Whamcloud - gitweb
add larry's changes that allow using mpirun or yod for running IOR.
[fs/lustre-release.git] / lustre / llite / llite_internal.h
index 4684383..bb7ea25 100644 (file)
 #ifndef LLITE_INTERNAL_H
 #define LLITE_INTERNAL_H
 
+#include <linux/lustre_debug.h>
+
+/* default to about 40meg of readahead on a given system.  That much tied
+ * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
+#define SBI_DEFAULT_RA_MAX ((40 << 20) >> PAGE_CACHE_SHIFT)
+
+enum ra_stat {
+        RA_STAT_HIT = 0,
+        RA_STAT_MISS,
+        RA_STAT_DISTANT_READPAGE,
+        RA_STAT_MISS_IN_WINDOW,
+        RA_STAT_FAILED_MATCH,
+        RA_STAT_DISCARDED,
+        RA_STAT_ZERO_LEN,
+        RA_STAT_ZERO_WINDOW,
+        RA_STAT_EOF,
+        RA_STAT_MAX_IN_FLIGHT,
+        _NR_RA_STAT,
+};
+
+struct ll_ra_info {
+        unsigned long             ra_cur_pages;
+        unsigned long             ra_max_pages;
+        unsigned long             ra_stats[_NR_RA_STAT];
+};
+
+/* after roughly how long should we remove an inactive mount? */
+#define GNS_MOUNT_TIMEOUT 120
+
+/* how often should the GNS timer look for mounts to cleanup? */
+#define GNS_TICK_TIMEOUT  1
+
+/* how many times GNS will try to wait for 1 second for mount */
+#define GNS_WAIT_ATTEMPTS 10
+
+struct ll_sb_info {
+        /* this protects pglist and max_r_a_pages.  It isn't safe to grab from
+         * interrupt contexts. */
+        spinlock_t                ll_lock;
+        
+        struct obd_uuid           ll_sb_uuid;
+        struct obd_export        *ll_md_exp;
+        struct obd_export        *ll_dt_exp;
+        struct lov_desc           ll_dt_desc;
+        struct proc_dir_entry    *ll_proc_root;
+        struct lustre_id          ll_rootid;     /* root lustre id */
+
+        struct lustre_mount_data *ll_lmd;
+        char                     *ll_instance;
+
+        int                       ll_flags;
+        struct list_head          ll_conn_chain; /* per-conn chain of SBs */
+
+        struct hlist_head         ll_orphan_dentry_list; /*please don't ask -p*/
+        struct ll_close_queue    *ll_lcq;
+
+        struct lprocfs_stats     *ll_stats;      /* lprocfs stats counter */
+
+        unsigned long             ll_pglist_gen;
+        struct list_head          ll_pglist;
+
+        struct ll_ra_info         ll_ra_info;
+
+        unsigned int              ll_remote;    /* remote client? */
+
+        /* times spent waiting for locks in each call site.  These are
+         * all protected by the ll_lock */
+        struct obd_service_time   ll_read_stime;
+        struct obd_service_time   ll_write_stime;
+        struct obd_service_time   ll_grouplock_stime;
+        struct obd_service_time   ll_seek_stime;
+        struct obd_service_time   ll_setattr_stime;
+        struct obd_service_time   ll_brw_stime;
+//      struct obd_service_time   ll_done_stime;
+
+        int                       ll_config_version; /* last-applied update */
+
+        /* list of GNS mounts; protected by the dcache_lock */
+        struct list_head          ll_mnt_list;
+
+        struct semaphore          ll_gns_sem;
+        spinlock_t                ll_gns_lock;
+        wait_queue_head_t         ll_gns_waitq;
+        atomic_t                  ll_gns_enabled;
+        int                       ll_gns_state;
+        struct timer_list         ll_gns_timer;
+        struct list_head          ll_gns_sbi_head;
+        struct completion         ll_gns_mount_finished;
+       struct dentry            *ll_gns_pending_dentry;
+
+        unsigned long             ll_gns_tick;
+        unsigned long             ll_gns_timeout;
+
+        /* path to upcall */
+        char                      ll_gns_upcall[PATH_MAX];
+
+        /* mount object entry name */
+        char                      ll_gns_oname[PATH_MAX];
+};
+
+struct ll_gns_ctl {
+        struct completion gc_starting;
+        struct completion gc_finishing;
+};
+
+/* mounting states */
+#define LL_GNS_IDLE               (1 << 0)
+#define LL_GNS_MOUNTING           (1 << 1)
+#define LL_GNS_FINISHED           (1 << 2)
+
+/* mounts checking flags */
