1 // SPDX-License-Identifier: GPL-2.0
3 * Simple file system for zoned block devices exposing zones as files.
5 * Copyright (C) 2019 Western Digital Corporation or its affiliates.
7 #include <linux/module.h>
9 #include <linux/magic.h>
10 #include <linux/iomap.h>
11 #include <linux/init.h>
12 #include <linux/slab.h>
13 #include <linux/blkdev.h>
14 #include <linux/statfs.h>
15 #include <linux/writeback.h>
16 #include <linux/quotaops.h>
17 #include <linux/seq_file.h>
18 #include <linux/parser.h>
19 #include <linux/uio.h>
20 #include <linux/mman.h>
21 #include <linux/sched/mm.h>
22 #include <linux/crc32.h>
23 #include <linux/task_io_accounting_ops.h>
27 #define CREATE_TRACE_POINTS
30 static inline int zonefs_zone_mgmt(struct inode *inode,
33 struct zonefs_inode_info *zi = ZONEFS_I(inode);
36 lockdep_assert_held(&zi->i_truncate_mutex);
38 trace_zonefs_zone_mgmt(inode, op);
39 ret = blkdev_zone_mgmt(inode->i_sb->s_bdev, op, zi->i_zsector,
40 zi->i_zone_size >> SECTOR_SHIFT, GFP_NOFS);
42 zonefs_err(inode->i_sb,
43 "Zone management operation %s at %llu failed %d\n",
44 blk_op_str(op), zi->i_zsector, ret);
51 static inline void zonefs_i_size_write(struct inode *inode, loff_t isize)
53 struct zonefs_inode_info *zi = ZONEFS_I(inode);
55 i_size_write(inode, isize);
57 * A full zone is no longer open/active and does not need
60 if (isize >= zi->i_max_size)
61 zi->i_flags &= ~ZONEFS_ZONE_OPEN;
64 static int zonefs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
65 unsigned int flags, struct iomap *iomap,
68 struct zonefs_inode_info *zi = ZONEFS_I(inode);
69 struct super_block *sb = inode->i_sb;
72 /* All I/Os should always be within the file maximum size */
73 if (WARN_ON_ONCE(offset + length > zi->i_max_size))
77 * Sequential zones can only accept direct writes. This is already
78 * checked when writes are issued, so warn if we see a page writeback
81 if (WARN_ON_ONCE(zi->i_ztype == ZONEFS_ZTYPE_SEQ &&
82 (flags & IOMAP_WRITE) && !(flags & IOMAP_DIRECT)))
86 * For conventional zones, all blocks are always mapped. For sequential
87 * zones, all blocks after always mapped below the inode size (zone
88 * write pointer) and unwriten beyond.
90 mutex_lock(&zi->i_truncate_mutex);
91 isize = i_size_read(inode);
93 iomap->type = IOMAP_UNWRITTEN;
95 iomap->type = IOMAP_MAPPED;
96 if (flags & IOMAP_WRITE)
97 length = zi->i_max_size - offset;
99 length = min(length, isize - offset);
100 mutex_unlock(&zi->i_truncate_mutex);
102 iomap->offset = ALIGN_DOWN(offset, sb->s_blocksize);
103 iomap->length = ALIGN(offset + length, sb->s_blocksize) - iomap->offset;
104 iomap->bdev = inode->i_sb->s_bdev;
105 iomap->addr = (zi->i_zsector << SECTOR_SHIFT) + iomap->offset;
107 trace_zonefs_iomap_begin(inode, iomap);
112 static const struct iomap_ops zonefs_iomap_ops = {
113 .iomap_begin = zonefs_iomap_begin,
116 static int zonefs_readpage(struct file *unused, struct page *page)
118 return iomap_readpage(page, &zonefs_iomap_ops);
121 static void zonefs_readahead(struct readahead_control *rac)
123 iomap_readahead(rac, &zonefs_iomap_ops);
127 * Map blocks for page writeback. This is used only on conventional zone files,
128 * which implies that the page range can only be within the fixed inode size.
130 static int zonefs_map_blocks(struct iomap_writepage_ctx *wpc,
131 struct inode *inode, loff_t offset)
133 struct zonefs_inode_info *zi = ZONEFS_I(inode);
135 if (WARN_ON_ONCE(zi->i_ztype != ZONEFS_ZTYPE_CNV))
137 if (WARN_ON_ONCE(offset >= i_size_read(inode)))
140 /* If the mapping is already OK, nothing needs to be done */
141 if (offset >= wpc->iomap.offset &&
142 offset < wpc->iomap.offset + wpc->iomap.length)
145 return zonefs_iomap_begin(inode, offset, zi->i_max_size - offset,
146 IOMAP_WRITE, &wpc->iomap, NULL);
149 static const struct iomap_writeback_ops zonefs_writeback_ops = {
150 .map_blocks = zonefs_map_blocks,
153 static int zonefs_writepage(struct page *page, struct writeback_control *wbc)
155 struct iomap_writepage_ctx wpc = { };
157 return iomap_writepage(page, wbc, &wpc, &zonefs_writeback_ops);
160 static int zonefs_writepages(struct address_space *mapping,
161 struct writeback_control *wbc)
163 struct iomap_writepage_ctx wpc = { };
165 return iomap_writepages(mapping, wbc, &wpc, &zonefs_writeback_ops);
168 static const struct address_space_operations zonefs_file_aops = {
169 .readpage = zonefs_readpage,
170 .readahead = zonefs_readahead,
171 .writepage = zonefs_writepage,
172 .writepages = zonefs_writepages,
173 .set_page_dirty = iomap_set_page_dirty,
174 .releasepage = iomap_releasepage,
175 .invalidatepage = iomap_invalidatepage,
176 .migratepage = iomap_migrate_page,
177 .is_partially_uptodate = iomap_is_partially_uptodate,
178 .error_remove_page = generic_error_remove_page,
179 .direct_IO = noop_direct_IO,
182 static void zonefs_update_stats(struct inode *inode, loff_t new_isize)
184 struct super_block *sb = inode->i_sb;
185 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
186 loff_t old_isize = i_size_read(inode);
189 if (new_isize == old_isize)
192 spin_lock(&sbi->s_lock);
195 * This may be called for an update after an IO error.
196 * So beware of the values seen.
198 if (new_isize < old_isize) {
199 nr_blocks = (old_isize - new_isize) >> sb->s_blocksize_bits;
200 if (sbi->s_used_blocks > nr_blocks)
201 sbi->s_used_blocks -= nr_blocks;
203 sbi->s_used_blocks = 0;
205 sbi->s_used_blocks +=
206 (new_isize - old_isize) >> sb->s_blocksize_bits;
207 if (sbi->s_used_blocks > sbi->s_blocks)
208 sbi->s_used_blocks = sbi->s_blocks;
211 spin_unlock(&sbi->s_lock);
215 * Check a zone condition and adjust its file inode access permissions for
216 * offline and readonly zones. Return the inode size corresponding to the
217 * amount of readable data in the zone.
