xref: /linux/fs/ceph/addr.c (revision 55d0969c451159cff86949b38c39171cab962069)
1 // SPDX-License-Identifier: GPL-2.0
2 #include <linux/ceph/ceph_debug.h>
3 
4 #include <linux/backing-dev.h>
5 #include <linux/fs.h>
6 #include <linux/mm.h>
7 #include <linux/swap.h>
8 #include <linux/pagemap.h>
9 #include <linux/slab.h>
10 #include <linux/pagevec.h>
11 #include <linux/task_io_accounting_ops.h>
12 #include <linux/signal.h>
13 #include <linux/iversion.h>
14 #include <linux/ktime.h>
15 #include <linux/netfs.h>
16 #include <trace/events/netfs.h>
17 
18 #include "super.h"
19 #include "mds_client.h"
20 #include "cache.h"
21 #include "metric.h"
22 #include "crypto.h"
23 #include <linux/ceph/osd_client.h>
24 #include <linux/ceph/striper.h>
25 
26 /*
27  * Ceph address space ops.
28  *
29  * There are a few funny things going on here.
30  *
31  * The page->private field is used to reference a struct
32  * ceph_snap_context for _every_ dirty page.  This indicates which
33  * snapshot the page was logically dirtied in, and thus which snap
34  * context needs to be associated with the osd write during writeback.
35  *
36  * Similarly, struct ceph_inode_info maintains a set of counters to
37  * count dirty pages on the inode.  In the absence of snapshots,
38  * i_wrbuffer_ref == i_wrbuffer_ref_head == the dirty page count.
39  *
40  * When a snapshot is taken (that is, when the client receives
41  * notification that a snapshot was taken), each inode with caps and
42  * with dirty pages (dirty pages implies there is a cap) gets a new
43  * ceph_cap_snap in the i_cap_snaps list (which is sorted in ascending
44  * order, new snaps go to the tail).  The i_wrbuffer_ref_head count is
45  * moved to capsnap->dirty. (Unless a sync write is currently in
46  * progress.  In that case, the capsnap is said to be "pending", new
47  * writes cannot start, and the capsnap isn't "finalized" until the
48  * write completes (or fails) and a final size/mtime for the inode for
49  * that snap can be settled upon.)  i_wrbuffer_ref_head is reset to 0.
50  *
51  * On writeback, we must submit writes to the osd IN SNAP ORDER.  So,
52  * we look for the first capsnap in i_cap_snaps and write out pages in
53  * that snap context _only_.  Then we move on to the next capsnap,
54  * eventually reaching the "live" or "head" context (i.e., pages that
55  * are not yet snapped) and are writing the most recently dirtied
56  * pages.
57  *
58  * Invalidate and so forth must take care to ensure the dirty page
59  * accounting is preserved.
60  */
61 
62 #define CONGESTION_ON_THRESH(congestion_kb) (congestion_kb >> (PAGE_SHIFT-10))
63 #define CONGESTION_OFF_THRESH(congestion_kb)				\
64 	(CONGESTION_ON_THRESH(congestion_kb) -				\
65 	 (CONGESTION_ON_THRESH(congestion_kb) >> 2))
66 
67 static int ceph_netfs_check_write_begin(struct file *file, loff_t pos, unsigned int len,
68 					struct folio **foliop, void **_fsdata);
69 
70 static inline struct ceph_snap_context *page_snap_context(struct page *page)
71 {
72 	if (PagePrivate(page))
73 		return (void *)page->private;
74 	return NULL;
75 }
76 
77 /*
78  * Dirty a page.  Optimistically adjust accounting, on the assumption
79  * that we won't race with invalidate.  If we do, readjust.
80  */
81 static bool ceph_dirty_folio(struct address_space *mapping, struct folio *folio)
82 {
83 	struct inode *inode = mapping->host;
84 	struct ceph_client *cl = ceph_inode_to_client(inode);
85 	struct ceph_inode_info *ci;
86 	struct ceph_snap_context *snapc;
87 
88 	if (folio_test_dirty(folio)) {
89 		doutc(cl, "%llx.%llx %p idx %lu -- already dirty\n",
90 		      ceph_vinop(inode), folio, folio->index);
91 		VM_BUG_ON_FOLIO(!folio_test_private(folio), folio);
92 		return false;
93 	}
94 
95 	ci = ceph_inode(inode);
96 
97 	/* dirty the head */
98 	spin_lock(&ci->i_ceph_lock);
99 	if (__ceph_have_pending_cap_snap(ci)) {
100 		struct ceph_cap_snap *capsnap =
101 				list_last_entry(&ci->i_cap_snaps,
102 						struct ceph_cap_snap,
103 						ci_item);
104 		snapc = ceph_get_snap_context(capsnap->context);
105 		capsnap->dirty_pages++;
106 	} else {
107 		BUG_ON(!ci->i_head_snapc);
108 		snapc = ceph_get_snap_context(ci->i_head_snapc);
109 		++ci->i_wrbuffer_ref_head;
110 	}
111 	if (ci->i_wrbuffer_ref == 0)
112 		ihold(inode);
113 	++ci->i_wrbuffer_ref;
114 	doutc(cl, "%llx.%llx %p idx %lu head %d/%d -> %d/%d "
115 	      "snapc %p seq %lld (%d snaps)\n",
116 	      ceph_vinop(inode), folio, folio->index,
117 	      ci->i_wrbuffer_ref-1, ci->i_wrbuffer_ref_head-1,
118 	      ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
119 	      snapc, snapc->seq, snapc->num_snaps);
120 	spin_unlock(&ci->i_ceph_lock);
121 
122 	/*
123 	 * Reference snap context in folio->private.  Also set
124 	 * PagePrivate so that we get invalidate_folio callback.
125 	 */
126 	VM_WARN_ON_FOLIO(folio->private, folio);
127 	folio_attach_private(folio, snapc);
128 
129 	return ceph_fscache_dirty_folio(mapping, folio);
130 }
131 
132 /*
133  * If we are truncating the full folio (i.e. offset == 0), adjust the
134  * dirty folio counters appropriately.  Only called if there is private
135  * data on the folio.
136  */
137 static void ceph_invalidate_folio(struct folio *folio, size_t offset,
138 				size_t length)
139 {
140 	struct inode *inode = folio->mapping->host;
141 	struct ceph_client *cl = ceph_inode_to_client(inode);
142 	struct ceph_inode_info *ci = ceph_inode(inode);
143 	struct ceph_snap_context *snapc;
144 
145 
146 	if (offset != 0 || length != folio_size(folio)) {
147 		doutc(cl, "%llx.%llx idx %lu partial dirty page %zu~%zu\n",
148 		      ceph_vinop(inode), folio->index, offset, length);
149 		return;
150 	}
151 
152 	WARN_ON(!folio_test_locked(folio));
153 	if (folio_test_private(folio)) {
154 		doutc(cl, "%llx.%llx idx %lu full dirty page\n",
155 		      ceph_vinop(inode), folio->index);
156 
157 		snapc = folio_detach_private(folio);
158 		ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
159 		ceph_put_snap_context(snapc);
160 	}
161 
162 	netfs_invalidate_folio(folio, offset, length);
163 }
164 
165 static void ceph_netfs_expand_readahead(struct netfs_io_request *rreq)
166 {
167 	struct inode *inode = rreq->inode;
168 	struct ceph_inode_info *ci = ceph_inode(inode);
169 	struct ceph_file_layout *lo = &ci->i_layout;
170 	unsigned long max_pages = inode->i_sb->s_bdi->ra_pages;
171 	loff_t end = rreq->start + rreq->len, new_end;
172 	struct ceph_netfs_request_data *priv = rreq->netfs_priv;
173 	unsigned long max_len;
174 	u32 blockoff;
175 
176 	if (priv) {
177 		/* Readahead is disabled by posix_fadvise POSIX_FADV_RANDOM */
178 		if (priv->file_ra_disabled)
179 			max_pages = 0;
180 		else
181 			max_pages = priv->file_ra_pages;
182 
183 	}
184 
185 	/* Readahead is disabled */
186 	if (!max_pages)
187 		return;
188 
189 	max_len = max_pages << PAGE_SHIFT;
190 
191 	/*
192 	 * Try to expand the length forward by rounding up it to the next
193 	 * block, but do not exceed the file size, unless the original
194 	 * request already exceeds it.
