Print this page
NEX-9200 Improve the scalability of attribute locking in zfs_zget
Reviewed by: Joyce McIntosh <joyce.mcintosh@nexenta.com>
Reviewed by: Sanjay Nadkarni <sanjay.nadkarni@nexenta.com>
| Split |
Close |
| Expand all |
| Collapse all |
--- old/usr/src/uts/common/fs/zfs/sys/zfs_znode.h
+++ new/usr/src/uts/common/fs/zfs/sys/zfs_znode.h
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21 /*
22 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 23 * Copyright (c) 2012, 2015 by Delphix. All rights reserved.
24 24 * Copyright (c) 2014 Integros [integros.com]
25 25 * Copyright 2016 Nexenta Systems, Inc. All rights reserved.
26 26 */
27 27
28 28 #ifndef _SYS_FS_ZFS_ZNODE_H
29 29 #define _SYS_FS_ZFS_ZNODE_H
30 30
31 31 #ifdef _KERNEL
32 32 #include <sys/isa_defs.h>
33 33 #include <sys/types32.h>
34 34 #include <sys/attr.h>
35 35 #include <sys/list.h>
36 36 #include <sys/dmu.h>
37 37 #include <sys/sa.h>
38 38 #include <sys/zfs_vfsops.h>
39 39 #include <sys/rrwlock.h>
40 40 #include <sys/zfs_sa.h>
41 41 #include <sys/zfs_stat.h>
42 42 #endif
43 43 #include <sys/zfs_acl.h>
44 44 #include <sys/zil.h>
45 45
46 46 #ifdef __cplusplus
47 47 extern "C" {
48 48 #endif
49 49
50 50 /*
51 51 * Additional file level attributes, that are stored
52 52 * in the upper half of zp_flags
53 53 */
54 54 #define ZFS_READONLY 0x0000000100000000
55 55 #define ZFS_HIDDEN 0x0000000200000000
56 56 #define ZFS_SYSTEM 0x0000000400000000
57 57 #define ZFS_ARCHIVE 0x0000000800000000
58 58 #define ZFS_IMMUTABLE 0x0000001000000000
59 59 #define ZFS_NOUNLINK 0x0000002000000000
60 60 #define ZFS_APPENDONLY 0x0000004000000000
61 61 #define ZFS_NODUMP 0x0000008000000000
62 62 #define ZFS_OPAQUE 0x0000010000000000
63 63 #define ZFS_AV_QUARANTINED 0x0000020000000000
64 64 #define ZFS_AV_MODIFIED 0x0000040000000000
65 65 #define ZFS_REPARSE 0x0000080000000000
66 66 #define ZFS_OFFLINE 0x0000100000000000
67 67 #define ZFS_SPARSE 0x0000200000000000
68 68
69 69 #define ZFS_ATTR_SET(zp, attr, value, pflags, tx) \
70 70 { \
71 71 if (value) \
72 72 pflags |= attr; \
73 73 else \
74 74 pflags &= ~attr; \
75 75 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_FLAGS(zp->z_zfsvfs), \
76 76 &pflags, sizeof (pflags), tx)); \
77 77 }
78 78
79 79 /*
80 80 * Define special zfs pflags
81 81 */
82 82 #define ZFS_XATTR 0x1 /* is an extended attribute */
83 83 #define ZFS_INHERIT_ACE 0x2 /* ace has inheritable ACEs */
84 84 #define ZFS_ACL_TRIVIAL 0x4 /* files ACL is trivial */
85 85 #define ZFS_ACL_OBJ_ACE 0x8 /* ACL has CMPLX Object ACE */
86 86 #define ZFS_ACL_PROTECTED 0x10 /* ACL protected */
87 87 #define ZFS_ACL_DEFAULTED 0x20 /* ACL should be defaulted */
