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 /* for MDS_CLOSE requests */
134 __u64 mcd_last_close_transno; /* last completed transaction ID */
135 __u64 mcd_last_close_xid; /* xid for the last transaction */
136 __u32 mcd_last_close_result; /* result from last RPC */
137 __u32 mcd_last_close_data; /* per-op data (disposition for open &c.) */
138 __u8 mcd_padding[MDS_LR_CLIENT_SIZE - 88];
141 /* uid/gid mapping support for remote client, some of them
142 * probably consume too much space??
144 #define MDS_IDMAP_HASHSIZE (32)
145 #define MDS_IDMAP_HASHFUNC(id) ((id) & (MDS_IDMAP_HASHSIZE - 1))
147 #define MDS_RMT_UIDMAP_IDX (0)
148 #define MDS_LCL_UIDMAP_IDX (1)
149 #define MDS_RMT_GIDMAP_IDX (2)
150 #define MDS_LCL_GIDMAP_IDX (3)
151 #define MDS_IDMAP_N_HASHES (4)
153 #define MDS_IDMAP_NOTFOUND (-1)
155 struct mds_idmap_entry {
156 struct list_head rmt_hash; /* hashed as rmt_id; */
157 struct list_head lcl_hash; /* hashed as lcl_id; */
159 uid_t rmt_id; /* remote uid/gid */
160 uid_t lcl_id; /* local uid/gid */
163 struct mds_idmap_table {
165 struct list_head mit_idmaps[MDS_IDMAP_N_HASHES][MDS_IDMAP_HASHSIZE];
168 /* file data for open files on MDS */
169 struct mds_file_data {
170 struct portals_handle mfd_handle; /* must be first */
171 atomic_t mfd_refcount;
172 struct list_head mfd_list;
175 struct dentry *mfd_dentry;
178 /* group hash table */
179 struct mds_grp_hash_entry {
180 struct list_head ge_hash;
181 struct group_info *ge_group_info;
184 atomic_t ge_refcount;
185 wait_queue_head_t ge_waitq;
186 long ge_acquisition_time;
187 unsigned long ge_acquire_expire;
188 unsigned long ge_expire;
191 #define MDSGRP_HASH_SIZE (128)
192 #define MDSGRP_HASH_INDEX(id) ((id) & (MDSGRP_HASH_SIZE - 1))
193 #define MDSGRP_UPCALL_MAXPATH (1024)
195 struct mds_grp_hash {
196 struct list_head gh_table[MDSGRP_HASH_SIZE];
198 char gh_upcall[MDSGRP_UPCALL_MAXPATH];
200 int gh_acquire_expire;
201 unsigned int gh_allow_setgroups:1;
207 #define PTL_NETID_ANY ((ptl_netid_t) -1)
209 #define LSD_PERM_SETUID 0x00000001
210 #define LSD_PERM_SETGID 0x00000002
211 #define LSD_PERM_SETGRP 0x00000004
213 struct lsd_permission {
219 /* lustre security descriptor */
220 struct lustre_sec_desc {
221 unsigned int lsd_invalid:1;
224 struct group_info *lsd_ginfo;
226 struct lsd_permission *lsd_perms;
229 struct lsd_cache_entry {
230 struct upcall_cache_entry base;
231 struct lustre_sec_desc lsd;
234 struct lsd_downcall_args {
241 struct lsd_permission *perms;
244 /* mds/mds_reint.c */
245 int mds_reint_rec(struct mds_update_record *r, int offset,
246 struct ptlrpc_request *req, struct lustre_handle *);
249 __u32 mds_lsd_get_perms(struct lustre_sec_desc *lsd, __u32 is_remote,
250 ptl_netid_t netid, ptl_nid_t nid);
255 mds_id2dentry(struct obd_device *obd, struct lustre_id *id,
256 struct vfsmount **mnt);
259 mds_id2locked_dentry(struct obd_device *obd, struct lustre_id *id,
260 struct vfsmount **mnt, int lock_mode,
261 struct lustre_handle *lockh, int *pmode,
262 char *name, int namelen, __u64 lockpart);
263 int mds_update_server_data(struct obd_device *, int force_sync);
264 int mds_update_last_fid(struct obd_device *obd, void *handle,
268 int mds_fs_setup(struct obd_device *obddev, struct vfsmount *mnt);
269 int mds_fs_cleanup(struct obd_device *obddev, int failover);
274 /* mdc/mdc_locks.c */
275 int it_disposition(struct lookup_intent *it, int flag);
276 void it_set_disposition(struct lookup_intent *it, int flag);
277 int it_open_error(int phase, struct lookup_intent *it);
278 int mdc_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data);
279 int mdc_change_cbdata(struct obd_export *exp, struct lustre_id *id,
280 ldlm_iterator_t it, void *data);
281 int mdc_intent_lock(struct obd_export *exp, struct lustre_id *parent,
282 const char *name, int len, void *lmm, int lmmsize,