219 static loff_t zonefs_check_zone_condition(struct inode *inode,
220 struct blk_zone *zone, bool warn,
223 struct zonefs_inode_info *zi = ZONEFS_I(inode);
225 switch (zone->cond) {
226 case BLK_ZONE_COND_OFFLINE:
228 * Dead zone: make the inode immutable, disable all accesses
229 * and set the file size to 0 (zone wp set to zone start).
232 zonefs_warn(inode->i_sb, "inode %lu: offline zone\n",
234 inode->i_flags |= S_IMMUTABLE;
235 inode->i_mode &= ~0777;
236 zone->wp = zone->start;
238 case BLK_ZONE_COND_READONLY:
240 * The write pointer of read-only zones is invalid. If such a
241 * zone is found during mount, the file size cannot be retrieved
242 * so we treat the zone as offline (mount == true case).
243 * Otherwise, keep the file size as it was when last updated
244 * so that the user can recover data. In both cases, writes are
245 * always disabled for the zone.
248 zonefs_warn(inode->i_sb, "inode %lu: read-only zone\n",
250 inode->i_flags |= S_IMMUTABLE;
252 zone->cond = BLK_ZONE_COND_OFFLINE;
253 inode->i_mode &= ~0777;
254 zone->wp = zone->start;
257 inode->i_mode &= ~0222;
258 return i_size_read(inode);
259 case BLK_ZONE_COND_FULL:
260 /* The write pointer of full zones is invalid. */
261 return zi->i_max_size;
263 if (zi->i_ztype == ZONEFS_ZTYPE_CNV)
264 return zi->i_max_size;
265 return (zone->wp - zone->start) << SECTOR_SHIFT;
269 struct zonefs_ioerr_data {
274 static int zonefs_io_error_cb(struct blk_zone *zone, unsigned int idx,
277 struct zonefs_ioerr_data *err = data;
278 struct inode *inode = err->inode;
279 struct zonefs_inode_info *zi = ZONEFS_I(inode);
280 struct super_block *sb = inode->i_sb;
281 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
282 loff_t isize, data_size;
285 * Check the zone condition: if the zone is not "bad" (offline or
286 * read-only), read errors are simply signaled to the IO issuer as long
287 * as there is no inconsistency between the inode size and the amount of
288 * data writen in the zone (data_size).
290 data_size = zonefs_check_zone_condition(inode, zone, true, false);
291 isize = i_size_read(inode);
292 if (zone->cond != BLK_ZONE_COND_OFFLINE &&
293 zone->cond != BLK_ZONE_COND_READONLY &&
294 !err->write && isize == data_size)
298 * At this point, we detected either a bad zone or an inconsistency
299 * between the inode size and the amount of data written in the zone.
300 * For the latter case, the cause may be a write IO error or an external
301 * action on the device. Two error patterns exist:
302 * 1) The inode size is lower than the amount of data in the zone:
303 * a write operation partially failed and data was writen at the end
304 * of the file. This can happen in the case of a large direct IO
305 * needing several BIOs and/or write requests to be processed.
306 * 2) The inode size is larger than the amount of data in the zone:
307 * this can happen with a deferred write error with the use of the
308 * device side write cache after getting successful write IO
309 * completions. Other possibilities are (a) an external corruption,
310 * e.g. an application reset the zone directly, or (b) the device
311 * has a serious problem (e.g. firmware bug).
313 * In all cases, warn about inode size inconsistency and handle the
314 * IO error according to the zone condition and to the mount options.
316 if (zi->i_ztype == ZONEFS_ZTYPE_SEQ && isize != data_size)
317 zonefs_warn(sb, "inode %lu: invalid size %lld (should be %lld)\n",
318 inode->i_ino, isize, data_size);
321 * First handle bad zones signaled by hardware. The mount options
322 * errors=zone-ro and errors=zone-offline result in changing the
323 * zone condition to read-only and offline respectively, as if the
324 * condition was signaled by the hardware.
326 if (zone->cond == BLK_ZONE_COND_OFFLINE ||
327 sbi->s_mount_opts & ZONEFS_MNTOPT_ERRORS_ZOL) {
328 zonefs_warn(sb, "inode %lu: read/write access disabled\n",
330 if (zone->cond != BLK_ZONE_COND_OFFLINE) {
331 zone->cond = BLK_ZONE_COND_OFFLINE;
332 data_size = zonefs_check_zone_condition(inode, zone,
335 } else if (zone->cond == BLK_ZONE_COND_READONLY ||
336 sbi->s_mount_opts & ZONEFS_MNTOPT_ERRORS_ZRO) {
337 zonefs_warn(sb, "inode %lu: write access disabled\n",
339 if (zone->cond != BLK_ZONE_COND_READONLY) {
340 zone->cond = BLK_ZONE_COND_READONLY;
341 data_size = zonefs_check_zone_condition(inode, zone,
347 * If the filesystem is mounted with the explicit-open mount option, we
348 * need to clear the ZONEFS_ZONE_OPEN flag if the zone transitioned to
349 * the read-only or offline condition, to avoid attempting an explicit
350 * close of the zone when the inode file is closed.
352 if ((sbi->s_mount_opts & ZONEFS_MNTOPT_EXPLICIT_OPEN) &&
353 (zone->cond == BLK_ZONE_COND_OFFLINE ||
354 zone->cond == BLK_ZONE_COND_READONLY))
355 zi->i_flags &= ~ZONEFS_ZONE_OPEN;
358 * If error=remount-ro was specified, any error result in remounting
359 * the volume as read-only.
361 if ((sbi->s_mount_opts & ZONEFS_MNTOPT_ERRORS_RO) && !sb_rdonly(sb)) {
362 zonefs_warn(sb, "remounting filesystem read-only\n");
363 sb->s_flags |= SB_RDONLY;
367 * Update block usage stats and the inode size to prevent access to
370 zonefs_update_stats(inode, data_size);
371 zonefs_i_size_write(inode, data_size);
372 zi->i_wpoffset = data_size;
378 * When an file IO error occurs, check the file zone to see if there is a change
379 * in the zone condition (e.g. offline or read-only). For a failed write to a
380 * sequential zone, the zone write pointer position must also be checked to
381 * eventually correct the file size and zonefs inode write pointer offset
382 * (which can be out of sync with the drive due to partial write failures).