195 	 */
196 	new_end = umin(round_up(end, lo->stripe_unit), rreq->i_size);
197 	if (new_end > end && new_end <= rreq->start + max_len)
198 		rreq->len = new_end - rreq->start;
199 
200 	/* Try to expand the start downward */
201 	div_u64_rem(rreq->start, lo->stripe_unit, &blockoff);
202 	if (rreq->len + blockoff <= max_len) {
203 		rreq->start -= blockoff;
204 		rreq->len += blockoff;
205 	}
206 }
207 
208 static void finish_netfs_read(struct ceph_osd_request *req)
209 {
210 	struct inode *inode = req->r_inode;
211 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
212 	struct ceph_client *cl = fsc->client;
213 	struct ceph_osd_data *osd_data = osd_req_op_extent_osd_data(req, 0);
214 	struct netfs_io_subrequest *subreq = req->r_priv;
215 	struct ceph_osd_req_op *op = &req->r_ops[0];
216 	int err = req->r_result;
217 	bool sparse = (op->op == CEPH_OSD_OP_SPARSE_READ);
218 
219 	ceph_update_read_metrics(&fsc->mdsc->metric, req->r_start_latency,
220 				 req->r_end_latency, osd_data->length, err);
221 
222 	doutc(cl, "result %d subreq->len=%zu i_size=%lld\n", req->r_result,
223 	      subreq->len, i_size_read(req->r_inode));
224 
225 	/* no object means success but no data */
226 	if (err == -ENOENT)
227 		err = 0;
228 	else if (err == -EBLOCKLISTED)
229 		fsc->blocklisted = true;
230 
231 	if (err >= 0) {
232 		if (sparse && err > 0)
233 			err = ceph_sparse_ext_map_end(op);
234 		if (err < subreq->len &&
235 		    subreq->rreq->origin != NETFS_DIO_READ)
236 			__set_bit(NETFS_SREQ_CLEAR_TAIL, &subreq->flags);
237 		if (IS_ENCRYPTED(inode) && err > 0) {
238 			err = ceph_fscrypt_decrypt_extents(inode,
239 					osd_data->pages, subreq->start,
240 					op->extent.sparse_ext,
241 					op->extent.sparse_ext_cnt);
242 			if (err > subreq->len)
243 				err = subreq->len;
244 		}
245 	}
246 
247 	if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGES) {
248 		ceph_put_page_vector(osd_data->pages,
249 				     calc_pages_for(osd_data->alignment,
250 					osd_data->length), false);
251 	}
252 	if (err > 0) {
253 		subreq->transferred = err;
254 		err = 0;
255 	}
256 	trace_netfs_sreq(subreq, netfs_sreq_trace_io_progress);
257 	netfs_read_subreq_terminated(subreq, err, false);
258 	iput(req->r_inode);
259 	ceph_dec_osd_stopping_blocker(fsc->mdsc);
260 }
261 
262 static bool ceph_netfs_issue_op_inline(struct netfs_io_subrequest *subreq)
263 {
264 	struct netfs_io_request *rreq = subreq->rreq;
265 	struct inode *inode = rreq->inode;
266 	struct ceph_mds_reply_info_parsed *rinfo;
267 	struct ceph_mds_reply_info_in *iinfo;
268 	struct ceph_mds_request *req;
269 	struct ceph_mds_client *mdsc = ceph_sb_to_mdsc(inode->i_sb);
270 	struct ceph_inode_info *ci = ceph_inode(inode);
271 	ssize_t err = 0;
272 	size_t len;
273 	int mode;
274 
275 	if (rreq->origin != NETFS_DIO_READ)
276 		__set_bit(NETFS_SREQ_CLEAR_TAIL, &subreq->flags);
277 	__clear_bit(NETFS_SREQ_COPY_TO_CACHE, &subreq->flags);
278 
279 	if (subreq->start >= inode->i_size)
280 		goto out;
281 
282 	/* We need to fetch the inline data. */
283 	mode = ceph_try_to_choose_auth_mds(inode, CEPH_STAT_CAP_INLINE_DATA);
284 	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, mode);
285 	if (IS_ERR(req)) {
286 		err = PTR_ERR(req);
287 		goto out;
288 	}
289 	req->r_ino1 = ci->i_vino;
290 	req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INLINE_DATA);
291 	req->r_num_caps = 2;
292 
293 	trace_netfs_sreq(subreq, netfs_sreq_trace_submit);
294 	err = ceph_mdsc_do_request(mdsc, NULL, req);
295 	if (err < 0)
296 		goto out;
297 
298 	rinfo = &req->r_reply_info;
299 	iinfo = &rinfo->targeti;
300 	if (iinfo->inline_version == CEPH_INLINE_NONE) {
301 		/* The data got uninlined */
302 		ceph_mdsc_put_request(req);
303 		return false;
304 	}
305 
306 	len = min_t(size_t, iinfo->inline_len - subreq->start, subreq->len);
307 	err = copy_to_iter(iinfo->inline_data + subreq->start, len, &subreq->io_iter);
308 	if (err == 0) {
309 		err = -EFAULT;
310 	} else {
311 		subreq->transferred += err;
312 		err = 0;
313 	}
314 
315 	ceph_mdsc_put_request(req);
316 out:
317 	netfs_read_subreq_terminated(subreq, err, false);
318 	return true;
319 }
320 
321 static int ceph_netfs_prepare_read(struct netfs_io_subrequest *subreq)
322 {
323 	struct netfs_io_request *rreq = subreq->rreq;
324 	struct inode *inode = rreq->inode;
325 	struct ceph_inode_info *ci = ceph_inode(inode);
326 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
327 	u64 objno, objoff;
328 	u32 xlen;
329 
330 	/* Truncate the extent at the end of the current block */
331 	ceph_calc_file_object_mapping(&ci->i_layout, subreq->start, subreq->len,
332 				      &objno, &objoff, &xlen);
333 	rreq->io_streams[0].sreq_max_len = umin(xlen, fsc->mount_options->rsize);
334 	return 0;
335 }
336 
337 static void ceph_netfs_issue_read(struct netfs_io_subrequest *subreq)
338 {
339 	struct netfs_io_request *rreq = subreq->rreq;
340 	struct inode *inode = rreq->inode;
341 	struct ceph_inode_info *ci = ceph_inode(inode);
342 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
343 	struct ceph_client *cl = fsc->client;
344 	struct ceph_osd_request *req = NULL;
345 	struct ceph_vino vino = ceph_vino(inode);
346 	int err;
347 	u64 len;
348 	bool sparse = IS_ENCRYPTED(inode) || ceph_test_mount_opt(fsc, SPARSEREAD);
349 	u64 off = subreq->start;
350 	int extent_cnt;
351 
352 	if (ceph_inode_is_shutdown(inode)) {
353 		err = -EIO;
354 		goto out;
355 	}
356 
357 	if (ceph_has_inline_data(ci) && ceph_netfs_issue_op_inline(subreq))
358 		return;
359 
360 	// TODO: This rounding here is slightly dodgy.  It *should* work, for
361 	// now, as the cache only deals in blocks that are a multiple of
362 	// PAGE_SIZE and fscrypt blocks are at most PAGE_SIZE.  What needs to
363 	// happen is for the fscrypt driving to be moved into netfslib and the
364 	// data in the cache also to be stored encrypted.
365 	len = subreq->len;
366 	ceph_fscrypt_adjust_off_and_len(inode, &off, &len);
367 
368 	req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout, vino,
369 			off, &len, 0, 1, sparse ? CEPH_OSD_OP_SPARSE_READ : CEPH_OSD_OP_READ,
370 			CEPH_OSD_FLAG_READ, NULL, ci->i_truncate_seq,
371 			ci->i_truncate_size, false);
372 	if (IS_ERR(req)) {
373 		err = PTR_ERR(req);
374 		req = NULL;
375 		goto out;
376 	}
377 
378 	if (sparse) {
379 		extent_cnt = __ceph_sparse_read_ext_count(inode, len);
380 		err = ceph_alloc_sparse_ext_map(&req->r_ops[0], extent_cnt);
381 		if (err)
382 			goto out;
383 	}
384 
385 	doutc(cl, "%llx.%llx pos=%llu orig_len=%zu len=%llu\n",
386 	      ceph_vinop(inode), subreq->start, subreq->len, len);
387 
388 	/*
389 	 * FIXME: For now, use CEPH_OSD_DATA_TYPE_PAGES instead of _ITER for
390 	 * encrypted inodes. We'd need infrastructure that handles an iov_iter
391 	 * instead of page arrays, and we don't have that as of yet. Once the
392 	 * dust settles on the write helpers and encrypt/decrypt routines for
393 	 * netfs, we should be able to rework this.
394 	 */
395 	if (IS_ENCRYPTED(inode)) {
396 		struct page **pages;
397 		size_t page_off;
398 
399 		err = iov_iter_get_pages_alloc2(&subreq->io_iter, &pages, len, &page_off);
400 		if (err < 0) {
401 			doutc(cl, "%llx.%llx failed to allocate pages, %d\n",
402 			      ceph_vinop(inode), err);
403 			goto out;
404 		}
405 
406 		/* should always give us a page-aligned read */
407 		WARN_ON_ONCE(page_off);
408 		len = err;
409 		err = 0;
410 
411 		osd_req_op_extent_osd_data_pages(req, 0, pages, len, 0, false,
412 						 false);
413 	} else {
414 		osd_req_op_extent_osd_iter(req, 0, &subreq->io_iter);
415 	}
416 	if (!ceph_inc_osd_stopping_blocker(fsc->mdsc)) {
417 		err = -EIO;
418 		goto out;
419 	}
420 	req->r_callback = finish_netfs_read;
421 	req->r_priv = subreq;
422 	req->r_inode = inode;
423 	ihold(inode);
424 
425 	trace_netfs_sreq(subreq, netfs_sreq_trace_submit);
426 	ceph_osdc_start_request(req->r_osdc, req);
427 out:
428 	ceph_osdc_put_request(req);
429 	if (err)
430 		netfs_read_subreq_terminated(subreq, err, false);
431 	doutc(cl, "%llx.%llx result %d\n", ceph_vinop(inode), err);
432 }
433 
434 static int ceph_init_request(struct netfs_io_request *rreq, struct file *file)
435 {
436 	struct inode *inode = rreq->inode;
437 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
438 	struct ceph_client *cl = ceph_inode_to_client(inode);
439 	int got = 0, want = CEPH_CAP_FILE_CACHE;
440 	struct ceph_netfs_request_data *priv;
441 	int ret = 0;
442 
443 	/* [DEPRECATED] Use PG_private_2 to mark folio being written to the cache. */
444 	__set_bit(NETFS_RREQ_USE_PGPRIV2, &rreq->flags);
445 
446 	if (rreq->origin != NETFS_READAHEAD)
447 		return 0;
448 
449 	priv = kzalloc(sizeof(*priv), GFP_NOFS);
450 	if (!priv)
451 		return -ENOMEM;
452 
453 	if (file) {
454 		struct ceph_rw_context *rw_ctx;
455 		struct ceph_file_info *fi = file->private_data;
456 
457 		priv->file_ra_pages = file->f_ra.ra_pages;
458 		priv->file_ra_disabled = file->f_mode & FMODE_RANDOM;
459 
460 		rw_ctx = ceph_find_rw_context(fi);
461 		if (rw_ctx) {
462 			rreq->netfs_priv = priv;
463 			return 0;
464 		}
465 	}
466 
467 	/*
468 	 * readahead callers do not necessarily hold Fcb caps
469 	 * (e.g. fadvise, madvise).
470 	 */
471 	ret = ceph_try_get_caps(inode, CEPH_CAP_FILE_RD, want, true, &got);
472 	if (ret < 0) {
473 		doutc(cl, "%llx.%llx, error getting cap\n", ceph_vinop(inode));
474 		goto out;
475 	}
476 
477 	if (!(got & want)) {
478 		doutc(cl, "%llx.%llx, no cache cap\n", ceph_vinop(inode));
479 		ret = -EACCES;
480 		goto out;
481 	}
482 	if (ret == 0) {
483 		ret = -EACCES;
484 		goto out;
485 	}
486 
487 	priv->caps = got;
488 	rreq->netfs_priv = priv;
489 	rreq->io_streams[0].sreq_max_len = fsc->mount_options->rsize;
490 
491 out:
492 	if (ret < 0) {
493 		if (got)
494 			ceph_put_cap_refs(ceph_inode(inode), got);
495 		kfree(priv);
496 	}
497 
498 	return ret;
499 }
500 
501 static void ceph_netfs_free_request(struct netfs_io_request *rreq)
502 {
503 	struct ceph_netfs_request_data *priv = rreq->netfs_priv;
504 
505 	if (!priv)
506 		return;
507 
508 	if (priv->caps)
509 		ceph_put_cap_refs(ceph_inode(rreq->inode), priv->caps);
510 	kfree(priv);
511 	rreq->netfs_priv = NULL;
512 }
513 
514 const struct netfs_request_ops ceph_netfs_ops = {
515 	.init_request		= ceph_init_request,
516 	.free_request		= ceph_netfs_free_request,
517 	.prepare_read		= ceph_netfs_prepare_read,
518 	.issue_read		= ceph_netfs_issue_read,
519 	.expand_readahead	= ceph_netfs_expand_readahead,
520 	.check_write_begin	= ceph_netfs_check_write_begin,
521 };
522 
523 #ifdef CONFIG_CEPH_FSCACHE
524 static void ceph_set_page_fscache(struct page *page)
525 {
526 	folio_start_private_2(page_folio(page)); /* [DEPRECATED] */
527 }
528 
529 static void ceph_fscache_write_terminated(void *priv, ssize_t error, bool was_async)
530 {
531 	struct inode *inode = priv;
532 
533 	if (IS_ERR_VALUE(error) && error != -ENOBUFS)
534 		ceph_fscache_invalidate(inode, false);
535 }
536 
537 static void ceph_fscache_write_to_cache(struct inode *inode, u64 off, u64 len, bool caching)
538 {
539 	struct ceph_inode_info *ci = ceph_inode(inode);
540 	struct fscache_cookie *cookie = ceph_fscache_cookie(ci);
541 
542 	fscache_write_to_cache(cookie, inode->i_mapping, off, len, i_size_read(inode),
543 			       ceph_fscache_write_terminated, inode, true, caching);
544 }
545 #else
546 static inline void ceph_set_page_fscache(struct page *page)
547 {
548 }
549 
550 static inline void ceph_fscache_write_to_cache(struct inode *inode, u64 off, u64 len, bool caching)
551 {
552 }
553 #endif /* CONFIG_CEPH_FSCACHE */
554 
555 struct ceph_writeback_ctl
556 {
557 	loff_t i_size;
558 	u64 truncate_size;
559 	u32 truncate_seq;
560 	bool size_stable;
561 	bool head_snapc;
562 };
563 
564 /*
565  * Get ref for the oldest snapc for an inode with dirty data... that is, the
566  * only snap context we are allowed to write back.
