1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * Copyright (C) 2001-2003 Cluster File Systems, Inc. <info@clusterfs.com>
6 * This file is part of Lustre, http://www.lustre.org.
8 * Lustre is free software; you can redistribute it and/or
9 * modify it under the terms of version 2 of the GNU General Public
10 * License as published by the Free Software Foundation.
12 * Lustre is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with Lustre; if not, write to the Free Software
19 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
21 * MDS data structures.
22 * See also lustre_idl.h for wire formats of requests.
30 # include <linux/fs.h>
31 # include <linux/dcache.h>
34 #include <linux/lustre_handles.h>
35 #include <libcfs/kp30.h>
36 #include <linux/lustre_idl.h>
37 #include <linux/lustre_lib.h>
38 #include <linux/lustre_dlm.h>
39 #include <linux/lustre_log.h>
40 #include <linux/lustre_export.h>
41 #include <linux/lustre_ucache.h>
43 struct ldlm_lock_desc;
45 struct ptlrpc_connection;
48 struct ptlrpc_request;
52 #define LUSTRE_MDS_NAME "mds"
53 #define LUSTRE_MDT_NAME "mdt"
54 #define LUSTRE_MDC_NAME "mdc"
56 struct mds_update_record {
58 struct lustre_id *ur_id1;
59 struct lustre_id *ur_id2;
68 int ur_cookielen; /* obsolete? */
69 struct llog_cookie *ur_logcookies; /* obsolete? */
70 struct iattr ur_iattr;
71 struct lvfs_ucred ur_uc;
78 #define ur_fsuid ur_uc.luc_fsuid
79 #define ur_fsgid ur_uc.luc_fsgid
80 #define ur_cap ur_uc.luc_cap
81 #define ur_uid ur_uc.luc_uid
84 #define MDS_LR_SERVER_SIZE 512
86 #define MDS_LR_CLIENT_START 8192
87 #define MDS_LR_CLIENT_SIZE 128
88 #if MDS_LR_CLIENT_START < MDS_LR_SERVER_SIZE
89 #error "Can't have MDS_LR_CLIENT_START < MDS_LR_SERVER_SIZE"
92 #define MDS_CLIENT_SLOTS 17
94 #define MDS_ROCOMPAT_LOVOBJID 0x00000001
95 #define MDS_ROCOMPAT_SUPP (MDS_ROCOMPAT_LOVOBJID)
97 #define MDS_INCOMPAT_SUPP (0)
99 #define MDS_MASTER_OBD 1
100 #define MDS_CACHE_OBD 0
102 /*flags for indicate the record are come from cmobd reint or
104 #define REC_REINT_CREATE 0x0001
106 /* Data stored per server at the head of the last_rcvd file. In le32 order.
107 * Try to keep this the same as fsd_server_data so we might one day merge. */
108 struct mds_server_data {
109 __u8 msd_uuid[40]; /* server UUID */
110 __u64 msd_last_transno; /* last completed transaction ID */
111 __u64 msd_mount_count; /* MDS incarnation number */
113 __u32 msd_feature_compat; /* compatible feature flags */
114 __u32 msd_feature_rocompat;/* read-only compatible feature flags */
115 __u32 msd_feature_incompat;/* incompatible feature flags */
116 __u32 msd_server_size; /* size of server data area */
117 __u32 msd_client_start; /* start of per-client data area */
118 __u16 msd_client_size; /* size of per-client data area */
119 __u16 msd_subdir_count; /* number of subdirectories for objects */
120 __u64 msd_catalog_oid; /* recovery catalog object id */
121 __u32 msd_catalog_ogen; /* recovery catalog inode generation */
122 __u8 msd_peeruuid[40]; /* UUID of LOV/OSC associated with MDS */
123 __u8 msd_padding[MDS_LR_SERVER_SIZE - 140];
126 /* Data stored per client in the last_rcvd file. In le32 order. */
127 struct mds_client_data {
128 __u8 mcd_uuid[40]; /* client UUID */
129 __u64 mcd_last_transno; /* last completed transaction ID */
