1 /* SPDX-License-Identifier: GPL-2.0 */
9 #include <linux/backing-dev.h>
29  * large volume sizes on 32-bit machines. */
44 #define CEPH_MOUNT_OPT_NOCOPYFROM      (1<<14) /* don't use RADOS 'copy-from' op */
55 	(fsc)->mount_options->flags |= CEPH_MOUNT_OPT_##opt
57 	(fsc)->mount_options->flags &= ~CEPH_MOUNT_OPT_##opt
59 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
173  * Each cap is referenced by the inode's i_caps rbtree and by per-mds
178 	struct rb_node ci_node;          /* per-ci cap tree */
180 	struct list_head session_caps;   /* per-session caplist */
183 		/* in-use caps */
219  * we first complete any in-process sync writes and writeback any dirty
253 	if (refcount_dec_and_test(&capsnap->nref)) {  in ceph_put_cap_snap()
254 		if (capsnap->xattr_blob)  in ceph_put_cap_snap()
255 			ceph_buffer_put(capsnap->xattr_blob);  in ceph_put_cap_snap()
278 	int mds;              /* -1 if same authority as parent */
331 	 * blob->vec.iov_len == 4 implies there are no xattrs; blob ==
383 	/* capabilities.  protected _both_ by i_ceph_lock and cap->session's
391 	 * is protected by the mdsc->cap_dirty_lock, but each individual item
393 	 * requires the mdsc->cap_dirty_lock. List presence for an item can
402 	 * or msdc->cap_dirty_lock. List presence can also be checked while
407 	/* we need to track cap writeback on a per-cap-bit basis, to allow
454 		struct ceph_snap_realm *i_snap_realm; /* snap realm (if caps) */
455 		struct ceph_snapid_map *i_snapid_map; /* snapid -> dev_t */
495 	return (struct ceph_fs_client *)inode->i_sb->s_fs_info;  in ceph_inode_to_fs_client()
501 	return (struct ceph_fs_client *)sb->s_fs_info;  in ceph_sb_to_fs_client()
507 	return (struct ceph_mds_client *)ceph_sb_to_fs_client(sb)->mdsc;  in ceph_sb_to_mdsc()
513 	return (struct ceph_client *)ceph_inode_to_fs_client(inode)->client;  in ceph_inode_to_client()
519 	return ceph_inode(inode)->i_vino;  in ceph_vino()
532  * Inode numbers in cephfs are 64 bits, but inode->i_ino is 32-bits on
544 /* for printf-style formatting */
545 #define ceph_vinop(i) ceph_inode(i)->i_vino.ino, ceph_inode(i)->i_vino.snap
549 	return ceph_inode(inode)->i_vino.ino;  in ceph_ino()
554 	return ceph_inode(inode)->i_vino.snap;  in ceph_snap()
558  * ceph_present_ino - format an inode number for presentation to userland
562  * If the user mounted with the ino32 option, then the 64-bit value needs
576 	return ceph_present_ino(inode->i_sb, ceph_ino(inode));  in ceph_present_inode()
583 	return ci->i_vino.ino == pvino->ino &&  in ceph_ino_compare()
584 		ci->i_vino.snap == pvino->snap;  in ceph_ino_compare()
622 	 * 32-bits first. Just use low-order bits on arches with 32-bit long.  in ceph_find_inode()
666 	if (!(READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE)) {  in ceph_set_error_write()
667 		spin_lock(&ci->i_ceph_lock);  in ceph_set_error_write()
668 		ci->i_ceph_flags |= CEPH_I_ERROR_WRITE;  in ceph_set_error_write()
669 		spin_unlock(&ci->i_ceph_lock);  in ceph_set_error_write()
675 	if (READ_ONCE(ci->i_ceph_flags) & CEPH_I_ERROR_WRITE) {  in ceph_clear_error_write()
676 		spin_lock(&ci->i_ceph_lock);  in ceph_clear_error_write()
677 		ci->i_ceph_flags &= ~CEPH_I_ERROR_WRITE;  in ceph_clear_error_write()
678 		spin_unlock(&ci->i_ceph_lock);  in ceph_clear_error_write()
692 	atomic64_set(&ci->i_complete_seq[0], release_count);  in __ceph_dir_set_complete()
693 	atomic64_set(&ci->i_complete_seq[1], ordered_count);  in __ceph_dir_set_complete()
698 	atomic64_inc(&ci->i_release_count);  in __ceph_dir_clear_complete()
