801 		mempool_t *pool = NULL;	/* Becomes non-null if mempool used */
802 		struct page *page;
803 		int want;
804 		u64 offset = 0, len = 0;
805 
806 		max_pages = max_pages_ever;
807 
808 get_more_pages:
809 		first = -1;
810 		want = min(end - index,
811 			   min((pgoff_t)PAGEVEC_SIZE,
812 			       max_pages - (pgoff_t)locked_pages) - 1)
813 			+ 1;
814 		pvec_pages = pagevec_lookup_tag(&pvec, mapping, &index,
815 						PAGECACHE_TAG_DIRTY,
816 						want);
817 		dout("pagevec_lookup_tag got %d\n", pvec_pages);
818 		if (!pvec_pages && !locked_pages)
819 			break;
820 		for (i = 0; i < pvec_pages && locked_pages < max_pages; i++) {
821 			page = pvec.pages[i];
822 			dout("? %p idx %lu\n", page, page->index);
823 			if (locked_pages == 0)
824 				lock_page(page);  /* first page */
825 			else if (!trylock_page(page))
826 				break;
827 
828 			/* only dirty pages, or our accounting breaks */
829 			if (unlikely(!PageDirty(page)) ||
830 			    unlikely(page->mapping != mapping)) {
831 				dout("!dirty or !mapping %p\n", page);
832 				unlock_page(page);
833 				break;
834 			}
835 			if (!wbc->range_cyclic && page->index > end) {
836 				dout("end of range %p\n", page);
837 				done = 1;
838 				unlock_page(page);
839 				break;
840 			}
841 			if (strip_unit_end && (page->index > strip_unit_end)) {
842 				dout("end of strip unit %p\n", page);
843 				unlock_page(page);
844 				break;
845 			}
846 			if (wbc->sync_mode != WB_SYNC_NONE) {
847 				dout("waiting on writeback %p\n", page);
848 				wait_on_page_writeback(page);
849 			}
850 			if (page_offset(page) >=
851 			    (snap_size == -1 ? i_size : snap_size)) {
852 				dout("%p page eof %llu\n", page,
853 				     (snap_size == -1 ? i_size : snap_size));
854 				done = 1;
855 				unlock_page(page);
856 				break;
857 			}
858 			if (PageWriteback(page)) {
859 				dout("%p under writeback\n", page);
860 				unlock_page(page);
861 				break;
862 			}
863 
864 			/* only if matching snap context */
865 			pgsnapc = page_snap_context(page);
866 			if (pgsnapc->seq > snapc->seq) {
867 				dout("page snapc %p %lld > oldest %p %lld\n",
868 				     pgsnapc, pgsnapc->seq, snapc, snapc->seq);
869 				unlock_page(page);
870 				if (!locked_pages)
871 					continue; /* keep looking for snap */
872 				break;
873 			}
874 
875 			if (!clear_page_dirty_for_io(page)) {
876 				dout("%p !clear_page_dirty_for_io\n", page);
877 				unlock_page(page);
878 				break;
879 			}
880 
881 			/*
882 			 * We have something to write.  If this is
883 			 * the first locked page this time through,
884 			 * calculate max possinle write size and
885 			 * allocate a page array
886 			 */
887 			if (locked_pages == 0) {
888 				u64 objnum;
889 				u64 objoff;
890 
891 				/* prepare async write request */
892 				offset = (u64)page_offset(page);
893 				len = wsize;
894 
895 				rc = ceph_calc_file_object_mapping(&ci->i_layout,
896 								offset, len,
897 								&objnum, &objoff,
898 								&len);
899 				if (rc < 0) {
900 					unlock_page(page);
901 					break;
902 				}
903 
904 				num_ops = 1 + do_sync;
905 				strip_unit_end = page->index +
906 					((len - 1) >> PAGE_SHIFT);
907 
908 				BUG_ON(pages);
909 				max_pages = calc_pages_for(0, (u64)len);
910 				pages = kmalloc(max_pages * sizeof (*pages),
911 						GFP_NOFS);
912 				if (!pages) {
913 					pool = fsc->wb_pagevec_pool;
914 					pages = mempool_alloc(pool, GFP_NOFS);
915 					BUG_ON(!pages);
916 				}
917 
918 				len = 0;
919 			} else if (page->index !=
920 				   (offset + len) >> PAGE_SHIFT) {
921 				if (num_ops >= (pool ?  CEPH_OSD_SLAB_OPS :
922 							CEPH_OSD_MAX_OPS)) {
923 					redirty_page_for_writepage(wbc, page);
924 					unlock_page(page);
925 					break;
926 				}
927 
928 				num_ops++;
929 				offset = (u64)page_offset(page);
930 				len = 0;
931 			}
932 
933 			/* note position of first page in pvec */
934 			if (first < 0)
935 				first = i;
936 			dout("%p will write page %p idx %lu\n",
937 			     inode, page, page->index);
938 
939 			if (atomic_long_inc_return(&fsc->writeback_count) >
940 			    CONGESTION_ON_THRESH(
941 				    fsc->mount_options->congestion_kb)) {
942 				set_bdi_congested(&fsc->backing_dev_info,
943 						  BLK_RW_ASYNC);
944 			}
945 
946 			pages[locked_pages] = page;
947 			locked_pages++;
948 			len += PAGE_SIZE;
949 		}
950 
951 		/* did we get anything? */
952 		if (!locked_pages)
953 			goto release_pvec_pages;
954 		if (i) {
955 			int j;
956 			BUG_ON(!locked_pages || first < 0);
957 
958 			if (pvec_pages && i == pvec_pages &&
959 			    locked_pages < max_pages) {
960 				dout("reached end pvec, trying for more\n");
961 				pagevec_reinit(&pvec);
962 				goto get_more_pages;
963 			}
964 
965 			/* shift unused pages over in the pvec...  we
