X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;ds=sidebyside;f=lustre%2Fllite%2Fllite_internal.h;h=8b6102d0282b5ed97a4d5c786a71d9593fd4d909;hb=8c8662fc548bb106e2efa59485d40fdc018efd9b;hp=2cca168841262d0d5206a943762a80885f26d999;hpb=bfdc6e992503046348306c24942efda3661e5c91;p=fs%2Flustre-release.git diff --git a/lustre/llite/llite_internal.h b/lustre/llite/llite_internal.h index 2cca168..8b6102d 100644 --- a/lustre/llite/llite_internal.h +++ b/lustre/llite/llite_internal.h @@ -10,19 +10,52 @@ #ifndef LLITE_INTERNAL_H #define LLITE_INTERNAL_H +#include + /* 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 */ + /* 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_mdc_exp; - struct obd_export *ll_osc_exp; - struct proc_dir_entry* ll_proc_root; - obd_id ll_rootino; /* number of root inode */ + 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; @@ -33,13 +66,22 @@ struct ll_sb_info { 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 */ + struct lprocfs_stats *ll_stats; /* lprocfs stats counter */ unsigned long ll_pglist_gen; struct list_head ll_pglist; - unsigned long ll_read_ahead_pages; - unsigned long ll_max_read_ahead_pages; + struct ll_ra_info ll_ra_info; + + /* 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 */ @@ -47,16 +89,36 @@ struct ll_sb_info { struct list_head ll_mnt_list; struct semaphore ll_gns_sem; + spinlock_t ll_gns_lock; wait_queue_head_t ll_gns_waitq; - struct completion ll_gns_completion; int ll_gns_state; struct timer_list ll_gns_timer; struct list_head ll_gns_sbi_head; + + unsigned long ll_gns_tick; + unsigned long ll_gns_timeout; + struct completion ll_gns_mount_finished; + + /* 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; }; -#define LL_GNS_STATE_IDLE 1100 -#define LL_GNS_STATE_MOUNTING 1101 -#define LL_GNS_STATE_FINISHED 1102 +/* 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; @@ -67,11 +129,12 @@ struct ll_readahead_state { }; extern kmem_cache_t *ll_file_data_slab; +extern kmem_cache_t *ll_intent_slab; struct lustre_handle; struct ll_file_data { - struct obd_client_handle fd_mds_och; struct ll_readahead_state fd_ras; __u32 fd_flags; + int fd_omode; struct lustre_handle fd_cwlockh; unsigned long fd_gid; }; @@ -111,18 +174,27 @@ struct ll_async_page { struct page *llap_page; struct list_head llap_pending_write; /* only trust these if the page lock is providing exclusion */ - int llap_write_queued:1, - llap_defer_uptodate:1; + unsigned llap_write_queued:1, + llap_defer_uptodate:1, + llap_origin:3, + llap_ra_used:1; + struct list_head llap_proc_item; }; -#define LL_CDEBUG_PAGE(mask, page, fmt, arg...) \ - CDEBUG(mask, "page %p map %p ind %lu priv %0lx: " fmt, \ - page, page->mapping, page->index, page->private, ## arg) +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 *osc, char *mdc); + struct super_block *sb, char *lov, + char *lmv); void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi); /* llite/dir.c */ @@ -130,7 +202,8 @@ 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 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 *); @@ -141,64 +214,92 @@ int ll_mdc_blocking_ast(struct ldlm_lock *, struct ldlm_lock_desc *, /* 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); +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; -extern int ll_inode_revalidate_it(struct dentry *, struct lookup_intent *); +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 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, struct ost_lvb *lvb); -int ll_local_open(struct file *file, struct lookup_intent *it); -int ll_mdc_close(struct obd_export *mdc_exp, struct inode *inode, - struct file *file); +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 lookup_intent *it, struct kstat *stat); +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_finish_gns(struct ll_sb_info *sbi); -int fill_page_with_path(struct dentry *, struct vfsmount *, char **pagep); -int ll_dir_process_mount_object(struct dentry *, struct vfsmount *); -int ll_gns_umount_all(struct ll_sb_info *sbi, int timeout); +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); -int ll_gns_start_thread(void); -void ll_gns_stop_thread(void); /* 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, int *flags); +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); @@ -257,6 +358,32 @@ 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 @@ -266,7 +393,6 @@ int ll_close_thread_start(struct ll_close_queue **lcq_ret); #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) -void __d_rehash(struct dentry * entry, int lock); static inline __u64 ll_ts2u64(struct timespec *time) { __u64 t = time->tv_sec; @@ -282,20 +408,20 @@ static inline __u64 ll_ts2u64(time_t *time) #endif /* don't need an addref as the sb_info should be holding one */ -static inline struct obd_export *ll_s2obdexp(struct super_block *sb) +static inline struct obd_export *ll_s2dtexp(struct super_block *sb) { - return ll_s2sbi(sb)->ll_osc_exp; + 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_s2mdcexp(struct super_block *sb) +static inline struct obd_export *ll_s2mdexp(struct super_block *sb) { - return ll_s2sbi(sb)->ll_mdc_exp; + return ll_s2sbi(sb)->ll_md_exp; } -static inline struct client_obd *sbi2mdc(struct ll_sb_info *sbi) +static inline struct client_obd *sbi2md(struct ll_sb_info *sbi) { - struct obd_device *obd = sbi->ll_mdc_exp->exp_obd; + struct obd_device *obd = sbi->ll_md_exp->exp_obd; if (obd == NULL) LBUG(); return &obd->u.cli; @@ -307,19 +433,19 @@ static inline struct ll_sb_info *ll_i2sbi(struct inode *inode) return ll_s2sbi(inode->i_sb); } -static inline struct obd_export *ll_i2obdexp(struct inode *inode) +static inline struct obd_export *ll_i2dtexp(struct inode *inode) { - return ll_s2obdexp(inode->i_sb); + return ll_s2dtexp(inode->i_sb); } -static inline struct obd_export *ll_i2mdcexp(struct inode *inode) +static inline struct obd_export *ll_i2mdexp(struct inode *inode) { - return ll_s2mdcexp(inode->i_sb); + return ll_s2mdexp(inode->i_sb); } static inline int ll_mds_max_easize(struct super_block *sb) { - return sbi2mdc(ll_s2sbi(sb))->cl_max_mds_easize; + return sbi2md(ll_s2sbi(sb))->cl_max_mds_easize; } static inline __u64 ll_file_maxbytes(struct inode *inode) @@ -327,4 +453,37 @@ 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); +} + #endif /* LLITE_INTERNAL_H */