*
* You should have received a copy of the GNU General Public License
* version 2 along with this program; If not, see
- * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
- *
- * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
- * CA 95054 USA or visit www.sun.com if you need additional information or
- * have any questions.
+ * http://www.gnu.org/licenses/gpl-2.0.html
*
* GPL HEADER END
*/
* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2015, Intel Corporation.
+ * Copyright (c) 2011, 2016, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#define OSD_STATFS_RESERVED (1ULL << 23) /* 8MB */
#define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
+/* check if ldiskfs support project quota */
+#ifndef LDISKFS_IOC_FSSETXATTR
+#undef HAVE_PROJECT_QUOTA
+#endif
+
+#define OBD_BRW_MAPPED OBD_BRW_LOCAL1
+
struct osd_directory {
struct iam_container od_container;
struct iam_descr od_descr;
extern const int osd_dto_credits_noquota[];
struct osd_object {
- struct dt_object oo_dt;
- /**
- * Inode for file system object represented by this osd_object. This
- * inode is pinned for the whole duration of lu_object life.
- *
- * Not modified concurrently (either setup early during object
- * creation, or assigned by osd_object_create() under write lock).
- */
- struct inode *oo_inode;
- /**
- * to protect index ops.
- */
- struct htree_lock_head *oo_hl_head;
+ struct dt_object oo_dt;
+ /**
+ * Inode for file system object represented by this osd_object. This
+ * inode is pinned for the whole duration of lu_object life.
+ *
+ * Not modified concurrently (either setup early during object
+ * creation, or assigned by osd_create() under write lock).
+ */
+ struct inode *oo_inode;
+ /**
+ * to protect index ops.
+ */
+ struct htree_lock_head *oo_hl_head;
struct rw_semaphore oo_ext_idx_sem;
struct rw_semaphore oo_sem;
struct osd_directory *oo_dir;
struct osd_mdobj_map {
struct dentry *omm_remote_parent;
};
-int osd_ldiskfs_add_entry(struct osd_thread_info *info,
+int osd_ldiskfs_add_entry(struct osd_thread_info *info, struct osd_device *osd,
handle_t *handle, struct dentry *child,
struct inode *inode, struct htree_lock *hlock);
struct osd_otable_it {
struct osd_device *ooi_dev;
struct osd_otable_cache ooi_cache;
+ struct osd_iit_param ooi_iit_param;
/* The following bits can be updated/checked w/o lock protection.
* If more bits will be introduced in the future and need lock to
#define FULL_SCRUB_THRESHOLD_RATE_DEFAULT 60
-/* There are at most 10 uid/gids are affected in a transaction, and
+/* There are at most 15 uid/gid/projids are affected in a transaction, and
* that's rename case:
- * - 2 for source parent uid & gid;
- * - 2 for source child uid & gid ('..' entry update when child is directory);
- * - 2 for target parent uid & gid;
- * - 2 for target child uid & gid (if the target child exists);
- * - 2 for root uid & gid (last_rcvd, llog, etc);
+ * - 3 for source parent uid & gid & projid;
+ * - 3 for source child uid & gid & projid ('..' entry update when
+ * child is directory);
+ * - 3 for target parent uid & gid & projid;
+ * - 3 for target child uid & gid & projid(if the target child exists);
+ * - 3 for root uid & gid(last_rcvd, llog, etc);
*
- * The 0 to (OSD_MAX_UGID_CNT - 1) bits of ot_id_type is for indicating
- * the id type of each id in the ot_id_array.
