#include <lustre_ver.h>
#include <lustre_disk.h> /* for s2sbi */
#include <lustre_eacl.h>
+#include <lustre_linkea.h>
/* for struct cl_lock_descr and struct cl_io */
#include <cl_object.h>
#include <lclient.h>
#include <lustre_lmv.h>
#include <lustre_mdc.h>
-#include <linux/lustre_intent.h>
+#include <lustre_intent.h>
#include <linux/compat.h>
+#include "range_lock.h"
+
#ifndef FMODE_EXEC
#define FMODE_EXEC 0
#endif
#define LL_DIR_END_OFF 0x7fffffffffffffffULL
#define LL_DIR_END_OFF_32BIT 0x7fffffffUL
+/* 4UL * 1024 * 1024 */
+#define LL_MAX_BLKSIZE_BITS 22
+
#define LL_IT2STR(it) ((it) ? ldlm_it2str((it)->it_op) : "0")
#define LUSTRE_FPRIVATE(file) ((file)->private_data)
struct inode lli_vfs_inode;
/* the most recent timestamps obtained from mds */
- struct ost_lvb lli_lvb;
+ obd_time lli_atime;
+ obd_time lli_mtime;
+ obd_time lli_ctime;
spinlock_t lli_agl_lock;
/* Try to make the d::member and f::member are aligned. Before using
/* for directory */
struct {
/* serialize normal readdir and statahead-readdir. */
- struct mutex d_readdir_mutex;
-
- /* metadata statahead */
- /* since parent-child threads can share the same @file
- * struct, "opendir_key" is the token when dir close for
- * case of parent exit before child -- it is me should
- * cleanup the dir readahead. */
- void *d_opendir_key;
- struct ll_statahead_info *d_sai;
- /* protect statahead stuff. */
- spinlock_t d_sa_lock;
+ struct mutex lli_readdir_mutex;
+
+ /* metadata statahead */
+ /* since parent-child threads can share the same @file
+ * struct, "opendir_key" is the token when dir close for
+ * case of parent exit before child -- it is me should
+ * cleanup the dir readahead. */
+ void *lli_opendir_key;
+ struct ll_statahead_info *lli_sai;
+ /* protect statahead stuff. */
+ spinlock_t lli_sa_lock;
/* "opendir_pid" is the token when lookup/revalid
* -- I am the owner of dir statahead. */
- pid_t d_opendir_pid;
+ pid_t lli_opendir_pid;
+ /* stat will try to access statahead entries or start
+ * statahead if this flag is set, and this flag will be
+ * set upon dir open, and cleared when dir is closed,
+ * statahead hit ratio is too low, or start statahead
+ * thread failed. */
+ unsigned int lli_sa_enabled:1;
+ /* generation for statahead */
+ unsigned int lli_sa_generation;
/* directory stripe information */
- struct lmv_stripe_md *d_lsm_md;
+ struct lmv_stripe_md *lli_lsm_md;
/* striped directory size */
- loff_t d_stripe_size;
+ loff_t lli_stripe_dir_size;
/* striped directory nlink */
- __u64 d_stripe_nlink;
- } d;
-
-#define lli_readdir_mutex u.d.d_readdir_mutex
-#define lli_opendir_key u.d.d_opendir_key
-#define lli_sai u.d.d_sai
-#define lli_sa_lock u.d.d_sa_lock
-#define lli_opendir_pid u.d.d_opendir_pid
-#define lli_lsm_md u.d.d_lsm_md
-#define lli_stripe_dir_size u.d.d_stripe_size
-#define lli_stripe_dir_nlink u.d.d_stripe_nlink
+ __u64 lli_stripe_dir_nlink;
+ };
/* for non-directory */
struct {
- struct mutex f_size_mutex;
- char *f_symlink_name;
- __u64 f_maxbytes;
+ struct mutex lli_size_mutex;
+ char *lli_symlink_name;
+ __u64 lli_maxbytes;
/*
* struct rw_semaphore {
* signed long count; // align d.d_def_acl
* struct list_head wait_list;
* }
*/
- struct rw_semaphore f_trunc_sem;
- struct mutex f_write_mutex;
+ struct rw_semaphore lli_trunc_sem;
+ struct range_lock_tree lli_write_tree;
- struct rw_semaphore f_glimpse_sem;
- cfs_time_t f_glimpse_time;
- struct list_head f_agl_list;
- __u64 f_agl_index;
+ struct rw_semaphore lli_glimpse_sem;
+ cfs_time_t lli_glimpse_time;
+ struct list_head lli_agl_list;
+ __u64 lli_agl_index;
/* for writepage() only to communicate to fsync */
- int f_async_rc;
+ int lli_async_rc;
/*
* whenever a process try to read/write the file, the
* so the read/write statistics for jobid will not be
* accurate if the file is shared by different jobs.