+#define LL_GNS_UMOUNT             (1 << 0)
+#define LL_GNS_CHECK              (1 << 1)
+
+struct ll_readahead_state {
+        spinlock_t      ras_lock;
+        unsigned long   ras_last_readpage, ras_consecutive;
+        unsigned long   ras_window_start, ras_window_len;
+        unsigned long   ras_next_readahead;
+
+};
+
+extern kmem_cache_t *ll_file_data_slab;
+extern kmem_cache_t *ll_intent_slab;
 struct lustre_handle;
+
+struct ll_file_data {
+        struct ll_readahead_state fd_ras;
+        __u32 fd_flags;
+        int fd_omode;
+        struct lustre_handle fd_cwlockh;
+        unsigned long fd_gid;
+};
+
 struct lov_stripe_md;
 
-int ll_mdc_cancel_unused(struct lustre_handle *conn, struct inode *inode,
-                         int flags, void *opaque);
-int ll_rd_dirty_pages(char *page, char **start, off_t off, int count,
-                      int *eof, void *data);
-int ll_rd_max_dirty_pages(char *page, char **start, off_t off, int count,
-                          int *eof, void *data);
-int ll_wr_max_dirty_pages(struct file *file, const char *buffer,
-                          unsigned long count, void *data);
-int ll_clear_dirty_pages(struct lustre_handle *conn, struct lov_stripe_md *lsm,
-                         unsigned long start, unsigned long end);
-int ll_mark_dirty_page(struct lustre_handle *conn, struct lov_stripe_md *lsm,
-                       unsigned long index);
+extern spinlock_t inode_lock;
+
+extern void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
+extern struct proc_dir_entry *proc_lustre_fs_root;
+
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,5,0))
+# define hlist_del_init list_del_init
+#endif
+
+static inline struct inode *ll_info2i(struct ll_inode_info *lli)
+{
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
+        return &lli->lli_vfs_inode;
+#else
+        return list_entry(lli, struct inode, u.generic_ip);
+#endif
+}
+
+
+struct it_cb_data {
+        struct inode *icbd_parent;
+        struct dentry **icbd_childp;
+        obd_id hash;
+};
+
+#define LLAP_MAGIC 98764321
+
+struct ll_async_page {
+        int             llap_magic;
+        void            *llap_cookie;
+        struct page     *llap_page;
+        struct list_head llap_pending_write;
+         /* only trust these if the page lock is providing exclusion */
+        unsigned         llap_write_queued:1,
+                         llap_defer_uptodate:1,
+                         llap_origin:3,
+                         llap_ra_used:1;
+
+        struct list_head llap_proc_item;
+};
+
+enum {
+        LLAP_ORIGIN_UNKNOWN = 0,
+        LLAP_ORIGIN_READPAGE,
+        LLAP_ORIGIN_READAHEAD,
+        LLAP_ORIGIN_COMMIT_WRITE,
+        LLAP_ORIGIN_WRITEPAGE,
+        LLAP__ORIGIN_MAX,
+};
+
+/* llite/lproc_llite.c */
+int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
+                                struct super_block *sb, char *lov,
+                                char *lmv);
+void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi);
+
+/* llite/dir.c */
+extern struct file_operations ll_dir_operations;
+extern struct inode_operations ll_dir_inode_operations;
+
+/* llite/namei.c */
+int ll_objects_destroy(struct ptlrpc_request *request, 
+                       struct inode *dir, int offset);
+struct inode *ll_iget(struct super_block *sb, ino_t hash,
+                      struct lustre_md *lic);
+struct dentry *ll_find_alias(struct inode *, struct dentry *);
+int ll_mdc_cancel_unused(struct lustre_handle *, struct inode *, int flags,
+                         void *opaque);
+int ll_mdc_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *,
+                        void *data, int flag);
+/* llite/rw.c */
+int ll_prepare_write(struct file *, struct page *, unsigned from, unsigned to);
+int ll_commit_write(struct file *, struct page *, unsigned from, unsigned to);
+int ll_writepage(struct page *page);
+void ll_inode_fill_obdo(struct inode *inode, int cmd, struct obdo *oa);
+void ll_ap_completion(void *data, int cmd, struct obdo *oa, int rc);
+void ll_removepage(struct page *page);
+int ll_readpage(struct file *file, struct page *page);
