4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful, but
11 * WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * General Public License version 2 for more details (a copy is included
14 * in the LICENSE file that accompanied this code).
16 * You should have received a copy of the GNU General Public License
17 * version 2 along with this program; if not, write to the
18 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19 * Boston, MA 021110-1307, USA
24 * Copyright (c) 2012, 2015, Intel Corporation.
25 * Use is subject to license terms.
27 * Author: Johann Lombardi <johann@whamcloud.com>
30 #include <lustre_quota.h>
32 #include "osd_internal.h"
35 * Helper function to retrieve DMU object id from fid for accounting object
37 uint64_t osd_quota_fid2dmu(const struct lu_fid *fid)
39 LASSERT(fid_is_acct(fid));
40 if (fid_oid(fid) == ACCT_GROUP_OID)
41 return DMU_GROUPUSED_OBJECT;
42 return DMU_USERUSED_OBJECT;
46 * Helper function to estimate the number of inodes in use for a give uid/gid
47 * from the block usage
49 static uint64_t osd_objset_user_iused(struct osd_device *osd, uint64_t uidbytes)
51 uint64_t refdbytes, availbytes, usedobjs, availobjs;
52 uint64_t uidobjs, bshift;
54 /* get fresh statfs info */
55 dmu_objset_space(osd->od_os, &refdbytes, &availbytes,
56 &usedobjs, &availobjs);
58 /* estimate the number of objects based on the disk usage */
59 bshift = fls64(osd->od_max_blksz) - 1;
60 uidobjs = osd_objs_count_estimate(refdbytes, usedobjs,
61 uidbytes >> bshift, bshift);
63 /* if we have at least 1 byte, we have at least one dnode ... */
64 uidobjs = max_t(uint64_t, uidobjs, 1);
70 * Space Accounting Management
74 * Return space usage consumed by a given uid or gid.
75 * Block usage is accurrate since it is maintained by DMU itself.
76 * However, DMU does not provide inode accounting, so the #inodes in use
77 * is estimated from the block usage and statfs information.
79 * \param env - is the environment passed by the caller
80 * \param dtobj - is the accounting object
81 * \param dtrec - is the record to fill with space usage information
82 * \param dtkey - is the id the of the user or group for which we would
83 * like to access disk usage.
85 * \retval +ve - success : exact match
86 * \retval -ve - failure
88 static int osd_acct_index_lookup(const struct lu_env *env,
89 struct dt_object *dtobj,
91 const struct dt_key *dtkey)
93 struct osd_thread_info *info = osd_oti_get(env);
94 char *buf = info->oti_buf;
95 size_t buflen = sizeof(info->oti_buf);
96 struct lquota_acct_rec *rec = (struct lquota_acct_rec *)dtrec;
97 struct osd_object *obj = osd_dt_obj(dtobj);
98 struct osd_device *osd = osd_obj2dev(obj);
103 rec->bspace = rec->ispace = 0;
105 /* convert the 64-bit uid/gid into a string */
106 snprintf(buf, buflen, "%llx", *((__u64 *)dtkey));
107 /* fetch DMU object ID (DMU_USERUSED_OBJECT/DMU_GROUPUSED_OBJECT) to be
109 oid = osd_quota_fid2dmu(lu_object_fid(&dtobj->do_lu));
111 /* disk usage (in bytes) is maintained by DMU.
112 * DMU_USERUSED_OBJECT/DMU_GROUPUSED_OBJECT are special objects which
113 * not associated with any dmu_but_t (see dnode_special_open()). */
114 rc = zap_lookup(osd->od_os, oid, buf, sizeof(uint64_t), 1,
117 /* user/group has not created anything yet */
118 CDEBUG(D_QUOTA, "%s: id %s not found in DMU accounting ZAP\n",
119 osd->od_svname, buf);
124 if (!osd_dmu_userobj_accounting_available(osd)) {
125 if (rec->bspace != 0)
126 /* estimate #inodes in use */
127 rec->ispace = osd_objset_user_iused(osd, rec->bspace);
130 snprintf(buf, buflen, OSD_DMU_USEROBJ_PREFIX "%llx",
132 rc = zap_lookup(osd->od_os, oid, buf, sizeof(uint64_t), 1,
136 "%s: id %s not found dnode accounting\n",
137 osd->od_svname, buf);
138 } else if (rc == 0) {
147 * Initialize osd Iterator for given osd index object.
149 * \param dt - osd index object
150 * \param attr - not used
152 static struct dt_it *osd_it_acct_init(const struct lu_env *env,
153 struct dt_object *dt,
156 struct osd_thread_info *info = osd_oti_get(env);
157 struct osd_it_quota *it;
158 struct lu_object *lo = &dt->do_lu;
159 struct osd_device *osd = osd_dev(lo->lo_dev);
163 LASSERT(lu_object_exists(lo));
166 RETURN(ERR_PTR(-ENOMEM));
170 RETURN(ERR_PTR(-ENOMEM));
172 memset(it, 0, sizeof(*it));
173 it->oiq_oid = osd_quota_fid2dmu(lu_object_fid(lo));
175 /* initialize zap cursor */
176 rc = osd_zap_cursor_init(&it->oiq_zc, osd->od_os, it->oiq_oid, 0);
182 /* take object reference */
184 it->oiq_obj = osd_dt_obj(dt);
187 RETURN((struct dt_it *)it);
191 * Free given iterator.
