/* LUSTRE_OSD_NAME */
#include <obd.h>
-/* class_register_type(), class_unregister_type(), class_get_type() */
+/* class_register_type(), class_unregister_type() */
#include <obd_class.h>
#include <lustre_disk.h>
#include <dt_object.h>
#define OSD_STATFS_RESERVED (1ULL << 23) /* 8MB */
#define OSD_STATFS_RESERVED_SHIFT (7) /* reserve 0.78% of all space */
+/* Default extent bytes when declaring write commit */
+#define OSD_DEFAULT_EXTENT_BYTES (1U << 20)
+
/* check if ldiskfs support project quota */
#ifndef LDISKFS_IOC_FSSETXATTR
#undef HAVE_PROJECT_QUOTA
struct list_head oo_xattr_list;
struct lu_object_header *oo_header;
+ __u64 oo_dirent_count;
};
struct osd_obj_seq {
od_check_ff:1,
od_is_ost:1,
od_in_init:1,
- od_index_in_idif:1;
+ od_index_in_idif:1,
+ /* Other flags */
+ od_read_cache:1,
+ od_writethrough_cache:1,
+ od_nonrotational:1;
+
__s64 od_auto_scrub_interval;
__u32 od_dirent_journal;
spinlock_t od_osfs_lock;
+ int od_fallocate_zero_blocks;
int od_connects;
struct lu_site od_site;
struct osd_obj_map *od_ost_map;
struct osd_mdobj_map *od_mdt_map;
+ /* objects with size > od_readcache_max_filesize will be
+ * served bypassing pagecache unless already cached */
unsigned long long od_readcache_max_filesize;
- int od_read_cache;
- int od_writethrough_cache;
+
+ /* reads > od_readcache_max_iosize will be
+ * served bypassing pagecache unless already cached */
+ unsigned long od_readcache_max_iosize;
+
+ /* writes > od_writethough_max_iosize will be
+ * served bypassing pagecache unless already cached */
+ unsigned long od_writethrough_max_iosize;
struct brw_stats od_brw_stats;
atomic_t od_r_in_flight;
/* service name associated with the osd device */
char od_svname[MAX_OBD_NAME];
char od_mntdev[MAX_OBD_NAME];
+ uuid_t od_uuid;
+
+ /* quota slave instance for inode */
+ struct qsd_instance *od_quota_slave_md;
- /* quota slave instance */
- struct qsd_instance *od_quota_slave;
+ /* quota slave instance for block */
+ struct qsd_instance *od_quota_slave_dt;
/* osd seq instance */
struct lu_client_seq *od_cl_seq;
int od_index_backup_stop;
/* T10PI type, zero if not supported */
enum osd_t10_type od_t10_type;
+ atomic_t od_commit_cb_in_flight;
+ wait_queue_head_t od_commit_cb_done;
+ unsigned int __percpu *od_extent_bytes_percpu;
};
+static inline struct qsd_instance *osd_def_qsd(struct osd_device *osd)
+{
+ if (osd->od_is_ost)
+ return osd->od_quota_slave_dt;
+ else
+ return osd->od_quota_slave_md;
+}
+
enum osd_full_scrub_ratio {
/* Trigger OI scrub to scan the whole device directly. */
OFSR_DIRECTLY = 0,
enum osd_op_type {
OSD_OT_ATTR_SET = 0,
- OSD_OT_PUNCH = 1,
- OSD_OT_XATTR_SET = 2,
- OSD_OT_CREATE = 3,
- OSD_OT_DESTROY = 4,
- OSD_OT_REF_ADD = 5,
- OSD_OT_REF_DEL = 6,
- OSD_OT_WRITE = 7,
- OSD_OT_INSERT = 8,
- OSD_OT_DELETE = 9,
- OSD_OT_QUOTA = 10,
- OSD_OT_MAX = 11
+ OSD_OT_PUNCH,
+ OSD_OT_XATTR_SET,
+ OSD_OT_CREATE,
+ OSD_OT_DESTROY,
+ OSD_OT_REF_ADD,
+ OSD_OT_REF_DEL,
+ OSD_OT_WRITE,