283 struct lustre_id *child, struct lookup_intent *, int,
284 struct ptlrpc_request **reqp,
285 ldlm_blocking_callback cb_blocking);
286 int mdc_enqueue(struct obd_export *exp,
288 struct lookup_intent *it,
290 struct mdc_op_data *data,
291 struct lustre_handle *lockh,
294 ldlm_completion_callback cb_completion,
295 ldlm_blocking_callback cb_blocking,
298 /* mdc/mdc_request.c */
299 int mdc_req2lustre_md(struct obd_export *exp_lmv, struct ptlrpc_request *req,
300 unsigned int offset, struct obd_export *exp_lov,
301 struct lustre_md *md);
302 int mdc_getstatus(struct obd_export *exp, struct lustre_id *rootid);
303 int mdc_getattr(struct obd_export *exp, struct lustre_id *id,
304 __u64 valid, const char *ea_name, int ea_namelen,
305 unsigned int ea_size, struct ptlrpc_request **request);
306 int mdc_getattr_lock(struct obd_export *exp, struct lustre_id *id,
307 char *filename, int namelen, __u64 valid,
308 unsigned int ea_size, struct ptlrpc_request **request);
309 int mdc_setattr(struct obd_export *exp, struct mdc_op_data *data,
310 struct iattr *iattr, void *ea, int ealen, void *ea2, int ea2len,
311 struct ptlrpc_request **request);
312 int mdc_open(struct obd_export *exp, obd_id ino, int type, int flags,
313 struct lov_mds_md *lmm, int lmm_size, struct lustre_handle *fh,
314 struct ptlrpc_request **);
316 struct obd_client_handle;
318 int mdc_set_open_replay_data(struct obd_export *exp,
319 struct obd_client_handle *och,
320 struct ptlrpc_request *open_req);
321 int mdc_clear_open_replay_data(struct obd_export *exp,
322 struct obd_client_handle *och);
323 int mdc_close(struct obd_export *, struct obdo *, struct obd_client_handle *,
324 struct ptlrpc_request **);
325 int mdc_readpage(struct obd_export *exp, struct lustre_id *id,
326 __u64, struct page *, struct ptlrpc_request **);
327 int mdc_create(struct obd_export *exp, struct mdc_op_data *op_data,
328 const void *data, int datalen, int mode, __u32 uid, __u32 gid,
329 __u64 rdev, struct ptlrpc_request **request);
330 int mdc_unlink(struct obd_export *exp, struct mdc_op_data *data,
331 struct ptlrpc_request **request);
332 int mdc_link(struct obd_export *exp, struct mdc_op_data *data,
333 struct ptlrpc_request **);
334 int mdc_rename(struct obd_export *exp, struct mdc_op_data *data,
335 const char *old, int oldlen, const char *new, int newlen,
336 struct ptlrpc_request **request);
337 int mdc_sync(struct obd_export *exp, struct lustre_id *id,
338 struct ptlrpc_request **);
339 int mdc_create_client(struct obd_uuid uuid, struct ptlrpc_client *cl);
341 /* store the generation of a newly-created inode in |req| for replay. */
342 int mdc_store_inode_generation(struct obd_export *exp,
343 struct ptlrpc_request *req,
344 int reqoff, int repoff);
346 int mdc_llog_process(struct obd_export *, char *, llog_cb_t,
349 int mdc_done_writing(struct obd_export *, struct obdo *);
351 /* ioctls for trying requests */
352 #define IOC_REQUEST_TYPE 'f'
353 #define IOC_REQUEST_MIN_NR 30
355 #define IOC_REQUEST_GETATTR _IOWR('f', 30, long)
356 #define IOC_REQUEST_READPAGE _IOWR('f', 31, long)
357 #define IOC_REQUEST_SETATTR _IOWR('f', 32, long)
358 #define IOC_REQUEST_CREATE _IOWR('f', 33, long)
359 #define IOC_REQUEST_OPEN _IOWR('f', 34, long)
360 #define IOC_REQUEST_CLOSE _IOWR('f', 35, long)
361 #define IOC_REQUEST_MAX_NR 35
363 #define MDS_CHECK_RESENT(req, reconstruct) \
365 if (lustre_msg_get_flags(req->rq_reqmsg) & MSG_RESENT) { \
366 struct mds_client_data *mcd = \
367 req->rq_export->exp_mds_data.med_mcd; \
369 if (le64_to_cpu(mcd->mcd_last_xid) == req->rq_xid) { \
371 RETURN(le32_to_cpu(mcd->mcd_last_result)); \
373 if (le64_to_cpu(mcd->mcd_last_close_xid) == req->rq_xid) { \
375 RETURN(le32_to_cpu(mcd->mcd_last_close_result));\
377 DEBUG_REQ(D_HA, req, "no reply for RESENT req" \
378 "(have "LPD64", and "LPD64")", \
379 mcd->mcd_last_xid, mcd->mcd_last_close_xid); \