193 * \param di - osd iterator
195 static void osd_it_acct_fini(const struct lu_env *env, struct dt_it *di)
197 struct osd_it_quota *it = (struct osd_it_quota *)di;
200 osd_zap_cursor_fini(it->oiq_zc);
201 osd_object_put(env, it->oiq_obj);
208 * Move on to the next valid entry.
210 * \param di - osd iterator
212 * \retval +ve - iterator reached the end
213 * \retval 0 - iterator has not reached the end yet
214 * \retval -ve - unexpected failure
216 static int osd_it_acct_next(const struct lu_env *env, struct dt_it *di)
218 struct osd_it_quota *it = (struct osd_it_quota *)di;
219 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
223 if (it->oiq_reset == 0)
224 zap_cursor_advance(it->oiq_zc);
226 rc = -zap_cursor_retrieve(it->oiq_zc, za);
227 if (rc == -ENOENT) /* reached the end */
233 * Return pointer to the key under iterator.
235 * \param di - osd iterator
237 static struct dt_key *osd_it_acct_key(const struct lu_env *env,
238 const struct dt_it *di)
240 struct osd_it_quota *it = (struct osd_it_quota *)di;
241 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
246 rc = -zap_cursor_retrieve(it->oiq_zc, za);
249 rc = kstrtoull(za->za_name, 16, &it->oiq_id);
251 RETURN((struct dt_key *) &it->oiq_id);
255 * Return size of key under iterator (in bytes)
257 * \param di - osd iterator
259 static int osd_it_acct_key_size(const struct lu_env *env,
260 const struct dt_it *di)
263 RETURN((int)sizeof(uint64_t));
267 * zap_cursor_retrieve read from current record.
268 * to read bytes we need to call zap_lookup explicitly.
270 static int osd_zap_cursor_retrieve_value(const struct lu_env *env,
271 zap_cursor_t *zc, char *buf,
272 int buf_size, int *bytes_read)
274 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
277 rc = -zap_cursor_retrieve(zc, za);
278 if (unlikely(rc != 0))
281 if (unlikely(za->za_integer_length <= 0))
284 actual_size = za->za_integer_length * za->za_num_integers;
286 if (actual_size > buf_size) {
287 actual_size = buf_size;
288 buf_size = actual_size / za->za_integer_length;
290 buf_size = za->za_num_integers;
293 rc = -zap_lookup(zc->zc_objset, zc->zc_zapobj,
294 za->za_name, za->za_integer_length,
298 *bytes_read = actual_size;
304 * Return pointer to the record under iterator.
306 * \param di - osd iterator
307 * \param attr - not used
309 static int osd_it_acct_rec(const struct lu_env *env,
310 const struct dt_it *di,
311 struct dt_rec *dtrec, __u32 attr)
313 struct osd_thread_info *info = osd_oti_get(env);
314 zap_attribute_t *za = &info->oti_za;
315 struct osd_it_quota *it = (struct osd_it_quota *)di;
316 struct lquota_acct_rec *rec = (struct lquota_acct_rec *)dtrec;
317 struct osd_object *obj = it->oiq_obj;
318 struct osd_device *osd = osd_obj2dev(obj);
324 rec->ispace = rec->bspace = 0;
326 /* retrieve block usage from the DMU accounting object */
327 rc = osd_zap_cursor_retrieve_value(env, it->oiq_zc,
328 (char *)&rec->bspace,
329 sizeof(uint64_t), &bytes_read);
333 if (!osd_dmu_userobj_accounting_available(osd)) {
334 if (rec->bspace != 0)
335 /* estimate #inodes in use */
336 rec->ispace = osd_objset_user_iused(osd, rec->bspace);
340 /* retrieve key associated with the current cursor */
341 rc = -zap_cursor_retrieve(it->oiq_zc, za);
342 if (unlikely(rc != 0))
345 /* inode accounting is not maintained by DMU, so we use our own ZAP to
346 * track inode usage */
347 rc = -zap_lookup(osd->od_os, it->oiq_obj->oo_dn->dn_object,
348 za->za_name, sizeof(uint64_t), 1, &rec->ispace);
350 /* user/group has not created any file yet */
351 CDEBUG(D_QUOTA, "%s: id %s not found in accounting ZAP\n",
352 osd->od_svname, za->za_name);
360 * Returns cookie for current Iterator position.