384 static void __zonefs_io_error(struct inode *inode, bool write)
386 struct zonefs_inode_info *zi = ZONEFS_I(inode);
387 struct super_block *sb = inode->i_sb;
388 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
389 unsigned int noio_flag;
390 unsigned int nr_zones =
391 zi->i_zone_size >> (sbi->s_zone_sectors_shift + SECTOR_SHIFT);
392 struct zonefs_ioerr_data err = {
399 * Memory allocations in blkdev_report_zones() can trigger a memory
400 * reclaim which may in turn cause a recursion into zonefs as well as
401 * struct request allocations for the same device. The former case may
402 * end up in a deadlock on the inode truncate mutex, while the latter
403 * may prevent IO forward progress. Executing the report zones under
404 * the GFP_NOIO context avoids both problems.
406 noio_flag = memalloc_noio_save();
407 ret = blkdev_report_zones(sb->s_bdev, zi->i_zsector, nr_zones,
408 zonefs_io_error_cb, &err);
410 zonefs_err(sb, "Get inode %lu zone information failed %d\n",
412 memalloc_noio_restore(noio_flag);
415 static void zonefs_io_error(struct inode *inode, bool write)
417 struct zonefs_inode_info *zi = ZONEFS_I(inode);
419 mutex_lock(&zi->i_truncate_mutex);
420 __zonefs_io_error(inode, write);
421 mutex_unlock(&zi->i_truncate_mutex);
424 static int zonefs_file_truncate(struct inode *inode, loff_t isize)
426 struct zonefs_inode_info *zi = ZONEFS_I(inode);
432 * Only sequential zone files can be truncated and truncation is allowed
433 * only down to a 0 size, which is equivalent to a zone reset, and to
434 * the maximum file size, which is equivalent to a zone finish.
436 if (zi->i_ztype != ZONEFS_ZTYPE_SEQ)
440 op = REQ_OP_ZONE_RESET;
441 else if (isize == zi->i_max_size)
442 op = REQ_OP_ZONE_FINISH;
446 inode_dio_wait(inode);
448 /* Serialize against page faults */
449 down_write(&zi->i_mmap_sem);
451 /* Serialize against zonefs_iomap_begin() */
452 mutex_lock(&zi->i_truncate_mutex);
454 old_isize = i_size_read(inode);
455 if (isize == old_isize)
458 ret = zonefs_zone_mgmt(inode, op);
463 * If the mount option ZONEFS_MNTOPT_EXPLICIT_OPEN is set,
464 * take care of open zones.
466 if (zi->i_flags & ZONEFS_ZONE_OPEN) {
468 * Truncating a zone to EMPTY or FULL is the equivalent of
469 * closing the zone. For a truncation to 0, we need to
470 * re-open the zone to ensure new writes can be processed.
471 * For a truncation to the maximum file size, the zone is
472 * closed and writes cannot be accepted anymore, so clear
476 ret = zonefs_zone_mgmt(inode, REQ_OP_ZONE_OPEN);
478 zi->i_flags &= ~ZONEFS_ZONE_OPEN;
481 zonefs_update_stats(inode, isize);
482 truncate_setsize(inode, isize);
483 zi->i_wpoffset = isize;
486 mutex_unlock(&zi->i_truncate_mutex);
487 up_write(&zi->i_mmap_sem);
492 static int zonefs_inode_setattr(struct user_namespace *mnt_userns,
493 struct dentry *dentry, struct iattr *iattr)
495 struct inode *inode = d_inode(dentry);
498 if (unlikely(IS_IMMUTABLE(inode)))
501 ret = setattr_prepare(&init_user_ns, dentry, iattr);
506 * Since files and directories cannot be created nor deleted, do not
507 * allow setting any write attributes on the sub-directories grouping
508 * files by zone type.
510 if ((iattr->ia_valid & ATTR_MODE) && S_ISDIR(inode->i_mode) &&
511 (iattr->ia_mode & 0222))
514 if (((iattr->ia_valid & ATTR_UID) &&
515 !uid_eq(iattr->ia_uid, inode->i_uid)) ||
516 ((iattr->ia_valid & ATTR_GID) &&
517 !gid_eq(iattr->ia_gid, inode->i_gid))) {
518 ret = dquot_transfer(inode, iattr);
523 if (iattr->ia_valid & ATTR_SIZE) {
524 ret = zonefs_file_truncate(inode, iattr->ia_size);
529 setattr_copy(&init_user_ns, inode, iattr);
534 static const struct inode_operations zonefs_file_inode_operations = {
535 .setattr = zonefs_inode_setattr,
538 static int zonefs_file_fsync(struct file *file, loff_t start, loff_t end,
541 struct inode *inode = file_inode(file);
544 if (unlikely(IS_IMMUTABLE(inode)))
548 * Since only direct writes are allowed in sequential files, page cache
549 * flush is needed only for conventional zone files.
551 if (ZONEFS_I(inode)->i_ztype == ZONEFS_ZTYPE_CNV)
552 ret = file_write_and_wait_range(file, start, end);
554 ret = blkdev_issue_flush(inode->i_sb->s_bdev);
557 zonefs_io_error(inode, true);
562 static vm_fault_t zonefs_filemap_fault(struct vm_fault *vmf)
564 struct zonefs_inode_info *zi = ZONEFS_I(file_inode(vmf->vma->vm_file));
567 down_read(&zi->i_mmap_sem);
568 ret = filemap_fault(vmf);
569 up_read(&zi->i_mmap_sem);
574 static vm_fault_t zonefs_filemap_page_mkwrite(struct vm_fault *vmf)
576 struct inode *inode = file_inode(vmf->vma->vm_file);
577 struct zonefs_inode_info *zi = ZONEFS_I(inode);
580 if (unlikely(IS_IMMUTABLE(inode)))
581 return VM_FAULT_SIGBUS;
584 * Sanity check: only conventional zone files can have shared
585 * writeable mappings.
587 if (WARN_ON_ONCE(zi->i_ztype != ZONEFS_ZTYPE_CNV))
588 return VM_FAULT_NOPAGE;
590 sb_start_pagefault(inode->i_sb);
591 file_update_time(vmf->vma->vm_file);
593 /* Serialize against truncates */
594 down_read(&zi->i_mmap_sem);
595 ret = iomap_page_mkwrite(vmf, &zonefs_iomap_ops);
596 up_read(&zi->i_mmap_sem);
598 sb_end_pagefault(inode->i_sb);
602 static const struct vm_operations_struct zonefs_file_vm_ops = {
603 .fault = zonefs_filemap_fault,
604 .map_pages = filemap_map_pages,
605 .page_mkwrite = zonefs_filemap_page_mkwrite,
608 static int zonefs_file_mmap(struct file *file, struct vm_area_struct *vma)
611 * Conventional zones accept random writes, so their files can support
612 * shared writable mappings. For sequential zone files, only read
613 * mappings are possible since there are no guarantees for write
614 * ordering between msync() and page cache writeback.