*/
-#define OSD_MAX_UGID_CNT 10
+#define OSD_MAX_UGID_CNT 15
-enum {
+enum osd_op_type {
OSD_OT_ATTR_SET = 0,
OSD_OT_PUNCH = 1,
OSD_OT_XATTR_SET = 2,
struct list_head ot_stop_dcb_list;
/* Link to the device, for debugging. */
struct lu_ref_link ot_dev_link;
- unsigned short ot_credits;
- unsigned short ot_id_cnt;
- unsigned short ot_id_type;
+ unsigned int ot_credits;
+ unsigned short ot_id_cnt;
+ __u8 ot_id_types[OSD_MAX_UGID_CNT];
unsigned int ot_remove_agents:1;
- uid_t ot_id_array[OSD_MAX_UGID_CNT];
+ uid_t ot_id_array[OSD_MAX_UGID_CNT];
struct lquota_trans *ot_quota_trans;
#if OSD_THANDLE_STATS
/** time when this handle was allocated */
- cfs_time_t oth_alloced;
+ ktime_t oth_alloced;
/** time when this thanle was started */
- cfs_time_t oth_started;
+ ktime_t oth_started;
#endif
};
* there would be one ext3 readdir for every mdd readdir page.
*/
-#define OSD_IT_EA_BUFSIZE (PAGE_CACHE_SIZE + PAGE_CACHE_SIZE/4)
+#define OSD_IT_EA_BUFSIZE (PAGE_SIZE + PAGE_SIZE/4)
/**
* This is iterator's in-memory data structure in interoperability
struct list_head oiq_list;
};
-#define MAX_BLOCKS_PER_PAGE (PAGE_CACHE_SIZE / 512)
+#define MAX_BLOCKS_PER_PAGE (PAGE_SIZE / 512)
struct osd_iobuf {
wait_queue_head_t dr_wait;
*/
struct timespec oti_time;
- /** osd_device reference, initialized in osd_trans_start() and
- used in osd_trans_stop() */
- struct osd_device *oti_dev;
-
/**
* following ipd and it structures are used for osd_index_iam_lookup()
* these are defined separately as we might do index operation
struct lu_buf oti_big_buf;
/** used in osd_ea_fid_set() to set fid into common ea */
union {
- struct lustre_mdt_attrs oti_mdt_attrs;
- /* old LMA for compatibility */
- char oti_mdt_attrs_old[LMA_OLD_SIZE];
+ struct lustre_ost_attrs oti_ost_attrs;
struct filter_fid_old oti_ff;
struct filter_fid oti_ff_new;
};
* cases where a large number of credits are being allocated for
* single transaction. */
unsigned int oti_credits_before;
- unsigned short oti_declare_ops[OSD_OT_MAX];
- unsigned short oti_declare_ops_cred[OSD_OT_MAX];
- unsigned short oti_declare_ops_used[OSD_OT_MAX];
+ unsigned int oti_declare_ops[OSD_OT_MAX];
+ unsigned int oti_declare_ops_cred[OSD_OT_MAX];
+ unsigned int oti_declare_ops_used[OSD_OT_MAX];
};
extern int ldiskfs_pdo;
int osd_ea_fid_set(struct osd_thread_info *info, struct inode *inode,
const struct lu_fid *fid, __u32 compat, __u32 incompat);
int osd_get_lma(struct osd_thread_info *info, struct inode *inode,
- struct dentry *dentry, struct lustre_mdt_attrs *lma);
+ struct dentry *dentry, struct lustre_ost_attrs *loa);
void osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
struct osd_inode_id *id, const struct lu_fid *fid);
int osd_get_idif(struct osd_thread_info *info, struct inode *inode,
struct lquota_id_info *qi, struct osd_object *obj,
bool enforce, int *flags);
int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
- long long space, struct osd_thandle *oh,
- struct osd_object *obj, bool is_blk, int *flags,
- bool force);
+ __u32 projid, long long space, struct osd_thandle *oh,
+ struct osd_object *obj, int *flags,
+ enum osd_qid_declare_flags);