966 			 * will need to release them below. */
967 			for (j = i; j < pvec_pages; j++) {
968 				dout(" pvec leftover page %p\n", pvec.pages[j]);
969 				pvec.pages[j-i+first] = pvec.pages[j];
970 			}
971 			pvec.nr -= i-first;
972 		}
973 
974 new_request:
975 		offset = page_offset(pages[0]);
976 		len = wsize;
977 
978 		req = ceph_osdc_new_request(&fsc->client->osdc,
979 					&ci->i_layout, vino,
980 					offset, &len, 0, num_ops,
981 					CEPH_OSD_OP_WRITE,
982 					CEPH_OSD_FLAG_WRITE |
983 					CEPH_OSD_FLAG_ONDISK,
984 					snapc, truncate_seq,
985 					truncate_size, false);
986 		if (IS_ERR(req)) {
987 			req = ceph_osdc_new_request(&fsc->client->osdc,
988 						&ci->i_layout, vino,
989 						offset, &len, 0,
990 						min(num_ops,
991 						    CEPH_OSD_SLAB_OPS),
992 						CEPH_OSD_OP_WRITE,
993 						CEPH_OSD_FLAG_WRITE |
994 						CEPH_OSD_FLAG_ONDISK,
995 						snapc, truncate_seq,
996 						truncate_size, true);
997 			BUG_ON(IS_ERR(req));
998 		}
999 		BUG_ON(len < page_offset(pages[locked_pages - 1]) +
1000 			     PAGE_SIZE - offset);
1001 
1002 		req->r_callback = writepages_finish;
1003 		req->r_inode = inode;
1004 
1005 		/* Format the osd request message and submit the write */
1006 		len = 0;
1007 		data_pages = pages;
1008 		op_idx = 0;
1009 		for (i = 0; i < locked_pages; i++) {
1010 			u64 cur_offset = page_offset(pages[i]);
1011 			if (offset + len != cur_offset) {
1012 				if (op_idx + do_sync + 1 == req->r_num_ops)
1013 					break;
1014 				osd_req_op_extent_dup_last(req, op_idx,
1015 							   cur_offset - offset);
1016 				dout("writepages got pages at %llu~%llu\n",
1017 				     offset, len);
1018 				osd_req_op_extent_osd_data_pages(req, op_idx,
1019 							data_pages, len, 0,
1020 							!!pool, false);
1021 				osd_req_op_extent_update(req, op_idx, len);
1022 
1023 				len = 0;
1024 				offset = cur_offset;
1025 				data_pages = pages + i;
1026 				op_idx++;
1027 			}
1028 
1029 			set_page_writeback(pages[i]);
1030 			len += PAGE_SIZE;
1031 		}
1032 
1033 		if (snap_size != -1) {
1034 			len = min(len, snap_size - offset);
1035 		} else if (i == locked_pages) {
1036 			/* writepages_finish() clears writeback pages
1037 			 * according to the data length, so make sure
1038 			 * data length covers all locked pages */
1039 			u64 min_len = len + 1 - PAGE_SIZE;
1040 			len = min(len, (u64)i_size_read(inode) - offset);
1041 			len = max(len, min_len);
1042 		}
1043 		dout("writepages got pages at %llu~%llu\n", offset, len);
1044 
1045 		osd_req_op_extent_osd_data_pages(req, op_idx, data_pages, len,
1046 						 0, !!pool, false);
1047 		osd_req_op_extent_update(req, op_idx, len);
1048 
1049 		if (do_sync) {
1050 			op_idx++;
1051 			osd_req_op_init(req, op_idx, CEPH_OSD_OP_STARTSYNC, 0);
1052 		}
1053 		BUG_ON(op_idx + 1 != req->r_num_ops);
1054 
1055 		pool = NULL;
1056 		if (i < locked_pages) {
1057 			BUG_ON(num_ops <= req->r_num_ops);
1058 			num_ops -= req->r_num_ops;
1059 			num_ops += do_sync;
1060 			locked_pages -= i;
1061 
1062 			/* allocate new pages array for next request */
1063 			data_pages = pages;
1064 			pages = kmalloc(locked_pages * sizeof (*pages),
1065 					GFP_NOFS);
1066 			if (!pages) {
1067 				pool = fsc->wb_pagevec_pool;
1068 				pages = mempool_alloc(pool, GFP_NOFS);
1069 				BUG_ON(!pages);
1070 			}
1071 			memcpy(pages, data_pages + i,
1072 			       locked_pages * sizeof(*pages));
1073 			memset(data_pages + i, 0,
1074 			       locked_pages * sizeof(*pages));
1075 		} else {
1076 			BUG_ON(num_ops != req->r_num_ops);
1077 			index = pages[i - 1]->index + 1;
1078 			/* request message now owns the pages array */
1079 			pages = NULL;
1080 		}
1081 
1082 		req->r_mtime = inode->i_mtime;
1083 		rc = ceph_osdc_start_request(&fsc->client->osdc, req, true);
1084 		BUG_ON(rc);
1085 		req = NULL;
1086 
1087 		wbc->nr_to_write -= i;
1088 		if (pages)
1089 			goto new_request;
1090 
1091 		if (wbc->nr_to_write <= 0)
1092 			done = 1;
1093 
1094 release_pvec_pages:
1095 		dout("pagevec_release on %d pages (%p)\n", (int)pvec.nr,
1096 		     pvec.nr ? pvec.pages[0] : NULL);
1097 		pagevec_release(&pvec);
1098 
1099 		if (locked_pages && !done)
1100 			goto retry;
1101 	}
1102 
1103 	if (should_loop && !done) {
1104 		/* more to do; loop back to beginning of file */
1105 		dout("writepages looping back to beginning of file\n");
1106 		should_loop = 0;
1107 		index = 0;
1108 		goto retry;
1109 	}
1110 
1111 	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
1112 		mapping->writeback_index = index;
1113 
1114 out:
1115 	ceph_osdc_put_request(req);
1116 	ceph_put_snap_context(snapc);
1117 	dout("writepages done, rc = %d\n", rc);
1118 	return rc;
1119 }
1120 
1121 
1122 
1123 /*
1124  * See if a given @snapc is either writeable, or already written.