703 	atomic64_inc(&ci->i_ordered_count);  in __ceph_dir_clear_ordered()
708 	return atomic64_read(&ci->i_complete_seq[0]) ==  in __ceph_dir_is_complete()
709 		atomic64_read(&ci->i_release_count);  in __ceph_dir_is_complete()
714 	return  atomic64_read(&ci->i_complete_seq[0]) ==  in __ceph_dir_is_complete_ordered()
715 		atomic64_read(&ci->i_release_count) &&  in __ceph_dir_is_complete_ordered()
716 		atomic64_read(&ci->i_complete_seq[1]) ==  in __ceph_dir_is_complete_ordered()
717 		atomic64_read(&ci->i_ordered_count);  in __ceph_dir_is_complete_ordered()
751 	return (struct ceph_dentry_info *)dentry->d_fsdata;  in ceph_dentry()
759 	return !RB_EMPTY_ROOT(&ci->i_caps);  in __ceph_is_any_real_caps()
772 	spin_lock(&ci->i_ceph_lock);  in ceph_caps_issued()
774 	spin_unlock(&ci->i_ceph_lock);  in ceph_caps_issued()
782 	spin_lock(&ci->i_ceph_lock);  in ceph_caps_issued_mask_metric()
784 	spin_unlock(&ci->i_ceph_lock);  in ceph_caps_issued_mask_metric()
790 	return ci->i_dirty_caps | ci->i_flushing_caps;  in __ceph_caps_dirty()
803 	return ci->i_nr_by_mode[0];  in __ceph_is_file_opened()
828  * we keep buffered readdir results attached to file->private_data
857 	/* used for -o dirstat read() on directory thing */
877 	spin_lock(&cf->rw_contexts_lock);  in ceph_add_rw_context()
878 	list_add(&ctx->list, &cf->rw_contexts);  in ceph_add_rw_context()
879 	spin_unlock(&cf->rw_contexts_lock);  in ceph_add_rw_context()
885 	spin_lock(&cf->rw_contexts_lock);  in ceph_del_rw_context()
886 	list_del(&ctx->list);  in ceph_del_rw_context()
887 	spin_unlock(&cf->rw_contexts_lock);  in ceph_del_rw_context()
894 	spin_lock(&cf->rw_contexts_lock);  in ceph_find_rw_context()
895 	list_for_each_entry(ctx, &cf->rw_contexts, list) {  in ceph_find_rw_context()
896 		if (ctx->thread == current) {  in ceph_find_rw_context()
901 	spin_unlock(&cf->rw_contexts_lock);  in ceph_find_rw_context()
912  * A "snap realm" describes a subset of the file hierarchy sharing
917  * All inodes within the realm that have capabilities are linked into a
918  * per-realm list.
932 	u64 *snaps;                   /* snaps specific to this realm */
941 	struct list_head dirty_item;     /* if realm needs new context */
945 	/* the current set of snaps for this realm */
974 	congestion_kb = (16*int_sqrt(totalram_pages())) << (PAGE_SHIFT-10);  in default_congestion_kb()
988 				struct ceph_snap_realm *realm);
990 				struct ceph_snap_realm *realm);
994 void ceph_change_snap_realm(struct inode *inode, struct ceph_snap_realm *realm);
1012  * a cap_snap is "pending" if it is still awaiting an in-progress
1017 	return !list_empty(&ci->i_cap_snaps) &&  in __ceph_have_pending_cap_snap()
1018 	       list_last_entry(&ci->i_cap_snaps, struct ceph_cap_snap,  in __ceph_have_pending_cap_snap()
1019 			       ci_item)->writing;  in __ceph_have_pending_cap_snap()
1113 	unsigned long flags = READ_ONCE(ceph_inode(inode)->i_ceph_flags);  in ceph_inode_is_shutdown()
1115 	int state = READ_ONCE(fsc->mount_state);  in ceph_inode_is_shutdown()
1297 	if (ci->i_inline_version == CEPH_INLINE_NONE ||  in ceph_has_inline_data()
1298 	    ci->i_inline_version == 1) /* initial version, no data */  in ceph_has_inline_data()
1377 		has_quota = !!ci->i_max_bytes;  in __ceph_has_quota()
1380 		has_quota = !!ci->i_max_files;  in __ceph_has_quota()
1383 		has_quota = !!(ci->i_max_files || ci->i_max_bytes);  in __ceph_has_quota()
1395 	ci->i_max_bytes = max_bytes;  in __ceph_update_quota()
1396 	ci->i_max_files = max_files;  in __ceph_update_quota()
1400 		ceph_adjust_quota_realms_count(&ci->netfs.inode, has_quota);  in __ceph_update_quota()