*/
- char f_jobid[JOBSTATS_JOBID_SIZE];
- } f;
-
-#define lli_size_mutex u.f.f_size_mutex
-#define lli_symlink_name u.f.f_symlink_name
-#define lli_maxbytes u.f.f_maxbytes
-#define lli_trunc_sem u.f.f_trunc_sem
-#define lli_write_mutex u.f.f_write_mutex
-#define lli_glimpse_sem u.f.f_glimpse_sem
-#define lli_glimpse_time u.f.f_glimpse_time
-#define lli_agl_list u.f.f_agl_list
-#define lli_agl_index u.f.f_agl_index
-#define lli_async_rc u.f.f_async_rc
-#define lli_jobid u.f.f_jobid
-
- } u;
+ char lli_jobid[LUSTRE_JOBID_SIZE];
+ };
+ };
/* XXX: For following frequent used members, although they maybe special
* used for non-directory object, it is some time-wasting to check
"nolck", \
"checksum", \
"flock", \
- "xattr", \
+ "user_xattr", \
"acl", \
"???", \
"rmt_client", \
"verbose", \
"layout", \
"user_fid2path",\
- "xattr", \
+ "xattr_cache", \
"norootsquash", \
}
struct list_head et_entries[EE_HASHES];
};
+
+/* This is embedded into llite super-blocks to keep track of connect
+ * flags (capabilities) supported by all imports given mount is
+ * connected to. */
+struct lustre_client_ocd {
+ /* This is conjunction of connect_flags across all imports
+ * (LOVs) this mount is connected to. This field is updated by
+ * cl_ocd_update() under ->lco_lock. */
+ __u64 lco_flags;
+ struct mutex lco_lock;
+ struct obd_export *lco_md_exp;
+ struct obd_export *lco_dt_exp;
+};
+
struct ll_sb_info {
struct list_head ll_list;
/* this protects pglist and ra_info. It isn't safe to
int ll_rw_stats_on;
/* metadata stat-ahead */
- unsigned int ll_sa_max; /* max statahead RPCs */
- atomic_t ll_sa_total; /* statahead thread started
+ unsigned int ll_sa_max; /* max statahead RPCs */
+ atomic_t ll_sa_total; /* statahead thread started
* count */
- atomic_t ll_sa_wrong; /* statahead thread stopped for
+ atomic_t ll_sa_wrong; /* statahead thread stopped for
* low hit ratio */
- atomic_t ll_agl_total; /* AGL thread started count */
+ atomic_t ll_sa_running; /* running statahead thread
+ * count */
+ atomic_t ll_agl_total; /* AGL thread started count */
- dev_t ll_sdev_orig; /* save s_dev before assign for
+ dev_t ll_sdev_orig; /* save s_dev before assign for
* clustred nfs */
- struct rmtacl_ctl_table ll_rct;
- struct eacl_table ll_et;
+ struct rmtacl_ctl_table ll_rct;
+ struct eacl_table ll_et;
/* root squash */
- struct root_squash_info ll_squash;
+ struct root_squash_info ll_squash;
};
#define LL_DEFAULT_MAX_RW_CHUNK (32 * 1024 * 1024)
static void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count) {}
#endif
+enum {
+ LPROC_LL_DIRTY_HITS,
+ LPROC_LL_DIRTY_MISSES,
+ LPROC_LL_READ_BYTES,
+ LPROC_LL_WRITE_BYTES,
+ LPROC_LL_BRW_READ,
+ LPROC_LL_BRW_WRITE,
+ LPROC_LL_OSC_READ,
+ LPROC_LL_OSC_WRITE,
+ LPROC_LL_IOCTL,
+ LPROC_LL_OPEN,
+ LPROC_LL_RELEASE,
+ LPROC_LL_MAP,
+ LPROC_LL_LLSEEK,
+ LPROC_LL_FSYNC,
+ LPROC_LL_READDIR,
+ LPROC_LL_SETATTR,
+ LPROC_LL_TRUNC,
+ LPROC_LL_FLOCK,
+ LPROC_LL_GETATTR,
+ LPROC_LL_CREATE,
+ LPROC_LL_LINK,
+ LPROC_LL_UNLINK,
+ LPROC_LL_SYMLINK,
+ LPROC_LL_MKDIR,
+ LPROC_LL_RMDIR,
+ LPROC_LL_MKNOD,
+ LPROC_LL_RENAME,
+ LPROC_LL_STAFS,
+ LPROC_LL_ALLOC_INODE,
+ LPROC_LL_SETXATTR,
+ LPROC_LL_GETXATTR,
+ LPROC_LL_GETXATTR_HITS,
+ LPROC_LL_LISTXATTR,
+ LPROC_LL_REMOVEXATTR,
+ LPROC_LL_INODE_PERM,
+ LPROC_LL_FILE_OPCODES
+};
/* llite/dir.c */
+struct ll_dir_chain {
+};
+
+static inline void ll_dir_chain_init(struct ll_dir_chain *chain)
+{
+}
+
+static inline void ll_dir_chain_fini(struct ll_dir_chain *chain)
+{
+}
+
extern const struct file_operations ll_dir_operations;
extern const struct inode_operations ll_dir_inode_operations;