1125  */
1126 static int context_is_writeable_or_written(struct inode *inode,
1127 					   struct ceph_snap_context *snapc)
1128 {
1129 	struct ceph_snap_context *oldest = get_oldest_context(inode, NULL);
1130 	int ret = !oldest || snapc->seq <= oldest->seq;
1131 
1132 	ceph_put_snap_context(oldest);
1133 	return ret;
1134 }
1135 
1136 /*
1137  * We are only allowed to write into/dirty the page if the page is
1138  * clean, or already dirty within the same snap context.
1139  *
1140  * called with page locked.
1141  * return success with page locked,
1142  * or any failure (incl -EAGAIN) with page unlocked.
1143  */
1144 static int ceph_update_writeable_page(struct file *file,
1145 			    loff_t pos, unsigned len,
1146 			    struct page *page)
1147 {
1148 	struct inode *inode = file_inode(file);
1149 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1150 	struct ceph_inode_info *ci = ceph_inode(inode);
1151 	loff_t page_off = pos & PAGE_MASK;
1152 	int pos_in_page = pos & ~PAGE_MASK;
1153 	int end_in_page = pos_in_page + len;
1154 	loff_t i_size;
1155 	int r;
1156 	struct ceph_snap_context *snapc, *oldest;
1157 
1158 	if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
1159 		dout(" page %p forced umount\n", page);
1160 		unlock_page(page);
1161 		return -EIO;
1162 	}
1163 
1164 retry_locked:
1165 	/* writepages currently holds page lock, but if we change that later, */
1166 	wait_on_page_writeback(page);
1167 
1168 	snapc = page_snap_context(page);
1169 	if (snapc && snapc != ci->i_head_snapc) {
1170 		/*
1171 		 * this page is already dirty in another (older) snap
1172 		 * context!  is it writeable now?
1173 		 */
1174 		oldest = get_oldest_context(inode, NULL);
1175 
1176 		if (snapc->seq > oldest->seq) {
1177 			ceph_put_snap_context(oldest);
1178 			dout(" page %p snapc %p not current or oldest\n",
1179 			     page, snapc);
1180 			/*
1181 			 * queue for writeback, and wait for snapc to
1182 			 * be writeable or written
1183 			 */
1184 			snapc = ceph_get_snap_context(snapc);
1185 			unlock_page(page);
1186 			ceph_queue_writeback(inode);
1187 			r = wait_event_killable(ci->i_cap_wq,
1188 			       context_is_writeable_or_written(inode, snapc));
1189 			ceph_put_snap_context(snapc);
1190 			if (r == -ERESTARTSYS)
1191 				return r;
1192 			return -EAGAIN;
1193 		}
1194 		ceph_put_snap_context(oldest);
1195 
1196 		/* yay, writeable, do it now (without dropping page lock) */
1197 		dout(" page %p snapc %p not current, but oldest\n",
1198 		     page, snapc);
1199 		if (!clear_page_dirty_for_io(page))
1200 			goto retry_locked;
1201 		r = writepage_nounlock(page, NULL);
1202 		if (r < 0)
1203 			goto fail_nosnap;
1204 		goto retry_locked;
1205 	}
1206 
1207 	if (PageUptodate(page)) {
1208 		dout(" page %p already uptodate\n", page);
1209 		return 0;
1210 	}
1211 
1212 	/* full page? */
1213 	if (pos_in_page == 0 && len == PAGE_SIZE)
1214 		return 0;
1215 
1216 	/* past end of file? */
1217 	i_size = i_size_read(inode);
1218 
1219 	if (page_off >= i_size ||
1220 	    (pos_in_page == 0 && (pos+len) >= i_size &&
1221 	     end_in_page - pos_in_page != PAGE_SIZE)) {
1222 		dout(" zeroing %p 0 - %d and %d - %d\n",
1223 		     page, pos_in_page, end_in_page, (int)PAGE_SIZE);
1224 		zero_user_segments(page,
1225 				   0, pos_in_page,
1226 				   end_in_page, PAGE_SIZE);
1227 		return 0;
1228 	}
1229 
1230 	/* we need to read it. */
1231 	r = readpage_nounlock(file, page);
1232 	if (r < 0)
1233 		goto fail_nosnap;
1234 	goto retry_locked;
1235 fail_nosnap:
1236 	unlock_page(page);
1237 	return r;
1238 }
1239 
1240 /*
1241  * We are only allowed to write into/dirty the page if the page is
1242  * clean, or already dirty within the same snap context.
