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 <linux/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>
42 struct ldlm_lock_desc;
44 struct ptlrpc_connection;
47 struct ptlrpc_request;
51 #define LUSTRE_MDS_NAME "mds"
52 #define LUSTRE_MDT_NAME "mdt"
53 #define LUSTRE_MDC_NAME "mdc"
55 struct mds_update_record {
57 struct ll_fid *ur_fid1;
58 struct ll_fid *ur_fid2;
66 struct llog_cookie *ur_logcookies;
67 struct iattr ur_iattr;
68 struct lvfs_ucred ur_uc;
75 #define _ur_fsuid ur_uc.luc_fsuid
76 #define _ur_fsgid ur_uc.luc_fsgid
77 #define _ur_cap ur_uc.luc_cap
78 #define _ur_uid ur_uc.luc_uid
80 /* i_attr_flags holds the open count in the inode in 2.4 */
81 //XXX Alex implement on 2.4 with i_attr_flags and find soln for 2.5 please
82 #if (LINUX_VERSION_CODE >= KERNEL_VERSION(2,5,0))
83 # define mds_inode_oatomic(inode) ((atomic_t *)&(inode)->i_cindex)
85 # define mds_inode_oatomic(inode) ((atomic_t *)&(inode)->i_attr_flags)
88 # define mds_open_orphan_count(inode) \
89 atomic_read(mds_inode_oatomic(inode))
90 # define mds_open_orphan_inc(inode) \
91 atomic_inc(mds_inode_oatomic(inode))
92 # define mds_open_orphan_dec_test(inode) \
93 atomic_dec_and_test(mds_inode_oatomic(inode))
95 #define mds_inode_is_orphan(inode) ((inode)->i_flags & 0x4000000)
96 #define mds_inode_set_orphan(inode) (inode)->i_flags |= 0x4000000
98 #define MDS_LR_SERVER_SIZE 512
100 #define MDS_LR_CLIENT_START 8192
101 #define MDS_LR_CLIENT_SIZE 128
102 #if MDS_LR_CLIENT_START < MDS_LR_SERVER_SIZE
103 #error "Can't have MDS_LR_CLIENT_START < MDS_LR_SERVER_SIZE"
106 #define MDS_CLIENT_SLOTS 17
108 #define MDS_ROCOMPAT_LOVOBJID 0x00000001
109 #define MDS_ROCOMPAT_SUPP (MDS_ROCOMPAT_LOVOBJID)
111 #define MDS_INCOMPAT_SUPP (0)
113 #define REAL_MDS_NUMBER 1
114 #define CACHE_MDS_NUMBER 0
116 /*flags for indicate the record are come from cmobd reint or
118 #define REC_REINT_CREATE 0x0001
120 /* Data stored per server at the head of the last_rcvd file. In le32 order.