130 __u64 mcd_last_xid; /* xid for the last transaction */
131 __u32 mcd_last_result; /* result from last RPC */
132 __u32 mcd_last_data; /* per-op data (disposition for open &c.) */
133 __u8 mcd_padding[MDS_LR_CLIENT_SIZE - 64];
136 /* simple uid/gid mapping hash table */
137 struct mds_idmap_item {
138 struct list_head hash;
143 #define MDS_IDMAP_HASHSIZE (32)
144 struct mds_idmap_table {
145 struct list_head uidmap[MDS_IDMAP_HASHSIZE];
146 struct list_head gidmap[MDS_IDMAP_HASHSIZE];
149 /* file data for open files on MDS */
150 struct mds_file_data {
151 struct portals_handle mfd_handle; /* must be first */
152 atomic_t mfd_refcount;
153 struct list_head mfd_list;
156 struct dentry *mfd_dentry;
159 /* group hash table */
160 struct mds_grp_hash_entry {
161 struct list_head ge_hash;
162 struct group_info *ge_group_info;
165 atomic_t ge_refcount;
166 wait_queue_head_t ge_waitq;
167 long ge_acquisition_time;
168 unsigned long ge_acquire_expire;
169 unsigned long ge_expire;
172 #define MDSGRP_HASH_SIZE (128)
173 #define MDSGRP_HASH_INDEX(id) ((id) & (MDSGRP_HASH_SIZE - 1))
174 #define MDSGRP_UPCALL_MAXPATH (1024)
176 struct mds_grp_hash {
177 struct list_head gh_table[MDSGRP_HASH_SIZE];
179 char gh_upcall[MDSGRP_UPCALL_MAXPATH];
181 int gh_acquire_expire;
182 unsigned int gh_allow_setgroups:1;
185 /* lustre security descriptor */
186 struct lustre_sec_desc {
189 struct group_info *lsd_ginfo;
190 unsigned int lsd_allow_setuid:1,
195 struct lsd_cache_entry {
196 struct upcall_cache_entry base;
197 struct lustre_sec_desc lsd;
200 struct lsd_downcall_args {
211 /* mds/mds_reint.c */
212 int mds_reint_rec(struct mds_update_record *r, int offset,
213 struct ptlrpc_request *req, struct lustre_handle *);
218 mds_id2dentry(struct obd_device *obd, struct lustre_id *id,
219 struct vfsmount **mnt);
222 mds_id2locked_dentry(struct obd_device *obd, struct lustre_id *id,
223 struct vfsmount **mnt, int lock_mode,
224 struct lustre_handle *lockh, int *pmode,
225 char *name, int namelen, __u64 lockpart);
226 int mds_update_server_data(struct obd_device *, int force_sync);
227 int mds_update_last_fid(struct obd_device *obd, void *handle,
231 int mds_fs_setup(struct obd_device *obddev, struct vfsmount *mnt);
232 int mds_fs_cleanup(struct obd_device *obddev, int failover);
237 /* mdc/mdc_locks.c */
238 int it_disposition(struct lookup_intent *it, int flag);
239 void it_set_disposition(struct lookup_intent *it, int flag);
240 int it_open_error(int phase, struct lookup_intent *it);
241 int mdc_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data);
242 int mdc_change_cbdata(struct obd_export *exp, struct lustre_id *id,
243 ldlm_iterator_t it, void *data);
244 int mdc_intent_lock(struct obd_export *exp, struct lustre_id *parent,
245 const char *name, int len, void *lmm, int lmmsize,
246 struct lustre_id *child, struct lookup_intent *, int,
247 struct ptlrpc_request **reqp,
248 ldlm_blocking_callback cb_blocking);
249 int mdc_enqueue(struct obd_export *exp,
251 struct lookup_intent *it,
253 struct mdc_op_data *data,
254 struct lustre_handle *lockh,
257 ldlm_completion_callback cb_completion,
258 ldlm_blocking_callback cb_blocking,
261 /* mdc/mdc_request.c */
262 int mdc_get_secdesc_size(void);
263 void mdc_pack_secdesc(struct ptlrpc_request *req, int size);