+struct ll_async_page *llap_from_cookie(void *cookie);
+struct ll_async_page *llap_from_page(struct page *page, unsigned origin);
+struct ll_async_page *llap_cast_private(struct page *page);
+void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras);
+
+void ll_ra_accounting(struct page *page, struct address_space *mapping);
+void ll_truncate(struct inode *inode);
+
+/* llite/file.c */
+extern struct file_operations ll_file_operations;
+extern struct inode_operations ll_file_inode_operations;
+int ll_md_real_close(struct obd_export *md_exp,
+                     struct inode *inode, int flags);
+extern int ll_inode_revalidate_it(struct dentry *);
+extern int ll_setxattr(struct dentry *, const char *, const void *,
+                       size_t, int);
+extern int ll_getxattr(struct dentry *, const char *, void *, size_t);
+extern int ll_listxattr(struct dentry *, char *, size_t);
+extern int ll_removexattr(struct dentry *, const char *);
+extern int ll_inode_permission(struct inode *, int, struct nameidata *);
+int ll_refresh_lsm(struct inode *inode, struct lov_stripe_md *lsm);
+int ll_extent_lock(struct ll_file_data *, struct inode *,
+                   struct lov_stripe_md *, int mode, ldlm_policy_data_t *,
+                   struct lustre_handle *, int ast_flags,
+                   struct obd_service_time *);
+int ll_extent_unlock(struct ll_file_data *, struct inode *,
+                     struct lov_stripe_md *, int mode, struct lustre_handle *);
+int ll_file_open(struct inode *inode, struct file *file);
+int ll_file_release(struct inode *inode, struct file *file);
+int ll_lsm_getattr(struct obd_export *, struct lov_stripe_md *, struct obdo *);
+int ll_glimpse_size(struct inode *inode);
+int ll_local_open(struct file *file, struct lookup_intent *it,
+                  struct obd_client_handle *och);
+int ll_md_close(struct obd_export *md_exp, struct inode *inode,
+                struct file *file);
+int ll_md_och_close(struct obd_export *md_exp, struct inode *inode,
+                    struct obd_client_handle *och);
+void ll_och_fill(struct inode *inode, struct lookup_intent *it,
+                 struct obd_client_handle *och);
+
+#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0))
+int ll_getattr(struct vfsmount *mnt, struct dentry *de, struct kstat *stat);
+#endif
+void ll_stime_record(struct ll_sb_info *sbi, struct timeval *start,
+                     struct obd_service_time *stime);
+
+/* llite/dcache.c */
+void ll_intent_drop_lock(struct lookup_intent *);
+void ll_intent_release(struct lookup_intent *);
+int ll_intent_alloc(struct lookup_intent *);
+void ll_intent_free(struct lookup_intent *it);
+extern void ll_set_dd(struct dentry *de);
+void ll_unhash_aliases(struct inode *);
+void ll_frob_intent(struct lookup_intent **itp, struct lookup_intent *deft);
+void ll_lookup_finish_locks(struct lookup_intent *it, struct dentry *dentry);
+int revalidate_it_finish(struct ptlrpc_request *request, int offset,
+                         struct lookup_intent *it, struct dentry *de);
+
+
+/* llite/llite_gns.c */
+int ll_gns_start_thread(void);
+void ll_gns_stop_thread(void);
+
+int ll_gns_mount_object(struct dentry *dentry,
+                        struct vfsmount *mnt);
+int ll_gns_umount_object(struct vfsmount *mnt);
+
+int ll_gns_check_mounts(struct ll_sb_info *sbi,
+                        int flags);
+
+void ll_gns_timer_callback(unsigned long data);
+void ll_gns_add_timer(struct ll_sb_info *sbi);
+void ll_gns_del_timer(struct ll_sb_info *sbi);
+
+/* llite/llite_lib.c */
+extern struct super_operations lustre_super_operations;
+
+char *ll_read_opt(const char *opt, char *data);
+int ll_set_opt(const char *opt, char *data, int fl);
+void ll_options(char *options, char **ost, char **mds, char **sec, 
+                int *async, int *flags);
+void ll_lli_init(struct ll_inode_info *lli);
+int ll_fill_super(struct super_block *sb, void *data, int silent);
+int lustre_fill_super(struct super_block *sb, void *data, int silent);
+void lustre_put_super(struct super_block *sb);
+struct inode *ll_inode_from_lock(struct ldlm_lock *lock);