88 88 #define ZFS_ACL_AUTO_INHERIT 0x40 /* ACL should be inherited */
89 89 #define ZFS_BONUS_SCANSTAMP 0x80 /* Scanstamp in bonus area */
90 90 #define ZFS_NO_EXECS_DENIED 0x100 /* exec was given to everyone */
91 91
92 92 #define SA_ZPL_ATIME(z) z->z_attr_table[ZPL_ATIME]
93 93 #define SA_ZPL_MTIME(z) z->z_attr_table[ZPL_MTIME]
94 94 #define SA_ZPL_CTIME(z) z->z_attr_table[ZPL_CTIME]
95 95 #define SA_ZPL_CRTIME(z) z->z_attr_table[ZPL_CRTIME]
96 96 #define SA_ZPL_GEN(z) z->z_attr_table[ZPL_GEN]
97 97 #define SA_ZPL_DACL_ACES(z) z->z_attr_table[ZPL_DACL_ACES]
98 98 #define SA_ZPL_XATTR(z) z->z_attr_table[ZPL_XATTR]
99 99 #define SA_ZPL_SYMLINK(z) z->z_attr_table[ZPL_SYMLINK]
100 100 #define SA_ZPL_RDEV(z) z->z_attr_table[ZPL_RDEV]
101 101 #define SA_ZPL_SCANSTAMP(z) z->z_attr_table[ZPL_SCANSTAMP]
102 102 #define SA_ZPL_UID(z) z->z_attr_table[ZPL_UID]
103 103 #define SA_ZPL_GID(z) z->z_attr_table[ZPL_GID]
104 104 #define SA_ZPL_PARENT(z) z->z_attr_table[ZPL_PARENT]
105 105 #define SA_ZPL_LINKS(z) z->z_attr_table[ZPL_LINKS]
106 106 #define SA_ZPL_MODE(z) z->z_attr_table[ZPL_MODE]
107 107 #define SA_ZPL_DACL_COUNT(z) z->z_attr_table[ZPL_DACL_COUNT]
108 108 #define SA_ZPL_FLAGS(z) z->z_attr_table[ZPL_FLAGS]
109 109 #define SA_ZPL_SIZE(z) z->z_attr_table[ZPL_SIZE]
110 110 #define SA_ZPL_ZNODE_ACL(z) z->z_attr_table[ZPL_ZNODE_ACL]
111 111 #define SA_ZPL_PAD(z) z->z_attr_table[ZPL_PAD]
112 112
113 113 /*
114 114 * Is ID ephemeral?
115 115 */
116 116 #define IS_EPHEMERAL(x) (x > MAXUID)
117 117
118 118 /*
119 119 * Should we use FUIDs?
120 120 */
121 121 #define USE_FUIDS(version, os) (version >= ZPL_VERSION_FUID && \
122 122 spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
123 123 #define USE_SA(version, os) (version >= ZPL_VERSION_SA && \
124 124 spa_version(dmu_objset_spa(os)) >= SPA_VERSION_SA)
125 125
126 126 #define MASTER_NODE_OBJ 1
127 127
128 128 /*
129 129 * Special attributes for master node.
130 130 * "userquota@" and "groupquota@" are also valid (from
131 131 * zfs_userquota_prop_prefixes[]).
132 132 */
133 133 #define ZFS_FSID "FSID"
134 134 #define ZFS_UNLINKED_SET "DELETE_QUEUE"
135 135 #define ZFS_ROOT_OBJ "ROOT"
136 136 #define ZPL_VERSION_STR "VERSION"
137 137 #define ZFS_FUID_TABLES "FUID"
138 138 #define ZFS_SHARES_DIR "SHARES"
139 139 #define ZFS_SA_ATTRS "SA_ATTRS"
140 140
141 141 /*
142 142 * Convert mode bits (zp_mode) to BSD-style DT_* values for storing in
143 143 * the directory entries.
144 144 */
145 145 #define IFTODT(mode) (((mode) & S_IFMT) >> 12)
146 146
147 147 /*
148 148 * The directory entry has the type (currently unused on Solaris) in the
149 149 * top 4 bits, and the object number in the low 48 bits. The "middle"
150 150 * 12 bits are unused.