567  */
568 static struct ceph_snap_context *
569 get_oldest_context(struct inode *inode, struct ceph_writeback_ctl *ctl,
570 		   struct ceph_snap_context *page_snapc)
571 {
572 	struct ceph_inode_info *ci = ceph_inode(inode);
573 	struct ceph_client *cl = ceph_inode_to_client(inode);
574 	struct ceph_snap_context *snapc = NULL;
575 	struct ceph_cap_snap *capsnap = NULL;
576 
577 	spin_lock(&ci->i_ceph_lock);
578 	list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
579 		doutc(cl, " capsnap %p snapc %p has %d dirty pages\n",
580 		      capsnap, capsnap->context, capsnap->dirty_pages);
581 		if (!capsnap->dirty_pages)
582 			continue;
583 
584 		/* get i_size, truncate_{seq,size} for page_snapc? */
585 		if (snapc && capsnap->context != page_snapc)
586 			continue;
587 
588 		if (ctl) {
589 			if (capsnap->writing) {
590 				ctl->i_size = i_size_read(inode);
591 				ctl->size_stable = false;
592 			} else {
593 				ctl->i_size = capsnap->size;
594 				ctl->size_stable = true;
595 			}
596 			ctl->truncate_size = capsnap->truncate_size;
597 			ctl->truncate_seq = capsnap->truncate_seq;
598 			ctl->head_snapc = false;
599 		}
600 
601 		if (snapc)
602 			break;
603 
604 		snapc = ceph_get_snap_context(capsnap->context);
605 		if (!page_snapc ||
606 		    page_snapc == snapc ||
607 		    page_snapc->seq > snapc->seq)
608 			break;
609 	}
610 	if (!snapc && ci->i_wrbuffer_ref_head) {
611 		snapc = ceph_get_snap_context(ci->i_head_snapc);
612 		doutc(cl, " head snapc %p has %d dirty pages\n", snapc,
613 		      ci->i_wrbuffer_ref_head);
614 		if (ctl) {
615 			ctl->i_size = i_size_read(inode);
616 			ctl->truncate_size = ci->i_truncate_size;
617 			ctl->truncate_seq = ci->i_truncate_seq;
618 			ctl->size_stable = false;
619 			ctl->head_snapc = true;
620 		}
621 	}
622 	spin_unlock(&ci->i_ceph_lock);
623 	return snapc;
624 }
625 
626 static u64 get_writepages_data_length(struct inode *inode,
627 				      struct page *page, u64 start)
628 {
629 	struct ceph_inode_info *ci = ceph_inode(inode);
630 	struct ceph_snap_context *snapc;
631 	struct ceph_cap_snap *capsnap = NULL;
632 	u64 end = i_size_read(inode);
633 	u64 ret;
634 
635 	snapc = page_snap_context(ceph_fscrypt_pagecache_page(page));
636 	if (snapc != ci->i_head_snapc) {
637 		bool found = false;
638 		spin_lock(&ci->i_ceph_lock);
639 		list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
640 			if (capsnap->context == snapc) {
641 				if (!capsnap->writing)
642 					end = capsnap->size;
643 				found = true;
644 				break;
645 			}
646 		}
647 		spin_unlock(&ci->i_ceph_lock);
648 		WARN_ON(!found);
649 	}
650 	if (end > ceph_fscrypt_page_offset(page) + thp_size(page))
651 		end = ceph_fscrypt_page_offset(page) + thp_size(page);
652 	ret = end > start ? end - start : 0;
653 	if (ret && fscrypt_is_bounce_page(page))
654 		ret = round_up(ret, CEPH_FSCRYPT_BLOCK_SIZE);
655 	return ret;
656 }
657 
658 /*
659  * Write a single page, but leave the page locked.
660  *
661  * If we get a write error, mark the mapping for error, but still adjust the
662  * dirty page accounting (i.e., page is no longer dirty).
663  */
664 static int writepage_nounlock(struct page *page, struct writeback_control *wbc)
665 {
666 	struct folio *folio = page_folio(page);
667 	struct inode *inode = page->mapping->host;
668 	struct ceph_inode_info *ci = ceph_inode(inode);
669 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
670 	struct ceph_client *cl = fsc->client;
671 	struct ceph_snap_context *snapc, *oldest;
672 	loff_t page_off = page_offset(page);
673 	int err;
674 	loff_t len = thp_size(page);
675 	loff_t wlen;
676 	struct ceph_writeback_ctl ceph_wbc;
677 	struct ceph_osd_client *osdc = &fsc->client->osdc;
678 	struct ceph_osd_request *req;
679 	bool caching = ceph_is_cache_enabled(inode);
680 	struct page *bounce_page = NULL;
681 
682 	doutc(cl, "%llx.%llx page %p idx %lu\n", ceph_vinop(inode), page,
683 	      page->index);
684 
685 	if (ceph_inode_is_shutdown(inode))
686 		return -EIO;
687 
688 	/* verify this is a writeable snap context */
689 	snapc = page_snap_context(page);
690 	if (!snapc) {
691 		doutc(cl, "%llx.%llx page %p not dirty?\n", ceph_vinop(inode),
692 		      page);
693 		return 0;
694 	}
695 	oldest = get_oldest_context(inode, &ceph_wbc, snapc);
696 	if (snapc->seq > oldest->seq) {
697 		doutc(cl, "%llx.%llx page %p snapc %p not writeable - noop\n",
698 		      ceph_vinop(inode), page, snapc);
699 		/* we should only noop if called by kswapd */
700 		WARN_ON(!(current->flags & PF_MEMALLOC));
701 		ceph_put_snap_context(oldest);
702 		redirty_page_for_writepage(wbc, page);
703 		return 0;
704 	}
705 	ceph_put_snap_context(oldest);
706 
707 	/* is this a partial page at end of file? */
708 	if (page_off >= ceph_wbc.i_size) {
709 		doutc(cl, "%llx.%llx folio at %lu beyond eof %llu\n",
710 		      ceph_vinop(inode), folio->index, ceph_wbc.i_size);
711 		folio_invalidate(folio, 0, folio_size(folio));
712 		return 0;
713 	}
714 
715 	if (ceph_wbc.i_size < page_off + len)
716 		len = ceph_wbc.i_size - page_off;
717 
718 	wlen = IS_ENCRYPTED(inode) ? round_up(len, CEPH_FSCRYPT_BLOCK_SIZE) : len;
719 	doutc(cl, "%llx.%llx page %p index %lu on %llu~%llu snapc %p seq %lld\n",
720 	      ceph_vinop(inode), page, page->index, page_off, wlen, snapc,
721 	      snapc->seq);
722 
723 	if (atomic_long_inc_return(&fsc->writeback_count) >
724 	    CONGESTION_ON_THRESH(fsc->mount_options->congestion_kb))
725 		fsc->write_congested = true;
726 
727 	req = ceph_osdc_new_request(osdc, &ci->i_layout, ceph_vino(inode),
728 				    page_off, &wlen, 0, 1, CEPH_OSD_OP_WRITE,
729 				    CEPH_OSD_FLAG_WRITE, snapc,
730 				    ceph_wbc.truncate_seq,
731 				    ceph_wbc.truncate_size, true);
732 	if (IS_ERR(req)) {
733 		redirty_page_for_writepage(wbc, page);
734 		return PTR_ERR(req);
735 	}
736 
737 	if (wlen < len)
738 		len = wlen;
739 
740 	set_page_writeback(page);
741 	if (caching)
742 		ceph_set_page_fscache(page);
743 	ceph_fscache_write_to_cache(inode, page_off, len, caching);
744 
745 	if (IS_ENCRYPTED(inode)) {
746 		bounce_page = fscrypt_encrypt_pagecache_blocks(page,
747 						    CEPH_FSCRYPT_BLOCK_SIZE, 0,
748 						    GFP_NOFS);
749 		if (IS_ERR(bounce_page)) {
750 			redirty_page_for_writepage(wbc, page);
751 			end_page_writeback(page);
752 			ceph_osdc_put_request(req);
753 			return PTR_ERR(bounce_page);
754 		}
755 	}
756 
757 	/* it may be a short write due to an object boundary */
758 	WARN_ON_ONCE(len > thp_size(page));
759 	osd_req_op_extent_osd_data_pages(req, 0,
760 			bounce_page ? &bounce_page : &page, wlen, 0,
761 			false, false);
762 	doutc(cl, "%llx.%llx %llu~%llu (%llu bytes, %sencrypted)\n",
763 	      ceph_vinop(inode), page_off, len, wlen,
764 	      IS_ENCRYPTED(inode) ? "" : "not ");
765 
766 	req->r_mtime = inode_get_mtime(inode);
767 	ceph_osdc_start_request(osdc, req);
768 	err = ceph_osdc_wait_request(osdc, req);
769 
770 	ceph_update_write_metrics(&fsc->mdsc->metric, req->r_start_latency,
771 				  req->r_end_latency, len, err);
772 	fscrypt_free_bounce_page(bounce_page);
773 	ceph_osdc_put_request(req);
774 	if (err == 0)
775 		err = len;
776 
777 	if (err < 0) {
778 		struct writeback_control tmp_wbc;
779 		if (!wbc)
780 			wbc = &tmp_wbc;
781 		if (err == -ERESTARTSYS) {
782 			/* killed by SIGKILL */
783 			doutc(cl, "%llx.%llx interrupted page %p\n",
784 			      ceph_vinop(inode), page);
785 			redirty_page_for_writepage(wbc, page);
786 			end_page_writeback(page);
787 			return err;
788 		}
789 		if (err == -EBLOCKLISTED)
790 			fsc->blocklisted = true;
791 		doutc(cl, "%llx.%llx setting page/mapping error %d %p\n",
792 		      ceph_vinop(inode), err, page);
793 		mapping_set_error(&inode->i_data, err);
794 		wbc->pages_skipped++;
795 	} else {
796 		doutc(cl, "%llx.%llx cleaned page %p\n",
797 		      ceph_vinop(inode), page);
798 		err = 0;  /* vfs expects us to return 0 */
799 	}
800 	oldest = detach_page_private(page);
801 	WARN_ON_ONCE(oldest != snapc);
802 	end_page_writeback(page);
803 	ceph_put_wrbuffer_cap_refs(ci, 1, snapc);
804 	ceph_put_snap_context(snapc);  /* page's reference */
805 
806 	if (atomic_long_dec_return(&fsc->writeback_count) <
807 	    CONGESTION_OFF_THRESH(fsc->mount_options->congestion_kb))
808 		fsc->write_congested = false;
809 
810 	return err;
811 }
812 
813 static int ceph_writepage(struct page *page, struct writeback_control *wbc)
814 {
815 	int err;
816 	struct inode *inode = page->mapping->host;
817 	BUG_ON(!inode);
818 	ihold(inode);
819 
820 	if (wbc->sync_mode == WB_SYNC_NONE &&
821 	    ceph_inode_to_fs_client(inode)->write_congested) {
822 		redirty_page_for_writepage(wbc, page);
823 		return AOP_WRITEPAGE_ACTIVATE;
824 	}
825 
826 	folio_wait_private_2(page_folio(page)); /* [DEPRECATED] */
827 
828 	err = writepage_nounlock(page, wbc);
829 	if (err == -ERESTARTSYS) {
830 		/* direct memory reclaimer was killed by SIGKILL. return 0
831 		 * to prevent caller from setting mapping/page error */
832 		err = 0;
833 	}
834 	unlock_page(page);
835 	iput(inode);
836 	return err;
837 }
838 
839 /*
840  * async writeback completion handler.
841  *
842  * If we get an error, set the mapping error bit, but not the individual
843  * page error bits.