+void ll_clear_inode(struct inode *inode);
+int ll_attr2inode(struct inode *inode, struct iattr *attr, int trunc);
+int ll_setattr_raw(struct inode *inode, struct iattr *attr);
+int ll_setattr(struct dentry *de, struct iattr *attr);
+int ll_statfs(struct super_block *sb, struct kstatfs *sfs);
+int ll_statfs_internal(struct super_block *sb, struct obd_statfs *osfs,
+                       unsigned long maxage);
+void ll_update_inode(struct inode *inode, struct lustre_md *);
+int it_disposition(struct lookup_intent *it, int flag);
+void it_set_disposition(struct lookup_intent *it, int flag);
+void ll_read_inode2(struct inode *inode, void *opaque);
+void ll_delete_inode(struct inode *inode);
+int ll_iocontrol(struct inode *inode, struct file *file,
+                 unsigned int cmd, unsigned long arg);
+void ll_umount_begin(struct super_block *sb);
+int ll_prep_inode(struct obd_export *, struct obd_export *, struct inode **inode,
+                  struct ptlrpc_request *req, int offset, struct super_block *);
+__u32 get_uuid2int(const char *name, int len);
+struct dentry *ll_fh_to_dentry(struct super_block *sb, __u32 *data, int len,
+                               int fhtype, int parent);
+int ll_dentry_to_fh(struct dentry *, __u32 *datap, int *lenp, int need_parent);
+int null_if_equal(struct ldlm_lock *lock, void *data);
+int ll_process_config_update(struct ll_sb_info *sbi, int clean);
+
+/* llite/special.c */
+extern struct inode_operations ll_special_inode_operations;
+extern struct file_operations ll_special_chr_inode_fops;
+extern struct file_operations ll_special_chr_file_fops;
+extern struct file_operations ll_special_blk_inode_fops;
+extern struct file_operations ll_special_fifo_inode_fops;
+extern struct file_operations ll_special_fifo_file_fops;
+extern struct file_operations ll_special_sock_inode_fops;
+
+/* llite/symlink.c */
+extern struct inode_operations ll_fast_symlink_inode_operations;
+
+/* llite/llite_close.c */
+struct ll_close_queue {
+        spinlock_t              lcq_lock;
+        struct list_head        lcq_list;
+        wait_queue_head_t       lcq_waitq;
+        struct completion       lcq_comp;
+};
+
+void llap_write_pending(struct inode *inode, struct ll_async_page *llap);
+void llap_write_complete(struct inode *inode, struct ll_async_page *llap);
+void ll_open_complete(struct inode *inode);
+int ll_is_inode_dirty(struct inode *inode);
+void ll_try_done_writing(struct inode *inode);
+void ll_queue_done_writing(struct inode *inode);
+void ll_close_thread_shutdown(struct ll_close_queue *lcq);
+int ll_close_thread_start(struct ll_close_queue **lcq_ret);
+
+
+/* llite/llite_mmap.c */
+#if  (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
+typedef struct rb_root  rb_root_t;
+typedef struct rb_node  rb_node_t;
+#endif
+
+struct ll_lock_tree_node;
+struct ll_lock_tree {
+        rb_root_t                       lt_root;
+        struct list_head                lt_locked_list;
+        struct ll_file_data             *lt_fd;
+};
+int ll_teardown_mmaps(struct address_space *mapping, __u64 first,
+                      __u64 last);
+int ll_file_mmap(struct file * file, struct vm_area_struct * vma);
+struct ll_lock_tree_node * ll_node_from_inode(struct inode *inode, __u64 start,
+                                              __u64 end, ldlm_mode_t mode);
+int ll_tree_lock(struct ll_lock_tree *tree,
+                 struct ll_lock_tree_node *first_node, struct inode *inode,
+                 const char *buf, size_t count, int ast_flags);
+int ll_tree_unlock(struct ll_lock_tree *tree, struct inode *inode);
+
+int ll_get_fid(struct obd_export *exp, struct lustre_id *idp,
+               char *filename, struct lustre_id *ret);
+
+/* generic */
+#define LL_SBI_NOLCK           0x1
+#define LL_SBI_READAHEAD       0x2
+#define LL_SUPER_MAGIC         0x0BD00BD0
+#define LL_MAX_BLKSIZE         (4UL * 1024 * 1024)
+
+#if  (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
+#define    ll_s2sbi(sb)        ((struct ll_sb_info *)((sb)->s_fs_info))
+#define    ll_set_sbi(sb, sbi) ((sb)->s_fs_info = sbi)
+static inline __u64 ll_ts2u64(struct timespec *time)