121 * Try to keep this the same as fsd_server_data so we might one day merge. */
122 struct mds_server_data {
123 __u8 msd_uuid[40]; /* server UUID */
124 __u64 msd_last_transno; /* last completed transaction ID */
125 __u64 msd_mount_count; /* MDS incarnation number */
127 __u32 msd_feature_compat; /* compatible feature flags */
128 __u32 msd_feature_rocompat;/* read-only compatible feature flags */
129 __u32 msd_feature_incompat;/* incompatible feature flags */
130 __u32 msd_server_size; /* size of server data area */
131 __u32 msd_client_start; /* start of per-client data area */
132 __u16 msd_client_size; /* size of per-client data area */
133 __u16 msd_subdir_count; /* number of subdirectories for objects */
134 __u64 msd_catalog_oid; /* recovery catalog object id */
135 __u32 msd_catalog_ogen; /* recovery catalog inode generation */
136 __u8 msd_peeruuid[40]; /* UUID of LOV/OSC associated with MDS */
137 __u8 msd_padding[MDS_LR_SERVER_SIZE - 140];
140 /* Data stored per client in the last_rcvd file. In le32 order. */
141 struct mds_client_data {
142 __u8 mcd_uuid[40]; /* client UUID */
143 __u64 mcd_last_transno; /* last completed transaction ID */
144 __u64 mcd_last_xid; /* xid for the last transaction */
145 __u32 mcd_last_result; /* result from last RPC */
146 __u32 mcd_last_data; /* per-op data (disposition for open &c.) */
147 __u8 mcd_padding[MDS_LR_CLIENT_SIZE - 64];
150 /* file data for open files on MDS */
151 struct mds_file_data {
152 struct portals_handle mfd_handle; /* must be first */
153 atomic_t mfd_refcount;
154 struct list_head mfd_list;
157 struct dentry *mfd_dentry;
160 /* group hash table */
161 struct mds_grp_hash_entry {
162 struct list_head ge_hash;
163 struct group_info *ge_group_info;
166 atomic_t ge_refcount;
167 wait_queue_head_t ge_waitq;
168 long ge_acquisition_time;
169 unsigned long ge_acquire_expire;
170 unsigned long ge_expire;
173 #define MDSGRP_HASH_SIZE (128)
174 #define MDSGRP_HASH_INDEX(id) ((id) & (MDSGRP_HASH_SIZE - 1))
175 #define MDSGRP_UPCALL_MAXPATH (1024)
177 struct mds_grp_hash {
178 struct list_head gh_table[MDSGRP_HASH_SIZE];
180 char gh_upcall[MDSGRP_UPCALL_MAXPATH];
182 int gh_acquire_expire;
183 unsigned int gh_allow_setgroups:1;
186 /* mds/mds_reint.c */
187 int mds_reint_rec(struct mds_update_record *r, int offset,
188 struct ptlrpc_request *req, struct lustre_handle *);
192 struct dentry *mds_fid2locked_dentry(struct obd_device *obd, struct ll_fid *fid,
193 struct vfsmount **mnt, int lock_mode,
194 struct lustre_handle *lockh, int *pmode,
195 char *name, int namelen, __u64 lockpart);
196 struct dentry *mds_fid2dentry(struct mds_obd *mds, struct ll_fid *fid,
197 struct vfsmount **mnt);
198 int mds_update_server_data(struct obd_device *, int force_sync);
201 int mds_fs_setup(struct obd_device *obddev, struct vfsmount *mnt);
202 int mds_fs_cleanup(struct obd_device *obddev, int failover);
207 /* mdc/mdc_locks.c */
208 int it_disposition(struct lookup_intent *it, int flag);
209 void it_set_disposition(struct lookup_intent *it, int flag);
210 int it_open_error(int phase, struct lookup_intent *it);
211 int mdc_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data);
212 int mdc_change_cbdata(struct obd_export *exp, struct ll_fid *fid,
213 ldlm_iterator_t it, void *data);
214 int mdc_intent_lock(struct obd_export *exp, struct ll_fid *parent,
215 const char *name, int len, void *lmm, int lmmsize,
216 struct ll_fid *child,
217 struct lookup_intent *, int,
218 struct ptlrpc_request **reqp,
219 ldlm_blocking_callback cb_blocking);
220 int mdc_enqueue(struct obd_export *exp,
222 struct lookup_intent *it,
224 struct mdc_op_data *data,
225 struct lustre_handle *lockh,