1243  */
1244 static int ceph_write_begin(struct file *file, struct address_space *mapping,
1245 			    loff_t pos, unsigned len, unsigned flags,
1246 			    struct page **pagep, void **fsdata)
1247 {
1248 	struct inode *inode = file_inode(file);
1249 	struct page *page;
1250 	pgoff_t index = pos >> PAGE_SHIFT;
1251 	int r;
1252 
1253 	do {
1254 		/* get a page */
1255 		page = grab_cache_page_write_begin(mapping, index, 0);
1256 		if (!page)
1257 			return -ENOMEM;
1258 
1259 		dout("write_begin file %p inode %p page %p %d~%d\n", file,
1260 		     inode, page, (int)pos, (int)len);
1261 
1262 		r = ceph_update_writeable_page(file, pos, len, page);
1263 		if (r < 0)
1264 			put_page(page);
1265 		else
1266 			*pagep = page;
1267 	} while (r == -EAGAIN);
1268 
1269 	return r;
1270 }
1271 
1272 /*
1273  * we don't do anything in here that simple_write_end doesn't do
1274  * except adjust dirty page accounting
1275  */
1276 static int ceph_write_end(struct file *file, struct address_space *mapping,
1277 			  loff_t pos, unsigned len, unsigned copied,
1278 			  struct page *page, void *fsdata)
1279 {
1280 	struct inode *inode = file_inode(file);
1281 	unsigned from = pos & (PAGE_SIZE - 1);
1282 	int check_cap = 0;
1283 
1284 	dout("write_end file %p inode %p page %p %d~%d (%d)\n", file,
1285 	     inode, page, (int)pos, (int)copied, (int)len);
1286 
1287 	/* zero the stale part of the page if we did a short copy */
1288 	if (copied < len)
1289 		zero_user_segment(page, from+copied, len);
1290 
1291 	/* did file size increase? */
1292 	if (pos+copied > i_size_read(inode))
1293 		check_cap = ceph_inode_set_size(inode, pos+copied);
1294 
1295 	if (!PageUptodate(page))
1296 		SetPageUptodate(page);
1297 
1298 	set_page_dirty(page);
1299 
1300 	unlock_page(page);
1301 	put_page(page);
1302 
1303 	if (check_cap)
1304 		ceph_check_caps(ceph_inode(inode), CHECK_CAPS_AUTHONLY, NULL);
1305 
1306 	return copied;
1307 }
1308 
1309 /*
1310  * we set .direct_IO to indicate direct io is supported, but since we
1311  * intercept O_DIRECT reads and writes early, this function should
1312  * never get called.
1313  */
1314 static ssize_t ceph_direct_io(struct kiocb *iocb, struct iov_iter *iter)
1315 {
1316 	WARN_ON(1);
1317 	return -EINVAL;
1318 }
1319 
1320 const struct address_space_operations ceph_aops = {
1321 	.readpage = ceph_readpage,
1322 	.readpages = ceph_readpages,
1323 	.writepage = ceph_writepage,
1324 	.writepages = ceph_writepages_start,
1325 	.write_begin = ceph_write_begin,
1326 	.write_end = ceph_write_end,
1327 	.set_page_dirty = ceph_set_page_dirty,
1328 	.invalidatepage = ceph_invalidatepage,
1329 	.releasepage = ceph_releasepage,
1330 	.direct_IO = ceph_direct_io,
1331 };
1332 
1333 static void ceph_block_sigs(sigset_t *oldset)
1334 {
1335 	sigset_t mask;
1336 	siginitsetinv(&mask, sigmask(SIGKILL));
1337 	sigprocmask(SIG_BLOCK, &mask, oldset);
1338 }
1339 
1340 static void ceph_restore_sigs(sigset_t *oldset)
1341 {
1342 	sigprocmask(SIG_SETMASK, oldset, NULL);
1343 }
1344 
1345 /*
1346  * vm ops
1347  */
1348 static int ceph_filemap_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
1349 {
1350 	struct inode *inode = file_inode(vma->vm_file);
1351 	struct ceph_inode_info *ci = ceph_inode(inode);
1352 	struct ceph_file_info *fi = vma->vm_file->private_data;
1353 	struct page *pinned_page = NULL;
1354 	loff_t off = vmf->pgoff << PAGE_SHIFT;
1355 	int want, got, ret;
1356 	sigset_t oldset;
1357 
1358 	ceph_block_sigs(&oldset);
1359 
1360 	dout("filemap_fault %p %llx.%llx %llu~%zd trying to get caps\n",
1361 	     inode, ceph_vinop(inode), off, (size_t)PAGE_SIZE);
1362 	if (fi->fmode & CEPH_FILE_MODE_LAZY)
1363 		want = CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO;
1364 	else
1365 		want = CEPH_CAP_FILE_CACHE;
1366 
1367 	got = 0;
1368 	ret = ceph_get_caps(ci, CEPH_CAP_FILE_RD, want, -1, &got, &pinned_page);
1369 	if (ret < 0)
1370 		goto out_restore;
1371 
1372 	dout("filemap_fault %p %llu~%zd got cap refs on %s\n",
1373 	     inode, off, (size_t)PAGE_SIZE, ceph_cap_string(got));
1374 
