* Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
* Use is subject to license terms.
*
- * Copyright (c) 2011, 2013, Intel Corporation.
+ * Copyright (c) 2011, 2014, Intel Corporation.
*/
/*
* This file is part of Lustre, http://www.lustre.org/
#ifndef _OSD_INTERNAL_H
#define _OSD_INTERNAL_H
-#if defined(__KERNEL__)
+/* struct mutex */
+#include <linux/mutex.h>
/* struct rw_semaphore */
#include <linux/rwsem.h>
/* struct dentry */
#include <linux/dcache.h>
/* struct dirent64 */
#include <linux/dirent.h>
-
+#include <linux/statfs.h>
#include <ldiskfs/ldiskfs.h>
#include <ldiskfs/ldiskfs_jbd2.h>
#include "osd_quota_fmt.h"
struct inode;
+extern struct kmem_cache *dynlock_cachep;
#define OSD_COUNTERS (0)
/* OI scrub should skip this inode. */
#define LDISKFS_STATE_LUSTRE_NOSCRUB 31
-/* Do not add OI mapping for this inode. */
-#define LDISKFS_STATE_LUSTRE_NO_OI 30
-
/** Enable thandle usage statistics */
#define OSD_THANDLE_STATS (0)
int oos_subdir_count; /* subdir count for each seq */
struct dentry *oos_root; /* O/<seq> */
struct dentry **oos_dirs; /* O/<seq>/d0-dXX */
- obd_seq oos_seq; /* seq number */
- cfs_list_t oos_seq_list; /* list to seq_list */
+ u64 oos_seq; /* seq number */
+ struct list_head oos_seq_list; /* list to seq_list */
};
struct osd_obj_map {
struct dentry *om_root; /* dentry for /O */
rwlock_t om_seq_list_lock; /* lock for seq_list */
- cfs_list_t om_seq_list; /* list head for seq */
+ struct list_head om_seq_list; /* list head for seq */
int om_subdir_count;
- struct semaphore om_dir_init_sem;
+ struct mutex om_dir_init_mutex;
};
struct osd_mdobj {
struct dentry *om_root; /* AGENT/<index> */
- obd_seq om_index; /* mdt index */
- cfs_list_t om_list; /* list to omm_list */
+ u64 om_index; /* mdt index */
+ struct list_head om_list; /* list to omm_list */
};
struct osd_mdobj_map {
struct dentry *omm_remote_parent;
};
-#define osd_ldiskfs_find_entry(dir, dentry, de, lock) \
- ll_ldiskfs_find_entry(dir, dentry, de, lock)
#define osd_ldiskfs_add_entry(handle, child, cinode, hlock) \
ldiskfs_add_entry(handle, child, cinode, hlock)
struct osd_inconsistent_item {
/* link into osd_scrub::os_inconsistent_items,
* protected by osd_scrub::os_lock. */
- cfs_list_t oii_list;
+ struct list_head oii_list;
/* The right FID <=> ino#/gen mapping. */
struct osd_idmap_cache oii_cache;
unsigned int od_fl_capa:1,
od_maybe_new:1,
od_noscrub:1,
- od_dirent_journal:1,
od_igif_inoi:1,
od_check_ff:1,
od_is_ost:1,
- od_lma_self_repair:1;
+ od_index_in_idif:1;
- unsigned long od_capa_timeout;
- __u32 od_capa_alg;
- struct lustre_capa_key *od_capa_keys;
- cfs_hlist_head_t *od_capa_hash;
+ unsigned long od_capa_timeout;
+ __u32 od_capa_alg;
+ __u32 od_dirent_journal;
+ struct lustre_capa_key *od_capa_keys;
+ struct hlist_head *od_capa_hash;
+
+ struct proc_dir_entry *od_proc_entry;
+ struct lprocfs_stats *od_stats;
- cfs_proc_dir_entry_t *od_proc_entry;
- struct lprocfs_stats *od_stats;
- /*
- * statfs optimization: we cache a bit.