+ OSD_OT_INSERT,
+ OSD_OT_DELETE,
+ OSD_OT_QUOTA,
+ OSD_OT_PREALLOC,
+ OSD_OT_MAX
};
struct osd_access_lock {
/* Link to the device, for debugging. */
struct lu_ref_link ot_dev_link;
unsigned int ot_credits;
+ unsigned int oh_declared_ext;
+
+ /* quota IDs related to the transaction */
unsigned short ot_id_cnt;
- __u8 ot_id_types[OSD_MAX_UGID_CNT];
- unsigned int ot_remove_agents:1;
+ __u8 ot_id_res[OSD_MAX_UGID_CNT];
+ __u8 ot_id_types[OSD_MAX_UGID_CNT];
uid_t ot_id_array[OSD_MAX_UGID_CNT];
struct lquota_trans *ot_quota_trans;
+
+ unsigned int ot_remove_agents:1;
#if OSD_THANDLE_STATS
/** time when this handle was allocated */
ktime_t oth_alloced;
unsigned int dr_init_at; /* the line iobuf was initialized */
};
+int osd_security_file_alloc(struct file *file);
+
+#define osd_dirty_inode(inode, flag) (inode)->i_sb->s_op->dirty_inode((inode), flag)
+
+#ifdef HAVE_INODE_TIMESPEC64
+# define osd_timespec timespec64
+#else
+# define osd_timespec timespec
+#endif
+
+static inline struct osd_timespec osd_inode_time(struct inode *inode,
+ s64 seconds)
+{
+ struct osd_timespec ts = { .tv_sec = seconds };
+
+ return ts;
+}
+
#define OSD_INS_CACHE_SIZE 8
struct osd_thread_info {
struct osd_inode_id oti_id3;
struct ost_id oti_ostid;
- /*
- * XXX temporary: for ->i_op calls.
- */
- struct timespec oti_time;
-
/**
* following ipd and it structures are used for osd_index_iam_lookup()
* these are defined separately as we might do index operation
struct osd_idmap_cache oti_cache;
/* dedicated OI cache for insert (which needs inum) */
- struct osd_idmap_cache *oti_ins_cache;
- int oti_ins_cache_size;
- int oti_ins_cache_used;
+ struct osd_idmap_cache *oti_ins_cache;
+ int oti_ins_cache_size;
+ int oti_ins_cache_used;
/* inc by osd_trans_create and dec by osd_trans_stop */
- int oti_ins_cache_depth;
-
- int oti_r_locks;
- int oti_w_locks;
- int oti_txns;
- /** used in osd_fid_set() to put xattr */
- struct lu_buf oti_buf;
- struct lu_buf oti_big_buf;
- /** used in osd_ea_fid_set() to set fid into common ea */
+ int oti_ins_cache_depth;
+
+ int oti_r_locks;
+ int oti_w_locks;
+ int oti_txns;
+ /** used in osd_fid_set() to put xattr */
+ struct lu_buf oti_buf;
+ struct lu_buf oti_big_buf;
+ /** used in osd_ea_fid_set() to set fid into common ea */
union {
struct lustre_ost_attrs oti_ost_attrs;
- struct filter_fid_old oti_ff;
- struct filter_fid oti_ff_new;
+ struct filter_fid_18_23 oti_ff_old;
+ struct filter_fid oti_ff;
};
/** 0-copy IO */
- struct osd_iobuf oti_iobuf;
+ struct osd_iobuf oti_iobuf;
/* used to access objects in /O */
- struct inode *oti_inode;
+ struct inode *oti_inode;
#define OSD_FID_REC_SZ 32
- char oti_ldp[OSD_FID_REC_SZ];
- char oti_ldp2[OSD_FID_REC_SZ];
+ char oti_ldp[OSD_FID_REC_SZ];
+ char oti_ldp2[OSD_FID_REC_SZ];
/* used by quota code */
union {
#if defined(HAVE_DQUOT_QC_DQBLK)
struct qc_dqblk oti_qdq;
-#elif defined(HAVE_DQUOT_FS_DISK_QUOTA)
- struct fs_disk_quota oti_fdq;
#else
- struct if_dqblk oti_dqblk;