1375 	if ((got & (CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO)) ||
1376 	    ci->i_inline_version == CEPH_INLINE_NONE)
1377 		ret = filemap_fault(vma, vmf);
1378 	else
1379 		ret = -EAGAIN;
1380 
1381 	dout("filemap_fault %p %llu~%zd dropping cap refs on %s ret %d\n",
1382 	     inode, off, (size_t)PAGE_SIZE, ceph_cap_string(got), ret);
1383 	if (pinned_page)
1384 		put_page(pinned_page);
1385 	ceph_put_cap_refs(ci, got);
1386 
1387 	if (ret != -EAGAIN)
1388 		goto out_restore;
1389 
1390 	/* read inline data */
1391 	if (off >= PAGE_SIZE) {
1392 		/* does not support inline data > PAGE_SIZE */
1393 		ret = VM_FAULT_SIGBUS;
1394 	} else {
1395 		int ret1;
1396 		struct address_space *mapping = inode->i_mapping;
1397 		struct page *page = find_or_create_page(mapping, 0,
1398 						mapping_gfp_constraint(mapping,
1399 						~__GFP_FS));
1400 		if (!page) {
1401 			ret = VM_FAULT_OOM;
1402 			goto out_inline;
1403 		}
1404 		ret1 = __ceph_do_getattr(inode, page,
1405 					 CEPH_STAT_CAP_INLINE_DATA, true);
1406 		if (ret1 < 0 || off >= i_size_read(inode)) {
1407 			unlock_page(page);
1408 			put_page(page);
1409 			if (ret1 < 0)
1410 				ret = ret1;
1411 			else
1412 				ret = VM_FAULT_SIGBUS;
1413 			goto out_inline;
1414 		}
1415 		if (ret1 < PAGE_SIZE)
1416 			zero_user_segment(page, ret1, PAGE_SIZE);
1417 		else
1418 			flush_dcache_page(page);
1419 		SetPageUptodate(page);
1420 		vmf->page = page;
1421 		ret = VM_FAULT_MAJOR | VM_FAULT_LOCKED;
1422 out_inline:
1423 		dout("filemap_fault %p %llu~%zd read inline data ret %d\n",
1424 		     inode, off, (size_t)PAGE_SIZE, ret);
1425 	}
1426 out_restore:
1427 	ceph_restore_sigs(&oldset);
1428 	if (ret < 0)
1429 		ret = (ret == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS;
1430 
1431 	return ret;
1432 }
1433 
1434 /*
1435  * Reuse write_begin here for simplicity.
1436  */
1437 static int ceph_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
1438 {
1439 	struct inode *inode = file_inode(vma->vm_file);
1440 	struct ceph_inode_info *ci = ceph_inode(inode);
1441 	struct ceph_file_info *fi = vma->vm_file->private_data;
1442 	struct ceph_cap_flush *prealloc_cf;
1443 	struct page *page = vmf->page;
1444 	loff_t off = page_offset(page);
1445 	loff_t size = i_size_read(inode);
1446 	size_t len;
1447 	int want, got, ret;
1448 	sigset_t oldset;
1449 
1450 	prealloc_cf = ceph_alloc_cap_flush();
1451 	if (!prealloc_cf)
1452 		return VM_FAULT_OOM;
1453 
1454 	ceph_block_sigs(&oldset);
1455 
1456 	if (ci->i_inline_version != CEPH_INLINE_NONE) {
1457 		struct page *locked_page = NULL;
1458 		if (off == 0) {
1459 			lock_page(page);
1460 			locked_page = page;
1461 		}
1462 		ret = ceph_uninline_data(vma->vm_file, locked_page);
1463 		if (locked_page)
1464 			unlock_page(locked_page);
1465 		if (ret < 0)
1466 			goto out_free;
1467 	}
1468 
1469 	if (off + PAGE_SIZE <= size)
1470 		len = PAGE_SIZE;
1471 	else
1472 		len = size & ~PAGE_MASK;
1473 
1474 	dout("page_mkwrite %p %llx.%llx %llu~%zd getting caps i_size %llu\n",
1475 	     inode, ceph_vinop(inode), off, len, size);
1476 	if (fi->fmode & CEPH_FILE_MODE_LAZY)
1477 		want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO;
1478 	else
1479 		want = CEPH_CAP_FILE_BUFFER;
1480 
1481 	got = 0;
1482 	ret = ceph_get_caps(ci, CEPH_CAP_FILE_WR, want, off + len,
1483 			    &got, NULL);
1484 	if (ret < 0)
1485 		goto out_free;
1486 
1487 	dout("page_mkwrite %p %llu~%zd got cap refs on %s\n",
1488 	     inode, off, len, ceph_cap_string(got));
1489 
1490 	/* Update time before taking page lock */
1491 	file_update_time(vma->vm_file);
1492 
1493 	do {
1494 		lock_page(page);
1495 
1496 		if ((off > size) || (page->mapping != inode->i_mapping)) {
1497 			unlock_page(page);
1498 			ret = VM_FAULT_NOPAGE;
1499 			break;
1500 		}
1501 
1502 		ret = ceph_update_writeable_page(vma->vm_file, off, len, page);
1503 		if (ret >= 0) {
1504 			/* success.  we'll keep the page locked. */
1505 			set_page_dirty(page);
1506 			ret = VM_FAULT_LOCKED;
1507 		}
1508 	} while (ret == -EAGAIN);
1509 