const struct dt_rec *osd_quota_pack(struct osd_object *obj,
const struct dt_rec *rec,
union lquota_rec *quota_rec);
void osd_quota_unpack(struct osd_object *obj, const struct dt_rec *rec);
-int osd_quota_migration(const struct lu_env *env, struct dt_object *dt);
#ifndef HAVE_I_UID_READ
static inline uid_t i_uid_read(const struct inode *inode)
}
#endif
+#ifdef HAVE_PROJECT_QUOTA
+static inline __u32 i_projid_read(struct inode *inode)
+{
+ return (__u32)from_kprojid(&init_user_ns, LDISKFS_I(inode)->i_projid);
+}
+
+static inline void i_projid_write(struct inode *inode, __u32 projid)
+{
+ kprojid_t kprojid;
+ kprojid = make_kprojid(&init_user_ns, (projid_t)projid);
+ LDISKFS_I(inode)->i_projid = kprojid;
+}
+#else
+static inline uid_t i_projid_read(struct inode *inode)
+{
+ return 0;
+}
+static inline void i_projid_write(struct inode *inode, __u32 projid)
+{
+ return;
+}
+#endif
+
+#ifdef HAVE_LDISKFS_INFO_JINODE
+# define osd_attach_jinode(inode) ldiskfs_inode_attach_jinode(inode)
+#else /* HAVE_LDISKFS_INFO_JINODE */
+# define osd_attach_jinode(inode) 0
+#endif /* HAVE_LDISKFS_INFO_JINODE */
+
#ifdef LDISKFS_HT_MISC
# define osd_journal_start_sb(sb, type, nblock) \
ldiskfs_journal_start_sb(sb, type, nblock)
-# define osd_ldiskfs_append(handle, inode, nblock) \
- ldiskfs_append(handle, inode, nblock)
+static inline struct buffer_head *osd_ldiskfs_append(handle_t *handle,
+ struct inode *inode,
+ ldiskfs_lblk_t *nblock)
+{
+ int rc;
+
+ rc = osd_attach_jinode(inode);
+ if (rc)
+ return ERR_PTR(rc);
+ return ldiskfs_append(handle, inode, nblock);
+}
# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
(__ldiskfs_find_entry(dir, name, de, inlined, lock) ?: \
ERR_PTR(-ENOENT))
return dev->od_mnt->mnt_sb;
}
+static inline const char *osd_dev2name(const struct osd_device *dev)
+{
+ return osd_sb(dev)->s_id;
+}
+
+static inline const char *osd_ino2name(const struct inode *inode)
+{
+ return inode->i_sb->s_id;
+}
+
+/**
+ * Put the osd object once done with it.
+ *
+ * \param obj osd object that needs to be put
+ */
+static inline void osd_object_put(const struct lu_env *env,
+ struct osd_object *obj)
+{
+ dt_object_put(env, &obj->oo_dt);
+}
+
static inline int osd_object_is_root(const struct osd_object *obj)
{
return osd_sb(osd_obj2dev(obj))->s_root->d_inode == obj->oo_inode;
static inline char *osd_name(struct osd_device *osd)
{
- return osd->od_dt_dev.dd_lu_dev.ld_obd->obd_name;
+ return osd->od_svname;
}
static inline bool osd_is_ea_inode(struct inode *inode)
static inline void osd_trans_declare_op(const struct lu_env *env,
struct osd_thandle *oh,
- unsigned int op, int credits)
+ enum osd_op_type op, int credits)
{
struct osd_thread_info *oti = osd_oti_get(env);
}
static inline void osd_trans_exec_op(const struct lu_env *env,
- struct thandle *th, unsigned int op)
+ struct thandle *th,
+ enum osd_op_type op)
{
struct osd_thread_info *oti = osd_oti_get(env);
struct osd_thandle *oh = container_of(th, struct osd_thandle,
if (unlikely(ldiskfs_track_declares_assert))
LASSERT(op < OSD_OT_MAX);
else {
- CWARN("%s: Invalid operation index %d\n",
- osd_name(osd_dt_dev(oh->ot_super.th_dev)), op);