+ struct fs_disk_quota oti_fdq;
#endif
struct if_dqinfo oti_dqinfo;
};
extern int ldiskfs_pdo;
-#ifndef HAVE_VFS_SETXATTR
-#define osd_setxattr(dentry, inode, name, buf, len, flag) \
- ((inode)->i_op->setxattr(dentry, name, buf, len, flag))
-#define osd_getxattr(dentry, inode, name, buf, len) \
- ((inode)->i_op->getxattr(dentry, name, buf, len))
-#define osd_removexattr(dentry, inode, name) \
- ((inode)->i_op->removexattr(dentry, name))
-#else /* HAVE_VFS_SETXATTR */
-#define osd_setxattr(dentry, inode, name, buf, len, flag) \
- __vfs_setxattr(dentry, inode, name, buf, len, flag)
-#define osd_getxattr(dentry, inode, name, buf, len) \
- __vfs_getxattr(dentry, inode, name, buf, len)
-#define osd_removexattr(dentry, inode, name) \
- __vfs_removexattr(dentry, name)
-#endif /* !HAVE_VFS_SETXATTR */
+/* autoconf test is in lustre-build-ldiskfs.m4 */
+#ifdef HAVE_BVEC_ITER_ALL
+#define DECLARE_BVEC_ITER_ALL(iter) struct bvec_iter_all iter
+#else
+#define DECLARE_BVEC_ITER_ALL(iter) int iter
+#endif
static inline int __osd_xattr_get(struct inode *inode, struct dentry *dentry,
const char *name, void *buf, int len)
dentry->d_inode = inode;
dentry->d_sb = inode->i_sb;
- return osd_getxattr(dentry, inode, name, buf, len);
+ return ll_vfs_getxattr(dentry, inode, name, buf, len);
}
static inline int __osd_xattr_set(struct osd_thread_info *info,
{
struct dentry *dentry = &info->oti_child_dentry;
- ll_vfs_dq_init(inode);
+ dquot_initialize(inode);
dentry->d_inode = inode;
dentry->d_sb = inode->i_sb;
- return osd_setxattr(dentry, inode, name, buf, buflen, fl);
+ return ll_vfs_setxattr(dentry, inode, name, buf, buflen, fl);
}
#ifdef CONFIG_PROC_FS
/* osd_lproc.c */
extern struct lprocfs_vars lprocfs_osd_obd_vars[];
-extern struct lprocfs_vars lprocfs_osd_module_vars[];
int osd_procfs_init(struct osd_device *osd, const char *name);
int osd_procfs_fini(struct osd_device *osd);
void osd_brw_stats_update(struct osd_device *osd, struct osd_iobuf *iobuf);
#endif
int osd_statfs(const struct lu_env *env, struct dt_device *dev,
- struct obd_statfs *sfs);
+ struct obd_statfs *sfs, struct obd_statfs_info *info);
struct inode *osd_iget(struct osd_thread_info *info, struct osd_device *dev,
struct osd_inode_id *id);
struct inode *
struct inode *src_parent, struct dentry *src_child,
const struct lu_fid *fid);
int osd_obj_spec_lookup(struct osd_thread_info *info, struct osd_device *osd,
- const struct lu_fid *fid, struct osd_inode_id *id);
+ const struct lu_fid *fid, struct osd_inode_id *id,
+ enum oi_check_flags flags);
int osd_obj_spec_insert(struct osd_thread_info *info, struct osd_device *osd,
const struct lu_fid *fid, const struct osd_inode_id *id,
handle_t *th);
struct osd_inode_id *id);
void osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
+struct dentry *osd_lookup_one_len_unlocked(struct osd_device *dev,
+ const char *name,
+ struct dentry *base, int len);
+struct dentry *osd_lookup_one_len(struct osd_device *dev, const char *name,
