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 (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
30 * Use is subject to license terms.
32 * Copyright (c) 2011, 2012, Whamcloud, Inc.
35 * This file is part of Lustre, http://www.lustre.org/
36 * Lustre is a trademark of Sun Microsystems, Inc.
38 * lustre/lvfs/lvfs_linux.c
40 * Author: Andreas Dilger <adilger@clusterfs.com>
44 # define EXPORT_SYMTAB
47 #define DEBUG_SUBSYSTEM S_FILTER
49 #include <linux/version.h>
51 #include <asm/unistd.h>
52 #include <linux/slab.h>
53 #include <linux/pagemap.h>
54 #include <linux/quotaops.h>
55 #include <linux/version.h>
56 #include <libcfs/libcfs.h>
57 #include <lustre_fsfilt.h>
59 #include <linux/module.h>
60 #include <linux/init.h>
61 #include <linux/lustre_compat25.h>
63 #include "lvfs_internal.h"
66 #include <lustre_lib.h>
67 #include <lustre_quota.h>
69 __u64 obd_max_pages = 0;
70 __u64 obd_max_alloc = 0;
71 struct lprocfs_stats *obd_memory = NULL;
72 cfs_spinlock_t obd_updatemax_lock = CFS_SPIN_LOCK_UNLOCKED;
73 /* refine later and change to seqlock or simlar from libcfs */
75 /* Debugging check only needed during development */
77 # define ASSERT_CTXT_MAGIC(magic) LASSERT((magic) == OBD_RUN_CTXT_MAGIC)
78 # define ASSERT_NOT_KERNEL_CTXT(msg) LASSERTF(!segment_eq(get_fs(), get_ds()),\
80 # define ASSERT_KERNEL_CTXT(msg) LASSERTF(segment_eq(get_fs(), get_ds()), msg)
82 # define ASSERT_CTXT_MAGIC(magic) do {} while(0)
83 # define ASSERT_NOT_KERNEL_CTXT(msg) do {} while(0)
84 # define ASSERT_KERNEL_CTXT(msg) do {} while(0)
87 static void push_group_info(struct lvfs_run_ctxt *save,
88 struct group_info *ginfo)
91 save->ngroups = current_ngroups;
96 save->group_info = current_cred()->group_info;
97 if ((cred = prepare_creds())) {
98 cred->group_info = ginfo;
101 task_unlock(current);
105 static void pop_group_info(struct lvfs_run_ctxt *save,
106 struct group_info *ginfo)
109 current_ngroups = save->ngroups;
113 if ((cred = prepare_creds())) {
114 cred->group_info = save->group_info;
117 task_unlock(current);
121 /* push / pop to root of obd store */
122 void push_ctxt(struct lvfs_run_ctxt *save, struct lvfs_run_ctxt *new_ctx,
123 struct lvfs_ucred *uc)
125 //ASSERT_NOT_KERNEL_CTXT("already in kernel context!\n");
126 ASSERT_CTXT_MAGIC(new_ctx->magic);
127 OBD_SET_CTXT_MAGIC(save);
130 LASSERT(cfs_atomic_read(&cfs_fs_pwd(current->fs)->d_count));
131 LASSERT(cfs_atomic_read(&new_ctx->pwd->d_count));
132 save->pwd = dget(cfs_fs_pwd(current->fs));
133 save->pwdmnt = mntget(cfs_fs_mnt(current->fs));
134 save->luc.luc_umask = cfs_curproc_umask();
135 save->ngroups = current_cred()->group_info->ngroups;
138 LASSERT(save->pwdmnt);
139 LASSERT(new_ctx->pwd);
140 LASSERT(new_ctx->pwdmnt);
144 save->luc.luc_uid = current_uid();
145 save->luc.luc_gid = current_gid();
146 save->luc.luc_fsuid = current_fsuid();
147 save->luc.luc_fsgid = current_fsgid();
148 save->luc.luc_cap = current_cap();
150 if ((cred = prepare_creds())) {
151 cred->uid = uc->luc_uid;
152 cred->gid = uc->luc_gid;
153 cred->fsuid = uc->luc_fsuid;
154 cred->fsgid = uc->luc_fsgid;
155 cred->cap_effective = uc->luc_cap;
159 push_group_info(save,
161 uc->luc_identity ? uc->luc_identity->mi_ginfo :
164 current->fs->umask = 0; /* umask already applied on client */
166 ll_set_fs_pwd(current->fs, new_ctx->pwdmnt, new_ctx->pwd);
168 EXPORT_SYMBOL(push_ctxt);
170 void pop_ctxt(struct lvfs_run_ctxt *saved, struct lvfs_run_ctxt *new_ctx,
171 struct lvfs_ucred *uc)
173 ASSERT_CTXT_MAGIC(saved->magic);
174 ASSERT_KERNEL_CTXT("popping non-kernel context!\n");
176 LASSERTF(cfs_fs_pwd(current->fs) == new_ctx->pwd, "%p != %p\n",
177 cfs_fs_pwd(current->fs), new_ctx->pwd);