228 ldlm_completion_callback cb_completion,
229 ldlm_blocking_callback cb_blocking,
232 /* mdc/mdc_request.c */
233 int mdc_get_secdesc_size(void);
234 void mdc_pack_secdesc(struct ptlrpc_request *req, int size);
235 int mdc_req2lustre_md(struct obd_export *exp_mdc, struct ptlrpc_request *req,
236 unsigned int offset, struct obd_export *exp_osc,
237 struct lustre_md *md);
238 int mdc_getstatus(struct obd_export *exp, struct ll_fid *rootfid);
239 int mdc_getattr(struct obd_export *exp, struct ll_fid *fid,
240 unsigned long valid, unsigned int ea_size,
241 struct ptlrpc_request **request);
242 int mdc_getattr_name(struct obd_export *exp, struct ll_fid *fid,
243 char *filename, int namelen, unsigned long valid,
244 unsigned int ea_size, struct ptlrpc_request **request);
245 int mdc_setattr(struct obd_export *exp, struct mdc_op_data *data,
246 struct iattr *iattr, void *ea, int ealen, void *ea2, int ea2len,
247 struct ptlrpc_request **request);
248 int mdc_open(struct obd_export *exp, obd_id ino, int type, int flags,
249 struct lov_mds_md *lmm, int lmm_size, struct lustre_handle *fh,
250 struct ptlrpc_request **);
251 struct obd_client_handle;
252 int mdc_set_open_replay_data(struct obd_export *exp,
253 struct obd_client_handle *och,
254 struct ptlrpc_request *open_req);
255 int mdc_clear_open_replay_data(struct obd_export *exp,
256 struct obd_client_handle *och);
257 int mdc_close(struct obd_export *, struct obdo *, struct obd_client_handle *,
258 struct ptlrpc_request **);
259 int mdc_readpage(struct obd_export *exp, struct ll_fid *mdc_fid,
260 __u64, struct page *, struct ptlrpc_request **);
261 int mdc_create(struct obd_export *exp, struct mdc_op_data *op_data,
262 const void *data, int datalen, int mode, __u32 uid, __u32 gid,
263 __u64 rdev, struct ptlrpc_request **request);
264 int mdc_unlink(struct obd_export *exp, struct mdc_op_data *data,
265 struct ptlrpc_request **request);
266 int mdc_link(struct obd_export *exp, struct mdc_op_data *data,
267 struct ptlrpc_request **);
268 int mdc_rename(struct obd_export *exp, struct mdc_op_data *data,
269 const char *old, int oldlen, const char *new, int newlen,
270 struct ptlrpc_request **request);
271 int mdc_sync(struct obd_export *exp, struct ll_fid *fid,
272 struct ptlrpc_request **);
273 int mdc_create_client(struct obd_uuid uuid, struct ptlrpc_client *cl);
275 /* Store the generation of a newly-created inode in |req| for replay. */
276 int mdc_store_inode_generation(struct obd_export *exp, struct ptlrpc_request *req,
277 int reqoff, int repoff);
278 int mdc_llog_process(struct obd_export *, char *logname, llog_cb_t, void *data);
279 int mdc_done_writing(struct obd_export *exp, struct obdo *);
280 int mdc_reint(struct ptlrpc_request *request, struct mdc_rpc_lock *rpc_lock,
282 static inline void mdc_pack_fid(struct ll_fid *fid, obd_id ino, __u32 gen,
286 fid->generation = gen;
290 /* ioctls for trying requests */
291 #define IOC_REQUEST_TYPE 'f'
292 #define IOC_REQUEST_MIN_NR 30
294 #define IOC_REQUEST_GETATTR _IOWR('f', 30, long)
295 #define IOC_REQUEST_READPAGE _IOWR('f', 31, long)
296 #define IOC_REQUEST_SETATTR _IOWR('f', 32, long)
297 #define IOC_REQUEST_CREATE _IOWR('f', 33, long)
298 #define IOC_REQUEST_OPEN _IOWR('f', 34, long)
299 #define IOC_REQUEST_CLOSE _IOWR('f', 35, long)
300 #define IOC_REQUEST_MAX_NR 35
302 #define MDS_CHECK_RESENT(req, reconstruct) \
304 if (lustre_msg_get_flags(req->rq_reqmsg) & MSG_RESENT) { \
305 struct mds_client_data *mcd = \
306 req->rq_export->exp_mds_data.med_mcd; \
307 if (mcd->mcd_last_xid == req->rq_xid) { \
311 DEBUG_REQ(D_HA, req, "no reply for RESENT req (have "LPD64")", \
312 mcd->mcd_last_xid); \