616 if (ZONEFS_I(file_inode(file))->i_ztype == ZONEFS_ZTYPE_SEQ &&
617 (vma->vm_flags & VM_SHARED) && (vma->vm_flags & VM_MAYWRITE))
621 vma->vm_ops = &zonefs_file_vm_ops;
626 static loff_t zonefs_file_llseek(struct file *file, loff_t offset, int whence)
628 loff_t isize = i_size_read(file_inode(file));
631 * Seeks are limited to below the zone size for conventional zones
632 * and below the zone write pointer for sequential zones. In both
633 * cases, this limit is the inode size.
635 return generic_file_llseek_size(file, offset, whence, isize, isize);
638 static int zonefs_file_write_dio_end_io(struct kiocb *iocb, ssize_t size,
639 int error, unsigned int flags)
641 struct inode *inode = file_inode(iocb->ki_filp);
642 struct zonefs_inode_info *zi = ZONEFS_I(inode);
645 zonefs_io_error(inode, true);
649 if (size && zi->i_ztype != ZONEFS_ZTYPE_CNV) {
651 * Note that we may be seeing completions out of order,
652 * but that is not a problem since a write completed
653 * successfully necessarily means that all preceding writes
654 * were also successful. So we can safely increase the inode
655 * size to the write end location.
657 mutex_lock(&zi->i_truncate_mutex);
658 if (i_size_read(inode) < iocb->ki_pos + size) {
659 zonefs_update_stats(inode, iocb->ki_pos + size);
660 zonefs_i_size_write(inode, iocb->ki_pos + size);
662 mutex_unlock(&zi->i_truncate_mutex);
668 static const struct iomap_dio_ops zonefs_write_dio_ops = {
669 .end_io = zonefs_file_write_dio_end_io,
672 static ssize_t zonefs_file_dio_append(struct kiocb *iocb, struct iov_iter *from)
674 struct inode *inode = file_inode(iocb->ki_filp);
675 struct zonefs_inode_info *zi = ZONEFS_I(inode);
676 struct block_device *bdev = inode->i_sb->s_bdev;
683 max = queue_max_zone_append_sectors(bdev_get_queue(bdev));
684 max = ALIGN_DOWN(max << SECTOR_SHIFT, inode->i_sb->s_blocksize);
685 iov_iter_truncate(from, max);
687 nr_pages = iov_iter_npages(from, BIO_MAX_VECS);
691 bio = bio_alloc(GFP_NOFS, nr_pages);
695 bio_set_dev(bio, bdev);
696 bio->bi_iter.bi_sector = zi->i_zsector;
697 bio->bi_write_hint = iocb->ki_hint;
698 bio->bi_ioprio = iocb->ki_ioprio;
699 bio->bi_opf = REQ_OP_ZONE_APPEND | REQ_SYNC | REQ_IDLE;
700 if (iocb->ki_flags & IOCB_DSYNC)
701 bio->bi_opf |= REQ_FUA;
703 ret = bio_iov_iter_get_pages(bio, from);
707 size = bio->bi_iter.bi_size;
708 task_io_account_write(size);
710 if (iocb->ki_flags & IOCB_HIPRI)
711 bio_set_polled(bio, iocb);
713 ret = submit_bio_wait(bio);
715 zonefs_file_write_dio_end_io(iocb, size, ret, 0);
716 trace_zonefs_file_dio_append(inode, size, ret);
719 bio_release_pages(bio, false);
723 iocb->ki_pos += size;
731 * Handle direct writes. For sequential zone files, this is the only possible
732 * write path. For these files, check that the user is issuing writes
733 * sequentially from the end of the file. This code assumes that the block layer
734 * delivers write requests to the device in sequential order. This is always the
735 * case if a block IO scheduler implementing the ELEVATOR_F_ZBD_SEQ_WRITE
736 * elevator feature is being used (e.g. mq-deadline). The block layer always
737 * automatically select such an elevator for zoned block devices during the
738 * device initialization.
740 static ssize_t zonefs_file_dio_write(struct kiocb *iocb, struct iov_iter *from)
742 struct inode *inode = file_inode(iocb->ki_filp);
743 struct zonefs_inode_info *zi = ZONEFS_I(inode);
744 struct super_block *sb = inode->i_sb;
745 bool sync = is_sync_kiocb(iocb);
751 * For async direct IOs to sequential zone files, refuse IOCB_NOWAIT
752 * as this can cause write reordering (e.g. the first aio gets EAGAIN
753 * on the inode lock but the second goes through but is now unaligned).
755 if (zi->i_ztype == ZONEFS_ZTYPE_SEQ && !sync &&
756 (iocb->ki_flags & IOCB_NOWAIT))
759 if (iocb->ki_flags & IOCB_NOWAIT) {
760 if (!inode_trylock(inode))
766 ret = generic_write_checks(iocb, from);
770 iov_iter_truncate(from, zi->i_max_size - iocb->ki_pos);
771 count = iov_iter_count(from);
773 if ((iocb->ki_pos | count) & (sb->s_blocksize - 1)) {
778 /* Enforce sequential writes (append only) in sequential zones */
779 if (zi->i_ztype == ZONEFS_ZTYPE_SEQ) {
780 mutex_lock(&zi->i_truncate_mutex);
781 if (iocb->ki_pos != zi->i_wpoffset) {
782 mutex_unlock(&zi->i_truncate_mutex);
786 mutex_unlock(&zi->i_truncate_mutex);
791 ret = zonefs_file_dio_append(iocb, from);
793 ret = iomap_dio_rw(iocb, from, &zonefs_iomap_ops,
794 &zonefs_write_dio_ops, 0);
795 if (zi->i_ztype == ZONEFS_ZTYPE_SEQ &&
796 (ret > 0 || ret == -EIOCBQUEUED)) {
799 mutex_lock(&zi->i_truncate_mutex);
800 zi->i_wpoffset += count;
801 mutex_unlock(&zi->i_truncate_mutex);
810 static ssize_t zonefs_file_buffered_write(struct kiocb *iocb,
811 struct iov_iter *from)
813 struct inode *inode = file_inode(iocb->ki_filp);
814 struct zonefs_inode_info *zi = ZONEFS_I(inode);
818 * Direct IO writes are mandatory for sequential zone files so that the
819 * write IO issuing order is preserved.