1510 	if (ret == VM_FAULT_LOCKED ||
1511 	    ci->i_inline_version != CEPH_INLINE_NONE) {
1512 		int dirty;
1513 		spin_lock(&ci->i_ceph_lock);
1514 		ci->i_inline_version = CEPH_INLINE_NONE;
1515 		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR,
1516 					       &prealloc_cf);
1517 		spin_unlock(&ci->i_ceph_lock);
1518 		if (dirty)
1519 			__mark_inode_dirty(inode, dirty);
1520 	}
1521 
1522 	dout("page_mkwrite %p %llu~%zd dropping cap refs on %s ret %d\n",
1523 	     inode, off, len, ceph_cap_string(got), ret);
1524 	ceph_put_cap_refs(ci, got);
1525 out_free:
1526 	ceph_restore_sigs(&oldset);
1527 	ceph_free_cap_flush(prealloc_cf);
1528 	if (ret < 0)
1529 		ret = (ret == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS;
1530 	return ret;
1531 }
1532 
1533 void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1534 			   char	*data, size_t len)
1535 {
1536 	struct address_space *mapping = inode->i_mapping;
1537 	struct page *page;
1538 
1539 	if (locked_page) {
1540 		page = locked_page;
1541 	} else {
1542 		if (i_size_read(inode) == 0)
1543 			return;
1544 		page = find_or_create_page(mapping, 0,
1545 					   mapping_gfp_constraint(mapping,
1546 					   ~__GFP_FS));
1547 		if (!page)
1548 			return;
1549 		if (PageUptodate(page)) {
1550 			unlock_page(page);
1551 			put_page(page);
1552 			return;
1553 		}
1554 	}
1555 
1556 	dout("fill_inline_data %p %llx.%llx len %zu locked_page %p\n",
1557 	     inode, ceph_vinop(inode), len, locked_page);
1558 
1559 	if (len > 0) {
1560 		void *kaddr = kmap_atomic(page);
1561 		memcpy(kaddr, data, len);
1562 		kunmap_atomic(kaddr);
1563 	}
1564 
1565 	if (page != locked_page) {
1566 		if (len < PAGE_SIZE)
1567 			zero_user_segment(page, len, PAGE_SIZE);
1568 		else
1569 			flush_dcache_page(page);
1570 
1571 		SetPageUptodate(page);
1572 		unlock_page(page);
1573 		put_page(page);
1574 	}
1575 }
1576 
1577 int ceph_uninline_data(struct file *filp, struct page *locked_page)
1578 {
1579 	struct inode *inode = file_inode(filp);
1580 	struct ceph_inode_info *ci = ceph_inode(inode);
1581 	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);
1582 	struct ceph_osd_request *req;
1583 	struct page *page = NULL;
1584 	u64 len, inline_version;
1585 	int err = 0;
1586 	bool from_pagecache = false;
1587 
1588 	spin_lock(&ci->i_ceph_lock);
1589 	inline_version = ci->i_inline_version;
1590 	spin_unlock(&ci->i_ceph_lock);
1591 
1592 	dout("uninline_data %p %llx.%llx inline_version %llu\n",
1593 	     inode, ceph_vinop(inode), inline_version);
1594 
1595 	if (inline_version == 1 || /* initial version, no data */
1596 	    inline_version == CEPH_INLINE_NONE)
1597 		goto out;
1598 
1599 	if (locked_page) {
1600 		page = locked_page;
1601 		WARN_ON(!PageUptodate(page));
1602 	} else if (ceph_caps_issued(ci) &
1603 		   (CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO)) {
1604 		page = find_get_page(inode->i_mapping, 0);
1605 		if (page) {
1606 			if (PageUptodate(page)) {
1607 				from_pagecache = true;
1608 				lock_page(page);
1609 			} else {
1610 				put_page(page);
1611 				page = NULL;
1612 			}
1613 		}
1614 	}
1615 
1616 	if (page) {
1617 		len = i_size_read(inode);
1618 		if (len > PAGE_SIZE)
1619 			len = PAGE_SIZE;
1620 	} else {
1621 		page = __page_cache_alloc(GFP_NOFS);
1622 		if (!page) {
1623 			err = -ENOMEM;
1624 			goto out;
1625 		}
1626 		err = __ceph_do_getattr(inode, page,
1627 					CEPH_STAT_CAP_INLINE_DATA, true);
1628 		if (err < 0) {
1629 			/* no inline data */
1630 			if (err == -ENODATA)
1631 				err = 0;
1632 			goto out;
1633 		}
1634 		len = err;
1635 	}
1636 
1637 	req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout,
1638 				    ceph_vino(inode), 0, &len, 0, 1,
1639 				    CEPH_OSD_OP_CREATE,
1640 				    CEPH_OSD_FLAG_ONDISK | CEPH_OSD_FLAG_WRITE,
1641 				    NULL, 0, 0, false);
1642 	if (IS_ERR(req)) {
1643 		err = PTR_ERR(req);
1644 		goto out;
1645 	}
1646 
1647 	req->r_mtime = inode->i_mtime;
1648 	err = ceph_osdc_start_request(&fsc->client->osdc, req, false);
1649 	if (!err)
1650 		err = ceph_osdc_wait_request(&fsc->client->osdc, req);
