#define LLITE_INTERNAL_H
#include <linux/lustre_debug.h>
+#include <linux/lustre_audit.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. */
/* 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;
struct ll_ra_info ll_ra_info;
- unsigned int ll_remote; /* remote client? */
-
+ unsigned int ll_remote; /* remote client? */
+
+ __u64 ll_audit_mask;
/* times spent waiting for locks in each call site. These are
* all protected by the ll_lock */
struct obd_service_time ll_read_stime;
/* mount object entry name */
char ll_gns_oname[PATH_MAX];
+ void *ll_crypto_info;
+
+ /* TODO: to support multi mount for capability */
+ struct list_head ll_capa_list;
+ struct timer_list ll_capa_timer;
};
struct ll_gns_ctl {
#define LL_GNS_UMOUNT (1 << 0)
#define LL_GNS_CHECK (1 << 1)
+/*
+ * per file-descriptor read-ahead data.
+ */
struct ll_readahead_state {
spinlock_t ras_lock;
- unsigned long ras_last_readpage, ras_consecutive;
+ /*
+ * index of the last page that read(2) needed and that wasn't in the
+ * cache. Used by ras_update() to detect seeks.
+ *
+ * XXX nikita: if access seeks into cached region, Lustre doesn't see
+ * this.
+ */
+ unsigned long ras_last_readpage;
+ /*
+ * number of pages read after last read-ahead window reset. As window
+ * is reset on each seek, this is effectively a number of consecutive
+ * accesses. Maybe ->ras_accessed_in_window is better name.
+ *
+ * XXX nikita: window is also reset (by ras_update()) when Lustre
+ * believes that memory pressure evicts read-ahead pages. In that
+ * case, it probably doesn't make sense to expand window to
+ * PTLRPC_MAX_BRW_PAGES on the third access.
+ */
+ unsigned long ras_consecutive;
+ /*
+ * Parameters of current read-ahead window. Handled by
+ * ras_update(). On the initial access to the file or after a seek,
+ * window is reset to 0. After 3 consecutive accesses, window is
+ * expanded to PTLRPC_MAX_BRW_PAGES. Afterwards, window is enlarged by
+ * PTLRPC_MAX_BRW_PAGES chunks up to ->ra_max_pages.
+ */
unsigned long ras_window_start, ras_window_len;
+ /*
+ * Where next read-ahead should start at. This lies within read-ahead
+ * window. Read-ahead window is read in pieces rather than at once
+ * because: 1. lustre limits total number of pages under read-ahead by
+ * ->ra_max_pages (see ll_ra_count_get()), 2. client cannot read pages
+ * not covered by DLM lock.
+ */
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 {
#define LLAP_MAGIC 98764321
struct ll_async_page {
- int llap_magic;
+ 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 */
+ /* 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;
+ /* used to find proper capability */
+ uid_t llap_fsuid;
+
struct list_head llap_proc_item;
};
LLAP__ORIGIN_MAX,
};
+/*
+ * remote ACL stuff
+ */
+#define REMOTE_ACL_HASHSIZE 16
+
+struct remote_acl {
+ struct list_head ra_perm_cache[REMOTE_ACL_HASHSIZE];
+ spinlock_t ra_lock;
+ /* we use one sem per inode, it's kind of coarse: one user must
+ * wait if another user is updating the perm on this inode. but
+ * I guess this is fine is real world usage.