821 if (zi->i_ztype != ZONEFS_ZTYPE_CNV)
824 if (iocb->ki_flags & IOCB_NOWAIT) {
825 if (!inode_trylock(inode))
831 ret = generic_write_checks(iocb, from);
835 iov_iter_truncate(from, zi->i_max_size - iocb->ki_pos);
837 ret = iomap_file_buffered_write(iocb, from, &zonefs_iomap_ops);
840 else if (ret == -EIO)
841 zonefs_io_error(inode, true);
846 ret = generic_write_sync(iocb, ret);
851 static ssize_t zonefs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
853 struct inode *inode = file_inode(iocb->ki_filp);
855 if (unlikely(IS_IMMUTABLE(inode)))
858 if (sb_rdonly(inode->i_sb))
861 /* Write operations beyond the zone size are not allowed */
862 if (iocb->ki_pos >= ZONEFS_I(inode)->i_max_size)
865 if (iocb->ki_flags & IOCB_DIRECT) {
866 ssize_t ret = zonefs_file_dio_write(iocb, from);
871 return zonefs_file_buffered_write(iocb, from);
874 static int zonefs_file_read_dio_end_io(struct kiocb *iocb, ssize_t size,
875 int error, unsigned int flags)
878 zonefs_io_error(file_inode(iocb->ki_filp), false);
885 static const struct iomap_dio_ops zonefs_read_dio_ops = {
886 .end_io = zonefs_file_read_dio_end_io,
889 static ssize_t zonefs_file_read_iter(struct kiocb *iocb, struct iov_iter *to)
891 struct inode *inode = file_inode(iocb->ki_filp);
892 struct zonefs_inode_info *zi = ZONEFS_I(inode);
893 struct super_block *sb = inode->i_sb;
897 /* Offline zones cannot be read */
898 if (unlikely(IS_IMMUTABLE(inode) && !(inode->i_mode & 0777)))
901 if (iocb->ki_pos >= zi->i_max_size)
904 if (iocb->ki_flags & IOCB_NOWAIT) {
905 if (!inode_trylock_shared(inode))
908 inode_lock_shared(inode);
911 /* Limit read operations to written data */
912 mutex_lock(&zi->i_truncate_mutex);
913 isize = i_size_read(inode);
914 if (iocb->ki_pos >= isize) {
915 mutex_unlock(&zi->i_truncate_mutex);
919 iov_iter_truncate(to, isize - iocb->ki_pos);
920 mutex_unlock(&zi->i_truncate_mutex);
922 if (iocb->ki_flags & IOCB_DIRECT) {
923 size_t count = iov_iter_count(to);
925 if ((iocb->ki_pos | count) & (sb->s_blocksize - 1)) {
929 file_accessed(iocb->ki_filp);
930 ret = iomap_dio_rw(iocb, to, &zonefs_iomap_ops,
931 &zonefs_read_dio_ops, 0);
933 ret = generic_file_read_iter(iocb, to);
935 zonefs_io_error(inode, false);
939 inode_unlock_shared(inode);
944 static inline bool zonefs_file_use_exp_open(struct inode *inode, struct file *file)
946 struct zonefs_inode_info *zi = ZONEFS_I(inode);
947 struct zonefs_sb_info *sbi = ZONEFS_SB(inode->i_sb);
949 if (!(sbi->s_mount_opts & ZONEFS_MNTOPT_EXPLICIT_OPEN))
952 if (zi->i_ztype != ZONEFS_ZTYPE_SEQ)
955 if (!(file->f_mode & FMODE_WRITE))
961 static int zonefs_open_zone(struct inode *inode)
963 struct zonefs_inode_info *zi = ZONEFS_I(inode);
964 struct zonefs_sb_info *sbi = ZONEFS_SB(inode->i_sb);
967 mutex_lock(&zi->i_truncate_mutex);
970 if (zi->i_wr_refcnt == 1) {
972 if (atomic_inc_return(&sbi->s_open_zones) > sbi->s_max_open_zones) {
973 atomic_dec(&sbi->s_open_zones);
978 if (i_size_read(inode) < zi->i_max_size) {
979 ret = zonefs_zone_mgmt(inode, REQ_OP_ZONE_OPEN);
982 atomic_dec(&sbi->s_open_zones);
985 zi->i_flags |= ZONEFS_ZONE_OPEN;
990 mutex_unlock(&zi->i_truncate_mutex);
995 static int zonefs_file_open(struct inode *inode, struct file *file)
999 ret = generic_file_open(inode, file);
1003 if (zonefs_file_use_exp_open(inode, file))
1004 return zonefs_open_zone(inode);
1009 static void zonefs_close_zone(struct inode *inode)
1011 struct zonefs_inode_info *zi = ZONEFS_I(inode);
1014 mutex_lock(&zi->i_truncate_mutex);
1016 if (!zi->i_wr_refcnt) {
1017 struct zonefs_sb_info *sbi = ZONEFS_SB(inode->i_sb);
1018 struct super_block *sb = inode->i_sb;
1021 * If the file zone is full, it is not open anymore and we only
1022 * need to decrement the open count.
1024 if (!(zi->i_flags & ZONEFS_ZONE_OPEN))
1027 ret = zonefs_zone_mgmt(inode, REQ_OP_ZONE_CLOSE);
1029 __zonefs_io_error(inode, false);
1031 * Leaving zones explicitly open may lead to a state
1032 * where most zones cannot be written (zone resources
1033 * exhausted). So take preventive action by remounting
1036 if (zi->i_flags & ZONEFS_ZONE_OPEN &&
1037 !(sb->s_flags & SB_RDONLY)) {
1038 zonefs_warn(sb, "closing zone failed, remounting filesystem read-only\n");
1039 sb->s_flags |= SB_RDONLY;
1042 zi->i_flags &= ~ZONEFS_ZONE_OPEN;
1044 atomic_dec(&sbi->s_open_zones);
1046 mutex_unlock(&zi->i_truncate_mutex);
1049 static int zonefs_file_release(struct inode *inode, struct file *file)
1052 * If we explicitly open a zone we must close it again as well, but the
1053 * zone management operation can fail (either due to an IO error or as
1054 * the zone has gone offline or read-only). Make sure we don't fail the
1055 * close(2) for user-space.