151 151 */
152 152 #define ZFS_DIRENT_TYPE(de) BF64_GET(de, 60, 4)
153 153 #define ZFS_DIRENT_OBJ(de) BF64_GET(de, 0, 48)
154 154
155 155 /*
156 156 * Directory entry locks control access to directory entries.
157 157 * They are used to protect creates, deletes, and renames.
158 158 * Each directory znode has a mutex and a list of locked names.
159 159 */
160 160 #ifdef _KERNEL
161 161 typedef struct zfs_dirlock {
162 162 char *dl_name; /* directory entry being locked */
163 163 uint32_t dl_sharecnt; /* 0 if exclusive, > 0 if shared */
164 164 uint8_t dl_namelock; /* 1 if z_name_lock is NOT held */
165 165 uint16_t dl_namesize; /* set if dl_name was allocated */
166 166 kcondvar_t dl_cv; /* wait for entry to be unlocked */
167 167 struct znode *dl_dzp; /* directory znode */
168 168 struct zfs_dirlock *dl_next; /* next in z_dirlocks list */
169 169 } zfs_dirlock_t;
170 170
171 171 typedef struct znode {
172 172 struct zfsvfs *z_zfsvfs;
173 173 vnode_t *z_vnode;
174 174 uint64_t z_id; /* object ID for this znode */
175 175 kmutex_t z_lock; /* znode modification lock */
176 176 krwlock_t z_parent_lock; /* parent lock for directories */
177 177 krwlock_t z_name_lock; /* "master" lock for dirent locks */
178 178 zfs_dirlock_t *z_dirlocks; /* directory entry lock list */
179 179 kmutex_t z_range_lock; /* protects changes to z_range_avl */
180 180 avl_tree_t z_range_avl; /* avl tree of file range locks */
181 181 uint8_t z_unlinked; /* file has been unlinked */
182 182 uint8_t z_atime_dirty; /* atime needs to be synced */
183 183 uint8_t z_zn_prefetch; /* Prefetch znodes? */
184 184 uint8_t z_moved; /* Has this znode been moved? */
185 185 uint_t z_blksz; /* block size in bytes */
186 186 uint_t z_seq; /* modification sequence number */
187 187 uint64_t z_mapcnt; /* number of pages mapped to file */
188 188 uint64_t z_gen; /* generation (cached) */
189 189 uint64_t z_size; /* file size (cached) */
190 190 uint64_t z_atime[2]; /* atime (cached) */
191 191 uint64_t z_links; /* file links (cached) */
192 192 uint64_t z_pflags; /* pflags (cached) */
193 193 uint64_t z_uid; /* uid fuid (cached) */
194 194 uint64_t z_gid; /* gid fuid (cached) */
195 195 mode_t z_mode; /* mode (cached) */
196 196 uint32_t z_sync_cnt; /* synchronous open count */
197 197 kmutex_t z_acl_lock; /* acl data lock */
198 198 zfs_acl_t *z_acl_cached; /* cached acl */
199 199 list_node_t z_link_node; /* all znodes in fs link */
200 200 sa_handle_t *z_sa_hdl; /* handle to sa data */
201 201 boolean_t z_is_sa; /* are we native sa? */
202 202 } znode_t;
203 203
204 204
205 205 /*
206 206 * Range locking rules
207 207 * --------------------
208 208 * 1. When truncating a file (zfs_create, zfs_setattr, zfs_space) the whole
209 209 * file range needs to be locked as RL_WRITER. Only then can the pages be
210 210 * freed etc and zp_size reset. zp_size must be set within range lock.
211 211 * 2. For writes and punching holes (zfs_write & zfs_space) just the range
212 212 * being written or freed needs to be locked as RL_WRITER.
213 213 * Multiple writes at the end of the file must coordinate zp_size updates
214 214 * to ensure data isn't lost. A compare and swap loop is currently used
215 215 * to ensure the file size is at least the offset last written.