178 LASSERTF(cfs_fs_mnt(current->fs) == new_ctx->pwdmnt, "%p != %p\n",
179 cfs_fs_mnt(current->fs), new_ctx->pwdmnt);
182 ll_set_fs_pwd(current->fs, saved->pwdmnt, saved->pwd);
185 mntput(saved->pwdmnt);
186 current->fs->umask = saved->luc.luc_umask;
189 if ((cred = prepare_creds())) {
190 cred->uid = saved->luc.luc_uid;
191 cred->gid = saved->luc.luc_gid;
192 cred->fsuid = saved->luc.luc_fsuid;
193 cred->fsgid = saved->luc.luc_fsgid;
194 cred->cap_effective = saved->luc.luc_cap;
198 pop_group_info(saved,
200 uc->luc_identity ? uc->luc_identity->mi_ginfo :
204 EXPORT_SYMBOL(pop_ctxt);
206 /* utility to make a file */
207 struct dentry *simple_mknod(struct dentry *dir, char *name, int mode, int fix)
209 struct dentry *dchild;
213 // ASSERT_KERNEL_CTXT("kernel doing mknod outside kernel context\n");
214 CDEBUG(D_INODE, "creating file %.*s\n", (int)strlen(name), name);
216 dchild = ll_lookup_one_len(name, dir, strlen(name));
218 GOTO(out_up, dchild);
220 if (dchild->d_inode) {
221 int old_mode = dchild->d_inode->i_mode;
222 if (!S_ISREG(old_mode))
223 GOTO(out_err, err = -EEXIST);
225 /* Fixup file permissions if necessary */
226 if (fix && (old_mode & S_IALLUGO) != (mode & S_IALLUGO)) {
227 CWARN("fixing permissions on %s from %o to %o\n",
228 name, old_mode, mode);
229 dchild->d_inode->i_mode = (mode & S_IALLUGO) |
230 (old_mode & ~S_IALLUGO);
231 mark_inode_dirty(dchild->d_inode);
233 GOTO(out_up, dchild);
236 err = ll_vfs_create(dir->d_inode, dchild, (mode & ~S_IFMT) | S_IFREG,
245 dchild = ERR_PTR(err);
249 EXPORT_SYMBOL(simple_mknod);
251 /* utility to make a directory */
252 struct dentry *simple_mkdir(struct dentry *dir, struct vfsmount *mnt,
253 const char *name, int mode, int fix)
255 struct dentry *dchild;
259 // ASSERT_KERNEL_CTXT("kernel doing mkdir outside kernel context\n");
260 CDEBUG(D_INODE, "creating directory %.*s\n", (int)strlen(name), name);
261 dchild = ll_lookup_one_len(name, dir, strlen(name));
263 GOTO(out_up, dchild);
265 if (dchild->d_inode) {
266 int old_mode = dchild->d_inode->i_mode;
267 if (!S_ISDIR(old_mode)) {
268 CERROR("found %s (%lu/%u) is mode %o\n", name,
269 dchild->d_inode->i_ino,
270 dchild->d_inode->i_generation, old_mode);
271 GOTO(out_err, err = -ENOTDIR);
274 /* Fixup directory permissions if necessary */
275 if (fix && (old_mode & S_IALLUGO) != (mode & S_IALLUGO)) {
277 "fixing permissions on %s from %o to %o\n",
278 name, old_mode, mode);
279 dchild->d_inode->i_mode = (mode & S_IALLUGO) |
280 (old_mode & ~S_IALLUGO);
281 mark_inode_dirty(dchild->d_inode);
283 GOTO(out_up, dchild);
286 err = ll_vfs_mkdir(dir->d_inode, dchild, mnt, mode);
294 dchild = ERR_PTR(err);
298 EXPORT_SYMBOL(simple_mkdir);
300 /* utility to rename a file */
301 int lustre_rename(struct dentry *dir, struct vfsmount *mnt,
302 char *oldname, char *newname)
304 struct dentry *dchild_old, *dchild_new;
308 ASSERT_KERNEL_CTXT("kernel doing rename outside kernel context\n");
309 CDEBUG(D_INODE, "renaming file %.*s to %.*s\n",
310 (int)strlen(oldname), oldname, (int)strlen(newname), newname);
312 dchild_old = ll_lookup_one_len(oldname, dir, strlen(oldname));
313 if (IS_ERR(dchild_old))
314 RETURN(PTR_ERR(dchild_old));
316 if (!dchild_old->d_inode)
317 GOTO(put_old, err = -ENOENT);
319 dchild_new = ll_lookup_one_len(newname, dir, strlen(newname));
320 if (IS_ERR(dchild_new))
321 GOTO(put_old, err = PTR_ERR(dchild_new));
323 err = ll_vfs_rename(dir->d_inode, dchild_old, mnt,
324 dir->d_inode, dchild_new, mnt);
331 EXPORT_SYMBOL(lustre_rename);
334 * Read a file from within kernel context. Prior to calling this
335 * function we should already have done a push_ctxt().