1057 if (zonefs_file_use_exp_open(inode, file))
1058 zonefs_close_zone(inode);
1063 static const struct file_operations zonefs_file_operations = {
1064 .open = zonefs_file_open,
1065 .release = zonefs_file_release,
1066 .fsync = zonefs_file_fsync,
1067 .mmap = zonefs_file_mmap,
1068 .llseek = zonefs_file_llseek,
1069 .read_iter = zonefs_file_read_iter,
1070 .write_iter = zonefs_file_write_iter,
1071 .splice_read = generic_file_splice_read,
1072 .splice_write = iter_file_splice_write,
1073 .iopoll = iomap_dio_iopoll,
1076 static struct kmem_cache *zonefs_inode_cachep;
1078 static struct inode *zonefs_alloc_inode(struct super_block *sb)
1080 struct zonefs_inode_info *zi;
1082 zi = kmem_cache_alloc(zonefs_inode_cachep, GFP_KERNEL);
1086 inode_init_once(&zi->i_vnode);
1087 mutex_init(&zi->i_truncate_mutex);
1088 init_rwsem(&zi->i_mmap_sem);
1089 zi->i_wr_refcnt = 0;
1091 return &zi->i_vnode;
1094 static void zonefs_free_inode(struct inode *inode)
1096 kmem_cache_free(zonefs_inode_cachep, ZONEFS_I(inode));
1102 static int zonefs_statfs(struct dentry *dentry, struct kstatfs *buf)
1104 struct super_block *sb = dentry->d_sb;
1105 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
1106 enum zonefs_ztype t;
1109 buf->f_type = ZONEFS_MAGIC;
1110 buf->f_bsize = sb->s_blocksize;
1111 buf->f_namelen = ZONEFS_NAME_MAX;
1113 spin_lock(&sbi->s_lock);
1115 buf->f_blocks = sbi->s_blocks;
1116 if (WARN_ON(sbi->s_used_blocks > sbi->s_blocks))
1119 buf->f_bfree = buf->f_blocks - sbi->s_used_blocks;
1120 buf->f_bavail = buf->f_bfree;
1122 for (t = 0; t < ZONEFS_ZTYPE_MAX; t++) {
1123 if (sbi->s_nr_files[t])
1124 buf->f_files += sbi->s_nr_files[t] + 1;
1128 spin_unlock(&sbi->s_lock);
1130 fsid = le64_to_cpup((void *)sbi->s_uuid.b) ^
1131 le64_to_cpup((void *)sbi->s_uuid.b + sizeof(u64));
1132 buf->f_fsid = u64_to_fsid(fsid);
1138 Opt_errors_ro, Opt_errors_zro, Opt_errors_zol, Opt_errors_repair,
1139 Opt_explicit_open, Opt_err,
1142 static const match_table_t tokens = {
1143 { Opt_errors_ro, "errors=remount-ro"},
1144 { Opt_errors_zro, "errors=zone-ro"},
1145 { Opt_errors_zol, "errors=zone-offline"},
1146 { Opt_errors_repair, "errors=repair"},
1147 { Opt_explicit_open, "explicit-open" },
1151 static int zonefs_parse_options(struct super_block *sb, char *options)
1153 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
1154 substring_t args[MAX_OPT_ARGS];
1160 while ((p = strsep(&options, ",")) != NULL) {
1166 token = match_token(p, tokens, args);
1169 sbi->s_mount_opts &= ~ZONEFS_MNTOPT_ERRORS_MASK;
1170 sbi->s_mount_opts |= ZONEFS_MNTOPT_ERRORS_RO;
1172 case Opt_errors_zro:
1173 sbi->s_mount_opts &= ~ZONEFS_MNTOPT_ERRORS_MASK;
1174 sbi->s_mount_opts |= ZONEFS_MNTOPT_ERRORS_ZRO;
1176 case Opt_errors_zol:
1177 sbi->s_mount_opts &= ~ZONEFS_MNTOPT_ERRORS_MASK;
1178 sbi->s_mount_opts |= ZONEFS_MNTOPT_ERRORS_ZOL;
1180 case Opt_errors_repair:
1181 sbi->s_mount_opts &= ~ZONEFS_MNTOPT_ERRORS_MASK;
1182 sbi->s_mount_opts |= ZONEFS_MNTOPT_ERRORS_REPAIR;
1184 case Opt_explicit_open:
1185 sbi->s_mount_opts |= ZONEFS_MNTOPT_EXPLICIT_OPEN;
1195 static int zonefs_show_options(struct seq_file *seq, struct dentry *root)
1197 struct zonefs_sb_info *sbi = ZONEFS_SB(root->d_sb);
1199 if (sbi->s_mount_opts & ZONEFS_MNTOPT_ERRORS_RO)
1200 seq_puts(seq, ",errors=remount-ro");
1201 if (sbi->s_mount_opts & ZONEFS_MNTOPT_ERRORS_ZRO)
1202 seq_puts(seq, ",errors=zone-ro");
1203 if (sbi->s_mount_opts & ZONEFS_MNTOPT_ERRORS_ZOL)
1204 seq_puts(seq, ",errors=zone-offline");
1205 if (sbi->s_mount_opts & ZONEFS_MNTOPT_ERRORS_REPAIR)
1206 seq_puts(seq, ",errors=repair");
1211 static int zonefs_remount(struct super_block *sb, int *flags, char *data)
1213 sync_filesystem(sb);
1215 return zonefs_parse_options(sb, data);
1218 static const struct super_operations zonefs_sops = {
1219 .alloc_inode = zonefs_alloc_inode,
1220 .free_inode = zonefs_free_inode,
1221 .statfs = zonefs_statfs,
1222 .remount_fs = zonefs_remount,
1223 .show_options = zonefs_show_options,
1226 static const struct inode_operations zonefs_dir_inode_operations = {
1227 .lookup = simple_lookup,
1228 .setattr = zonefs_inode_setattr,
1231 static void zonefs_init_dir_inode(struct inode *parent, struct inode *inode,
1232 enum zonefs_ztype type)
1234 struct super_block *sb = parent->i_sb;
1236 inode->i_ino = blkdev_nr_zones(sb->s_bdev->bd_disk) + type + 1;
1237 inode_init_owner(&init_user_ns, inode, parent, S_IFDIR | 0555);
1238 inode->i_op = &zonefs_dir_inode_operations;
1239 inode->i_fop = &simple_dir_operations;
1240 set_nlink(inode, 2);
1244 static void zonefs_init_file_inode(struct inode *inode, struct blk_zone *zone,
1245 enum zonefs_ztype type)
1247 struct super_block *sb = inode->i_sb;
1248 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
1249 struct zonefs_inode_info *zi = ZONEFS_I(inode);
1251 inode->i_ino = zone->start >> sbi->s_zone_sectors_shift;
1252 inode->i_mode = S_IFREG | sbi->s_perm;
1255 zi->i_zsector = zone->start;
1256 zi->i_zone_size = zone->len << SECTOR_SHIFT;
1258 zi->i_max_size = min_t(loff_t, MAX_LFS_FILESIZE,
1259 zone->capacity << SECTOR_SHIFT);
1260 zi->i_wpoffset = zonefs_check_zone_condition(inode, zone, true, true);
1262 inode->i_uid = sbi->s_uid;
1263 inode->i_gid = sbi->s_gid;
1264 inode->i_size = zi->i_wpoffset;
1265 inode->i_blocks = zi->i_max_size >> SECTOR_SHIFT;
1267 inode->i_op = &zonefs_file_inode_operations;
1268 inode->i_fop = &zonefs_file_operations;
1269 inode->i_mapping->a_ops = &zonefs_file_aops;
1271 sb->s_maxbytes = max(zi->i_max_size, sb->s_maxbytes);
1272 sbi->s_blocks += zi->i_max_size >> sb->s_blocksize_bits;
1273 sbi->s_used_blocks += zi->i_wpoffset >> sb->s_blocksize_bits;
1276 static struct dentry *zonefs_create_inode(struct dentry *parent,
1277 const char *name, struct blk_zone *zone,
1278 enum zonefs_ztype type)
1280 struct inode *dir = d_inode(parent);
1281 struct dentry *dentry;
1282 struct inode *inode;
1284 dentry = d_alloc_name(parent, name);
1288 inode = new_inode(parent->d_sb);
1292 inode->i_ctime = inode->i_mtime = inode->i_atime = dir->i_ctime;
1294 zonefs_init_file_inode(inode, zone, type);
1296 zonefs_init_dir_inode(dir, inode, type);
1297 d_add(dentry, inode);
1308 struct zonefs_zone_data {
1309 struct super_block *sb;
1310 unsigned int nr_zones[ZONEFS_ZTYPE_MAX];
1311 struct blk_zone *zones;
1315 * Create a zone group and populate it with zone files.