#ifdef HAVE_DIR_CONTEXT
/* llite/namei.c */
extern const struct inode_operations ll_special_inode_operations;
-int ll_objects_destroy(struct ptlrpc_request *request,
- struct inode *dir);
struct inode *ll_iget(struct super_block *sb, ino_t hash,
struct lustre_md *lic);
int ll_test_inode_by_fid(struct inode *inode, void *opaque);
#else
int ll_fsync(struct file *file, struct dentry *dentry, int data);
#endif
-int ll_merge_lvb(const struct lu_env *env, struct inode *inode);
+int ll_merge_attr(const struct lu_env *env, struct inode *inode);
int ll_fid2path(struct inode *inode, void __user *arg);
int ll_data_version(struct inode *inode, __u64 *data_version, int flags);
int ll_hsm_release(struct inode *inode);
int ll_prep_inode(struct inode **inode, struct ptlrpc_request *req,
struct super_block *, struct lookup_intent *);
void lustre_dump_dentry(struct dentry *, int recur);
-int ll_obd_statfs(struct inode *inode, void *arg);
+int ll_obd_statfs(struct inode *inode, void __user *arg);
int ll_get_max_mdsize(struct ll_sb_info *sbi, int *max_mdsize);
int ll_get_default_mdsize(struct ll_sb_info *sbi, int *default_mdsize);
int ll_get_max_cookiesize(struct ll_sb_info *sbi, int *max_cookiesize);
int ll_get_default_cookiesize(struct ll_sb_info *sbi, int *default_cookiesize);
int ll_process_config(struct lustre_cfg *lcfg);
struct md_op_data *ll_prep_md_op_data(struct md_op_data *op_data,
- struct inode *i1, struct inode *i2,
- const char *name, int namelen,
- int mode, __u32 opc, void *data);
+ struct inode *i1, struct inode *i2,
+ const char *name, size_t namelen,
+ __u32 mode, __u32 opc, void *data);
void ll_finish_md_op_data(struct md_op_data *op_data);
int ll_get_obd_name(struct inode *inode, unsigned int cmd, unsigned long arg);
char *ll_get_fsname(struct super_block *sb, char *buf, int buflen);
void ll_compute_rootsquash_state(struct ll_sb_info *sbi);
+ssize_t ll_copy_user_md(const struct lov_user_md __user *md,
+ struct lov_user_md **kbuf);
+
+/* Compute expected user md size when passing in a md from user space */
+static inline ssize_t ll_lov_user_md_size(const struct lov_user_md *lum)
+{
+ switch (lum->lmm_magic) {
+ case LOV_USER_MAGIC_V1:
+ return sizeof(struct lov_user_md_v1);
+ case LOV_USER_MAGIC_V3:
+ return sizeof(struct lov_user_md_v3);
+ case LOV_USER_MAGIC_SPECIFIC:
+ if (lum->lmm_stripe_count > LOV_MAX_STRIPE_COUNT)
+ return -EINVAL;
+
+ return lov_user_md_size(lum->lmm_stripe_count,
+ LOV_USER_MAGIC_SPECIFIC);
+ }
+
+ return -EINVAL;
+}
/* llite/llite_nfs.c */
extern struct export_operations lustre_export_operations;
struct ll_cl_context vti_io_ctx;
};
+extern struct lu_context_key vvp_key;
+
static inline struct vvp_thread_info *vvp_env_info(const struct lu_env *env)
{
- extern struct lu_context_key vvp_key;
struct vvp_thread_info *info;
info = lu_context_key_get(&env->le_ctx, &vvp_key);
struct vvp_io vs_ios;
};
+extern struct lu_context_key vvp_session_key;
+
static inline struct vvp_session *vvp_env_session(const struct lu_env *env)
{
- extern struct lu_context_key vvp_session_key;
struct vvp_session *ses;
ses = lu_context_key_get(env->le_ses, &vvp_session_key);
/* per inode struct, for dir only */
struct ll_statahead_info {
- struct inode *sai_inode;
- atomic_t sai_refcount; /* when access this struct, hold
+ struct dentry *sai_dentry;
+ atomic_t sai_refcount; /* when access this struct, hold
* refcount */
- unsigned int sai_generation; /* generation for statahead */
- unsigned int sai_max; /* max ahead of lookup */
- __u64 sai_sent; /* stat requests sent count */
- __u64 sai_replied; /* stat requests which received