+ struct dentry *base, int len);
+
int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
u64 seq, struct lu_seq_range *range);
/* osd_quota.c */
int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
struct lquota_id_info *qi, struct osd_object *obj,
- bool enforce, int *flags);
+ bool enforce, enum osd_quota_local_flags *local_flags);
int osd_declare_inode_qid(const struct lu_env *env, qid_t uid, qid_t gid,
__u32 projid, long long space, struct osd_thandle *oh,
- struct osd_object *obj, int *flags,
+ struct osd_object *obj,
+ enum osd_quota_local_flags *local_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);
-#ifndef HAVE_I_UID_READ
-static inline uid_t i_uid_read(const struct inode *inode)
-{
- return inode->i_uid;
-}
-
-static inline gid_t i_gid_read(const struct inode *inode)
-{
- return inode->i_gid;
-}
-
-static inline void i_uid_write(struct inode *inode, uid_t uid)
-{
- inode->i_uid = uid;
-}
-
-static inline void i_gid_write(struct inode *inode, gid_t gid)
-{
- inode->i_gid = gid;
-}
-#endif
-
#ifdef HAVE_PROJECT_QUOTA
static inline __u32 i_projid_read(struct inode *inode)
{
}
#endif
+#ifdef HAVE_LDISKFS_IGET_WITH_FLAGS
+# define osd_ldiskfs_iget(sb, ino) \
+ ldiskfs_iget((sb), (ino), \
+ LDISKFS_IGET_HANDLE | LDISKFS_IGET_SPECIAL)
+#else
+# define osd_ldiskfs_iget(sb, ino) ldiskfs_iget((sb), (ino))
+#endif
+
#ifdef HAVE_LDISKFS_INFO_JINODE
# define osd_attach_jinode(inode) ldiskfs_inode_attach_jinode(inode)
#else /* HAVE_LDISKFS_INFO_JINODE */
return ERR_PTR(rc);
return ldiskfs_append(handle, inode, nblock);
}
-# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
+
+# ifdef HAVE___LDISKFS_FIND_ENTRY
+# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
(__ldiskfs_find_entry(dir, name, de, inlined, lock) ?: \
ERR_PTR(-ENOENT))
+# else
+# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
+ (ldiskfs_find_entry_locked(dir, name, de, inlined, lock) ?: \
+ ERR_PTR(-ENOENT))
+# endif
+
# define osd_journal_start(inode, type, nblocks) \
ldiskfs_journal_start(inode, type, nblocks)
# define osd_transaction_size(dev) \
(osd_journal(dev)->j_max_transaction_buffers / 2)
-#else
+#else /* ! defined LDISKFS_HT_MISC */
# define LDISKFS_HT_MISC 0
# define osd_journal_start_sb(sb, type, nblock) \
ldiskfs_journal_start_sb(sb, nblock)
return bh;
}
-# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
+# ifdef HAVE___LDISKFS_FIND_ENTRY
+# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
(__ldiskfs_find_entry(dir, name, de, lock) ?: \
ERR_PTR(-ENOENT))
+# else
+# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
+ (ldiskfs_find_entry_locked(dir, name, de, lock) ?: \
+ ERR_PTR(-ENOENT))
+# endif
# define osd_journal_start(inode, type, nblocks) \
ldiskfs_journal_start(inode, nblocks)
# define osd_transaction_size(dev) \
(osd_journal(dev)->j_max_transaction_buffers)
+#endif /* LDISKFS_HT_MISC */
+
+#ifndef HAVE___LDISKFS_FIND_ENTRY
+# define __ldiskfs_add_entry(handle, child, inode, hlock) \