1651 	ceph_osdc_put_request(req);
1652 	if (err < 0)
1653 		goto out;
1654 
1655 	req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout,
1656 				    ceph_vino(inode), 0, &len, 1, 3,
1657 				    CEPH_OSD_OP_WRITE,
1658 				    CEPH_OSD_FLAG_ONDISK | CEPH_OSD_FLAG_WRITE,
1659 				    NULL, ci->i_truncate_seq,
1660 				    ci->i_truncate_size, false);
1661 	if (IS_ERR(req)) {
1662 		err = PTR_ERR(req);
1663 		goto out;
1664 	}
1665 
1666 	osd_req_op_extent_osd_data_pages(req, 1, &page, len, 0, false, false);
1667 
1668 	{
1669 		__le64 xattr_buf = cpu_to_le64(inline_version);
1670 		err = osd_req_op_xattr_init(req, 0, CEPH_OSD_OP_CMPXATTR,
1671 					    "inline_version", &xattr_buf,
1672 					    sizeof(xattr_buf),
1673 					    CEPH_OSD_CMPXATTR_OP_GT,
1674 					    CEPH_OSD_CMPXATTR_MODE_U64);
1675 		if (err)
1676 			goto out_put;
1677 	}
1678 
1679 	{
1680 		char xattr_buf[32];
1681 		int xattr_len = snprintf(xattr_buf, sizeof(xattr_buf),
1682 					 "%llu", inline_version);
1683 		err = osd_req_op_xattr_init(req, 2, CEPH_OSD_OP_SETXATTR,
1684 					    "inline_version",
1685 					    xattr_buf, xattr_len, 0, 0);
1686 		if (err)
1687 			goto out_put;
1688 	}
1689 
1690 	req->r_mtime = inode->i_mtime;
1691 	err = ceph_osdc_start_request(&fsc->client->osdc, req, false);
1692 	if (!err)
1693 		err = ceph_osdc_wait_request(&fsc->client->osdc, req);
1694 out_put:
1695 	ceph_osdc_put_request(req);
1696 	if (err == -ECANCELED)
1697 		err = 0;
1698 out:
1699 	if (page && page != locked_page) {
1700 		if (from_pagecache) {
1701 			unlock_page(page);
1702 			put_page(page);
1703 		} else
1704 			__free_pages(page, 0);
1705 	}
1706 
1707 	dout("uninline_data %p %llx.%llx inline_version %llu = %d\n",
1708 	     inode, ceph_vinop(inode), inline_version, err);
1709 	return err;
1710 }
1711 
1712 static const struct vm_operations_struct ceph_vmops = {
1713 	.fault		= ceph_filemap_fault,
1714 	.page_mkwrite	= ceph_page_mkwrite,
1715 };
1716 
1717 int ceph_mmap(struct file *file, struct vm_area_struct *vma)
1718 {
1719 	struct address_space *mapping = file->f_mapping;
1720 
1721 	if (!mapping->a_ops->readpage)
1722 		return -ENOEXEC;
1723 	file_accessed(file);
1724 	vma->vm_ops = &ceph_vmops;
1725 	return 0;
1726 }
1727 
1728 enum {
1729 	POOL_READ	= 1,
1730 	POOL_WRITE	= 2,
1731 };
1732 
1733 static int __ceph_pool_perm_get(struct ceph_inode_info *ci, u32 pool)
1734 {
1735 	struct ceph_fs_client *fsc = ceph_inode_to_client(&ci->vfs_inode);
1736 	struct ceph_mds_client *mdsc = fsc->mdsc;
1737 	struct ceph_osd_request *rd_req = NULL, *wr_req = NULL;
1738 	struct rb_node **p, *parent;
1739 	struct ceph_pool_perm *perm;
1740 	struct page **pages;
1741 	int err = 0, err2 = 0, have = 0;
1742 
1743 	down_read(&mdsc->pool_perm_rwsem);
1744 	p = &mdsc->pool_perm_tree.rb_node;
1745 	while (*p) {
1746 		perm = rb_entry(*p, struct ceph_pool_perm, node);
1747 		if (pool < perm->pool)
1748 			p = &(*p)->rb_left;
1749 		else if (pool > perm->pool)
1750 			p = &(*p)->rb_right;
1751 		else {
1752 			have = perm->perm;
1753 			break;
1754 		}
1755 	}
1756 	up_read(&mdsc->pool_perm_rwsem);
1757 	if (*p)
1758 		goto out;
1759 
1760 	dout("__ceph_pool_perm_get pool %u no perm cached\n", pool);
1761 
1762 	down_write(&mdsc->pool_perm_rwsem);
1763 	parent = NULL;
1764 	while (*p) {
1765 		parent = *p;
1766 		perm = rb_entry(parent, struct ceph_pool_perm, node);
1767 		if (pool < perm->pool)
1768 			p = &(*p)->rb_left;
1769 		else if (pool > perm->pool)
1770 			p = &(*p)->rb_right;
1771 		else {
1772 			have = perm->perm;
1773 			break;
1774 		}
1775 	}
1776 	if (*p) {
1777 		up_write(&mdsc->pool_perm_rwsem);
1778 		goto out;
1779 	}
1780 
1781 	rd_req = ceph_osdc_alloc_request(&fsc->client->osdc, NULL,
1782 					 1, false, GFP_NOFS);
1783 	if (!rd_req) {
1784 		err = -ENOMEM;
1785 		goto out_unlock;
1786 	}
1787 
1788 	rd_req->r_flags = CEPH_OSD_FLAG_READ;
1789 	osd_req_op_init(rd_req, 0, CEPH_OSD_OP_STAT, 0);
1790 	rd_req->r_base_oloc.pool = pool;
1791 	ceph_oid_printf(&rd_req->r_base_oid, "%llx.00000000", ci->i_vino.ino);