+ unsigned int sai_max; /* max ahead of lookup */
+ __u64 sai_sent; /* stat requests sent count */
+ __u64 sai_replied; /* stat requests which received
* reply */
- __u64 sai_index; /* index of statahead entry */
- __u64 sai_index_wait; /* index of entry which is the
+ __u64 sai_index; /* index of statahead entry */
+ __u64 sai_index_wait; /* index of entry which is the
* caller is waiting for */
- __u64 sai_hit; /* hit count */
- __u64 sai_miss; /* miss count:
- * for "ls -al" case, it
- * includes hidden dentry miss;
+ __u64 sai_hit; /* hit count */
+ __u64 sai_miss; /* miss count:
+ * for "ls -al" case, includes
+ * hidden dentry miss;
* for "ls -l" case, it does not
* include hidden dentry miss.
* "sai_miss_hidden" is used for
* the later case.
*/
- unsigned int sai_consecutive_miss; /* consecutive miss */
- unsigned int sai_miss_hidden;/* "ls -al", but first dentry
- * is not a hidden one */
- unsigned int sai_skip_hidden;/* skipped hidden dentry count */
- unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
+ unsigned int sai_consecutive_miss; /* consecutive miss */
+ unsigned int sai_miss_hidden;/* "ls -al", but first dentry
+ * is not a hidden one */
+ unsigned int sai_skip_hidden;/* skipped hidden dentry count
+ */
+ unsigned int sai_ls_all:1, /* "ls -al", do stat-ahead for
* hidden entries */
sai_agl_valid:1,/* AGL is valid for the dir */
sai_in_readpage:1;/* statahead is in readdir()*/
wait_queue_head_t sai_waitq; /* stat-ahead wait queue */
struct ptlrpc_thread sai_thread; /* stat-ahead thread */
struct ptlrpc_thread sai_agl_thread; /* AGL thread */
- struct list_head sai_entries; /* entry list */
- struct list_head sai_entries_received; /* entries returned */
- struct list_head sai_entries_stated; /* entries stated */
- struct list_head sai_entries_agl; /* AGL entries to be sent */
+ struct list_head sai_interim_entries; /* entries which got async
+ * stat reply, but not
+ * instantiated */
+ struct list_head sai_entries; /* completed entries */
+ struct list_head sai_agls; /* AGLs to be sent */
struct list_head sai_cache[LL_SA_CACHE_SIZE];
spinlock_t sai_cache_lock[LL_SA_CACHE_SIZE];
atomic_t sai_cache_count; /* entry count in cache */
};
-int do_statahead_enter(struct inode *dir, struct dentry **dentry,
- int only_unplug);
-void ll_stop_statahead(struct inode *dir, void *key);
+int ll_statahead(struct inode *dir, struct dentry **dentry, bool unplug);
+void ll_authorize_statahead(struct inode *dir, void *key);
+void ll_deauthorize_statahead(struct inode *dir, void *key);
static inline int ll_glimpse_size(struct inode *inode)
{
return rc;
}
-static inline void
-ll_statahead_mark(struct inode *dir, struct dentry *dentry)
-{
- struct ll_inode_info *lli = ll_i2info(dir);
- struct ll_statahead_info *sai = lli->lli_sai;
- struct ll_dentry_data *ldd = ll_d2d(dentry);
-
- /* not the same process, don't mark */
- if (lli->lli_opendir_pid != current_pid())
- return;
-
- LASSERT(ldd != NULL);
- if (sai != NULL)
- ldd->lld_sa_generation = sai->sai_generation;
-}
-
-static inline int
-d_need_statahead(struct inode *dir, struct dentry *dentryp)
+/* dentry may statahead when statahead is enabled and current process has opened
+ * parent directory, and this dentry hasn't accessed statahead cache before */
+static inline bool
+dentry_may_statahead(struct inode *dir, struct dentry *dentry)
{
struct ll_inode_info *lli;
struct ll_dentry_data *ldd;
if (ll_i2sbi(dir)->ll_sa_max == 0)
- return -EAGAIN;
+ return false;
lli = ll_i2info(dir);
+
+ /* statahead is not allowed for this dir, there may be three causes:
+ * 1. dir is not opened.