216 216 * 3. For reads (zfs_read, zfs_get_data & zfs_putapage) just the range being
217 217 * read needs to be locked as RL_READER. A check against zp_size can then
218 218 * be made for reading beyond end of file.
219 219 */
220 220
221 221 /*
222 222 * Convert between znode pointers and vnode pointers
223 223 */
224 224 #define ZTOV(ZP) ((ZP)->z_vnode)
225 225 #define VTOZ(VP) ((znode_t *)(VP)->v_data)
226 226
227 227 /* Called on entry to each ZFS vnode and vfs operation */
228 228 #define ZFS_ENTER(zfsvfs) \
229 229 { \
230 230 rrm_enter_read(&(zfsvfs)->z_teardown_lock, FTAG); \
231 231 if ((zfsvfs)->z_unmounted) { \
232 232 ZFS_EXIT(zfsvfs); \
233 233 return (EIO); \
234 234 } \
235 235 }
236 236
237 237 /* Must be called before exiting the vop */
238 238 #define ZFS_EXIT(zfsvfs) rrm_exit(&(zfsvfs)->z_teardown_lock, FTAG)
239 239
|
↓ open down ↓ |
239 lines elided |
↑ open up ↑ |
240 240 /* Verifies the znode is valid */
241 241 #define ZFS_VERIFY_ZP(zp) \
242 242 if ((zp)->z_sa_hdl == NULL) { \
243 243 ZFS_EXIT((zp)->z_zfsvfs); \
244 244 return (EIO); \
245 245 } \
246 246
247 247 /*
248 248 * Macros for dealing with dmu_buf_hold
249 249 */
250 -#define ZFS_OBJ_HASH(obj_num) ((obj_num) & (ZFS_OBJ_MTX_SZ - 1))
250 +#define ZFS_OBJ_HASH(obj_num, table_sz) ((obj_num) & (table_sz - 1))
251 251 #define ZFS_OBJ_MUTEX(zfsvfs, obj_num) \
252 - (&(zfsvfs)->z_hold_mtx[ZFS_OBJ_HASH(obj_num)])
252 + (&(zfsvfs)->z_hold_mtx[ZFS_OBJ_HASH(obj_num, (zfsvfs)->z_hold_mtx_sz)])
253 253 #define ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num) \
254 254 mutex_enter(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
255 255 #define ZFS_OBJ_HOLD_TRYENTER(zfsvfs, obj_num) \
256 256 mutex_tryenter(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
257 257 #define ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num) \
258 258 mutex_exit(ZFS_OBJ_MUTEX((zfsvfs), (obj_num)))
259 259
260 260 /* Encode ZFS stored time values from a struct timespec */
261 261 #define ZFS_TIME_ENCODE(tp, stmp) \
262 262 { \
263 263 (stmp)[0] = (uint64_t)(tp)->tv_sec; \
264 264 (stmp)[1] = (uint64_t)(tp)->tv_nsec; \
265 265 }
266 266
267 267 /* Decode ZFS stored time values to a struct timespec */
268 268 #define ZFS_TIME_DECODE(tp, stmp) \
269 269 { \
270 270 (tp)->tv_sec = (time_t)(stmp)[0]; \
271 271 (tp)->tv_nsec = (long)(stmp)[1]; \
272 272 }
273 273
274 274 /*
275 275 * Timestamp defines
276 276 */
277 277 #define ACCESSED (AT_ATIME)
278 278 #define STATE_CHANGED (AT_CTIME)
279 279 #define CONTENT_MODIFIED (AT_MTIME | AT_CTIME)
280 280
281 281 #define ZFS_ACCESSTIME_STAMP(zfsvfs, zp) \
282 282 if ((zfsvfs)->z_atime && !((zfsvfs)->z_vfs->vfs_flag & VFS_RDONLY)) \
283 283 zfs_tstamp_update_setup(zp, ACCESSED, NULL, NULL, B_FALSE);
284 284
285 285 extern int zfs_init_fs(zfsvfs_t *, znode_t **);
286 286 extern void zfs_set_dataprop(objset_t *);
287 287 extern void zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *,
288 288 dmu_tx_t *tx);
289 289 extern void zfs_tstamp_update_setup(znode_t *, uint_t, uint64_t [2],
290 290 uint64_t [2], boolean_t);
291 291 extern void zfs_grow_blocksize(znode_t *, uint64_t, dmu_tx_t *);