844  */
845 static void writepages_finish(struct ceph_osd_request *req)
846 {
847 	struct inode *inode = req->r_inode;
848 	struct ceph_inode_info *ci = ceph_inode(inode);
849 	struct ceph_client *cl = ceph_inode_to_client(inode);
850 	struct ceph_osd_data *osd_data;
851 	struct page *page;
852 	int num_pages, total_pages = 0;
853 	int i, j;
854 	int rc = req->r_result;
855 	struct ceph_snap_context *snapc = req->r_snapc;
856 	struct address_space *mapping = inode->i_mapping;
857 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
858 	unsigned int len = 0;
859 	bool remove_page;
860 
861 	doutc(cl, "%llx.%llx rc %d\n", ceph_vinop(inode), rc);
862 	if (rc < 0) {
863 		mapping_set_error(mapping, rc);
864 		ceph_set_error_write(ci);
865 		if (rc == -EBLOCKLISTED)
866 			fsc->blocklisted = true;
867 	} else {
868 		ceph_clear_error_write(ci);
869 	}
870 
871 	/*
872 	 * We lost the cache cap, need to truncate the page before
873 	 * it is unlocked, otherwise we'd truncate it later in the
874 	 * page truncation thread, possibly losing some data that
875 	 * raced its way in
876 	 */
877 	remove_page = !(ceph_caps_issued(ci) &
878 			(CEPH_CAP_FILE_CACHE|CEPH_CAP_FILE_LAZYIO));
879 
880 	/* clean all pages */
881 	for (i = 0; i < req->r_num_ops; i++) {
882 		if (req->r_ops[i].op != CEPH_OSD_OP_WRITE) {
883 			pr_warn_client(cl,
884 				"%llx.%llx incorrect op %d req %p index %d tid %llu\n",
885 				ceph_vinop(inode), req->r_ops[i].op, req, i,
886 				req->r_tid);
887 			break;
888 		}
889 
890 		osd_data = osd_req_op_extent_osd_data(req, i);
891 		BUG_ON(osd_data->type != CEPH_OSD_DATA_TYPE_PAGES);
892 		len += osd_data->length;
893 		num_pages = calc_pages_for((u64)osd_data->alignment,
894 					   (u64)osd_data->length);
895 		total_pages += num_pages;
896 		for (j = 0; j < num_pages; j++) {
897 			page = osd_data->pages[j];
898 			if (fscrypt_is_bounce_page(page)) {
899 				page = fscrypt_pagecache_page(page);
900 				fscrypt_free_bounce_page(osd_data->pages[j]);
901 				osd_data->pages[j] = page;
902 			}
903 			BUG_ON(!page);
904 			WARN_ON(!PageUptodate(page));
905 
906 			if (atomic_long_dec_return(&fsc->writeback_count) <
907 			     CONGESTION_OFF_THRESH(
908 					fsc->mount_options->congestion_kb))
909 				fsc->write_congested = false;
910 
911 			ceph_put_snap_context(detach_page_private(page));
912 			end_page_writeback(page);
913 			doutc(cl, "unlocking %p\n", page);
914 
915 			if (remove_page)
916 				generic_error_remove_folio(inode->i_mapping,
917 							  page_folio(page));
918 
919 			unlock_page(page);
920 		}
921 		doutc(cl, "%llx.%llx wrote %llu bytes cleaned %d pages\n",
922 		      ceph_vinop(inode), osd_data->length,
923 		      rc >= 0 ? num_pages : 0);
924 
925 		release_pages(osd_data->pages, num_pages);
926 	}
927 
928 	ceph_update_write_metrics(&fsc->mdsc->metric, req->r_start_latency,
929 				  req->r_end_latency, len, rc);
930 
931 	ceph_put_wrbuffer_cap_refs(ci, total_pages, snapc);
932 
933 	osd_data = osd_req_op_extent_osd_data(req, 0);
934 	if (osd_data->pages_from_pool)
935 		mempool_free(osd_data->pages, ceph_wb_pagevec_pool);
936 	else
937 		kfree(osd_data->pages);
938 	ceph_osdc_put_request(req);
939 	ceph_dec_osd_stopping_blocker(fsc->mdsc);
940 }
941 
942 /*
943  * initiate async writeback
944  */
945 static int ceph_writepages_start(struct address_space *mapping,
946 				 struct writeback_control *wbc)
947 {
948 	struct inode *inode = mapping->host;
949 	struct ceph_inode_info *ci = ceph_inode(inode);
950 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
951 	struct ceph_client *cl = fsc->client;
952 	struct ceph_vino vino = ceph_vino(inode);
953 	pgoff_t index, start_index, end = -1;
954 	struct ceph_snap_context *snapc = NULL, *last_snapc = NULL, *pgsnapc;
955 	struct folio_batch fbatch;
956 	int rc = 0;
957 	unsigned int wsize = i_blocksize(inode);
958 	struct ceph_osd_request *req = NULL;
959 	struct ceph_writeback_ctl ceph_wbc;
960 	bool should_loop, range_whole = false;
961 	bool done = false;
962 	bool caching = ceph_is_cache_enabled(inode);
963 	xa_mark_t tag;
964 
965 	if (wbc->sync_mode == WB_SYNC_NONE &&
966 	    fsc->write_congested)
967 		return 0;
968 
969 	doutc(cl, "%llx.%llx (mode=%s)\n", ceph_vinop(inode),
970 	      wbc->sync_mode == WB_SYNC_NONE ? "NONE" :
971 	      (wbc->sync_mode == WB_SYNC_ALL ? "ALL" : "HOLD"));
972 
973 	if (ceph_inode_is_shutdown(inode)) {
974 		if (ci->i_wrbuffer_ref > 0) {
975 			pr_warn_ratelimited_client(cl,
976 				"%llx.%llx %lld forced umount\n",
977 				ceph_vinop(inode), ceph_ino(inode));
978 		}
979 		mapping_set_error(mapping, -EIO);
980 		return -EIO; /* we're in a forced umount, don't write! */
981 	}
982 	if (fsc->mount_options->wsize < wsize)
983 		wsize = fsc->mount_options->wsize;
984 
985 	folio_batch_init(&fbatch);
986 
987 	start_index = wbc->range_cyclic ? mapping->writeback_index : 0;
988 	index = start_index;
989 
990 	if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages) {
991 		tag = PAGECACHE_TAG_TOWRITE;
992 	} else {
993 		tag = PAGECACHE_TAG_DIRTY;
994 	}
995 retry:
996 	/* find oldest snap context with dirty data */
997 	snapc = get_oldest_context(inode, &ceph_wbc, NULL);
998 	if (!snapc) {
999 		/* hmm, why does writepages get called when there
1000 		   is no dirty data? */
1001 		doutc(cl, " no snap context with dirty data?\n");
1002 		goto out;
1003 	}
1004 	doutc(cl, " oldest snapc is %p seq %lld (%d snaps)\n", snapc,
1005 	      snapc->seq, snapc->num_snaps);
1006 
1007 	should_loop = false;
1008 	if (ceph_wbc.head_snapc && snapc != last_snapc) {
1009 		/* where to start/end? */
1010 		if (wbc->range_cyclic) {
1011 			index = start_index;
1012 			end = -1;
1013 			if (index > 0)
1014 				should_loop = true;
1015 			doutc(cl, " cyclic, start at %lu\n", index);
1016 		} else {
1017 			index = wbc->range_start >> PAGE_SHIFT;
1018 			end = wbc->range_end >> PAGE_SHIFT;
1019 			if (wbc->range_start == 0 && wbc->range_end == LLONG_MAX)
1020 				range_whole = true;
1021 			doutc(cl, " not cyclic, %lu to %lu\n", index, end);
1022 		}
1023 	} else if (!ceph_wbc.head_snapc) {
1024 		/* Do not respect wbc->range_{start,end}. Dirty pages
1025 		 * in that range can be associated with newer snapc.
1026 		 * They are not writeable until we write all dirty pages
1027 		 * associated with 'snapc' get written */
1028 		if (index > 0)
1029 			should_loop = true;
1030 		doutc(cl, " non-head snapc, range whole\n");
1031 	}
1032 
1033 	if (wbc->sync_mode == WB_SYNC_ALL || wbc->tagged_writepages)
1034 		tag_pages_for_writeback(mapping, index, end);
1035 
1036 	ceph_put_snap_context(last_snapc);
1037 	last_snapc = snapc;
1038 
1039 	while (!done && index <= end) {
1040 		int num_ops = 0, op_idx;
1041 		unsigned i, nr_folios, max_pages, locked_pages = 0;
1042 		struct page **pages = NULL, **data_pages;
1043 		struct page *page;
1044 		pgoff_t strip_unit_end = 0;
1045 		u64 offset = 0, len = 0;
1046 		bool from_pool = false;
1047 
1048 		max_pages = wsize >> PAGE_SHIFT;
1049 
1050 get_more_pages:
1051 		nr_folios = filemap_get_folios_tag(mapping, &index,
1052 						   end, tag, &fbatch);
1053 		doutc(cl, "pagevec_lookup_range_tag got %d\n", nr_folios);
1054 		if (!nr_folios && !locked_pages)
1055 			break;
1056 		for (i = 0; i < nr_folios && locked_pages < max_pages; i++) {
1057 			page = &fbatch.folios[i]->page;
1058 			doutc(cl, "? %p idx %lu\n", page, page->index);
1059 			if (locked_pages == 0)
1060 				lock_page(page);  /* first page */
1061 			else if (!trylock_page(page))
1062 				break;
1063 
1064 			/* only dirty pages, or our accounting breaks */
1065 			if (unlikely(!PageDirty(page)) ||
1066 			    unlikely(page->mapping != mapping)) {
1067 				doutc(cl, "!dirty or !mapping %p\n", page);
1068 				unlock_page(page);
1069 				continue;
1070 			}
1071 			/* only if matching snap context */
1072 			pgsnapc = page_snap_context(page);
1073 			if (pgsnapc != snapc) {
1074 				doutc(cl, "page snapc %p %lld != oldest %p %lld\n",
1075 				      pgsnapc, pgsnapc->seq, snapc, snapc->seq);
1076 				if (!should_loop &&
1077 				    !ceph_wbc.head_snapc &&
1078 				    wbc->sync_mode != WB_SYNC_NONE)
1079 					should_loop = true;
1080 				unlock_page(page);
1081 				continue;
1082 			}
1083 			if (page_offset(page) >= ceph_wbc.i_size) {
1084 				struct folio *folio = page_folio(page);
1085 
1086 				doutc(cl, "folio at %lu beyond eof %llu\n",
1087 				      folio->index, ceph_wbc.i_size);
1088 				if ((ceph_wbc.size_stable ||
1089 				    folio_pos(folio) >= i_size_read(inode)) &&
1090 				    folio_clear_dirty_for_io(folio))
1091 					folio_invalidate(folio, 0,
1092 							folio_size(folio));
1093 				folio_unlock(folio);
1094 				continue;
1095 			}
1096 			if (strip_unit_end && (page->index > strip_unit_end)) {
1097 				doutc(cl, "end of strip unit %p\n", page);
1098 				unlock_page(page);
1099 				break;
1100 			}
1101 			if (PageWriteback(page) ||
1102 			    PagePrivate2(page) /* [DEPRECATED] */) {
1103 				if (wbc->sync_mode == WB_SYNC_NONE) {
1104 					doutc(cl, "%p under writeback\n", page);
1105 					unlock_page(page);
1106 					continue;
1107 				}
1108 				doutc(cl, "waiting on writeback %p\n", page);
1109 				wait_on_page_writeback(page);
1110 				folio_wait_private_2(page_folio(page)); /* [DEPRECATED] */
1111 			}
1112 
1113 			if (!clear_page_dirty_for_io(page)) {
1114 				doutc(cl, "%p !clear_page_dirty_for_io\n", page);
1115 				unlock_page(page);
1116 				continue;
1117 			}
1118 
1119 			/*
1120 			 * We have something to write.  If this is
1121 			 * the first locked page this time through,
1122 			 * calculate max possinle write size and