- */
- cfs_time_t od_osfs_age;
- struct obd_statfs od_statfs;
spinlock_t od_osfs_lock;
int od_connects;
struct osd_obj_map *od_ost_map;
struct osd_mdobj_map *od_mdt_map;
- unsigned long long od_readcache_max_filesize;
- int od_read_cache;
- int od_writethrough_cache;
+ unsigned long long od_readcache_max_filesize;
+ int od_read_cache;
+ int od_writethrough_cache;
- struct brw_stats od_brw_stats;
- cfs_atomic_t od_r_in_flight;
- cfs_atomic_t od_w_in_flight;
+ struct brw_stats od_brw_stats;
+ atomic_t od_r_in_flight;
+ atomic_t od_w_in_flight;
struct mutex od_otable_mutex;
struct osd_otable_it *od_otable_it;
struct osd_scrub od_scrub;
- cfs_list_t od_ios_list;
+ struct list_head od_ios_list;
/* service name associated with the osd device */
char od_svname[MAX_OBD_NAME];
/* quota slave instance */
struct qsd_instance *od_quota_slave;
+
+ /* osd seq instance */
+ struct lu_client_seq *od_cl_seq;
+ /* If the ratio of "the total OI mappings count" vs
+ * "the bad OI mappings count" is lower than the
+ * osd_device::od_full_scrub_ratio, then trigger
+ * OI scrub to scan the whole the device. */
+ __u64 od_full_scrub_ratio;
+ /* If the speed of found bad OI mappings (per minute)
+ * exceeds the osd_device::od_full_scrub_threshold_rate,
+ * then trigger OI scrub to scan the whole device. */
+ __u64 od_full_scrub_threshold_rate;
};
+enum osd_full_scrub_ratio {
+ /* Trigger OI scrub to scan the whole device directly. */
+ OFSR_DIRECTLY = 0,
+
+ /* Because the bad OI mappings count cannot be larger than
+ * the total OI mappints count, then setting OFSR_NEVER means
+ * that the whole device scanning cannot be triggered by auto
+ * detected bad OI mappings during the RPC services. */
+ OFSR_NEVER = 1,
+ OFSR_DEFAULT = 10000,
+};
+
+#define FULL_SCRUB_THRESHOLD_RATE_DEFAULT 60
+
/* There are at most 10 uid/gids are affected in a transaction, and
* that's rename case:
* - 2 for source parent uid & gid;
struct thandle ot_super;
handle_t *ot_handle;
struct ldiskfs_journal_cb_entry ot_jcb;
- cfs_list_t ot_dcb_list;
+ struct list_head ot_dcb_list;
/* Link to the device, for debugging. */
struct lu_ref_link ot_dev_link;
unsigned short ot_credits;
DTO_OBJECT_DELETE,
DTO_ATTR_SET_BASE,
DTO_XATTR_SET,
- DTO_LOG_REC, /**< XXX temporary: dt layer knows nothing about llog. */
DTO_WRITE_BASE,
DTO_WRITE_BLOCK,
DTO_ATTR_SET_CHOWN,
* osd dev stats
*/
-#ifdef LPROCFS
+#ifdef CONFIG_PROC_FS
enum {
LPROC_OSD_READ_BYTES = 0,
LPROC_OSD_WRITE_BYTES = 1,
* mode (i.e. iterator over ldiskfs style directory)
*/
struct osd_it_ea {
- struct osd_object *oie_obj;
- /** used in ldiskfs iterator, to stored file pointer */
- struct file oie_file;
- /** how many entries have been read-cached from storage */
- int oie_rd_dirent;
- /** current entry is being iterated by caller */
- int oie_it_dirent;
- /** current processing entry */
- struct osd_it_ea_dirent *oie_dirent;
- /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
- void *oie_buf;
+ struct osd_object *oie_obj;
+ /** used in ldiskfs iterator, to stored file pointer */
+ struct file oie_file;
+ /** how many entries have been read-cached from storage */
+ int oie_rd_dirent;
+ /** current entry is being iterated by caller */
+ int oie_it_dirent;
+ /** current processing entry */
+ struct osd_it_ea_dirent *oie_dirent;
+ /** buffer to hold entries, size == OSD_IT_EA_BUFSIZE */
+ void *oie_buf;
+ struct dentry oie_dentry;
};
/**
};
struct osd_quota_leaf {
- cfs_list_t oql_link;
+ struct list_head oql_link;
uint oql_blk;
};
/** the record index in the leaf/index block */
uint oiq_index[LUSTRE_DQTREEDEPTH + 1];
/** list of already processed leaf blocks */
- cfs_list_t oiq_list;
+ struct list_head oiq_list;
};
#define MAX_BLOCKS_PER_PAGE (PAGE_CACHE_SIZE / 512)
struct osd_iobuf {
wait_queue_head_t dr_wait;
- cfs_atomic_t dr_numreqs; /* number of reqs being processed */
+ atomic_t dr_numreqs; /* number of reqs being processed */
int dr_max_pages;
int dr_npages;
int dr_error;
* XXX temporary: for ->i_op calls.
*/
struct timespec oti_time;
- /*
- * XXX temporary: fake struct file for osd_object_sync
- */
- struct file oti_file;
+ /*
+ * XXX temporary: fake struct file for osd_object_sync
+ */
+ struct file oti_file;
/*
* XXX temporary: for capa operations.
*/
* in open iterator session.