+ */
+ struct semaphore ra_update_sem;
+};
+
+struct lustre_remote_perm {
+ struct list_head lrp_list;
+ uid_t lrp_auth_uid; /* authenticated uid */
+ gid_t lrp_auth_gid; /* authenticated gid */
+ uint16_t lrp_perm; /* permission bits */
+ uint16_t lrp_valid:1, /* lrp_perm is valid */
+ lrp_setuid:1, /* allow setuid */
+ lrp_setgid:1; /* allow setgid */
+ struct list_head lrp_setxid_perms; /* setxid perms list */
+};
+
+struct remote_perm_setxid {
+ struct list_head list; /* permission list */
+ uid_t uid;
+ gid_t gid;
+ uint16_t perm;
+};
+
/* llite/lproc_llite.c */
int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
struct super_block *sb, char *lov,
extern struct inode_operations ll_dir_inode_operations;
/* llite/namei.c */
-int ll_objects_destroy(struct ptlrpc_request *request,
+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);
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);
+struct lpage_data *lpd_cast_private(struct page *page);
+
void ll_ra_accounting(struct page *page, struct address_space *mapping);
void ll_truncate(struct inode *inode);
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 *);
+int ll_setxattr(struct dentry *, const char *, const void *,
+ size_t, int);
+int ll_getxattr(struct dentry *, const char *, void *, size_t);
+int ll_listxattr(struct dentry *, char *, size_t);
+int ll_removexattr(struct dentry *, const char *);
extern int ll_inode_permission(struct inode *, int, struct nameidata *);
+int ll_get_acl(struct inode *, struct posix_acl **acl,
+ struct ptlrpc_request **req);
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 *,
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);
-
+ struct obd_client_handle *och, int dirty);
+int ll_och_fill(struct inode *inode, struct lookup_intent *it,
+ struct obd_client_handle *och);
+int ll_set_audit(struct inode *, __u64);
+int ll_audit_log(struct inode *, audit_op, int);
+
+int ll_getxattr_internal(struct inode *inode, const char *name,
+ void *value, size_t size, __u64 valid);
#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_capa.c */
+void ll_capa_timer_callback(unsigned long unused);
+int ll_capa_start_thread(void);
+void ll_capa_stop_thread(void);
+int ll_set_och_capa(struct inode *inode, struct lookup_intent *it,
+ struct obd_client_handle *och);
+
/* llite/dcache.c */
void ll_intent_drop_lock(struct lookup_intent *);
void ll_intent_release(struct lookup_intent *);
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_options(char *options, char **ost, char **mds, char **gss,
+ char **mds_sec, char **oss_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);
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 ll_fetch_remote_perm(struct inode *inode, struct ptlrpc_request *req,
+ uint16_t *perm);
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);
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);
+int ll_show_options(struct seq_file *m, struct vfsmount *mnt);
+int ll_flush_cred(struct inode *inode);
+
+int ll_remote_acl_permission(struct inode *inode, int mode);
+int ll_remote_acl_update(struct inode *inode, struct mds_remote_perm *perm);
+void ll_inode_invalidate_acl(struct inode *inode);
/* llite/special.c */
extern struct inode_operations ll_special_inode_operations;
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);
+void ll_close_thread_stop(struct ll_close_queue *lcq);
int ll_close_thread_start(struct ll_close_queue **lcq_ret);
typedef struct rb_node rb_node_t;
#endif
-struct ll_lock_tree_node;
+struct ll_lock_tree_node {
+ rb_node_t lt_node;
+ struct list_head lt_locked_item;
+ __u64 lt_oid;
+ ldlm_policy_data_t lt_policy;
+ struct lustre_handle lt_lockh;
+ ldlm_mode_t lt_mode;
+};
+
struct ll_lock_tree {
rb_root_t lt_root;
struct list_head lt_locked_list;
{
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;
return ll_i2info(inode)->lli_maxbytes;
}
+
static inline void