362 * \param di - osd iterator
364 static __u64 osd_it_acct_store(const struct lu_env *env,
365 const struct dt_it *di)
367 struct osd_it_quota *it = (struct osd_it_quota *)di;
370 RETURN(osd_zap_cursor_serialize(it->oiq_zc));
374 * Restore iterator from cookie. if the \a hash isn't found,
375 * restore the first valid record.
377 * \param di - osd iterator
378 * \param hash - iterator location cookie
380 * \retval +ve - di points to exact matched key
381 * \retval 0 - di points to the first valid record
382 * \retval -ve - failure
384 static int osd_it_acct_load(const struct lu_env *env,
385 const struct dt_it *di, __u64 hash)
387 struct osd_it_quota *it = (struct osd_it_quota *)di;
388 struct osd_device *osd = osd_obj2dev(it->oiq_obj);
389 zap_attribute_t *za = &osd_oti_get(env)->oti_za;
394 /* create new cursor pointing to the new hash */
395 rc = osd_zap_cursor_init(&zc, osd->od_os, it->oiq_oid, hash);
398 osd_zap_cursor_fini(it->oiq_zc);
402 rc = -zap_cursor_retrieve(it->oiq_zc, za);
405 else if (rc == -ENOENT)
412 * Move Iterator to record specified by \a key, if the \a key isn't found,
413 * move to the first valid record.
415 * \param di - osd iterator
416 * \param key - uid or gid
418 * \retval +ve - di points to exact matched key
419 * \retval 0 - di points to the first valid record
420 * \retval -ve - failure
422 static int osd_it_acct_get(const struct lu_env *env, struct dt_it *di,
423 const struct dt_key *key)
427 /* XXX: like osd_zap_it_get(), API is currently broken */
428 LASSERT(*((__u64 *)key) == 0);
430 RETURN(osd_it_acct_load(env, di, 0));
436 * \param di - osd iterator
438 static void osd_it_acct_put(const struct lu_env *env, struct dt_it *di)
443 * Index and Iterator operations for accounting objects
445 const struct dt_index_operations osd_acct_index_ops = {
446 .dio_lookup = osd_acct_index_lookup,
448 .init = osd_it_acct_init,
449 .fini = osd_it_acct_fini,
450 .get = osd_it_acct_get,
451 .put = osd_it_acct_put,
452 .next = osd_it_acct_next,
453 .key = osd_it_acct_key,
454 .key_size = osd_it_acct_key_size,
455 .rec = osd_it_acct_rec,
456 .store = osd_it_acct_store,
457 .load = osd_it_acct_load
462 * Quota Enforcement Management
466 * Wrapper for qsd_op_begin().
468 * \param env - the environment passed by the caller
469 * \param osd - is the osd_device
470 * \param uid - user id of the inode
471 * \param gid - group id of the inode
472 * \param space - how many blocks/inodes will be consumed/released
473 * \param oh - osd transaction handle
474 * \param is_blk - block quota or inode quota?
475 * \param flags - if the operation is write, return no user quota, no
476 * group quota, or sync commit flags to the caller
477 * \param force - set to 1 when changes are performed by root user and thus
478 * can't failed with EDQUOT
480 * \retval 0 - success
481 * \retval -ve - failure
483 int osd_declare_quota(const struct lu_env *env, struct osd_device *osd,
484 qid_t uid, qid_t gid, long long space,
485 struct osd_thandle *oh, bool is_blk, int *flags,
488 struct osd_thread_info *info = osd_oti_get(env);
489 struct lquota_id_info *qi = &info->oti_qi;
490 struct qsd_instance *qsd = osd->od_quota_slave;
491 int rcu, rcg; /* user & group rc */
494 if (unlikely(qsd == NULL))
495 /* quota slave instance hasn't been allocated yet */
498 /* let's start with user quota */
499 qi->lqi_id.qid_uid = uid;
500 qi->lqi_type = USRQUOTA;
501 qi->lqi_space = space;
502 qi->lqi_is_blk = is_blk;
503 rcu = qsd_op_begin(env, qsd, &oh->ot_quota_trans, qi, flags);
505 if (force && (rcu == -EDQUOT || rcu == -EINPROGRESS))
506 /* ignore EDQUOT & EINPROGRESS when changes are done by root */
509 /* For non-fatal error, we want to continue to get the noquota flags
510 * for group id. This is only for commit write, which has @flags passed
511 * in. See osd_declare_write_commit().
512 * When force is set to true, we also want to proceed with the gid */
513 if (rcu && (rcu != -EDQUOT || flags == NULL))
516 /* and now group quota */
517 qi->lqi_id.qid_gid = gid;
518 qi->lqi_type = GRPQUOTA;
519 rcg = qsd_op_begin(env, qsd, &oh->ot_quota_trans, qi, flags);
521 if (force && (rcg == -EDQUOT || rcg == -EINPROGRESS))
522 /* as before, ignore EDQUOT & EINPROGRESS for root */
525 RETURN(rcu ? rcu : rcg);