1317 static int zonefs_create_zgroup(struct zonefs_zone_data *zd,
1318 enum zonefs_ztype type)
1320 struct super_block *sb = zd->sb;
1321 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
1322 struct blk_zone *zone, *next, *end;
1323 const char *zgroup_name;
1329 /* If the group is empty, there is nothing to do */
1330 if (!zd->nr_zones[type])
1333 file_name = kmalloc(ZONEFS_NAME_MAX, GFP_KERNEL);
1337 if (type == ZONEFS_ZTYPE_CNV)
1338 zgroup_name = "cnv";
1340 zgroup_name = "seq";
1342 dir = zonefs_create_inode(sb->s_root, zgroup_name, NULL, type);
1349 * The first zone contains the super block: skip it.
1351 end = zd->zones + blkdev_nr_zones(sb->s_bdev->bd_disk);
1352 for (zone = &zd->zones[1]; zone < end; zone = next) {
1355 if (zonefs_zone_type(zone) != type)
1359 * For conventional zones, contiguous zones can be aggregated
1360 * together to form larger files. Note that this overwrites the
1361 * length of the first zone of the set of contiguous zones
1362 * aggregated together. If one offline or read-only zone is
1363 * found, assume that all zones aggregated have the same
1366 if (type == ZONEFS_ZTYPE_CNV &&
1367 (sbi->s_features & ZONEFS_F_AGGRCNV)) {
1368 for (; next < end; next++) {
1369 if (zonefs_zone_type(next) != type)
1371 zone->len += next->len;
1372 zone->capacity += next->capacity;
1373 if (next->cond == BLK_ZONE_COND_READONLY &&
1374 zone->cond != BLK_ZONE_COND_OFFLINE)
1375 zone->cond = BLK_ZONE_COND_READONLY;
1376 else if (next->cond == BLK_ZONE_COND_OFFLINE)
1377 zone->cond = BLK_ZONE_COND_OFFLINE;
1379 if (zone->capacity != zone->len) {
1380 zonefs_err(sb, "Invalid conventional zone capacity\n");
1387 * Use the file number within its group as file name.
1389 snprintf(file_name, ZONEFS_NAME_MAX - 1, "%u", n);
1390 if (!zonefs_create_inode(dir, file_name, zone, type)) {
1398 zonefs_info(sb, "Zone group \"%s\" has %u file%s\n",
1399 zgroup_name, n, n > 1 ? "s" : "");
1401 sbi->s_nr_files[type] = n;
1410 static int zonefs_get_zone_info_cb(struct blk_zone *zone, unsigned int idx,
1413 struct zonefs_zone_data *zd = data;
1416 * Count the number of usable zones: the first zone at index 0 contains
1417 * the super block and is ignored.
1419 switch (zone->type) {
1420 case BLK_ZONE_TYPE_CONVENTIONAL:
1421 zone->wp = zone->start + zone->len;
1423 zd->nr_zones[ZONEFS_ZTYPE_CNV]++;
1425 case BLK_ZONE_TYPE_SEQWRITE_REQ:
1426 case BLK_ZONE_TYPE_SEQWRITE_PREF:
1428 zd->nr_zones[ZONEFS_ZTYPE_SEQ]++;
1431 zonefs_err(zd->sb, "Unsupported zone type 0x%x\n",
1436 memcpy(&zd->zones[idx], zone, sizeof(struct blk_zone));
1441 static int zonefs_get_zone_info(struct zonefs_zone_data *zd)
1443 struct block_device *bdev = zd->sb->s_bdev;
1446 zd->zones = kvcalloc(blkdev_nr_zones(bdev->bd_disk),
1447 sizeof(struct blk_zone), GFP_KERNEL);
1451 /* Get zones information from the device */
1452 ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES,
1453 zonefs_get_zone_info_cb, zd);
1455 zonefs_err(zd->sb, "Zone report failed %d\n", ret);
1459 if (ret != blkdev_nr_zones(bdev->bd_disk)) {
1460 zonefs_err(zd->sb, "Invalid zone report (%d/%u zones)\n",
1461 ret, blkdev_nr_zones(bdev->bd_disk));
1468 static inline void zonefs_cleanup_zone_info(struct zonefs_zone_data *zd)
1474 * Read super block information from the device.