+ CWARN("%s: opcode %u: invalid value >= %u\n",
+ osd_name(osd_dt_dev(oh->ot_super.th_dev)),
+ op, OSD_OT_MAX);
libcfs_debug_dumpstack(NULL);
return;
}
if (op == OSD_OT_REF_ADD &&
oti->oti_declare_ops_cred[OSD_OT_DESTROY] > 0)
goto proceed;
+ CWARN("%s: opcode %u: credits = 0, rollback = %u\n",
+ osd_name(osd_dt_dev(oh->ot_super.th_dev)), op, rb);
osd_trans_dump_creds(env, th);
- CERROR("%s: op = %d, rb = %d\n",
- osd_name(osd_dt_dev(oh->ot_super.th_dev)), op, rb);
- if (unlikely(ldiskfs_track_declares_assert))
- LBUG();
+ LASSERT(!ldiskfs_track_declares_assert);
}
proceed:
oti->oti_credits_before = oh->ot_handle->h_buffer_credits;
left = oti->oti_declare_ops_cred[op] - oti->oti_declare_ops_used[op];
if (unlikely(oti->oti_credits_before < left)) {
+ CWARN("%s: opcode %u: before %u < left %u, rollback = %u\n",
+ osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
+ oti->oti_credits_before, left, rb);
osd_trans_dump_creds(env, th);
- CERROR("%s: op = %d, rb = %d\n",
- osd_name(osd_dt_dev(oh->ot_super.th_dev)), op, rb);
/* on a very small fs (testing?) it's possible that
* the transaction can't fit 1/4 of journal, so we
* just request less credits (see osd_trans_start()).
* ignore the same case here */
rb = osd_transaction_size(osd_dt_dev(th->th_dev));
- if (unlikely(oh->ot_credits < rb)) {
- if (unlikely(ldiskfs_track_declares_assert))
- LBUG();
- }
+ if (unlikely(oh->ot_credits < rb))
+ LASSERT(!ldiskfs_track_declares_assert);
}
}
static inline void osd_trans_exec_check(const struct lu_env *env,
struct thandle *th,
- unsigned int op)
+ enum osd_op_type op)
{
struct osd_thread_info *oti = osd_oti_get(env);
struct osd_thandle *oh = container_of(th, struct osd_thandle,
oti->oti_declare_ops_used[OSD_OT_QUOTA] += over;
oti->oti_declare_ops_used[op] -= over;
} else {
- CWARN("op %d: used %u, used now %u, reserved %u\n",
- op, oti->oti_declare_ops_used[op], used,
+ CWARN("%s: opcode %d: used %u, used now %u, reserved %u\n",
+ osd_name(osd_dt_dev(oh->ot_super.th_dev)), op,
+ oti->oti_declare_ops_used[op], used,
oti->oti_declare_ops_cred[op]);
osd_trans_dump_creds(env, th);
if (unlikely(ldiskfs_track_declares_assert))
static inline int is_32bit_api(void)
{
#ifdef CONFIG_COMPAT
- return is_compat_task();
+ return in_compat_syscall();
#else
return (BITS_PER_LONG == 32);
#endif
ldiskfs_lblk_t block,
int create)
{
-#ifdef HAVE_EXT4_BREAD_4ARGS
- return ldiskfs_bread(handle, inode, block, create);
-#else
+ int rc = 0;
struct buffer_head *bh;
- int error = 0;
- bh = ldiskfs_bread(handle, inode, block, create, &error);
- if (bh == NULL && error != 0)
- bh = ERR_PTR(error);
+ if (create) {
+ rc = osd_attach_jinode(inode);
+ if (rc)
+ return ERR_PTR(rc);
+ }
+#ifdef HAVE_EXT4_BREAD_4ARGS
+ bh = ldiskfs_bread(handle, inode, block, create);
+#else
- return bh;
+ bh = ldiskfs_bread(handle, inode, block, create, &rc);
+ if (bh == NULL && rc != 0)
+ bh = ERR_PTR(rc);
#endif
+ return bh;
}
void ldiskfs_inc_count(handle_t *handle, struct inode *inode);
void osd_fini_iobuf(struct osd_device *d, struct osd_iobuf *iobuf);
+
#endif /* _OSD_INTERNAL_H */
git://git.whamcloud.com / fs / lustre-release.git / blobdiff