+ ldiskfs_add_entry_locked(handle, child, inode, hlock)
#endif
/*
#endif
#define OSD_MAX_CACHE_SIZE OBD_OBJECT_EOF
+#define OSD_READCACHE_MAX_IO_MB 8
+#define OSD_WRITECACHE_MAX_IO_MB 8
extern const struct dt_index_operations osd_otable_ops;
static inline struct osd_device *osd_dt_dev(const struct dt_device *d)
{
- LASSERT(lu_device_is_osd(&d->dd_lu_dev));
- return container_of0(d, struct osd_device, od_dt_dev);
+ LASSERT(lu_device_is_osd(&d->dd_lu_dev));
+ return container_of(d, struct osd_device, od_dt_dev);
}
static inline struct osd_device *osd_dev(const struct lu_device *d)
{
- LASSERT(lu_device_is_osd(d));
- return osd_dt_dev(container_of0(d, struct dt_device, dd_lu_dev));
+ LASSERT(lu_device_is_osd(d));
+ return osd_dt_dev(container_of(d, struct dt_device, dd_lu_dev));
}
static inline struct osd_device *osd_obj2dev(const struct osd_object *o)
return dev->od_mnt->mnt_sb;
}
+static inline const char *osd_sb2name(const struct super_block *sb)
+{
+ /* this is LDISKFS_SB(sb), but preserves "const" */
+ const struct ldiskfs_sb_info *sbi = sb->s_fs_info;
+
+ return sbi->s_es->s_volume_name;
+}
+
static inline const char *osd_dev2name(const struct osd_device *dev)
{
- return osd_sb(dev)->s_id;
+ return osd_sb2name(osd_sb(dev));
}
static inline const char *osd_ino2name(const struct inode *inode)
{
- return inode->i_sb->s_id;
+ return osd_sb2name(inode->i_sb);
}
/**
static inline struct osd_object *osd_obj(const struct lu_object *o)
{
- LASSERT(lu_device_is_osd(o->lo_dev));
- return container_of0(o, struct osd_object, oo_dt.do_lu);
+ LASSERT(lu_device_is_osd(o->lo_dev));
+ return container_of(o, struct osd_object, oo_dt.do_lu);
}
static inline struct osd_object *osd_dt_obj(const struct dt_object *d)
const int blocks);
int osd_ldiskfs_read(struct inode *inode, void *buf, int size, loff_t *offs);
-int osd_ldiskfs_write_record(struct inode *inode, void *buf, int bufsize,
- int write_NUL, loff_t *offs, handle_t *handle);
static inline
struct dentry *osd_child_dentry_by_inode(const struct lu_env *env,
return child_dentry;
}
+/* build quasi file structure when it is needed to call an inode i_fop */
+static inline struct file *osd_quasi_file_init(const struct lu_env *env,
+ struct dentry *dentry,
+ struct inode *inode)
+{
+ struct osd_thread_info *info = osd_oti_get(env);
+
+ info->oti_file.f_path.dentry = dentry;
+ info->oti_file.f_mapping = inode->i_mapping;
+ info->oti_file.f_op = inode->i_fop;
+ info->oti_file.f_inode = inode;
+ info->oti_file.f_pos = 0;
+ info->oti_file.private_data = NULL;
+ info->oti_file.f_cred = current_cred();
+ info->oti_file.f_flags = O_NOATIME;
+ info->oti_file.f_mode = FMODE_64BITHASH | FMODE_NONOTIFY;
+
+ return &info->oti_file;
+}
+
+static inline struct file *osd_quasi_file(const struct lu_env *env,
+ struct inode *inode)
+{
+ struct osd_thread_info *info = osd_oti_get(env);
+
+ info->oti_obj_dentry.d_inode = inode;
+ info->oti_obj_dentry.d_sb = inode->i_sb;
+