1123 			 * allocate a page array
1124 			 */
1125 			if (locked_pages == 0) {
1126 				u64 objnum;
1127 				u64 objoff;
1128 				u32 xlen;
1129 
1130 				/* prepare async write request */
1131 				offset = (u64)page_offset(page);
1132 				ceph_calc_file_object_mapping(&ci->i_layout,
1133 							      offset, wsize,
1134 							      &objnum, &objoff,
1135 							      &xlen);
1136 				len = xlen;
1137 
1138 				num_ops = 1;
1139 				strip_unit_end = page->index +
1140 					((len - 1) >> PAGE_SHIFT);
1141 
1142 				BUG_ON(pages);
1143 				max_pages = calc_pages_for(0, (u64)len);
1144 				pages = kmalloc_array(max_pages,
1145 						      sizeof(*pages),
1146 						      GFP_NOFS);
1147 				if (!pages) {
1148 					from_pool = true;
1149 					pages = mempool_alloc(ceph_wb_pagevec_pool, GFP_NOFS);
1150 					BUG_ON(!pages);
1151 				}
1152 
1153 				len = 0;
1154 			} else if (page->index !=
1155 				   (offset + len) >> PAGE_SHIFT) {
1156 				if (num_ops >= (from_pool ?  CEPH_OSD_SLAB_OPS :
1157 							     CEPH_OSD_MAX_OPS)) {
1158 					redirty_page_for_writepage(wbc, page);
1159 					unlock_page(page);
1160 					break;
1161 				}
1162 
1163 				num_ops++;
1164 				offset = (u64)page_offset(page);
1165 				len = 0;
1166 			}
1167 
1168 			/* note position of first page in fbatch */
1169 			doutc(cl, "%llx.%llx will write page %p idx %lu\n",
1170 			      ceph_vinop(inode), page, page->index);
1171 
1172 			if (atomic_long_inc_return(&fsc->writeback_count) >
1173 			    CONGESTION_ON_THRESH(
1174 				    fsc->mount_options->congestion_kb))
1175 				fsc->write_congested = true;
1176 
1177 			if (IS_ENCRYPTED(inode)) {
1178 				pages[locked_pages] =
1179 					fscrypt_encrypt_pagecache_blocks(page,
1180 						PAGE_SIZE, 0,
1181 						locked_pages ? GFP_NOWAIT : GFP_NOFS);
1182 				if (IS_ERR(pages[locked_pages])) {
1183 					if (PTR_ERR(pages[locked_pages]) == -EINVAL)
1184 						pr_err_client(cl,
1185 							"inode->i_blkbits=%hhu\n",
1186 							inode->i_blkbits);
1187 					/* better not fail on first page! */
1188 					BUG_ON(locked_pages == 0);
1189 					pages[locked_pages] = NULL;
1190 					redirty_page_for_writepage(wbc, page);
1191 					unlock_page(page);
1192 					break;
1193 				}
1194 				++locked_pages;
1195 			} else {
1196 				pages[locked_pages++] = page;
1197 			}
1198 
1199 			fbatch.folios[i] = NULL;
1200 			len += thp_size(page);
1201 		}
1202 
1203 		/* did we get anything? */
1204 		if (!locked_pages)
1205 			goto release_folios;
1206 		if (i) {
1207 			unsigned j, n = 0;
1208 			/* shift unused page to beginning of fbatch */
1209 			for (j = 0; j < nr_folios; j++) {
1210 				if (!fbatch.folios[j])
1211 					continue;
1212 				if (n < j)
1213 					fbatch.folios[n] = fbatch.folios[j];
1214 				n++;
1215 			}
1216 			fbatch.nr = n;
1217 
1218 			if (nr_folios && i == nr_folios &&
1219 			    locked_pages < max_pages) {
1220 				doutc(cl, "reached end fbatch, trying for more\n");
1221 				folio_batch_release(&fbatch);
1222 				goto get_more_pages;
1223 			}
1224 		}
1225 
1226 new_request:
1227 		offset = ceph_fscrypt_page_offset(pages[0]);
1228 		len = wsize;
1229 
1230 		req = ceph_osdc_new_request(&fsc->client->osdc,
1231 					&ci->i_layout, vino,
1232 					offset, &len, 0, num_ops,
1233 					CEPH_OSD_OP_WRITE, CEPH_OSD_FLAG_WRITE,
1234 					snapc, ceph_wbc.truncate_seq,
1235 					ceph_wbc.truncate_size, false);
1236 		if (IS_ERR(req)) {
1237 			req = ceph_osdc_new_request(&fsc->client->osdc,
1238 						&ci->i_layout, vino,
1239 						offset, &len, 0,
1240 						min(num_ops,
1241 						    CEPH_OSD_SLAB_OPS),
1242 						CEPH_OSD_OP_WRITE,
1243 						CEPH_OSD_FLAG_WRITE,
1244 						snapc, ceph_wbc.truncate_seq,
1245 						ceph_wbc.truncate_size, true);
1246 			BUG_ON(IS_ERR(req));
1247 		}
1248 		BUG_ON(len < ceph_fscrypt_page_offset(pages[locked_pages - 1]) +
1249 			     thp_size(pages[locked_pages - 1]) - offset);
1250 
1251 		if (!ceph_inc_osd_stopping_blocker(fsc->mdsc)) {
1252 			rc = -EIO;
1253 			goto release_folios;
1254 		}
1255 		req->r_callback = writepages_finish;
1256 		req->r_inode = inode;
1257 
1258 		/* Format the osd request message and submit the write */
1259 		len = 0;
1260 		data_pages = pages;
1261 		op_idx = 0;
1262 		for (i = 0; i < locked_pages; i++) {
1263 			struct page *page = ceph_fscrypt_pagecache_page(pages[i]);
1264 
1265 			u64 cur_offset = page_offset(page);
1266 			/*
1267 			 * Discontinuity in page range? Ceph can handle that by just passing
1268 			 * multiple extents in the write op.
1269 			 */
1270 			if (offset + len != cur_offset) {
1271 				/* If it's full, stop here */
1272 				if (op_idx + 1 == req->r_num_ops)
1273 					break;
1274 
1275 				/* Kick off an fscache write with what we have so far. */
1276 				ceph_fscache_write_to_cache(inode, offset, len, caching);
1277 
1278 				/* Start a new extent */
1279 				osd_req_op_extent_dup_last(req, op_idx,
1280 							   cur_offset - offset);
1281 				doutc(cl, "got pages at %llu~%llu\n", offset,
1282 				      len);
1283 				osd_req_op_extent_osd_data_pages(req, op_idx,
1284 							data_pages, len, 0,
1285 							from_pool, false);
1286 				osd_req_op_extent_update(req, op_idx, len);
1287 
1288 				len = 0;
1289 				offset = cur_offset;
1290 				data_pages = pages + i;
1291 				op_idx++;
1292 			}
1293 
1294 			set_page_writeback(page);
1295 			if (caching)
1296 				ceph_set_page_fscache(page);
1297 			len += thp_size(page);
1298 		}
1299 		ceph_fscache_write_to_cache(inode, offset, len, caching);
1300 
1301 		if (ceph_wbc.size_stable) {
1302 			len = min(len, ceph_wbc.i_size - offset);
1303 		} else if (i == locked_pages) {
1304 			/* writepages_finish() clears writeback pages
1305 			 * according to the data length, so make sure
1306 			 * data length covers all locked pages */
1307 			u64 min_len = len + 1 - thp_size(page);
1308 			len = get_writepages_data_length(inode, pages[i - 1],
1309 							 offset);
1310 			len = max(len, min_len);
1311 		}
1312 		if (IS_ENCRYPTED(inode))
1313 			len = round_up(len, CEPH_FSCRYPT_BLOCK_SIZE);
1314 
1315 		doutc(cl, "got pages at %llu~%llu\n", offset, len);
1316 
1317 		if (IS_ENCRYPTED(inode) &&
1318 		    ((offset | len) & ~CEPH_FSCRYPT_BLOCK_MASK))
1319 			pr_warn_client(cl,
1320 				"bad encrypted write offset=%lld len=%llu\n",
1321 				offset, len);
1322 
1323 		osd_req_op_extent_osd_data_pages(req, op_idx, data_pages, len,
1324 						 0, from_pool, false);
1325 		osd_req_op_extent_update(req, op_idx, len);
1326 
1327 		BUG_ON(op_idx + 1 != req->r_num_ops);
1328 
1329 		from_pool = false;
1330 		if (i < locked_pages) {
1331 			BUG_ON(num_ops <= req->r_num_ops);
1332 			num_ops -= req->r_num_ops;
1333 			locked_pages -= i;
1334 
1335 			/* allocate new pages array for next request */
1336 			data_pages = pages;
1337 			pages = kmalloc_array(locked_pages, sizeof(*pages),
1338 					      GFP_NOFS);
1339 			if (!pages) {
1340 				from_pool = true;
1341 				pages = mempool_alloc(ceph_wb_pagevec_pool, GFP_NOFS);
1342 				BUG_ON(!pages);
1343 			}
1344 			memcpy(pages, data_pages + i,
1345 			       locked_pages * sizeof(*pages));
1346 			memset(data_pages + i, 0,
1347 			       locked_pages * sizeof(*pages));
1348 		} else {
1349 			BUG_ON(num_ops != req->r_num_ops);
1350 			index = pages[i - 1]->index + 1;
1351 			/* request message now owns the pages array */
1352 			pages = NULL;
1353 		}
1354 
1355 		req->r_mtime = inode_get_mtime(inode);
1356 		ceph_osdc_start_request(&fsc->client->osdc, req);
1357 		req = NULL;
1358 
1359 		wbc->nr_to_write -= i;
1360 		if (pages)
1361 			goto new_request;
1362 
1363 		/*
1364 		 * We stop writing back only if we are not doing
1365 		 * integrity sync. In case of integrity sync we have to
1366 		 * keep going until we have written all the pages
1367 		 * we tagged for writeback prior to entering this loop.
1368 		 */
1369 		if (wbc->nr_to_write <= 0 && wbc->sync_mode == WB_SYNC_NONE)
1370 			done = true;
1371 
1372 release_folios:
1373 		doutc(cl, "folio_batch release on %d folios (%p)\n",
1374 		      (int)fbatch.nr, fbatch.nr ? fbatch.folios[0] : NULL);
1375 		folio_batch_release(&fbatch);
1376 	}
1377 
1378 	if (should_loop && !done) {
1379 		/* more to do; loop back to beginning of file */
1380 		doutc(cl, "looping back to beginning of file\n");
1381 		end = start_index - 1; /* OK even when start_index == 0 */
1382 
1383 		/* to write dirty pages associated with next snapc,
1384 		 * we need to wait until current writes complete */
1385 		if (wbc->sync_mode != WB_SYNC_NONE &&
1386 		    start_index == 0 && /* all dirty pages were checked */
1387 		    !ceph_wbc.head_snapc) {
1388 			struct page *page;
1389 			unsigned i, nr;
1390 			index = 0;
1391 			while ((index <= end) &&
1392 			       (nr = filemap_get_folios_tag(mapping, &index,
1393 						(pgoff_t)-1,
1394 						PAGECACHE_TAG_WRITEBACK,
1395 						&fbatch))) {
1396 				for (i = 0; i < nr; i++) {
1397 					page = &fbatch.folios[i]->page;
1398 					if (page_snap_context(page) != snapc)
1399 						continue;
1400 					wait_on_page_writeback(page);
1401 				}
1402 				folio_batch_release(&fbatch);
1403 				cond_resched();
1404 			}
1405 		}
1406 
1407 		start_index = 0;
1408 		index = 0;
1409 		goto retry;
1410 	}
1411 
1412 	if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
1413 		mapping->writeback_index = index;
1414 
1415 out:
1416 	ceph_osdc_put_request(req);
1417 	ceph_put_snap_context(last_snapc);
1418 	doutc(cl, "%llx.%llx dend - startone, rc = %d\n", ceph_vinop(inode),
1419 	      rc);
1420 	return rc;
1421 }
1422 
1423 
1424 
1425 /*
1426  * See if a given @snapc is either writeable, or already written.