264 int mdc_req2lustre_md(struct obd_export *exp_lmv, struct ptlrpc_request *req,
265 unsigned int offset, struct obd_export *exp_lov,
266 struct lustre_md *md);
267 int mdc_getstatus(struct obd_export *exp, struct lustre_id *rootid);
268 int mdc_getattr(struct obd_export *exp, struct lustre_id *id,
269 __u64 valid, const char *ea_name, int ea_namelen,
270 unsigned int ea_size, struct ptlrpc_request **request);
271 int mdc_getattr_lock(struct obd_export *exp, struct lustre_id *id,
272 char *filename, int namelen, __u64 valid,
273 unsigned int ea_size, struct ptlrpc_request **request);
274 int mdc_setattr(struct obd_export *exp, struct mdc_op_data *data,
275 struct iattr *iattr, void *ea, int ealen, void *ea2, int ea2len,
276 struct ptlrpc_request **request);
277 int mdc_open(struct obd_export *exp, obd_id ino, int type, int flags,
278 struct lov_mds_md *lmm, int lmm_size, struct lustre_handle *fh,
279 struct ptlrpc_request **);
281 struct obd_client_handle;
283 int mdc_set_open_replay_data(struct obd_export *exp,
284 struct obd_client_handle *och,
285 struct ptlrpc_request *open_req);
286 int mdc_clear_open_replay_data(struct obd_export *exp,
287 struct obd_client_handle *och);
288 int mdc_close(struct obd_export *, struct obdo *, struct obd_client_handle *,
289 struct ptlrpc_request **);
290 int mdc_readpage(struct obd_export *exp, struct lustre_id *id,
291 __u64, struct page *, struct ptlrpc_request **);
292 int mdc_create(struct obd_export *exp, struct mdc_op_data *op_data,
293 const void *data, int datalen, int mode, __u32 uid, __u32 gid,
294 __u64 rdev, struct ptlrpc_request **request);
295 int mdc_unlink(struct obd_export *exp, struct mdc_op_data *data,
296 struct ptlrpc_request **request);
297 int mdc_link(struct obd_export *exp, struct mdc_op_data *data,
298 struct ptlrpc_request **);
299 int mdc_rename(struct obd_export *exp, struct mdc_op_data *data,
300 const char *old, int oldlen, const char *new, int newlen,
301 struct ptlrpc_request **request);
302 int mdc_sync(struct obd_export *exp, struct lustre_id *id,
303 struct ptlrpc_request **);
304 int mdc_create_client(struct obd_uuid uuid, struct ptlrpc_client *cl);
306 /* store the generation of a newly-created inode in |req| for replay. */
307 int mdc_store_inode_generation(struct obd_export *exp,
308 struct ptlrpc_request *req,
309 int reqoff, int repoff);
311 int mdc_llog_process(struct obd_export *, char *, llog_cb_t,
314 int mdc_done_writing(struct obd_export *, struct obdo *);
316 /* ioctls for trying requests */
317 #define IOC_REQUEST_TYPE 'f'
318 #define IOC_REQUEST_MIN_NR 30
320 #define IOC_REQUEST_GETATTR _IOWR('f', 30, long)
321 #define IOC_REQUEST_READPAGE _IOWR('f', 31, long)
322 #define IOC_REQUEST_SETATTR _IOWR('f', 32, long)
323 #define IOC_REQUEST_CREATE _IOWR('f', 33, long)
324 #define IOC_REQUEST_OPEN _IOWR('f', 34, long)
325 #define IOC_REQUEST_CLOSE _IOWR('f', 35, long)
326 #define IOC_REQUEST_MAX_NR 35
328 #define MDS_CHECK_RESENT(req, reconstruct) \
330 if (lustre_msg_get_flags(req->rq_reqmsg) & MSG_RESENT) { \
331 struct mds_client_data *mcd = \
332 req->rq_export->exp_mds_data.med_mcd; \
333 if (mcd->mcd_last_xid == req->rq_xid) { \
335 RETURN(req->rq_repmsg->status); \
337 DEBUG_REQ(D_HA, req, "no reply for RESENT req (have "LPD64")", \
338 mcd->mcd_last_xid); \