+{
+        __u64 t = time->tv_sec;
+        return t;
+}
+#else  /* 2.4 here */
+#define    ll_s2sbi(sb)        ((struct ll_sb_info *)((sb)->u.generic_sbp))
+#define    ll_set_sbi(sb, sbi) ((sb)->u.generic_sbp = sbi)
+static inline __u64 ll_ts2u64(time_t *time)
+{
+        return *time;
+}
+#endif
+
+/* don't need an addref as the sb_info should be holding one */
+static inline struct obd_export *ll_s2dtexp(struct super_block *sb)
+{
+        return ll_s2sbi(sb)->ll_dt_exp;
+}
+
+/* don't need an addref as the sb_info should be holding one */
+static inline struct obd_export *ll_s2mdexp(struct super_block *sb)
+{
+        return ll_s2sbi(sb)->ll_md_exp;
+}
+
+static inline struct client_obd *sbi2md(struct ll_sb_info *sbi)
+{
+        struct obd_device *obd = sbi->ll_md_exp->exp_obd;
+        if (obd == NULL)
+                LBUG();
+        return &obd->u.cli;
+}
+
+// FIXME: replace the name of this with LL_SB to conform to kernel stuff
+static inline struct ll_sb_info *ll_i2sbi(struct inode *inode)
+{
+        return ll_s2sbi(inode->i_sb);
+}
+
+static inline struct obd_export *ll_i2dtexp(struct inode *inode)
+{
+        return ll_s2dtexp(inode->i_sb);
+}
+
+static inline struct obd_export *ll_i2mdexp(struct inode *inode)
+{
+        return ll_s2mdexp(inode->i_sb);
+}
+
+static inline int ll_mds_max_easize(struct super_block *sb)
+{
+        return sbi2md(ll_s2sbi(sb))->cl_max_mds_easize;
+}
+
+static inline __u64 ll_file_maxbytes(struct inode *inode)
+{
+        return ll_i2info(inode)->lli_maxbytes;
+}
+
+static inline void
+ll_inode2id(struct lustre_id *id, struct inode *inode)
+{
+        struct lustre_id *lid = &ll_i2info(inode)->lli_id;
+
+        mdc_pack_id(id, inode->i_ino, inode->i_generation,
+                    (inode->i_mode & S_IFMT), id_group(lid),
+                    id_fid(lid));
+}
+
+static inline void 
+ll_prepare_mdc_data(struct mdc_op_data *data, struct inode *i1,
+                    struct inode *i2, const char *name, int namelen,
+                    int mode)
+{
+        LASSERT(i1);
+        ll_inode2id(&data->id1, i1);
+
+        /* it could be directory with mea */
+        data->mea1 = ll_i2info(i1)->lli_mea;
+
+        if (i2) {
+                ll_inode2id(&data->id2, i2);
+                data->mea2 = ll_i2info(i2)->lli_mea;
+        }
+
+       data->valid = 0;
+        data->name = name;
+        data->namelen = namelen;
+        data->create_mode = mode;
+        data->mod_time = LTIME_S(CURRENT_TIME);
+}
+
+#if 0
+/* 
+ * this was needed for catching correct calling place of ll_intent_alloc() with
+ * missed ll_intent_free() causing memory leak. --umka
+ */
+#define ll_intent_alloc(it)                                             \
+        ({                                                              \
+                int err;                                                \
+                OBD_SLAB_ALLOC((it)->d.fs_data, ll_intent_slab, SLAB_KERNEL, \
+                               sizeof(struct lustre_intent_data));      \
+                if (!(it)->d.fs_data) {                                 \
+                        err = -ENOMEM;                                  \
+                } else {                                                \
+                        err = 0;                                        \
+                }                                                       \
+                (it)->it_op_release = ll_intent_release;                \
+                err;                                                    \
+        })
+
+#define ll_intent_free(it)                                      \
+        do {                                                    \
+                if ((it)->d.fs_data) {                                  \
+                        OBD_SLAB_FREE((it)->d.fs_data, ll_intent_slab,  \
+                                      sizeof(struct lustre_intent_data)); \
+                        (it)->d.fs_data = NULL;                         \
+                }                                                       \
+        } while (0)
+#endif
 
 #endif /* LLITE_INTERNAL_H */