ll_inode2id(struct lustre_id *id, struct inode *inode)
{
id_fid(lid));
}
-static inline void
+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)
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
+struct crypto_helper_ops {
+ int (*init_it_key)(struct inode *inode, struct lookup_intent *it);
+ int (*create_key)(struct inode *dir, mode_t mode, void **key,
+ int* key_size);
+ int (*get_mac)(struct inode *inode, struct iattr *iattr, void*value,
+ int size, void **key, int* key_size);
+ int (*decrypt_key)(struct inode *inode, struct lookup_intent *it);
+ int (*init_inode_key)(struct inode *inode, void *md_key);
+ int (*destroy_key)(struct inode *inode);
+};
+
+/*llite crypto ops for crypto api*/
+struct ll_crypto_info {
+ struct obd_export *ll_gt_exp;
+ struct list_head ll_cops_list;
+ struct crypto_helper_ops *ll_cops;
+ int ll_c_flags;
+};
+#define ll_page2key(page) ((ll_i2info(page->mapping->host))->lli_key_info)
+static inline struct ll_crypto_info* ll_s2crpi(struct super_block *sb)
+{
+ return (struct ll_crypto_info*)ll_s2sbi(sb)->ll_crypto_info;
+}
+static inline struct ll_crypto_info* ll_i2crpi(struct inode *inode)
+{
+ return (struct ll_crypto_info*)ll_i2sbi(inode)->ll_crypto_info;
+}
+
+static inline struct crypto_helper_ops* ll_i2crpops(struct inode *inode)
+{
+ return ll_i2crpi(inode)->ll_cops;
+}
+
+static inline struct crypto_helper_ops* ll_s2crpops(struct super_block *sb)
+{
+ return ll_s2crpi(sb)->ll_cops;
+}
+
+static inline struct obd_export *ll_s2gsexp(struct super_block *sb)
+{
+ struct ll_crypto_info *llci = ll_s2crpi(sb);
+ if (llci)
+ return llci->ll_gt_exp;
+ return NULL;
+}
+static inline struct obd_export *ll_i2gsexp(struct inode *inode)
+{
+ return ll_s2gsexp(inode->i_sb);
+}
+
+static inline
+int ll_crypto_init_it_key(struct inode *inode, struct lookup_intent *it)
+{
+ struct ll_crypto_info *lci = ll_i2crpi(inode);
+ if (lci) {
+ struct crypto_helper_ops *ops = lci->ll_cops;;
+ if (ops && ops->init_it_key)
+ return ops->init_it_key(inode, it);
+ }
+ return 0;
+}
+
+static inline
+int ll_crypto_create_key(struct inode *inode, mode_t mode, void **key,
+ int* key_size)
+{
+ struct ll_crypto_info *lci = ll_i2crpi(inode);
+ if (lci) {
+ struct crypto_helper_ops *ops = lci->ll_cops;;
+ if (ops && ops->create_key)
+ return ops->create_key(inode, mode, key, key_size);
+ }
+ return 0;
+}
+
+static inline
+int ll_crypto_get_mac(struct inode *inode, struct iattr *attr, void *acl,
+ int acl_size, void **mac, int *mac_size)
+{
+ struct ll_crypto_info *lci = ll_i2crpi(inode);
+ if (lci) {
+ struct crypto_helper_ops *ops = lci->ll_cops;
+ if (ops && ops->get_mac)
+ return ops->get_mac(inode, attr, acl, acl_size,
+ mac, mac_size);
+ }
+ return 0;
+}
+
+static inline
+int ll_crypto_decrypt_key(struct inode *inode, struct lookup_intent *it)
+{
+ struct ll_crypto_info *lci = ll_i2crpi(inode);
+ if (lci) {
+ struct crypto_helper_ops *ops = lci->ll_cops;
+ if (ops && ops->decrypt_key)
+ return ops->decrypt_key(inode, it);
+ }
+ return 0;
+}
+
+static inline
+int ll_crypto_init_inode_key(struct inode *inode, void *md_key)
+{
+ struct ll_crypto_info *lci = ll_i2crpi(inode);
+ if (lci) {
+ struct crypto_helper_ops *ops = lci->ll_cops;
+ if (ops && ops->init_inode_key)
+ return ops->init_inode_key(inode, md_key);
+ }
+ return 0;
+}
+
+static inline
+int ll_crypto_destroy_inode_key(struct inode *inode)
+{
+ struct ll_crypto_info *lci = ll_i2crpi(inode);
+ if (lci) {
+ struct crypto_helper_ops *ops = lci->ll_cops;
+ if (ops && ops->destroy_key)
+ return ops->destroy_key(inode);
+ }
+ return 0;
+}
+int lustre_init_crypto(struct super_block *sb, char *gkc,
+ struct obd_connect_data *data, int async);
+int lustre_destroy_crypto(struct super_block *sb);
+int ll_set_sb_gksinfo(struct super_block *sb, char *type);
+/* pass this flag to ll_md_real_close() to send close rpc right away */
+#define FMODE_SYNC 00000010
#endif /* LLITE_INTERNAL_H */