1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * linux/mds/mds_reint.c
6 * Lustre Metadata Server (mds) reintegration routines
8 * Copyright (C) 2002 Cluster File Systems, Inc.
9 * author: Peter Braam <braam@clusterfs.com>
11 * This code is issued under the GNU General Public License.
12 * See the file COPYING in this distribution
16 // XXX - add transaction sequence numbers
20 #include <linux/version.h>
21 #include <linux/module.h>
23 #include <linux/stat.h>
24 #include <linux/locks.h>
25 #include <linux/quotaops.h>
26 #include <asm/unistd.h>
27 #include <asm/uaccess.h>
29 #define DEBUG_SUBSYSTEM S_MDS
31 #include <linux/obd_support.h>
32 #include <linux/obd_class.h>
33 #include <linux/obd.h>
34 #include <linux/lustre_lib.h>
35 #include <linux/lustre_idl.h>
36 #include <linux/lustre_mds.h>
37 #include <linux/obd_class.h>
39 struct mds_client_info *mds_uuid_to_mci(struct mds_obd *mds, __u8 *uuid)
46 list_for_each(p, &mds->mds_client_info) {
47 struct mds_client_info *mci;
49 mci = list_entry(p, struct mds_client_info, mci_list);
50 CDEBUG(D_INFO, "checking client UUID '%s'\n",
51 mci->mci_mcd->mcd_uuid);
52 if (!strncmp(mci->mci_mcd->mcd_uuid, uuid,
53 sizeof(mci->mci_mcd->mcd_uuid)))
56 CDEBUG(D_INFO, "no client UUID found for '%s'\n", uuid);
60 int mds_update_last_rcvd(struct mds_obd *mds, void *handle,
61 struct ptlrpc_request *req)
63 /* get from req->rq_connection-> or req->rq_client */
64 struct mds_client_info *mci;
68 mci = mds_uuid_to_mci(mds, req->rq_connection->c_remote_uuid);
70 CERROR("unable to locate MDS client data for UUID '%s'\n",
71 ptlrpc_req_to_uuid(req));
72 /* This will be a real error once everything is working */
77 off = MDS_LR_CLIENT + mci->mci_off * MDS_LR_SIZE;
79 ++mds->mds_last_rcvd; /* lock this, or make it an LDLM function? */
80 mci->mci_mcd->mcd_last_rcvd = cpu_to_le64(mds->mds_last_rcvd);
81 mci->mci_mcd->mcd_mount_count = cpu_to_le64(mds->mds_mount_count);
82 mci->mci_mcd->mcd_last_xid = cpu_to_le32(req->rq_reqmsg->xid);
84 mds_fs_set_last_rcvd(mds, handle);
85 rc = lustre_fwrite(mds->mds_rcvd_filp, (char *)mci->mci_mcd,
86 sizeof(*mci->mci_mcd), &off);
87 CDEBUG(D_INODE, "wrote trans #%Ld for client '%s' at #%d: rc = %d\n",
88 mds->mds_last_rcvd, mci->mci_mcd->mcd_uuid, mci->mci_off, rc);
89 // store new value and last committed value in req struct
91 if (rc == sizeof(mci->mci_mcd))
99 static int mds_reint_setattr(struct mds_update_record *rec,
100 struct ptlrpc_request *req)
102 struct mds_obd *mds = &req->rq_obd->u.mds;
107 de = mds_fid2dentry(mds, rec->ur_fid1, NULL);
108 if (IS_ERR(de) || OBD_FAIL_CHECK(OBD_FAIL_MDS_REINT_SETATTR)) {
109 GOTO(out_setattr, rc = -ESTALE);
112 CDEBUG(D_INODE, "ino %ld\n", de->d_inode->i_ino);
114 OBD_FAIL_WRITE(OBD_FAIL_MDS_REINT_SETATTR_WRITE,
115 de->d_inode->i_sb->s_dev);
117 handle = mds_fs_start(mds, de->d_inode, MDS_FSOP_SETATTR);
119 GOTO(out_setattr_de, rc = PTR_ERR(handle));
120 rc = mds_fs_setattr(mds, de, handle, &rec->ur_iattr);
123 rc = mds_update_last_rcvd(mds, handle, req);
124 /* FIXME: need to return last_rcvd, last_committed */
128 /* FIXME: keep rc intact */
129 rc = mds_fs_commit(mds, de->d_inode, handle);
137 static int mds_reint_create(struct mds_update_record *rec,
138 struct ptlrpc_request *req)
140 struct dentry *de = NULL;
141 struct mds_obd *mds = &req->rq_obd->u.mds;
142 struct dentry *dchild = NULL;
145 int rc = 0, type = rec->ur_mode & S_IFMT;
148 de = mds_fid2dentry(mds, rec->ur_fid1, NULL);
149 if (IS_ERR(de) || OBD_FAIL_CHECK(OBD_FAIL_MDS_REINT_CREATE)) {
151 GOTO(out_create_de, rc = -ESTALE);
154 CDEBUG(D_INODE, "parent ino %ld\n", dir->i_ino);
157 dchild = lookup_one_len(rec->ur_name, de, rec->ur_namelen - 1);
158 if (IS_ERR(dchild)) {
159 CERROR("child lookup error %ld\n", PTR_ERR(dchild));
162 GOTO(out_create_dchild, rc = -ESTALE);
165 if (dchild->d_inode) {
166 CERROR("child exists (dir %ld, name %s, ino %ld)\n",
167 dir->i_ino, rec->ur_name, dchild->d_inode->i_ino);
169 GOTO(out_create_dchild, rc = -EEXIST);
172 OBD_FAIL_WRITE(OBD_FAIL_MDS_REINT_CREATE_WRITE, dir->i_sb->s_dev);
176 handle = mds_fs_start(mds, dir, MDS_FSOP_CREATE);
178 GOTO(out_create_dchild, PTR_ERR(handle));
179 rc = vfs_create(dir, dchild, rec->ur_mode);
184 handle = mds_fs_start(mds, dir, MDS_FSOP_MKDIR);
186 GOTO(out_create_dchild, PTR_ERR(handle));
187 rc = vfs_mkdir(dir, dchild, rec->ur_mode);
192 handle = mds_fs_start(mds, dir, MDS_FSOP_SYMLINK);
194 GOTO(out_create_dchild, PTR_ERR(handle));
195 rc = vfs_symlink(dir, dchild, rec->ur_tgt);
203 int rdev = rec->ur_id;
204 handle = mds_fs_start(mds, dir, MDS_FSOP_MKNOD);
206 GOTO(out_create_dchild, PTR_ERR(handle));
207 rc = vfs_mknod(dir, dchild, rec->ur_mode, rdev);
212 CERROR("bad file type %d for create of %s\n",type,rec->ur_name);
213 GOTO(out_create_dchild, rc = -EINVAL);
217 CERROR("error during create: %d\n", rc);
219 GOTO(out_create_commit, rc);
222 struct inode *inode = dchild->d_inode;
223 struct mds_body *body;
225 CDEBUG(D_INODE, "created ino %ld\n", dchild->d_inode->i_ino);
226 if (type == S_IFREG) {
227 rc = mds_fs_set_objid(mds, inode, handle, rec->ur_id);
229 CERROR("error %d setting objid for %ld\n",
233 iattr.ia_atime = rec->ur_time;
234 iattr.ia_ctime = rec->ur_time;
235 iattr.ia_mtime = rec->ur_time;
236 iattr.ia_uid = rec->ur_uid;
237 iattr.ia_gid = rec->ur_gid;
238 iattr.ia_valid = ATTR_UID | ATTR_GID | ATTR_ATIME |
239 ATTR_MTIME | ATTR_CTIME;
241 rc = mds_fs_setattr(mds, dchild, handle, &iattr);
242 /* XXX should we abort here in case of error? */
245 rc = mds_update_last_rcvd(mds, handle, req);
247 body = lustre_msg_buf(req->rq_repmsg, 0);
248 body->ino = inode->i_ino;
249 body->generation = inode->i_generation;
250 body->last_rcvd = mds->mds_last_rcvd;
251 body->last_committed = mds->mds_last_committed;
255 /* FIXME: keep rc intact */
256 rc = mds_fs_commit(mds, dir, handle);
266 static int mds_reint_unlink(struct mds_update_record *rec,
267 struct ptlrpc_request *req)
269 struct dentry *de = NULL;
270 struct dentry *dchild = NULL;
271 struct mds_obd *mds = &req->rq_obd->u.mds;
272 struct inode *dir, *inode;
277 de = mds_fid2dentry(mds, rec->ur_fid1, NULL);
278 if (IS_ERR(de) || OBD_FAIL_CHECK(OBD_FAIL_MDS_REINT_UNLINK)) {
280 GOTO(out_unlink, rc = -ESTALE);
283 CDEBUG(D_INODE, "parent ino %ld\n", dir->i_ino);
286 dchild = lookup_one_len(rec->ur_name, de, rec->ur_namelen - 1);
287 if (IS_ERR(dchild)) {
288 CERROR("child lookup error %ld\n", PTR_ERR(dchild));
290 GOTO(out_unlink_de, rc = -ESTALE);
293 inode = dchild->d_inode;
295 CERROR("child doesn't exist (dir %ld, name %s\n",
296 dir->i_ino, rec->ur_name);
298 GOTO(out_unlink_dchild, rc = -ESTALE);
301 if (inode->i_ino != rec->ur_fid2->id) {
302 CERROR("inode and FID ID do not match (%ld != %Ld)\n",
303 inode->i_ino, rec->ur_fid2->id);
305 GOTO(out_unlink_dchild, rc = -ESTALE);
307 if (inode->i_generation != rec->ur_fid2->generation) {
308 CERROR("inode and FID GENERATION do not match (%d != %d)\n",
309 inode->i_generation, rec->ur_fid2->generation);
311 GOTO(out_unlink_dchild, rc = -ESTALE);
314 OBD_FAIL_WRITE(OBD_FAIL_MDS_REINT_UNLINK_WRITE, dir->i_sb->s_dev);
316 switch (dchild->d_inode->i_mode & S_IFMT) {
318 handle = mds_fs_start(mds, dir, MDS_FSOP_RMDIR);
320 GOTO(out_unlink_dchild, rc = PTR_ERR(handle));
321 rc = vfs_rmdir(dir, dchild);
324 handle = mds_fs_start(mds, dir, MDS_FSOP_UNLINK);
326 GOTO(out_unlink_dchild, rc = PTR_ERR(handle));
327 rc = vfs_unlink(dir, dchild);
332 rc = mds_update_last_rcvd(mds, handle, req);
333 /* FIXME: need to return last_rcvd, last_committed */
334 /* FIXME: keep rc intact */
335 rc = mds_fs_commit(mds, dir, handle);
348 static int mds_reint_link(struct mds_update_record *rec,
349 struct ptlrpc_request *req)
351 struct dentry *de_src = NULL;
352 struct dentry *de_tgt_dir = NULL;
353 struct dentry *dchild = NULL;
354 struct mds_obd *mds = &req->rq_obd->u.mds;
359 de_src = mds_fid2dentry(mds, rec->ur_fid1, NULL);
360 if (IS_ERR(de_src) || OBD_FAIL_CHECK(OBD_FAIL_MDS_REINT_LINK)) {
361 GOTO(out_link, rc = -ESTALE);
364 de_tgt_dir = mds_fid2dentry(mds, rec->ur_fid2, NULL);
365 if (IS_ERR(de_tgt_dir)) {
366 GOTO(out_link_de_src, rc = -ESTALE);
369 down(&de_tgt_dir->d_inode->i_sem);
370 dchild = lookup_one_len(rec->ur_name, de_tgt_dir, rec->ur_namelen - 1);
371 if (IS_ERR(dchild)) {
372 CERROR("child lookup error %ld\n", PTR_ERR(dchild));
373 GOTO(out_link_de_tgt_dir, rc = -ESTALE);
376 if (dchild->d_inode) {
377 CERROR("child exists (dir %ld, name %s\n",
378 de_tgt_dir->d_inode->i_ino, rec->ur_name);
379 GOTO(out_link_dchild, rc = -EEXIST);
382 OBD_FAIL_WRITE(OBD_FAIL_MDS_REINT_LINK_WRITE,
383 dchild->d_inode->i_sb->s_dev);
385 handle = mds_fs_start(mds, de_tgt_dir->d_inode, MDS_FSOP_LINK);
387 GOTO(out_link_dchild, rc = PTR_ERR(handle));
389 rc = vfs_link(de_src, de_tgt_dir->d_inode, dchild);
392 rc = mds_update_last_rcvd(mds, handle, req);
394 /* FIXME: need to return last_rcvd, last_committed */
395 /* FIXME: keep rc intact */
396 rc = mds_fs_commit(mds, de_tgt_dir->d_inode, handle);
402 up(&de_tgt_dir->d_inode->i_sem);
411 static int mds_reint_rename(struct mds_update_record *rec,
412 struct ptlrpc_request *req)
414 struct dentry *de_srcdir = NULL;
415 struct dentry *de_tgtdir = NULL;
416 struct dentry *de_old = NULL;
417 struct dentry *de_new = NULL;
418 struct mds_obd *mds = &req->rq_obd->u.mds;
423 de_srcdir = mds_fid2dentry(mds, rec->ur_fid1, NULL);
424 if (IS_ERR(de_srcdir)) {
425 GOTO(out_rename, rc = -ESTALE);
428 de_tgtdir = mds_fid2dentry(mds, rec->ur_fid2, NULL);
429 if (IS_ERR(de_tgtdir)) {
430 GOTO(out_rename_srcdir, rc = -ESTALE);
433 de_old = lookup_one_len(rec->ur_name, de_srcdir, rec->ur_namelen - 1);
434 if (IS_ERR(de_old)) {
435 CERROR("old child lookup error %ld\n", PTR_ERR(de_old));
436 GOTO(out_rename_tgtdir, rc = -ESTALE);
439 de_new = lookup_one_len(rec->ur_tgt, de_tgtdir, rec->ur_tgtlen - 1);
440 if (IS_ERR(de_new)) {
441 CERROR("new child lookup error %ld\n", PTR_ERR(de_new));
442 GOTO(out_rename_deold, rc = -ESTALE);
445 OBD_FAIL_WRITE(OBD_FAIL_MDS_REINT_RENAME_WRITE,
446 de_srcdir->d_inode->i_sb->s_dev);
448 handle = mds_fs_start(mds, de_tgtdir->d_inode, MDS_FSOP_RENAME);
450 GOTO(out_rename_denew, rc = PTR_ERR(handle));
451 rc = vfs_rename(de_srcdir->d_inode, de_old, de_tgtdir->d_inode, de_new);
454 rc = mds_update_last_rcvd(mds, handle, req);
456 /* FIXME: need to return last_rcvd, last_committed */
457 /* FIXME: keep rc intact */
458 rc = mds_fs_commit(mds, de_tgtdir->d_inode, handle);
474 typedef int (*mds_reinter)(struct mds_update_record *, struct ptlrpc_request*);
476 static mds_reinter reinters[REINT_MAX+1] = {
477 [REINT_SETATTR] mds_reint_setattr,
478 [REINT_CREATE] mds_reint_create,
479 [REINT_UNLINK] mds_reint_unlink,
480 [REINT_LINK] mds_reint_link,
481 [REINT_RENAME] mds_reint_rename,
484 int mds_reint_rec(struct mds_update_record *rec, struct ptlrpc_request *req)
486 int rc, size = sizeof(struct mds_body);
488 if (rec->ur_opcode < 1 || rec->ur_opcode > REINT_MAX) {
489 CERROR("opcode %d not valid\n", rec->ur_opcode);
490 rc = req->rq_status = -EINVAL;
494 rc = lustre_pack_msg(1, &size, NULL, &req->rq_replen, &req->rq_repmsg);
496 CERROR("mds: out of memory\n");
497 rc = req->rq_status = -ENOMEM;
501 rc = reinters[rec->ur_opcode](rec, req);