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 OBD_MDS_DEVICENAME "mds"
53 #define OBD_MDT_DEVICENAME "mdt"
54 #define OBD_MDC_DEVICENAME "mdc"
56 struct mds_update_record {
58 struct lustre_id *ur_id1;
59 struct lustre_id *ur_id2;
70 int ur_cookielen; /* obsolete? */
71 struct llog_cookie *ur_logcookies; /* obsolete? */
72 struct iattr ur_iattr;
73 struct lvfs_ucred ur_uc;
80 #define ur_fsuid ur_uc.luc_fsuid
81 #define ur_fsgid ur_uc.luc_fsgid
82 #define ur_cap ur_uc.luc_cap
83 #define ur_uid ur_uc.luc_uid
86 #define MDS_LR_SERVER_SIZE 512
88 #define MDS_LR_CLIENT_START 8192
89 #define MDS_LR_CLIENT_SIZE 128
90 #if MDS_LR_CLIENT_START < MDS_LR_SERVER_SIZE
91 #error "Can't have MDS_LR_CLIENT_START < MDS_LR_SERVER_SIZE"
94 #define MDS_CLIENT_SLOTS 17
96 #define MDS_ROCOMPAT_LOVOBJID 0x00000001
97 #define MDS_ROCOMPAT_SUPP (MDS_ROCOMPAT_LOVOBJID)
99 #define MDS_INCOMPAT_SUPP (0)
101 #define MDS_MASTER_OBD 1
102 #define MDS_CACHE_OBD 0
104 /*flags for indicate the record are come from cmobd reint or
106 #define REC_REINT_CREATE 0x0001
108 /* Data stored per server at the head of the last_rcvd file. In le32 order.
109 * Try to keep this the same as fsd_server_data so we might one day merge. */
110 struct mds_server_data {
111 __u8 msd_uuid[40]; /* server UUID */
112 __u64 msd_last_transno; /* last completed transaction ID */
113 __u64 msd_mount_count; /* MDS incarnation number */
115 __u32 msd_feature_compat; /* compatible feature flags */
116 __u32 msd_feature_rocompat;/* read-only compatible feature flags */
117 __u32 msd_feature_incompat;/* incompatible feature flags */
118 __u32 msd_server_size; /* size of server data area */
119 __u32 msd_client_start; /* start of per-client data area */
120 __u16 msd_client_size; /* size of per-client data area */
121 __u16 msd_subdir_count; /* number of subdirectories for objects */
122 __u64 msd_catalog_oid; /* recovery catalog object id */
123 __u32 msd_catalog_ogen; /* recovery catalog inode generation */
124 __u8 msd_peeruuid[40]; /* UUID of LOV/OSC associated with MDS */
125 __u8 msd_padding[MDS_LR_SERVER_SIZE - 140];
128 /* Data stored per client in the last_rcvd file. In le32 order. */
129 struct mds_client_data {
130 __u8 mcd_uuid[40]; /* client UUID */
131 __u64 mcd_last_transno; /* last completed transaction ID */
132 __u64 mcd_last_xid; /* xid for the last transaction */
133 __u32 mcd_last_result; /* result from last RPC */
134 __u32 mcd_last_data; /* per-op data (disposition for open &c.) */
135 /* for MDS_CLOSE requests */
136 __u64 mcd_last_close_transno; /* last completed transaction ID */
137 __u64 mcd_last_close_xid; /* xid for the last transaction */
138 __u32 mcd_last_close_result; /* result from last RPC */
139 __u32 mcd_last_close_data; /* per-op data (disposition for open &c.) */
140 __u8 mcd_padding[MDS_LR_CLIENT_SIZE - 88];
143 /* uid/gid mapping support for remote client, some of them
144 * probably consume too much space??
146 #define MDS_IDMAP_HASHSIZE (32)
147 #define MDS_IDMAP_HASHFUNC(id) ((id) & (MDS_IDMAP_HASHSIZE - 1))
149 #define MDS_RMT_UIDMAP_IDX (0)
150 #define MDS_LCL_UIDMAP_IDX (1)
151 #define MDS_RMT_GIDMAP_IDX (2)
152 #define MDS_LCL_GIDMAP_IDX (3)
153 #define MDS_IDMAP_N_HASHES (4)
155 #define MDS_IDMAP_NOTFOUND (-1)
157 struct mds_idmap_entry {
158 struct list_head rmt_hash; /* hashed as rmt_id; */
159 struct list_head lcl_hash; /* hashed as lcl_id; */
161 uid_t rmt_id; /* remote uid/gid */
162 uid_t lcl_id; /* local uid/gid */
165 struct mds_idmap_table {
167 struct list_head mit_idmaps[MDS_IDMAP_N_HASHES][MDS_IDMAP_HASHSIZE];
170 /* file data for open files on MDS */
171 struct mds_file_data {
172 struct portals_handle mfd_handle; /* must be first */
173 atomic_t mfd_refcount;
174 struct list_head mfd_list;
177 struct dentry *mfd_dentry;
180 /* group hash table */
181 struct mds_grp_hash_entry {
182 struct list_head ge_hash;
183 struct group_info *ge_group_info;
186 atomic_t ge_refcount;
187 wait_queue_head_t ge_waitq;
188 long ge_acquisition_time;
189 unsigned long ge_acquire_expire;
190 unsigned long ge_expire;
193 #define MDSGRP_HASH_SIZE (128)
194 #define MDSGRP_HASH_INDEX(id) ((id) & (MDSGRP_HASH_SIZE - 1))
195 #define MDSGRP_UPCALL_MAXPATH (1024)
197 struct mds_grp_hash {
198 struct list_head gh_table[MDSGRP_HASH_SIZE];
200 char gh_upcall[MDSGRP_UPCALL_MAXPATH];
202 int gh_acquire_expire;
203 unsigned int gh_allow_setgroups:1;
209 #define PTL_NETID_ANY ((ptl_netid_t) -1)
211 #define LSD_PERM_SETUID 0x00000001
212 #define LSD_PERM_SETGID 0x00000002
213 #define LSD_PERM_SETGRP 0x00000004
215 struct lsd_permission {
221 /* lustre security descriptor */
222 struct lustre_sec_desc {
223 unsigned int lsd_invalid:1;
226 struct group_info *lsd_ginfo;
228 struct lsd_permission *lsd_perms;
231 struct lsd_cache_entry {
232 struct upcall_cache_entry base;
233 struct lustre_sec_desc lsd;
236 struct lsd_downcall_args {
243 struct lsd_permission *perms;
246 /* remote acl upcall */
247 struct rmtacl_upcall_desc {
248 int status; /* helper execution status */
249 int upcall_status; /* error in upcall itself */
250 int get; /* is getfacl */
251 char *cmd; /* cmdline (up) */
252 __u32 cmdlen; /* cmdline length (up) */
253 char *res; /* output (down) */
254 __u32 reslen; /* output length (down) */
255 /* upcall internal use */
260 struct rmtacl_upcall_entry {
261 struct upcall_cache_entry base;
262 struct rmtacl_upcall_desc *desc;
265 struct rmtacl_downcall_args {
267 char *res; /* output text */
268 __u32 reslen; /* output text length */
269 int status; /* helper exit code */
272 /* mds/mds_reint.c */
273 int mds_reint_rec(struct mds_update_record *r, int offset,
274 struct ptlrpc_request *req, struct lustre_handle *);
277 __u32 mds_lsd_get_perms(struct lustre_sec_desc *lsd, __u32 is_remote,
278 ptl_netid_t netid, ptl_nid_t nid);
283 mds_id2dentry(struct obd_device *obd, struct lustre_id *id,
284 struct vfsmount **mnt);
287 mds_id2locked_dentry(struct obd_device *obd, struct lustre_id *id,
288 struct vfsmount **mnt, int lock_mode,
289 struct lustre_handle *lockh, int *pmode,
290 char *name, int namelen, __u64 lockpart);
291 int mds_update_server_data(struct obd_device *, int force_sync);
292 int mds_update_last_fid(struct obd_device *obd, void *handle,
296 int mds_fs_setup(struct obd_device *obddev, struct vfsmount *mnt);
297 int mds_fs_cleanup(struct obd_device *obddev, int failover);
302 /* mdc/mdc_locks.c */
303 int it_disposition(struct lookup_intent *it, int flag);
304 void it_set_disposition(struct lookup_intent *it, int flag);
305 int it_open_error(int phase, struct lookup_intent *it);
306 int mdc_set_lock_data(struct obd_export *exp, __u64 *lockh, void *data);
307 int mdc_change_cbdata(struct obd_export *exp, struct lustre_id *id,
308 ldlm_iterator_t it, void *data);
309 int mdc_intent_lock(struct obd_export *exp, struct lustre_id *parent,
310 const char *name, int len, void *lmm, int lmmsize,
311 struct lustre_id *child, struct lookup_intent *, int,
312 struct ptlrpc_request **reqp,
313 ldlm_blocking_callback cb_blocking);
314 int mdc_enqueue(struct obd_export *exp,
316 struct lookup_intent *it,
318 struct mdc_op_data *data,
319 struct lustre_handle *lockh,
322 ldlm_completion_callback cb_completion,
323 ldlm_blocking_callback cb_blocking,
326 /* mdc/mdc_request.c */
327 int mdc_req2lustre_md(struct obd_export *exp_lmv, struct ptlrpc_request *req,
328 unsigned int offset, struct obd_export *exp_lov,
329 struct lustre_md *md);
330 int mdc_getstatus(struct obd_export *exp, struct lustre_id *rootid);
331 int mdc_getattr(struct obd_export *exp, struct lustre_id *id,
332 __u64 valid, const char *xattr_name,
333 const void *xattr_data, unsigned int xattr_datalen,
334 unsigned int ea_size, struct ptlrpc_request **request);
335 int mdc_getattr_lock(struct obd_export *exp, struct lustre_id *id,
336 char *filename, int namelen, __u64 valid,
337 unsigned int ea_size, struct ptlrpc_request **request);
338 int mdc_setattr(struct obd_export *exp, struct mdc_op_data *data,
339 struct iattr *iattr, void *ea, int ealen, void *ea2, int ea2len,
340 void *ea3, int ea3len, struct ptlrpc_request **request);
341 int mdc_open(struct obd_export *exp, obd_id ino, int type, int flags,
342 struct lov_mds_md *lmm, int lmm_size, struct lustre_handle *fh,
343 struct ptlrpc_request **);
345 struct obd_client_handle;
347 int mdc_set_open_replay_data(struct obd_export *exp,
348 struct obd_client_handle *och,
349 struct ptlrpc_request *open_req);
350 int mdc_clear_open_replay_data(struct obd_export *exp,
351 struct obd_client_handle *och);
352 int mdc_close(struct obd_export *, struct obdo *, struct obd_client_handle *,
353 struct ptlrpc_request **);
354 int mdc_readpage(struct obd_export *exp, struct lustre_id *id,
355 __u64, struct page *, struct ptlrpc_request **);
356 int mdc_create(struct obd_export *exp, struct mdc_op_data *op_data,
357 const void *data, int datalen, int mode, __u32 uid, __u32 gid,
358 __u64 rdev, struct ptlrpc_request **request);
359 int mdc_unlink(struct obd_export *exp, struct mdc_op_data *data,
360 struct ptlrpc_request **request);
361 int mdc_link(struct obd_export *exp, struct mdc_op_data *data,
362 struct ptlrpc_request **);
363 int mdc_rename(struct obd_export *exp, struct mdc_op_data *data,
364 const char *old, int oldlen, const char *new, int newlen,
365 struct ptlrpc_request **request);
366 int mdc_sync(struct obd_export *exp, struct lustre_id *id,
367 struct ptlrpc_request **);
368 int mdc_create_client(struct obd_uuid uuid, struct ptlrpc_client *cl);
370 /* store the generation of a newly-created inode in |req| for replay. */
371 int mdc_store_inode_generation(struct obd_export *exp,
372 struct ptlrpc_request *req,
373 int reqoff, int repoff);
375 int mdc_llog_process(struct obd_export *, char *, llog_cb_t,
378 int mdc_done_writing(struct obd_export *, struct obdo *);
380 /* ioctls for trying requests */
381 #define IOC_REQUEST_TYPE 'f'
382 #define IOC_REQUEST_MIN_NR 30
384 #define IOC_REQUEST_GETATTR _IOWR('f', 30, long)
385 #define IOC_REQUEST_READPAGE _IOWR('f', 31, long)
386 #define IOC_REQUEST_SETATTR _IOWR('f', 32, long)
387 #define IOC_REQUEST_CREATE _IOWR('f', 33, long)
388 #define IOC_REQUEST_OPEN _IOWR('f', 34, long)
389 #define IOC_REQUEST_CLOSE _IOWR('f', 35, long)
390 #define IOC_REQUEST_MAX_NR 35
392 #define MDS_CHECK_RESENT(req, reconstruct) \
394 if (lustre_msg_get_flags(req->rq_reqmsg) & MSG_RESENT) { \
395 struct mds_client_data *mcd = \
396 req->rq_export->exp_mds_data.med_mcd; \
398 if (le64_to_cpu(mcd->mcd_last_xid) == req->rq_xid) { \
400 RETURN(le32_to_cpu(mcd->mcd_last_result)); \
402 if (le64_to_cpu(mcd->mcd_last_close_xid) == req->rq_xid) { \
404 RETURN(le32_to_cpu(mcd->mcd_last_close_result));\
406 DEBUG_REQ(D_HA, req, "no reply for RESENT req" \
407 "(have "LPD64", and "LPD64")", \
408 mcd->mcd_last_xid, mcd->mcd_last_close_xid); \