*/
- /** osd iterator context used for iterator session */
-
- union {
- struct osd_it_iam oti_it;
- /* ldiskfs iterator data structure,
- * see osd_it_ea_{init, fini} */
- struct osd_it_ea oti_it_ea;
- struct osd_it_quota oti_it_quota;
- };
-
/** pre-allocated buffer used by oti_it_ea, size OSD_IT_EA_BUFSIZE */
void *oti_it_ea_buf;
+ unsigned int oti_it_ea_buf_used:1;
struct kstatfs oti_ksfs;
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_mdt_attrs oti_mdt_attrs;
return inode->i_op->setxattr(dentry, name, buf, buflen, fl);
}
-#ifdef LPROCFS
+#ifdef CONFIG_PROC_FS
/* osd_lproc.c */
-void lprocfs_osd_init_vars(struct lprocfs_static_vars *lvars);
+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);
+#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(3, 0, 52, 0)
+int osd_register_proc_index_in_idif(struct osd_device *osd);
+#endif
#endif
int osd_statfs(const struct lu_env *env, struct dt_device *dev,
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);
-int osd_add_oi_cache(struct osd_thread_info *info, struct osd_device *osd,
- struct osd_inode_id *id, const struct lu_fid *fid);
+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 dentry *dentry, struct lu_fid *fid);
const struct lu_fid *fid, struct osd_inode_id *id);
int osd_obj_map_insert(struct osd_thread_info *info, struct osd_device *osd,
const struct lu_fid *fid, const struct osd_inode_id *id,
- struct thandle *th);
+ handle_t *th);
int osd_obj_map_delete(struct osd_thread_info *info, struct osd_device *osd,
- const struct lu_fid *fid, struct thandle *th);
+ const struct lu_fid *fid, handle_t *th);
int osd_obj_map_update(struct osd_thread_info *info, struct osd_device *osd,
const struct lu_fid *fid, const struct osd_inode_id *id,
- struct thandle *th);
+ handle_t *th);
int osd_obj_map_recover(struct osd_thread_info *info, struct osd_device *osd,
struct inode *src_parent, struct dentry *src_child,
const struct lu_fid *fid);
const struct lu_fid *fid, struct osd_inode_id *id);
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,
- struct thandle *th);
+ handle_t *th);
int osd_obj_spec_update(struct osd_thread_info *info, struct osd_device *osd,
const struct lu_fid *fid, const struct osd_inode_id *id,
- struct thandle *th);
+ handle_t *th);
void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags);
int osd_scrub_file_store(struct osd_scrub *scrub);
char *osd_lf_fid2name(const struct lu_fid *fid);
-int osd_scrub_start(struct osd_device *dev);
+int osd_scrub_start(struct osd_device *dev, __u32 flags);
int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev);
void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev);
int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
int insert);
int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
struct osd_inode_id *id);
-int osd_scrub_dump(struct osd_device *dev, char *buf, int len);
+int osd_scrub_dump(struct seq_file *m, struct osd_device *dev);
int osd_fld_lookup(const struct lu_env *env, struct osd_device *osd,
- obd_seq seq, struct lu_seq_range *range);
+ u64 seq, struct lu_seq_range *range);
int osd_delete_from_remote_parent(const struct lu_env *env,
struct osd_device *osd,
struct osd_it_quota *it);
/* osd_quota.c */
int osd_declare_qid(const struct lu_env *env, struct osd_thandle *oh,
- struct lquota_id_info *qi, bool allocated, int *flags);
+ 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,
- bool is_blk, bool allocated, int *flags, bool force);
+ struct osd_object *obj, bool is_blk, int *flags,
+ bool force);
const struct dt_rec *osd_quota_pack(struct osd_object *obj,
const struct dt_rec *rec,
union lquota_rec *quota_rec);
feat == &dt_quota_bgrp_features) ? true : false;
}
+#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
+
/*
* Invariants, assertions.
*/
LASSERT(oh->ot_handle == NULL);
if (unlikely(op >= OSD_OT_MAX)) {
- if (unlikely(ldiskfs_track_declares_assert))
+ if (unlikely(ldiskfs_track_declares_assert)) {
LASSERT(op < OSD_OT_MAX);
- else {
+ } else {
CWARN("%s: Invalid operation index %d\n",
osd_name(oti->oti_dev), op);
libcfs_debug_dumpstack(NULL);
{
return (!fid_is_namespace_visible(fid) && !fid_is_idif(fid));
}
-#endif /* __KERNEL__ */
+
+static inline unsigned long osd_remote_parent_ino(struct osd_device *dev)
+{
+ return dev->od_mdt_map->omm_remote_parent->d_inode->i_ino;
+}
+
+#ifdef JOURNAL_START_HAS_3ARGS
+# define osd_journal_start_sb(sb, type, nblock) \
+ ldiskfs_journal_start_sb(sb, type, nblock)
+# define osd_ldiskfs_append(handle, inode, nblock, err) \
+ ldiskfs_append(handle, inode, nblock)
+# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
+ ldiskfs_find_entry(dir, name, de, inlined, lock)
+# 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
+# define LDISKFS_HT_MISC 0
+# define osd_journal_start_sb(sb, type, nblock) \
+ ldiskfs_journal_start_sb(sb, nblock)
+# define osd_ldiskfs_append(handle, inode, nblock, err) \
+ ldiskfs_append(handle, inode, nblock, err)
+# define osd_ldiskfs_find_entry(dir, name, de, inlined, lock) \
+ ldiskfs_find_entry(dir, name, de, lock)
+# 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
+
+void ldiskfs_inc_count(handle_t *handle, struct inode *inode);
+void ldiskfs_dec_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