1792 
1793 	err = ceph_osdc_alloc_messages(rd_req, GFP_NOFS);
1794 	if (err)
1795 		goto out_unlock;
1796 
1797 	wr_req = ceph_osdc_alloc_request(&fsc->client->osdc, NULL,
1798 					 1, false, GFP_NOFS);
1799 	if (!wr_req) {
1800 		err = -ENOMEM;
1801 		goto out_unlock;
1802 	}
1803 
1804 	wr_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ACK;
1805 	osd_req_op_init(wr_req, 0, CEPH_OSD_OP_CREATE, CEPH_OSD_OP_FLAG_EXCL);
1806 	ceph_oloc_copy(&wr_req->r_base_oloc, &rd_req->r_base_oloc);
1807 	ceph_oid_copy(&wr_req->r_base_oid, &rd_req->r_base_oid);
1808 
1809 	err = ceph_osdc_alloc_messages(wr_req, GFP_NOFS);
1810 	if (err)
1811 		goto out_unlock;
1812 
1813 	/* one page should be large enough for STAT data */
1814 	pages = ceph_alloc_page_vector(1, GFP_KERNEL);
1815 	if (IS_ERR(pages)) {
1816 		err = PTR_ERR(pages);
1817 		goto out_unlock;
1818 	}
1819 
1820 	osd_req_op_raw_data_in_pages(rd_req, 0, pages, PAGE_SIZE,
1821 				     0, false, true);
1822 	err = ceph_osdc_start_request(&fsc->client->osdc, rd_req, false);
1823 
1824 	wr_req->r_mtime = ci->vfs_inode.i_mtime;
1825 	err2 = ceph_osdc_start_request(&fsc->client->osdc, wr_req, false);
1826 
1827 	if (!err)
1828 		err = ceph_osdc_wait_request(&fsc->client->osdc, rd_req);
1829 	if (!err2)
1830 		err2 = ceph_osdc_wait_request(&fsc->client->osdc, wr_req);
1831 
1832 	if (err >= 0 || err == -ENOENT)
1833 		have |= POOL_READ;
1834 	else if (err != -EPERM)
1835 		goto out_unlock;
1836 
1837 	if (err2 == 0 || err2 == -EEXIST)
1838 		have |= POOL_WRITE;
1839 	else if (err2 != -EPERM) {
1840 		err = err2;
1841 		goto out_unlock;
1842 	}
1843 
1844 	perm = kmalloc(sizeof(*perm), GFP_NOFS);
1845 	if (!perm) {
1846 		err = -ENOMEM;
1847 		goto out_unlock;
1848 	}
1849 
1850 	perm->pool = pool;
1851 	perm->perm = have;
1852 	rb_link_node(&perm->node, parent, p);
1853 	rb_insert_color(&perm->node, &mdsc->pool_perm_tree);
1854 	err = 0;
1855 out_unlock:
1856 	up_write(&mdsc->pool_perm_rwsem);
1857 
1858 	ceph_osdc_put_request(rd_req);
1859 	ceph_osdc_put_request(wr_req);
1860 out:
1861 	if (!err)
1862 		err = have;
1863 	dout("__ceph_pool_perm_get pool %u result = %d\n", pool, err);
1864 	return err;
1865 }
1866 
1867 int ceph_pool_perm_check(struct ceph_inode_info *ci, int need)
1868 {
1869 	u32 pool;
1870 	int ret, flags;
1871 
1872 	/* does not support pool namespace yet */
1873 	if (ci->i_pool_ns_len)
1874 		return -EIO;
1875 
1876 	if (ceph_test_mount_opt(ceph_inode_to_client(&ci->vfs_inode),
1877 				NOPOOLPERM))
1878 		return 0;
1879 
1880 	spin_lock(&ci->i_ceph_lock);
1881 	flags = ci->i_ceph_flags;
1882 	pool = ceph_file_layout_pg_pool(ci->i_layout);
1883 	spin_unlock(&ci->i_ceph_lock);
1884 check:
1885 	if (flags & CEPH_I_POOL_PERM) {
1886 		if ((need & CEPH_CAP_FILE_RD) && !(flags & CEPH_I_POOL_RD)) {
1887 			dout("ceph_pool_perm_check pool %u no read perm\n",
1888 			     pool);
1889 			return -EPERM;
1890 		}
1891 		if ((need & CEPH_CAP_FILE_WR) && !(flags & CEPH_I_POOL_WR)) {
1892 			dout("ceph_pool_perm_check pool %u no write perm\n",
1893 			     pool);
1894 			return -EPERM;
1895 		}
1896 		return 0;
1897 	}
1898 
1899 	ret = __ceph_pool_perm_get(ci, pool);
1900 	if (ret < 0)
1901 		return ret;
1902 
1903 	flags = CEPH_I_POOL_PERM;
1904 	if (ret & POOL_READ)
1905 		flags |= CEPH_I_POOL_RD;
1906 	if (ret & POOL_WRITE)
1907 		flags |= CEPH_I_POOL_WR;
1908 
1909 	spin_lock(&ci->i_ceph_lock);
1910 	if (pool == ceph_file_layout_pg_pool(ci->i_layout)) {
1911 		ci->i_ceph_flags = flags;
1912         } else {
1913 		pool = ceph_file_layout_pg_pool(ci->i_layout);
1914 		flags = ci->i_ceph_flags;
1915 	}
1916 	spin_unlock(&ci->i_ceph_lock);
1917 	goto check;
1918 }
1919 
1920 void ceph_pool_perm_destroy(struct ceph_mds_client *mdsc)
1921 {
1922 	struct ceph_pool_perm *perm;
1923 	struct rb_node *n;
1924 
1925 	while (!RB_EMPTY_ROOT(&mdsc->pool_perm_tree)) {
1926 		n = rb_first(&mdsc->pool_perm_tree);
1927 		perm = rb_entry(n, struct ceph_pool_perm, node);
1928 		rb_erase(n, &mdsc->pool_perm_tree);
1929 		kfree(perm);
1930 	}
1931 }
1932