1427  */
1428 static int context_is_writeable_or_written(struct inode *inode,
1429 					   struct ceph_snap_context *snapc)
1430 {
1431 	struct ceph_snap_context *oldest = get_oldest_context(inode, NULL, NULL);
1432 	int ret = !oldest || snapc->seq <= oldest->seq;
1433 
1434 	ceph_put_snap_context(oldest);
1435 	return ret;
1436 }
1437 
1438 /**
1439  * ceph_find_incompatible - find an incompatible context and return it
1440  * @page: page being dirtied
1441  *
1442  * We are only allowed to write into/dirty a page if the page is
1443  * clean, or already dirty within the same snap context. Returns a
1444  * conflicting context if there is one, NULL if there isn't, or a
1445  * negative error code on other errors.
1446  *
1447  * Must be called with page lock held.
1448  */
1449 static struct ceph_snap_context *
1450 ceph_find_incompatible(struct page *page)
1451 {
1452 	struct inode *inode = page->mapping->host;
1453 	struct ceph_client *cl = ceph_inode_to_client(inode);
1454 	struct ceph_inode_info *ci = ceph_inode(inode);
1455 
1456 	if (ceph_inode_is_shutdown(inode)) {
1457 		doutc(cl, " %llx.%llx page %p is shutdown\n",
1458 		      ceph_vinop(inode), page);
1459 		return ERR_PTR(-ESTALE);
1460 	}
1461 
1462 	for (;;) {
1463 		struct ceph_snap_context *snapc, *oldest;
1464 
1465 		wait_on_page_writeback(page);
1466 
1467 		snapc = page_snap_context(page);
1468 		if (!snapc || snapc == ci->i_head_snapc)
1469 			break;
1470 
1471 		/*
1472 		 * this page is already dirty in another (older) snap
1473 		 * context!  is it writeable now?
1474 		 */
1475 		oldest = get_oldest_context(inode, NULL, NULL);
1476 		if (snapc->seq > oldest->seq) {
1477 			/* not writeable -- return it for the caller to deal with */
1478 			ceph_put_snap_context(oldest);
1479 			doutc(cl, " %llx.%llx page %p snapc %p not current or oldest\n",
1480 			      ceph_vinop(inode), page, snapc);
1481 			return ceph_get_snap_context(snapc);
1482 		}
1483 		ceph_put_snap_context(oldest);
1484 
1485 		/* yay, writeable, do it now (without dropping page lock) */
1486 		doutc(cl, " %llx.%llx page %p snapc %p not current, but oldest\n",
1487 		      ceph_vinop(inode), page, snapc);
1488 		if (clear_page_dirty_for_io(page)) {
1489 			int r = writepage_nounlock(page, NULL);
1490 			if (r < 0)
1491 				return ERR_PTR(r);
1492 		}
1493 	}
1494 	return NULL;
1495 }
1496 
1497 static int ceph_netfs_check_write_begin(struct file *file, loff_t pos, unsigned int len,
1498 					struct folio **foliop, void **_fsdata)
1499 {
1500 	struct inode *inode = file_inode(file);
1501 	struct ceph_inode_info *ci = ceph_inode(inode);
1502 	struct ceph_snap_context *snapc;
1503 
1504 	snapc = ceph_find_incompatible(folio_page(*foliop, 0));
1505 	if (snapc) {
1506 		int r;
1507 
1508 		folio_unlock(*foliop);
1509 		folio_put(*foliop);
1510 		*foliop = NULL;
1511 		if (IS_ERR(snapc))
1512 			return PTR_ERR(snapc);
1513 
1514 		ceph_queue_writeback(inode);
1515 		r = wait_event_killable(ci->i_cap_wq,
1516 					context_is_writeable_or_written(inode, snapc));
1517 		ceph_put_snap_context(snapc);
1518 		return r == 0 ? -EAGAIN : r;
1519 	}
1520 	return 0;
1521 }
1522 
1523 /*
1524  * We are only allowed to write into/dirty the page if the page is
1525  * clean, or already dirty within the same snap context.
1526  */
1527 static int ceph_write_begin(struct file *file, struct address_space *mapping,
1528 			    loff_t pos, unsigned len,
1529 			    struct folio **foliop, void **fsdata)
1530 {
1531 	struct inode *inode = file_inode(file);
1532 	struct ceph_inode_info *ci = ceph_inode(inode);
1533 	int r;
1534 
1535 	r = netfs_write_begin(&ci->netfs, file, inode->i_mapping, pos, len, foliop, NULL);
1536 	if (r < 0)
1537 		return r;
1538 
1539 	folio_wait_private_2(*foliop); /* [DEPRECATED] */
1540 	WARN_ON_ONCE(!folio_test_locked(*foliop));
1541 	return 0;
1542 }
1543 
1544 /*
1545  * we don't do anything in here that simple_write_end doesn't do
1546  * except adjust dirty page accounting
1547  */
1548 static int ceph_write_end(struct file *file, struct address_space *mapping,
1549 			  loff_t pos, unsigned len, unsigned copied,
1550 			  struct folio *folio, void *fsdata)
1551 {
1552 	struct inode *inode = file_inode(file);
1553 	struct ceph_client *cl = ceph_inode_to_client(inode);
1554 	bool check_cap = false;
1555 
1556 	doutc(cl, "%llx.%llx file %p folio %p %d~%d (%d)\n", ceph_vinop(inode),
1557 	      file, folio, (int)pos, (int)copied, (int)len);
1558 
1559 	if (!folio_test_uptodate(folio)) {
1560 		/* just return that nothing was copied on a short copy */
1561 		if (copied < len) {
1562 			copied = 0;
1563 			goto out;
1564 		}
1565 		folio_mark_uptodate(folio);
1566 	}
1567 
1568 	/* did file size increase? */
1569 	if (pos+copied > i_size_read(inode))
1570 		check_cap = ceph_inode_set_size(inode, pos+copied);
1571 
1572 	folio_mark_dirty(folio);
1573 
1574 out:
1575 	folio_unlock(folio);
1576 	folio_put(folio);
1577 
1578 	if (check_cap)
1579 		ceph_check_caps(ceph_inode(inode), CHECK_CAPS_AUTHONLY);
1580 
1581 	return copied;
1582 }
1583 
1584 const struct address_space_operations ceph_aops = {
1585 	.read_folio = netfs_read_folio,
1586 	.readahead = netfs_readahead,
1587 	.writepage = ceph_writepage,
1588 	.writepages = ceph_writepages_start,
1589 	.write_begin = ceph_write_begin,
1590 	.write_end = ceph_write_end,
1591 	.dirty_folio = ceph_dirty_folio,
1592 	.invalidate_folio = ceph_invalidate_folio,
1593 	.release_folio = netfs_release_folio,
1594 	.direct_IO = noop_direct_IO,
1595 };
1596 
1597 static void ceph_block_sigs(sigset_t *oldset)
1598 {
1599 	sigset_t mask;
1600 	siginitsetinv(&mask, sigmask(SIGKILL));
1601 	sigprocmask(SIG_BLOCK, &mask, oldset);
1602 }
1603 
1604 static void ceph_restore_sigs(sigset_t *oldset)
1605 {
1606 	sigprocmask(SIG_SETMASK, oldset, NULL);
1607 }
1608 
1609 /*
1610  * vm ops
1611  */
1612 static vm_fault_t ceph_filemap_fault(struct vm_fault *vmf)
1613 {
1614 	struct vm_area_struct *vma = vmf->vma;
1615 	struct inode *inode = file_inode(vma->vm_file);
1616 	struct ceph_inode_info *ci = ceph_inode(inode);
1617 	struct ceph_client *cl = ceph_inode_to_client(inode);
1618 	struct ceph_file_info *fi = vma->vm_file->private_data;
1619 	loff_t off = (loff_t)vmf->pgoff << PAGE_SHIFT;
1620 	int want, got, err;
1621 	sigset_t oldset;
1622 	vm_fault_t ret = VM_FAULT_SIGBUS;
1623 
1624 	if (ceph_inode_is_shutdown(inode))
1625 		return ret;
1626 
1627 	ceph_block_sigs(&oldset);
1628 
1629 	doutc(cl, "%llx.%llx %llu trying to get caps\n",
1630 	      ceph_vinop(inode), off);
1631 	if (fi->fmode & CEPH_FILE_MODE_LAZY)
1632 		want = CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO;
1633 	else
1634 		want = CEPH_CAP_FILE_CACHE;
1635 
1636 	got = 0;
1637 	err = ceph_get_caps(vma->vm_file, CEPH_CAP_FILE_RD, want, -1, &got);
1638 	if (err < 0)
1639 		goto out_restore;
1640 
1641 	doutc(cl, "%llx.%llx %llu got cap refs on %s\n", ceph_vinop(inode),
1642 	      off, ceph_cap_string(got));
1643 
1644 	if ((got & (CEPH_CAP_FILE_CACHE | CEPH_CAP_FILE_LAZYIO)) ||
1645 	    !ceph_has_inline_data(ci)) {
1646 		CEPH_DEFINE_RW_CONTEXT(rw_ctx, got);
1647 		ceph_add_rw_context(fi, &rw_ctx);
1648 		ret = filemap_fault(vmf);
1649 		ceph_del_rw_context(fi, &rw_ctx);
1650 		doutc(cl, "%llx.%llx %llu drop cap refs %s ret %x\n",
1651 		      ceph_vinop(inode), off, ceph_cap_string(got), ret);
1652 	} else
1653 		err = -EAGAIN;
1654 
1655 	ceph_put_cap_refs(ci, got);
1656 
1657 	if (err != -EAGAIN)
1658 		goto out_restore;
1659 
1660 	/* read inline data */
1661 	if (off >= PAGE_SIZE) {
1662 		/* does not support inline data > PAGE_SIZE */
1663 		ret = VM_FAULT_SIGBUS;
1664 	} else {
1665 		struct address_space *mapping = inode->i_mapping;
1666 		struct page *page;
1667 
1668 		filemap_invalidate_lock_shared(mapping);
1669 		page = find_or_create_page(mapping, 0,
1670 				mapping_gfp_constraint(mapping, ~__GFP_FS));
1671 		if (!page) {
1672 			ret = VM_FAULT_OOM;
1673 			goto out_inline;
1674 		}
1675 		err = __ceph_do_getattr(inode, page,
1676 					 CEPH_STAT_CAP_INLINE_DATA, true);
1677 		if (err < 0 || off >= i_size_read(inode)) {
1678 			unlock_page(page);
1679 			put_page(page);
1680 			ret = vmf_error(err);
1681 			goto out_inline;
1682 		}
1683 		if (err < PAGE_SIZE)
1684 			zero_user_segment(page, err, PAGE_SIZE);
1685 		else
1686 			flush_dcache_page(page);
1687 		SetPageUptodate(page);
1688 		vmf->page = page;