+ return osd_quasi_file_init(env, &info->oti_obj_dentry, inode);
+}
+
+static inline struct file *osd_quasi_file_by_dentry(const struct lu_env *env,
+ struct dentry *dentry)
+{
+ return osd_quasi_file_init(env, dentry, dentry->d_inode);
+}
+
extern int osd_trans_declare_op2rb[];
extern int ldiskfs_track_declares_assert;
void osd_trans_dump_creds(const struct lu_env *env, struct thandle *th);
oh->ot_credits += credits;
}
+/* linux: v5.4-rc3-21-g933f1c1e0b75 renamed h_buffer_credits */
+#ifdef HAVE_JOURNAL_TOTAL_CREDITS
+#define h_buffer_credits h_total_credits
+#endif
+
static inline void osd_trans_exec_op(const struct lu_env *env,
struct thandle *th,
enum osd_op_type op)
#endif
#ifdef HAVE_BI_BDEV
-# define bio_set_dev(bio, bdev) ((bio)->bi_bdev = bdev)
-# define bio_get_queue(bio) bdev_get_queue((bio)->bi_bdev)
+# define bio_get_dev(bio) ((bio)->bi_bdev)
+# define bio_get_disk(bio) (bio_get_dev(bio)->bd_disk)
+# define bio_get_queue(bio) bdev_get_queue(bio_get_dev(bio))
+# define bio_set_dev(bio, bdev) (bio_get_dev(bio) = (bdev))
#else
-# define bio_get_queue(bio) ((bio)->bi_disk->queue)
+# define bio_get_disk(bio) ((bio)->bi_disk)
+# define bio_get_queue(bio) (bio_get_disk(bio)->queue)
#endif
void ldiskfs_inc_count(handle_t *handle, struct inode *inode);
int osd_trunc_lock(struct osd_object *obj, struct osd_thandle *oh,
bool shared);
-void osd_trunc_unlock_all(struct list_head *list);
+void osd_trunc_unlock_all(const struct lu_env *env, struct list_head *list);
void osd_process_truncates(struct list_head *list);
void osd_execute_truncate(struct osd_object *obj);
+#ifdef HAVE_BIO_ENDIO_USES_ONE_ARG
+#define osd_dio_complete_routine(bio, error) dio_complete_routine(bio)
+#else
+#define osd_dio_complete_routine(bio, error) dio_complete_routine(bio, error)
+#endif
+
+#ifndef HAVE___BI_CNT
+#define __bi_cnt bi_cnt
+#endif
+
+#ifndef HAVE_BI_OPF
+#define bi_opf bi_rw
+#endif
+
+#ifndef HAVE_CLEAN_BDEV_ALIASES
+#define clean_bdev_aliases(bdev, block, len) \
+ unmap_underlying_metadata((bdev), (block))
+#endif
+
+#ifndef HAVE_BI_STATUS
+#define bi_status bi_error
+#endif
+
/*
* Maximum size of xattr attributes for FEATURE_INCOMPAT_EA_INODE 1Mb
* This limit is arbitrary, but is reasonable for the xattr API.
int osd_get_integrity_profile(struct osd_device *osd,
integrity_gen_fn **generate_fn,
integrity_vrfy_fn **verify_fn);
+#else
+#define integrity_gen_fn void
+#define integrity_vrfy_fn int
+static inline int osd_get_integrity_profile(struct osd_device *osd,
+ integrity_gen_fn **generate_fn,
+ integrity_vrfy_fn **verify_fn)
+{
+ return 0;
+}
+
+static inline bool bio_integrity_prep_fn(struct bio *bio,
+ integrity_gen_fn *generate_fn,
+ integrity_vrfy_fn *verify_fn)
+{
+ return bio_integrity_prep(bio);
+}
#endif
+#ifdef HAVE_BIO_BI_PHYS_SEGMENTS
+#define osd_bio_nr_segs(bio) ((bio)->bi_phys_segments)
+#else
+#define osd_bio_nr_segs(bio) bio_segments((bio))
+#endif /* HAVE_BIO_BI_PHYS_SEGMENTS */
+
#endif /* _OSD_INTERNAL_H */