292 292 extern int zfs_freesp(znode_t *, uint64_t, uint64_t, int, boolean_t);
293 293 extern void zfs_znode_init(void);
294 294 extern void zfs_znode_fini(void);
295 295 extern int zfs_zget(zfsvfs_t *, uint64_t, znode_t **);
296 296 extern int zfs_rezget(znode_t *);
297 297 extern void zfs_zinactive(znode_t *);
298 298 extern void zfs_znode_delete(znode_t *, dmu_tx_t *);
299 299 extern void zfs_znode_free(znode_t *);
300 300 extern void zfs_remove_op_tables();
301 301 extern int zfs_create_op_tables();
302 302 extern int zfs_sync(vfs_t *vfsp, short flag, cred_t *cr);
303 303 extern dev_t zfs_cmpldev(uint64_t);
304 304 extern int zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value);
305 305 extern int zfs_get_stats(objset_t *os, nvlist_t *nv);
306 306 extern boolean_t zfs_get_vfs_flag_unmounted(objset_t *os);
307 307 extern void zfs_znode_dmu_fini(znode_t *);
308 308
309 309 extern void zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
310 310 znode_t *dzp, znode_t *zp, char *name, vsecattr_t *, zfs_fuid_info_t *,
311 311 vattr_t *vap);
312 312 extern int zfs_log_create_txtype(zil_create_t, vsecattr_t *vsecp,
313 313 vattr_t *vap);
314 314 extern void zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
315 315 znode_t *dzp, char *name, uint64_t foid);
316 316 #define ZFS_NO_OBJECT 0 /* no object id */
317 317 extern void zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
318 318 znode_t *dzp, znode_t *zp, char *name);
319 319 extern void zfs_log_symlink(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
320 320 znode_t *dzp, znode_t *zp, char *name, char *link);
321 321 extern void zfs_log_rename(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
322 322 znode_t *sdzp, char *sname, znode_t *tdzp, char *dname, znode_t *szp);
323 323 extern void zfs_log_write(zilog_t *zilog, dmu_tx_t *tx, int txtype,
324 324 znode_t *zp, offset_t off, ssize_t len, int ioflag);
325 325 extern void zfs_log_truncate(zilog_t *zilog, dmu_tx_t *tx, int txtype,
326 326 znode_t *zp, uint64_t off, uint64_t len);
327 327 extern void zfs_log_setattr(zilog_t *zilog, dmu_tx_t *tx, int txtype,
328 328 znode_t *zp, vattr_t *vap, uint_t mask_applied, zfs_fuid_info_t *fuidp);
329 329 extern void zfs_log_acl(zilog_t *zilog, dmu_tx_t *tx, znode_t *zp,
330 330 vsecattr_t *vsecp, zfs_fuid_info_t *fuidp);
331 331 extern void zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx);
332 332 extern void zfs_upgrade(zfsvfs_t *zfsvfs, dmu_tx_t *tx);
333 333 extern int zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx);
334 334
335 335 extern caddr_t zfs_map_page(page_t *, enum seg_rw);
336 336 extern void zfs_unmap_page(page_t *, caddr_t);
337 337
338 338 extern zil_get_data_t zfs_get_data;
339 339 extern zil_replay_func_t *zfs_replay_vector[TX_MAX_TYPE];
340 340 extern int zfsfstype;
341 341
342 342 #endif /* _KERNEL */
343 343
344 344 extern int zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len);
345 345
346 346 #ifdef __cplusplus
347 347 }
348 348 #endif
349 349
350 350 #endif /* _SYS_FS_ZFS_ZNODE_H */
|
↓ open down ↓ |
88 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX