1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 * lustre/lvfs/lvfs_linux.c
38 * Author: Andreas Dilger <adilger@clusterfs.com>
42 # define EXPORT_SYMTAB
45 #define DEBUG_SUBSYSTEM S_FILTER
47 #include <linux/version.h>
49 #include <asm/unistd.h>
50 #include <linux/slab.h>
51 #include <linux/pagemap.h>
52 #include <linux/quotaops.h>
53 #include <linux/version.h>
54 #include <libcfs/libcfs.h>
55 #include <lustre_fsfilt.h>
57 #include <linux/module.h>
58 #include <linux/init.h>
59 #include <linux/lustre_compat25.h>
61 #include "lvfs_internal.h"
64 #include <lustre_lib.h>
65 #include <lustre_quota.h>
67 __u64 obd_max_pages = 0;
68 __u64 obd_max_alloc = 0;
69 struct lprocfs_stats *obd_memory = NULL;
70 spinlock_t obd_updatemax_lock = SPIN_LOCK_UNLOCKED;
71 /* refine later and change to seqlock or simlar from libcfs */
73 /* Debugging check only needed during development */
75 # define ASSERT_CTXT_MAGIC(magic) LASSERT((magic) == OBD_RUN_CTXT_MAGIC)
76 # define ASSERT_NOT_KERNEL_CTXT(msg) LASSERTF(!segment_eq(get_fs(), get_ds()),\
78 # define ASSERT_KERNEL_CTXT(msg) LASSERTF(segment_eq(get_fs(), get_ds()), msg)
80 # define ASSERT_CTXT_MAGIC(magic) do {} while(0)
81 # define ASSERT_NOT_KERNEL_CTXT(msg) do {} while(0)
82 # define ASSERT_KERNEL_CTXT(msg) do {} while(0)
85 static void push_group_info(struct lvfs_run_ctxt *save,
86 struct group_info *ginfo)
89 save->ngroups = current_ngroups;
93 save->group_info = current->group_info;
94 current->group_info = ginfo;
99 static void pop_group_info(struct lvfs_run_ctxt *save,
100 struct group_info *ginfo)
103 current_ngroups = save->ngroups;
106 current->group_info = save->group_info;
107 task_unlock(current);
111 /* push / pop to root of obd store */
112 void push_ctxt(struct lvfs_run_ctxt *save, struct lvfs_run_ctxt *new_ctx,
113 struct lvfs_ucred *uc)
115 //ASSERT_NOT_KERNEL_CTXT("already in kernel context!\n");
116 ASSERT_CTXT_MAGIC(new_ctx->magic);
117 OBD_SET_CTXT_MAGIC(save);
120 LASSERT(atomic_read(¤t->fs->pwd->d_count));
121 LASSERT(atomic_read(&new_ctx->pwd->d_count));
122 save->pwd = dget(current->fs->pwd);
123 save->pwdmnt = mntget(current->fs->pwdmnt);
124 save->luc.luc_umask = current->fs->umask;
125 save->ngroups = current->group_info->ngroups;
128 LASSERT(save->pwdmnt);
129 LASSERT(new_ctx->pwd);
130 LASSERT(new_ctx->pwdmnt);
133 save->luc.luc_uid = current->uid;
134 save->luc.luc_gid = current->gid;
135 save->luc.luc_fsuid = current->fsuid;
136 save->luc.luc_fsgid = current->fsgid;
137 save->luc.luc_cap = current->cap_effective;
139 current->uid = uc->luc_uid;
140 current->gid = uc->luc_gid;
141 current->fsuid = uc->luc_fsuid;
142 current->fsgid = uc->luc_fsgid;
143 current->cap_effective = uc->luc_cap;
145 push_group_info(save,
147 uc->luc_identity ? uc->luc_identity->mi_ginfo :
150 current->fs->umask = 0; /* umask already applied on client */
152 ll_set_fs_pwd(current->fs, new_ctx->pwdmnt, new_ctx->pwd);
154 EXPORT_SYMBOL(push_ctxt);
156 void pop_ctxt(struct lvfs_run_ctxt *saved, struct lvfs_run_ctxt *new_ctx,
157 struct lvfs_ucred *uc)
159 ASSERT_CTXT_MAGIC(saved->magic);
160 ASSERT_KERNEL_CTXT("popping non-kernel context!\n");
162 LASSERTF(current->fs->pwd == new_ctx->pwd, "%p != %p\n",
163 current->fs->pwd, new_ctx->pwd);
164 LASSERTF(current->fs->pwdmnt == new_ctx->pwdmnt, "%p != %p\n",
165 current->fs->pwdmnt, new_ctx->pwdmnt);
168 ll_set_fs_pwd(current->fs, saved->pwdmnt, saved->pwd);
171 mntput(saved->pwdmnt);
172 current->fs->umask = saved->luc.luc_umask;
174 current->uid = saved->luc.luc_uid;
175 current->gid = saved->luc.luc_gid;
176 current->fsuid = saved->luc.luc_fsuid;
177 current->fsgid = saved->luc.luc_fsgid;
178 current->cap_effective = saved->luc.luc_cap;
179 pop_group_info(saved,
181 uc->luc_identity ? uc->luc_identity->mi_ginfo :
185 EXPORT_SYMBOL(pop_ctxt);
187 /* utility to make a file */
188 struct dentry *simple_mknod(struct dentry *dir, char *name, int mode, int fix)
190 struct dentry *dchild;
194 // ASSERT_KERNEL_CTXT("kernel doing mknod outside kernel context\n");
195 CDEBUG(D_INODE, "creating file %.*s\n", (int)strlen(name), name);
197 dchild = ll_lookup_one_len(name, dir, strlen(name));
199 GOTO(out_up, dchild);
201 if (dchild->d_inode) {
202 int old_mode = dchild->d_inode->i_mode;
203 if (!S_ISREG(old_mode))
204 GOTO(out_err, err = -EEXIST);
206 /* Fixup file permissions if necessary */
207 if (fix && (old_mode & S_IALLUGO) != (mode & S_IALLUGO)) {
208 CWARN("fixing permissions on %s from %o to %o\n",
209 name, old_mode, mode);
210 dchild->d_inode->i_mode = (mode & S_IALLUGO) |
211 (old_mode & ~S_IALLUGO);
212 mark_inode_dirty(dchild->d_inode);
214 GOTO(out_up, dchild);
217 err = ll_vfs_create(dir->d_inode, dchild, (mode & ~S_IFMT) | S_IFREG,
226 dchild = ERR_PTR(err);
230 EXPORT_SYMBOL(simple_mknod);
232 /* utility to make a directory */
233 struct dentry *simple_mkdir(struct dentry *dir, struct vfsmount *mnt,
234 const char *name, int mode, int fix)
236 struct dentry *dchild;
240 // ASSERT_KERNEL_CTXT("kernel doing mkdir outside kernel context\n");
241 CDEBUG(D_INODE, "creating directory %.*s\n", (int)strlen(name), name);
242 dchild = ll_lookup_one_len(name, dir, strlen(name));
244 GOTO(out_up, dchild);
246 if (dchild->d_inode) {
247 int old_mode = dchild->d_inode->i_mode;
248 if (!S_ISDIR(old_mode)) {
249 CERROR("found %s (%lu/%u) is mode %o\n", name,
250 dchild->d_inode->i_ino,
251 dchild->d_inode->i_generation, old_mode);
252 GOTO(out_err, err = -ENOTDIR);
255 /* Fixup directory permissions if necessary */
256 if (fix && (old_mode & S_IALLUGO) != (mode & S_IALLUGO)) {
258 "fixing permissions on %s from %o to %o\n",
259 name, old_mode, mode);
260 dchild->d_inode->i_mode = (mode & S_IALLUGO) |
261 (old_mode & ~S_IALLUGO);
262 mark_inode_dirty(dchild->d_inode);
264 GOTO(out_up, dchild);
267 err = ll_vfs_mkdir(dir->d_inode, dchild, mnt, mode);
275 dchild = ERR_PTR(err);
279 EXPORT_SYMBOL(simple_mkdir);
281 /* utility to rename a file */
282 int lustre_rename(struct dentry *dir, struct vfsmount *mnt,
283 char *oldname, char *newname)
285 struct dentry *dchild_old, *dchild_new;
289 ASSERT_KERNEL_CTXT("kernel doing rename outside kernel context\n");
290 CDEBUG(D_INODE, "renaming file %.*s to %.*s\n",
291 (int)strlen(oldname), oldname, (int)strlen(newname), newname);
293 dchild_old = ll_lookup_one_len(oldname, dir, strlen(oldname));
294 if (IS_ERR(dchild_old))
295 RETURN(PTR_ERR(dchild_old));
297 if (!dchild_old->d_inode)
298 GOTO(put_old, err = -ENOENT);
300 dchild_new = ll_lookup_one_len(newname, dir, strlen(newname));
301 if (IS_ERR(dchild_new))
302 GOTO(put_old, err = PTR_ERR(dchild_new));
304 err = ll_vfs_rename(dir->d_inode, dchild_old, mnt,
305 dir->d_inode, dchild_new, mnt);
312 EXPORT_SYMBOL(lustre_rename);
315 * Read a file from within kernel context. Prior to calling this
316 * function we should already have done a push_ctxt().
318 int lustre_fread(struct file *file, void *buf, int len, loff_t *off)
320 ASSERT_KERNEL_CTXT("kernel doing read outside kernel context\n");
321 if (!file || !file->f_op || !file->f_op->read || !off)
324 return file->f_op->read(file, buf, len, off);
326 EXPORT_SYMBOL(lustre_fread);
329 * Write a file from within kernel context. Prior to calling this
330 * function we should already have done a push_ctxt().
332 int lustre_fwrite(struct file *file, const void *buf, int len, loff_t *off)
335 ASSERT_KERNEL_CTXT("kernel doing write outside kernel context\n");
343 if (!file->f_op->write)
346 RETURN(file->f_op->write(file, buf, len, off));
348 EXPORT_SYMBOL(lustre_fwrite);
351 * Sync a file from within kernel context. Prior to calling this
352 * function we should already have done a push_ctxt().
354 int lustre_fsync(struct file *file)
357 ASSERT_KERNEL_CTXT("kernel doing sync outside kernel context\n");
358 if (!file || !file->f_op || !file->f_op->fsync)
361 RETURN(file->f_op->fsync(file, file->f_dentry, 0));
363 EXPORT_SYMBOL(lustre_fsync);
365 struct l_file *l_dentry_open(struct lvfs_run_ctxt *ctxt, struct l_dentry *de,
368 mntget(ctxt->pwdmnt);
369 return dentry_open(de, ctxt->pwdmnt, flags);
371 EXPORT_SYMBOL(l_dentry_open);
373 #ifdef HAVE_VFS_READDIR_U64_INO
374 static int l_filldir(void *__buf, const char *name, int namlen, loff_t offset,
375 u64 ino, unsigned int d_type)
377 static int l_filldir(void *__buf, const char *name, int namlen, loff_t offset,
378 ino_t ino, unsigned int d_type)
381 struct l_linux_dirent *dirent;
382 struct l_readdir_callback *buf = (struct l_readdir_callback *)__buf;
384 dirent = buf->lrc_dirent;
386 dirent->lld_off = offset;
388 OBD_ALLOC(dirent, sizeof(*dirent));
393 list_add_tail(&dirent->lld_list, buf->lrc_list);
395 buf->lrc_dirent = dirent;
396 dirent->lld_ino = ino;
397 LASSERT(sizeof(dirent->lld_name) >= namlen + 1);
398 memcpy(dirent->lld_name, name, namlen);
403 long l_readdir(struct file *file, struct list_head *dentry_list)
405 struct l_linux_dirent *lastdirent;
406 struct l_readdir_callback buf;
409 buf.lrc_dirent = NULL;
410 buf.lrc_list = dentry_list;
412 error = vfs_readdir(file, l_filldir, &buf);
416 lastdirent = buf.lrc_dirent;
418 lastdirent->lld_off = file->f_pos;
422 EXPORT_SYMBOL(l_readdir);
424 int l_notify_change(struct vfsmount *mnt, struct dentry *dchild,
425 struct iattr *newattrs)
429 LOCK_INODE_MUTEX(dchild->d_inode);
430 #ifdef HAVE_SECURITY_PLUG
431 rc = notify_change(dchild, mnt, newattrs);
433 rc = notify_change(dchild, newattrs);
435 UNLOCK_INODE_MUTEX(dchild->d_inode);
438 EXPORT_SYMBOL(l_notify_change);
440 /* utility to truncate a file */
441 int simple_truncate(struct dentry *dir, struct vfsmount *mnt,
442 char *name, loff_t length)
444 struct dentry *dchild;
445 struct iattr newattrs;
449 CDEBUG(D_INODE, "truncating file %.*s to %lld\n", (int)strlen(name),
450 name, (long long)length);
451 dchild = ll_lookup_one_len(name, dir, strlen(name));
453 GOTO(out, err = PTR_ERR(dchild));
455 if (dchild->d_inode) {
456 int old_mode = dchild->d_inode->i_mode;
457 if (S_ISDIR(old_mode)) {
458 CERROR("found %s (%lu/%u) is mode %o\n", name,
459 dchild->d_inode->i_ino,
460 dchild->d_inode->i_generation, old_mode);
461 GOTO(out_dput, err = -EISDIR);
464 newattrs.ia_size = length;
465 newattrs.ia_valid = ATTR_SIZE;
466 err = l_notify_change(mnt, dchild, &newattrs);
474 EXPORT_SYMBOL(simple_truncate);
476 #ifdef LUSTRE_KERNEL_VERSION
477 #ifndef HAVE_CLEAR_RDONLY_ON_PUT
478 #error rdonly patchset must be updated [cfs bz11248]
480 void dev_set_rdonly(lvfs_sbdev_type dev);
481 int dev_check_rdonly(lvfs_sbdev_type dev);
483 void __lvfs_set_rdonly(lvfs_sbdev_type dev, lvfs_sbdev_type jdev)
485 if (jdev && (jdev != dev)) {
486 CDEBUG(D_IOCTL | D_HA, "set journal dev %lx rdonly\n",
488 dev_set_rdonly(jdev);
490 CDEBUG(D_IOCTL | D_HA, "set dev %lx rdonly\n", (long)dev);
494 int lvfs_check_rdonly(lvfs_sbdev_type dev)
496 return dev_check_rdonly(dev);
499 EXPORT_SYMBOL(__lvfs_set_rdonly);
500 EXPORT_SYMBOL(lvfs_check_rdonly);
502 int lvfs_check_io_health(struct obd_device *obd, struct file *file)
504 char *write_page = NULL;
509 OBD_ALLOC(write_page, CFS_PAGE_SIZE);
513 rc = fsfilt_write_record(obd, file, write_page, CFS_PAGE_SIZE, &offset, 1);
515 OBD_FREE(write_page, CFS_PAGE_SIZE);
517 CDEBUG(D_INFO, "write 1 page synchronously for checking io rc %d\n",rc);
520 EXPORT_SYMBOL(lvfs_check_io_health);
521 #endif /* LUSTRE_KERNEL_VERSION */
523 void obd_update_maxusage()
527 max1 = obd_pages_sum();
528 max2 = obd_memory_sum();
530 spin_lock(&obd_updatemax_lock);
531 if (max1 > obd_max_pages)
532 obd_max_pages = max1;
533 if (max2 > obd_max_alloc)
534 obd_max_alloc = max2;
535 spin_unlock(&obd_updatemax_lock);
539 __u64 obd_memory_max(void)
543 spin_lock(&obd_updatemax_lock);
545 spin_unlock(&obd_updatemax_lock);
550 __u64 obd_pages_max(void)
554 spin_lock(&obd_updatemax_lock);
556 spin_unlock(&obd_updatemax_lock);
561 EXPORT_SYMBOL(obd_update_maxusage);
562 EXPORT_SYMBOL(obd_pages_max);
563 EXPORT_SYMBOL(obd_memory_max);
564 EXPORT_SYMBOL(obd_memory);
567 __s64 lprocfs_read_helper(struct lprocfs_counter *lc,
568 enum lprocfs_fields_flags field)
576 centry = atomic_read(&lc->lc_cntl.la_entry);
579 case LPROCFS_FIELDS_FLAGS_CONFIG:
582 case LPROCFS_FIELDS_FLAGS_SUM:
585 case LPROCFS_FIELDS_FLAGS_MIN:
588 case LPROCFS_FIELDS_FLAGS_MAX:
591 case LPROCFS_FIELDS_FLAGS_AVG:
592 ret = (lc->lc_max - lc->lc_min)/2;
594 case LPROCFS_FIELDS_FLAGS_SUMSQUARE:
595 ret = lc->lc_sumsquare;
597 case LPROCFS_FIELDS_FLAGS_COUNT:
603 } while (centry != atomic_read(&lc->lc_cntl.la_entry) &&
604 centry != atomic_read(&lc->lc_cntl.la_exit));
608 EXPORT_SYMBOL(lprocfs_read_helper);
611 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
612 MODULE_DESCRIPTION("Lustre VFS Filesystem Helper v0.1");
613 MODULE_LICENSE("GPL");