1689 		ret = VM_FAULT_MAJOR | VM_FAULT_LOCKED;
1690 out_inline:
1691 		filemap_invalidate_unlock_shared(mapping);
1692 		doutc(cl, "%llx.%llx %llu read inline data ret %x\n",
1693 		      ceph_vinop(inode), off, ret);
1694 	}
1695 out_restore:
1696 	ceph_restore_sigs(&oldset);
1697 	if (err < 0)
1698 		ret = vmf_error(err);
1699 
1700 	return ret;
1701 }
1702 
1703 static vm_fault_t ceph_page_mkwrite(struct vm_fault *vmf)
1704 {
1705 	struct vm_area_struct *vma = vmf->vma;
1706 	struct inode *inode = file_inode(vma->vm_file);
1707 	struct ceph_client *cl = ceph_inode_to_client(inode);
1708 	struct ceph_inode_info *ci = ceph_inode(inode);
1709 	struct ceph_file_info *fi = vma->vm_file->private_data;
1710 	struct ceph_cap_flush *prealloc_cf;
1711 	struct page *page = vmf->page;
1712 	loff_t off = page_offset(page);
1713 	loff_t size = i_size_read(inode);
1714 	size_t len;
1715 	int want, got, err;
1716 	sigset_t oldset;
1717 	vm_fault_t ret = VM_FAULT_SIGBUS;
1718 
1719 	if (ceph_inode_is_shutdown(inode))
1720 		return ret;
1721 
1722 	prealloc_cf = ceph_alloc_cap_flush();
1723 	if (!prealloc_cf)
1724 		return VM_FAULT_OOM;
1725 
1726 	sb_start_pagefault(inode->i_sb);
1727 	ceph_block_sigs(&oldset);
1728 
1729 	if (off + thp_size(page) <= size)
1730 		len = thp_size(page);
1731 	else
1732 		len = offset_in_thp(page, size);
1733 
1734 	doutc(cl, "%llx.%llx %llu~%zd getting caps i_size %llu\n",
1735 	      ceph_vinop(inode), off, len, size);
1736 	if (fi->fmode & CEPH_FILE_MODE_LAZY)
1737 		want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO;
1738 	else
1739 		want = CEPH_CAP_FILE_BUFFER;
1740 
1741 	got = 0;
1742 	err = ceph_get_caps(vma->vm_file, CEPH_CAP_FILE_WR, want, off + len, &got);
1743 	if (err < 0)
1744 		goto out_free;
1745 
1746 	doutc(cl, "%llx.%llx %llu~%zd got cap refs on %s\n", ceph_vinop(inode),
1747 	      off, len, ceph_cap_string(got));
1748 
1749 	/* Update time before taking page lock */
1750 	file_update_time(vma->vm_file);
1751 	inode_inc_iversion_raw(inode);
1752 
1753 	do {
1754 		struct ceph_snap_context *snapc;
1755 
1756 		lock_page(page);
1757 
1758 		if (page_mkwrite_check_truncate(page, inode) < 0) {
1759 			unlock_page(page);
1760 			ret = VM_FAULT_NOPAGE;
1761 			break;
1762 		}
1763 
1764 		snapc = ceph_find_incompatible(page);
1765 		if (!snapc) {
1766 			/* success.  we'll keep the page locked. */
1767 			set_page_dirty(page);
1768 			ret = VM_FAULT_LOCKED;
1769 			break;
1770 		}
1771 
1772 		unlock_page(page);
1773 
1774 		if (IS_ERR(snapc)) {
1775 			ret = VM_FAULT_SIGBUS;
1776 			break;
1777 		}
1778 
1779 		ceph_queue_writeback(inode);
1780 		err = wait_event_killable(ci->i_cap_wq,
1781 				context_is_writeable_or_written(inode, snapc));
1782 		ceph_put_snap_context(snapc);
1783 	} while (err == 0);
1784 
1785 	if (ret == VM_FAULT_LOCKED) {
1786 		int dirty;
1787 		spin_lock(&ci->i_ceph_lock);
1788 		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR,
1789 					       &prealloc_cf);
1790 		spin_unlock(&ci->i_ceph_lock);
1791 		if (dirty)
1792 			__mark_inode_dirty(inode, dirty);
1793 	}
1794 
1795 	doutc(cl, "%llx.%llx %llu~%zd dropping cap refs on %s ret %x\n",
1796 	      ceph_vinop(inode), off, len, ceph_cap_string(got), ret);
1797 	ceph_put_cap_refs_async(ci, got);
1798 out_free:
1799 	ceph_restore_sigs(&oldset);
1800 	sb_end_pagefault(inode->i_sb);
1801 	ceph_free_cap_flush(prealloc_cf);
1802 	if (err < 0)
1803 		ret = vmf_error(err);
1804 	return ret;
1805 }
1806 
1807 void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
1808 			   char	*data, size_t len)
1809 {
1810 	struct ceph_client *cl = ceph_inode_to_client(inode);
1811 	struct address_space *mapping = inode->i_mapping;
1812 	struct page *page;
1813 
1814 	if (locked_page) {
1815 		page = locked_page;
1816 	} else {
1817 		if (i_size_read(inode) == 0)
1818 			return;
1819 		page = find_or_create_page(mapping, 0,
1820 					   mapping_gfp_constraint(mapping,
1821 					   ~__GFP_FS));
1822 		if (!page)
1823 			return;
1824 		if (PageUptodate(page)) {
1825 			unlock_page(page);
1826 			put_page(page);
1827 			return;
1828 		}
1829 	}
1830 
1831 	doutc(cl, "%p %llx.%llx len %zu locked_page %p\n", inode,
1832 	      ceph_vinop(inode), len, locked_page);
1833 
1834 	if (len > 0) {
1835 		void *kaddr = kmap_atomic(page);
1836 		memcpy(kaddr, data, len);
1837 		kunmap_atomic(kaddr);
1838 	}
1839 
1840 	if (page != locked_page) {
1841 		if (len < PAGE_SIZE)
1842 			zero_user_segment(page, len, PAGE_SIZE);
1843 		else
1844 			flush_dcache_page(page);
1845 
1846 		SetPageUptodate(page);
1847 		unlock_page(page);
1848 		put_page(page);
1849 	}
1850 }
1851 
1852 int ceph_uninline_data(struct file *file)
1853 {
1854 	struct inode *inode = file_inode(file);
1855 	struct ceph_inode_info *ci = ceph_inode(inode);
1856 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(inode);
1857 	struct ceph_client *cl = fsc->client;
1858 	struct ceph_osd_request *req = NULL;
1859 	struct ceph_cap_flush *prealloc_cf = NULL;
1860 	struct folio *folio = NULL;
1861 	u64 inline_version = CEPH_INLINE_NONE;
1862 	struct page *pages[1];
1863 	int err = 0;
1864 	u64 len;
1865 
1866 	spin_lock(&ci->i_ceph_lock);
1867 	inline_version = ci->i_inline_version;
1868 	spin_unlock(&ci->i_ceph_lock);
1869 
1870 	doutc(cl, "%llx.%llx inline_version %llu\n", ceph_vinop(inode),
1871 	      inline_version);
1872 
1873 	if (ceph_inode_is_shutdown(inode)) {
1874 		err = -EIO;
1875 		goto out;
1876 	}
1877 
1878 	if (inline_version == CEPH_INLINE_NONE)
1879 		return 0;
1880 
1881 	prealloc_cf = ceph_alloc_cap_flush();
1882 	if (!prealloc_cf)
1883 		return -ENOMEM;
1884 
1885 	if (inline_version == 1) /* initial version, no data */
1886 		goto out_uninline;
1887 
1888 	folio = read_mapping_folio(inode->i_mapping, 0, file);
1889 	if (IS_ERR(folio)) {
1890 		err = PTR_ERR(folio);
1891 		goto out;
1892 	}
1893 
1894 	folio_lock(folio);
1895 
1896 	len = i_size_read(inode);
1897 	if (len > folio_size(folio))
1898 		len = folio_size(folio);
1899 
1900 	req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout,
1901 				    ceph_vino(inode), 0, &len, 0, 1,
1902 				    CEPH_OSD_OP_CREATE, CEPH_OSD_FLAG_WRITE,
1903 				    NULL, 0, 0, false);
1904 	if (IS_ERR(req)) {
1905 		err = PTR_ERR(req);
1906 		goto out_unlock;
1907 	}
1908 
1909 	req->r_mtime = inode_get_mtime(inode);
1910 	ceph_osdc_start_request(&fsc->client->osdc, req);
1911 	err = ceph_osdc_wait_request(&fsc->client->osdc, req);
1912 	ceph_osdc_put_request(req);
1913 	if (err < 0)
1914 		goto out_unlock;
1915 
1916 	req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout,
1917 				    ceph_vino(inode), 0, &len, 1, 3,
1918 				    CEPH_OSD_OP_WRITE, CEPH_OSD_FLAG_WRITE,
1919 				    NULL, ci->i_truncate_seq,
1920 				    ci->i_truncate_size, false);
1921 	if (IS_ERR(req)) {
1922 		err = PTR_ERR(req);
1923 		goto out_unlock;
1924 	}
1925 
1926 	pages[0] = folio_page(folio, 0);
1927 	osd_req_op_extent_osd_data_pages(req, 1, pages, len, 0, false, false);
1928 
1929 	{
1930 		__le64 xattr_buf = cpu_to_le64(inline_version);
1931 		err = osd_req_op_xattr_init(req, 0, CEPH_OSD_OP_CMPXATTR,
1932 					    "inline_version", &xattr_buf,
1933 					    sizeof(xattr_buf),
1934 					    CEPH_OSD_CMPXATTR_OP_GT,
1935 					    CEPH_OSD_CMPXATTR_MODE_U64);
1936 		if (err)
1937 			goto out_put_req;
1938 	}
1939 
1940 	{
1941 		char xattr_buf[32];
1942 		int xattr_len = snprintf(xattr_buf, sizeof(xattr_buf),
1943 					 "%llu", inline_version);
1944 		err = osd_req_op_xattr_init(req, 2, CEPH_OSD_OP_SETXATTR,
1945 					    "inline_version",
1946 					    xattr_buf, xattr_len, 0, 0);
1947 		if (err)
1948 			goto out_put_req;
1949 	}
1950 
1951 	req->r_mtime = inode_get_mtime(inode);
1952 	ceph_osdc_start_request(&fsc->client->osdc, req);
1953 	err = ceph_osdc_wait_request(&fsc->client->osdc, req);
1954 
1955 	ceph_update_write_metrics(&fsc->mdsc->metric, req->r_start_latency,
1956 				  req->r_end_latency, len, err);
1957 
1958 out_uninline:
1959 	if (!err) {
1960 		int dirty;
1961 
1962 		/* Set to CAP_INLINE_NONE and dirty the caps */
1963 		down_read(&fsc->mdsc->snap_rwsem);
1964 		spin_lock(&ci->i_ceph_lock);
1965 		ci->i_inline_version = CEPH_INLINE_NONE;
1966 		dirty = __ceph_mark_dirty_caps(ci, CEPH_CAP_FILE_WR, &prealloc_cf);
1967 		spin_unlock(&ci->i_ceph_lock);
1968 		up_read(&fsc->mdsc->snap_rwsem);
1969 		if (dirty)
1970 			__mark_inode_dirty(inode, dirty);
1971 	}
1972 out_put_req:
1973 	ceph_osdc_put_request(req);
1974 	if (err == -ECANCELED)
1975 		err = 0;
1976 out_unlock:
1977 	if (folio) {
1978 		folio_unlock(folio);
1979 		folio_put(folio);
1980 	}
1981 out:
1982 	ceph_free_cap_flush(prealloc_cf);
1983 	doutc(cl, "%llx.%llx inline_version %llu = %d\n",
1984 	      ceph_vinop(inode), inline_version, err);
1985 	return err;
1986 }
1987 
1988 static const struct vm_operations_struct ceph_vmops = {
1989 	.fault		= ceph_filemap_fault,
1990 	.page_mkwrite	= ceph_page_mkwrite,
1991 };
1992 
1993 int ceph_mmap(struct file *file, struct vm_area_struct *vma)
1994 {
1995 	struct address_space *mapping = file->f_mapping;
1996 
1997 	if (!mapping->a_ops->read_folio)
1998 		return -ENOEXEC;
1999 	vma->vm_ops = &ceph_vmops;
2000 	return 0;
2001 }
2002 
2003 enum {
2004 	POOL_READ	= 1,
2005 	POOL_WRITE	= 2,
2006 };
2007 
2008 static int __ceph_pool_perm_get(struct ceph_inode_info *ci,
2009 				s64 pool, struct ceph_string *pool_ns)
2010 {
2011 	struct ceph_fs_client *fsc = ceph_inode_to_fs_client(&ci->netfs.inode);
2012 	struct ceph_mds_client *mdsc = fsc->mdsc;
2013 	struct ceph_client *cl = fsc->client;
2014 	struct ceph_osd_request *rd_req = NULL, *wr_req = NULL;
2015 	struct rb_node **p, *parent;
2016 	struct ceph_pool_perm *perm;
2017 	struct page **pages;
2018 	size_t pool_ns_len;
2019 	int err = 0, err2 = 0, have = 0;
2020 
2021 	down_read(&mdsc->pool_perm_rwsem);
2022 	p = &mdsc->pool_perm_tree.rb_node;
2023 	while (*p) {
2024 		perm = rb_entry(*p, struct ceph_pool_perm, node);
2025 		if (pool < perm->pool)
2026 			p = &(*p)->rb_left;
2027 		else if (pool > perm->pool)
2028 			p = &(*p)->rb_right;
2029 		else {
2030 			int ret = ceph_compare_string(pool_ns,
2031 						perm->pool_ns,
2032 						perm->pool_ns_len);
2033 			if (ret < 0)
2034 				p = &(*p)->rb_left;
2035 			else if (ret > 0)
2036 				p = &(*p)->rb_right;
2037 			else {
2038 				have = perm->perm;
2039 				break;
2040 			}
2041 		}
2042 	}
2043 	up_read(&mdsc->pool_perm_rwsem);
2044 	if (*p)
2045 		goto out;
2046 
2047 	if (pool_ns)
2048 		doutc(cl, "pool %lld ns %.*s no perm cached\n", pool,
2049 		      (int)pool_ns->len, pool_ns->str);
2050 	else
2051 		doutc(cl, "pool %lld no perm cached\n", pool);
2052 
2053 	down_write(&mdsc->pool_perm_rwsem);
2054 	p = &mdsc->pool_perm_tree.rb_node;
2055 	parent = NULL;
2056 	while (*p) {
2057 		parent = *p;
2058 		perm = rb_entry(parent, struct ceph_pool_perm, node);
2059 		if (pool < perm->pool)
2060 			p = &(*p)->rb_left;
2061 		else if (pool > perm->pool)
2062 			p = &(*p)->rb_right;
2063 		else {
2064 			int ret = ceph_compare_string(pool_ns,
2065 						perm->pool_ns,
2066 						perm->pool_ns_len);
2067 			if (ret < 0)
2068 				p = &(*p)->rb_left;
2069 			else if (ret > 0)
2070 				p = &(*p)->rb_right;
2071 			else {
2072 				have = perm->perm;
2073 				break;
2074 			}
2075 		}
2076 	}
2077 	if (*p) {
2078 		up_write(&mdsc->pool_perm_rwsem);
2079 		goto out;
2080 	}
2081 
2082 	rd_req = ceph_osdc_alloc_request(&fsc->client->osdc, NULL,
2083 					 1, false, GFP_NOFS);
2084 	if (!rd_req) {
2085 		err = -ENOMEM;
2086 		goto out_unlock;
2087 	}
2088 
2089 	rd_req->r_flags = CEPH_OSD_FLAG_READ;
2090 	osd_req_op_init(rd_req, 0, CEPH_OSD_OP_STAT, 0);
2091 	rd_req->r_base_oloc.pool = pool;
2092 	if (pool_ns)
2093 		rd_req->r_base_oloc.pool_ns = ceph_get_string(pool_ns);
2094 	ceph_oid_printf(&rd_req->r_base_oid, "%llx.00000000", ci->i_vino.ino);
2095 
2096 	err = ceph_osdc_alloc_messages(rd_req, GFP_NOFS);
2097 	if (err)
2098 		goto out_unlock;
2099 
2100 	wr_req = ceph_osdc_alloc_request(&fsc->client->osdc, NULL,
2101 					 1, false, GFP_NOFS);
2102 	if (!wr_req) {
2103 		err = -ENOMEM;
2104 		goto out_unlock;
2105 	}
2106 
2107 	wr_req->r_flags = CEPH_OSD_FLAG_WRITE;
2108 	osd_req_op_init(wr_req, 0, CEPH_OSD_OP_CREATE, CEPH_OSD_OP_FLAG_EXCL);
2109 	ceph_oloc_copy(&wr_req->r_base_oloc, &rd_req->r_base_oloc);
2110 	ceph_oid_copy(&wr_req->r_base_oid, &rd_req->r_base_oid);
2111 
2112 	err = ceph_osdc_alloc_messages(wr_req, GFP_NOFS);
2113 	if (err)
2114 		goto out_unlock;
2115 
2116 	/* one page should be large enough for STAT data */
2117 	pages = ceph_alloc_page_vector(1, GFP_KERNEL);
2118 	if (IS_ERR(pages)) {
2119 		err = PTR_ERR(pages);
2120 		goto out_unlock;
2121 	}
2122 
2123 	osd_req_op_raw_data_in_pages(rd_req, 0, pages, PAGE_SIZE,
2124 				     0, false, true);
2125 	ceph_osdc_start_request(&fsc->client->osdc, rd_req);
2126 
2127 	wr_req->r_mtime = inode_get_mtime(&ci->netfs.inode);
2128 	ceph_osdc_start_request(&fsc->client->osdc, wr_req);
2129 
2130 	err = ceph_osdc_wait_request(&fsc->client->osdc, rd_req);
2131 	err2 = ceph_osdc_wait_request(&fsc->client->osdc, wr_req);
2132 
2133 	if (err >= 0 || err == -ENOENT)
2134 		have |= POOL_READ;
2135 	else if (err != -EPERM) {
2136 		if (err == -EBLOCKLISTED)
2137 			fsc->blocklisted = true;
2138 		goto out_unlock;
2139 	}
2140 
2141 	if (err2 == 0 || err2 == -EEXIST)
2142 		have |= POOL_WRITE;
2143 	else if (err2 != -EPERM) {
2144 		if (err2 == -EBLOCKLISTED)
2145 			fsc->blocklisted = true;
2146 		err = err2;
2147 		goto out_unlock;
2148 	}
2149 
2150 	pool_ns_len = pool_ns ? pool_ns->len : 0;
2151 	perm = kmalloc(struct_size(perm, pool_ns, pool_ns_len + 1), GFP_NOFS);
2152 	if (!perm) {
2153 		err = -ENOMEM;
2154 		goto out_unlock;
2155 	}
2156 
2157 	perm->pool = pool;
2158 	perm->perm = have;
2159 	perm->pool_ns_len = pool_ns_len;
2160 	if (pool_ns_len > 0)
2161 		memcpy(perm->pool_ns, pool_ns->str, pool_ns_len);
2162 	perm->pool_ns[pool_ns_len] = 0;
2163 
2164 	rb_link_node(&perm->node, parent, p);
2165 	rb_insert_color(&perm->node, &mdsc->pool_perm_tree);
2166 	err = 0;
2167 out_unlock:
2168 	up_write(&mdsc->pool_perm_rwsem);
2169 
2170 	ceph_osdc_put_request(rd_req);
2171 	ceph_osdc_put_request(wr_req);
2172 out:
2173 	if (!err)
2174 		err = have;
2175 	if (pool_ns)
2176 		doutc(cl, "pool %lld ns %.*s result = %d\n", pool,
2177 		      (int)pool_ns->len, pool_ns->str, err);
2178 	else
2179 		doutc(cl, "pool %lld result = %d\n", pool, err);
2180 	return err;
2181 }
2182 
2183 int ceph_pool_perm_check(struct inode *inode, int need)
2184 {
2185 	struct ceph_client *cl = ceph_inode_to_client(inode);
2186 	struct ceph_inode_info *ci = ceph_inode(inode);
2187 	struct ceph_string *pool_ns;
2188 	s64 pool;
2189 	int ret, flags;
2190 
2191 	/* Only need to do this for regular files */
2192 	if (!S_ISREG(inode->i_mode))
2193 		return 0;
2194 
2195 	if (ci->i_vino.snap != CEPH_NOSNAP) {
2196 		/*
2197 		 * Pool permission check needs to write to the first object.
2198 		 * But for snapshot, head of the first object may have alread
2199 		 * been deleted. Skip check to avoid creating orphan object.
2200 		 */
2201 		return 0;
2202 	}
2203 
2204 	if (ceph_test_mount_opt(ceph_inode_to_fs_client(inode),
2205 				NOPOOLPERM))
2206 		return 0;
2207 
2208 	spin_lock(&ci->i_ceph_lock);
2209 	flags = ci->i_ceph_flags;
2210 	pool = ci->i_layout.pool_id;
2211 	spin_unlock(&ci->i_ceph_lock);
2212 check:
2213 	if (flags & CEPH_I_POOL_PERM) {
2214 		if ((need & CEPH_CAP_FILE_RD) && !(flags & CEPH_I_POOL_RD)) {
2215 			doutc(cl, "pool %lld no read perm\n", pool);
2216 			return -EPERM;
2217 		}
2218 		if ((need & CEPH_CAP_FILE_WR) && !(flags & CEPH_I_POOL_WR)) {
2219 			doutc(cl, "pool %lld no write perm\n", pool);
2220 			return -EPERM;
2221 		}
2222 		return 0;
2223 	}
2224 
2225 	pool_ns = ceph_try_get_string(ci->i_layout.pool_ns);
2226 	ret = __ceph_pool_perm_get(ci, pool, pool_ns);
2227 	ceph_put_string(pool_ns);
2228 	if (ret < 0)
2229 		return ret;
2230 
2231 	flags = CEPH_I_POOL_PERM;
2232 	if (ret & POOL_READ)
2233 		flags |= CEPH_I_POOL_RD;
2234 	if (ret & POOL_WRITE)
2235 		flags |= CEPH_I_POOL_WR;
2236 
2237 	spin_lock(&ci->i_ceph_lock);
2238 	if (pool == ci->i_layout.pool_id &&
2239 	    pool_ns == rcu_dereference_raw(ci->i_layout.pool_ns)) {
2240 		ci->i_ceph_flags |= flags;
2241         } else {
2242 		pool = ci->i_layout.pool_id;
2243 		flags = ci->i_ceph_flags;
2244 	}
2245 	spin_unlock(&ci->i_ceph_lock);
2246 	goto check;
2247 }
2248 
2249 void ceph_pool_perm_destroy(struct ceph_mds_client *mdsc)
2250 {
2251 	struct ceph_pool_perm *perm;
2252 	struct rb_node *n;
2253 
2254 	while (!RB_EMPTY_ROOT(&mdsc->pool_perm_tree)) {
2255 		n = rb_first(&mdsc->pool_perm_tree);
2256 		perm = rb_entry(n, struct ceph_pool_perm, node);
2257 		rb_erase(n, &mdsc->pool_perm_tree);
2258 		kfree(perm);
2259 	}
2260 }
2261