1476 static int zonefs_read_super(struct super_block *sb)
1478 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
1479 struct zonefs_super *super;
1480 u32 crc, stored_crc;
1482 struct bio_vec bio_vec;
1486 page = alloc_page(GFP_KERNEL);
1490 bio_init(&bio, &bio_vec, 1);
1491 bio.bi_iter.bi_sector = 0;
1492 bio.bi_opf = REQ_OP_READ;
1493 bio_set_dev(&bio, sb->s_bdev);
1494 bio_add_page(&bio, page, PAGE_SIZE, 0);
1496 ret = submit_bio_wait(&bio);
1503 if (le32_to_cpu(super->s_magic) != ZONEFS_MAGIC)
1506 stored_crc = le32_to_cpu(super->s_crc);
1508 crc = crc32(~0U, (unsigned char *)super, sizeof(struct zonefs_super));
1509 if (crc != stored_crc) {
1510 zonefs_err(sb, "Invalid checksum (Expected 0x%08x, got 0x%08x)",
1515 sbi->s_features = le64_to_cpu(super->s_features);
1516 if (sbi->s_features & ~ZONEFS_F_DEFINED_FEATURES) {
1517 zonefs_err(sb, "Unknown features set 0x%llx\n",
1522 if (sbi->s_features & ZONEFS_F_UID) {
1523 sbi->s_uid = make_kuid(current_user_ns(),
1524 le32_to_cpu(super->s_uid));
1525 if (!uid_valid(sbi->s_uid)) {
1526 zonefs_err(sb, "Invalid UID feature\n");
1531 if (sbi->s_features & ZONEFS_F_GID) {
1532 sbi->s_gid = make_kgid(current_user_ns(),
1533 le32_to_cpu(super->s_gid));
1534 if (!gid_valid(sbi->s_gid)) {
1535 zonefs_err(sb, "Invalid GID feature\n");
1540 if (sbi->s_features & ZONEFS_F_PERM)
1541 sbi->s_perm = le32_to_cpu(super->s_perm);
1543 if (memchr_inv(super->s_reserved, 0, sizeof(super->s_reserved))) {
1544 zonefs_err(sb, "Reserved area is being used\n");
1548 import_uuid(&sbi->s_uuid, super->s_uuid);
1560 * Check that the device is zoned. If it is, get the list of zones and create
1561 * sub-directories and files according to the device zone configuration and
1564 static int zonefs_fill_super(struct super_block *sb, void *data, int silent)
1566 struct zonefs_zone_data zd;
1567 struct zonefs_sb_info *sbi;
1568 struct inode *inode;
1569 enum zonefs_ztype t;
1572 if (!bdev_is_zoned(sb->s_bdev)) {
1573 zonefs_err(sb, "Not a zoned block device\n");
1578 * Initialize super block information: the maximum file size is updated
1579 * when the zone files are created so that the format option
1580 * ZONEFS_F_AGGRCNV which increases the maximum file size of a file
1581 * beyond the zone size is taken into account.
1583 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1587 spin_lock_init(&sbi->s_lock);
1588 sb->s_fs_info = sbi;
1589 sb->s_magic = ZONEFS_MAGIC;
1591 sb->s_op = &zonefs_sops;
1592 sb->s_time_gran = 1;
1595 * The block size is set to the device zone write granularity to ensure
1596 * that write operations are always aligned according to the device
1597 * interface constraints.
1599 sb_set_blocksize(sb, bdev_zone_write_granularity(sb->s_bdev));
1600 sbi->s_zone_sectors_shift = ilog2(bdev_zone_sectors(sb->s_bdev));
1601 sbi->s_uid = GLOBAL_ROOT_UID;
1602 sbi->s_gid = GLOBAL_ROOT_GID;
1604 sbi->s_mount_opts = ZONEFS_MNTOPT_ERRORS_RO;
1605 sbi->s_max_open_zones = bdev_max_open_zones(sb->s_bdev);
1606 atomic_set(&sbi->s_open_zones, 0);
1607 if (!sbi->s_max_open_zones &&
1608 sbi->s_mount_opts & ZONEFS_MNTOPT_EXPLICIT_OPEN) {
1609 zonefs_info(sb, "No open zones limit. Ignoring explicit_open mount option\n");
1610 sbi->s_mount_opts &= ~ZONEFS_MNTOPT_EXPLICIT_OPEN;
1613 ret = zonefs_read_super(sb);
1617 ret = zonefs_parse_options(sb, data);
1621 memset(&zd, 0, sizeof(struct zonefs_zone_data));
1623 ret = zonefs_get_zone_info(&zd);
1627 zonefs_info(sb, "Mounting %u zones",
1628 blkdev_nr_zones(sb->s_bdev->bd_disk));
1630 /* Create root directory inode */
1632 inode = new_inode(sb);
1636 inode->i_ino = blkdev_nr_zones(sb->s_bdev->bd_disk);
1637 inode->i_mode = S_IFDIR | 0555;
1638 inode->i_ctime = inode->i_mtime = inode->i_atime = current_time(inode);
1639 inode->i_op = &zonefs_dir_inode_operations;
1640 inode->i_fop = &simple_dir_operations;
1641 set_nlink(inode, 2);
1643 sb->s_root = d_make_root(inode);
1647 /* Create and populate files in zone groups directories */
1648 for (t = 0; t < ZONEFS_ZTYPE_MAX; t++) {
1649 ret = zonefs_create_zgroup(&zd, t);
1655 zonefs_cleanup_zone_info(&zd);
1660 static struct dentry *zonefs_mount(struct file_system_type *fs_type,
1661 int flags, const char *dev_name, void *data)
1663 return mount_bdev(fs_type, flags, dev_name, data, zonefs_fill_super);
1666 static void zonefs_kill_super(struct super_block *sb)
1668 struct zonefs_sb_info *sbi = ZONEFS_SB(sb);
1671 d_genocide(sb->s_root);
1672 kill_block_super(sb);
1677 * File system definition and registration.
1679 static struct file_system_type zonefs_type = {
1680 .owner = THIS_MODULE,
1682 .mount = zonefs_mount,
1683 .kill_sb = zonefs_kill_super,
1684 .fs_flags = FS_REQUIRES_DEV,
1687 static int __init zonefs_init_inodecache(void)
1689 zonefs_inode_cachep = kmem_cache_create("zonefs_inode_cache",
1690 sizeof(struct zonefs_inode_info), 0,
1691 (SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD | SLAB_ACCOUNT),
1693 if (zonefs_inode_cachep == NULL)
1698 static void zonefs_destroy_inodecache(void)
1701 * Make sure all delayed rcu free inodes are flushed before we
1702 * destroy the inode cache.
1705 kmem_cache_destroy(zonefs_inode_cachep);
1708 static int __init zonefs_init(void)
1712 BUILD_BUG_ON(sizeof(struct zonefs_super) != ZONEFS_SUPER_SIZE);
1714 ret = zonefs_init_inodecache();
1718 ret = register_filesystem(&zonefs_type);
1720 zonefs_destroy_inodecache();
1727 static void __exit zonefs_exit(void)
1729 zonefs_destroy_inodecache();
1730 unregister_filesystem(&zonefs_type);
1733 MODULE_AUTHOR("Damien Le Moal");
1734 MODULE_DESCRIPTION("Zone file system for zoned block devices");
1735 MODULE_LICENSE("GPL");
1736 module_init(zonefs_init);
1737 module_exit(zonefs_exit);