+ * 2. statahead hit ratio is too low.
+ * 3. previous stat started statahead thread failed. */
+ if (!lli->lli_sa_enabled)
+ return false;
+
/* not the same process, don't statahead */
if (lli->lli_opendir_pid != current_pid())
- return -EAGAIN;
+ return false;
- /* statahead has been stopped */
- if (lli->lli_opendir_key == NULL)
- return -EAGAIN;
-
- ldd = ll_d2d(dentryp);
/*
- * When stats a dentry, the system trigger more than once "revalidate"
- * or "lookup", for "getattr", for "getxattr", and maybe for others.
- * Under patchless client mode, the operation intent is not accurate,
- * which maybe misguide the statahead thread. For example:
- * The "revalidate" call for "getattr" and "getxattr" of a dentry maybe
- * have the same operation intent -- "IT_GETATTR".
- * In fact, one dentry should has only one chance to interact with the
- * statahead thread, otherwise the statahead windows will be confused.
+ * When stating a dentry, kernel may trigger 'revalidate' or 'lookup'
+ * multiple times, eg. for 'getattr', 'getxattr' and etc.
+ * For patchless client, lookup intent is not accurate, which may
+ * misguide statahead. For example:
+ * The 'revalidate' call for 'getattr' and 'getxattr' of a dentry will
+ * have the same intent -- IT_GETATTR, while one dentry should access
+ * statahead cache once, otherwise statahead windows is messed up.
* The solution is as following:
- * Assign "lld_sa_generation" with "sai_generation" when a dentry
- * "IT_GETATTR" for the first time, and the subsequent "IT_GETATTR"
- * will bypass interacting with statahead thread for checking:
- * "lld_sa_generation == lli_sai->sai_generation"
+ * Assign 'lld_sa_generation' with 'lli_sa_generation' when a dentry
+ * IT_GETATTR for the first time, and subsequent IT_GETATTR will
+ * bypass interacting with statahead cache by checking
+ * 'lld_sa_generation == lli->lli_sa_generation'.
*/
- if (ldd && lli->lli_sai &&
- ldd->lld_sa_generation == lli->lli_sai->sai_generation)
- return -EAGAIN;
+ ldd = ll_d2d(dentry);
+ if (ldd != NULL && ldd->lld_sa_generation == lli->lli_sa_generation)
+ return false;
- return 1;
-}
-
-static inline int
-ll_statahead_enter(struct inode *dir, struct dentry **dentryp, int only_unplug)
-{
- int ret;
-
- ret = d_need_statahead(dir, *dentryp);
- if (ret <= 0)
- return ret;
-
- return do_statahead_enter(dir, dentryp, only_unplug);
+ return true;
}
/* llite ioctl register support rountine */
#define cl_isize_read(inode) i_size_read(inode)
-static inline int cl_merge_lvb(const struct lu_env *env, struct inode *inode)
-{
- return ll_merge_lvb(env, inode);
-}
-
#define cl_inode_atime(inode) LTIME_S((inode)->i_atime)
#define cl_inode_ctime(inode) LTIME_S((inode)->i_ctime)
#define cl_inode_mtime(inode) LTIME_S((inode)->i_mtime)
int ll_page_sync_io(const struct lu_env *env, struct cl_io *io,
struct cl_page *page, enum cl_req_type crt);
+int ll_getparent(struct file *file, struct getparent __user *arg);
+
#endif /* LLITE_INTERNAL_H */