337 int lustre_fread(struct file *file, void *buf, int len, loff_t *off)
339 ASSERT_KERNEL_CTXT("kernel doing read outside kernel context\n");
340 if (!file || !file->f_op || !file->f_op->read || !off)
343 return file->f_op->read(file, buf, len, off);
345 EXPORT_SYMBOL(lustre_fread);
348 * Write a file from within kernel context. Prior to calling this
349 * function we should already have done a push_ctxt().
351 int lustre_fwrite(struct file *file, const void *buf, int len, loff_t *off)
354 ASSERT_KERNEL_CTXT("kernel doing write outside kernel context\n");
362 if (!file->f_op->write)
365 RETURN(file->f_op->write(file, buf, len, off));
367 EXPORT_SYMBOL(lustre_fwrite);
370 * Sync a file from within kernel context. Prior to calling this
371 * function we should already have done a push_ctxt().
373 int lustre_fsync(struct file *file)
376 ASSERT_KERNEL_CTXT("kernel doing sync outside kernel context\n");
377 if (!file || !file->f_op || !file->f_op->fsync)
380 RETURN(cfs_do_fsync(file, 0));
382 EXPORT_SYMBOL(lustre_fsync);
384 /* Note: dput(dchild) will be called if there is an error */
385 struct l_file *l_dentry_open(struct lvfs_run_ctxt *ctxt, struct l_dentry *de,
388 mntget(ctxt->pwdmnt);
389 return ll_dentry_open(de, ctxt->pwdmnt, flags, current_cred());
391 EXPORT_SYMBOL(l_dentry_open);
393 static int l_filldir(void *__buf, const char *name, int namlen, loff_t offset,
394 u64 ino, unsigned int d_type)
396 struct l_linux_dirent *dirent;
397 struct l_readdir_callback *buf = (struct l_readdir_callback *)__buf;
399 dirent = buf->lrc_dirent;
401 dirent->lld_off = offset;
403 OBD_ALLOC(dirent, sizeof(*dirent));
408 cfs_list_add_tail(&dirent->lld_list, buf->lrc_list);
410 buf->lrc_dirent = dirent;
411 dirent->lld_ino = ino;
412 LASSERT(sizeof(dirent->lld_name) >= namlen + 1);
413 memcpy(dirent->lld_name, name, namlen);
418 long l_readdir(struct file *file, cfs_list_t *dentry_list)
420 struct l_linux_dirent *lastdirent;
421 struct l_readdir_callback buf;
424 buf.lrc_dirent = NULL;
425 buf.lrc_list = dentry_list;
427 error = vfs_readdir(file, l_filldir, &buf);
431 lastdirent = buf.lrc_dirent;
433 lastdirent->lld_off = file->f_pos;
437 EXPORT_SYMBOL(l_readdir);
439 int l_notify_change(struct vfsmount *mnt, struct dentry *dchild,
440 struct iattr *newattrs)
444 LOCK_INODE_MUTEX(dchild->d_inode);
445 #ifdef HAVE_SECURITY_PLUG
446 rc = notify_change(dchild, mnt, newattrs);
448 rc = notify_change(dchild, newattrs);
450 UNLOCK_INODE_MUTEX(dchild->d_inode);
453 EXPORT_SYMBOL(l_notify_change);
455 /* utility to truncate a file */
456 int simple_truncate(struct dentry *dir, struct vfsmount *mnt,
457 char *name, loff_t length)
459 struct dentry *dchild;
460 struct iattr newattrs;
464 CDEBUG(D_INODE, "truncating file %.*s to %lld\n", (int)strlen(name),
465 name, (long long)length);
466 dchild = ll_lookup_one_len(name, dir, strlen(name));
468 GOTO(out, err = PTR_ERR(dchild));
470 if (dchild->d_inode) {
471 int old_mode = dchild->d_inode->i_mode;
472 if (S_ISDIR(old_mode)) {
473 CERROR("found %s (%lu/%u) is mode %o\n", name,
474 dchild->d_inode->i_ino,
475 dchild->d_inode->i_generation, old_mode);
476 GOTO(out_dput, err = -EISDIR);
479 newattrs.ia_size = length;
480 newattrs.ia_valid = ATTR_SIZE;
481 err = l_notify_change(mnt, dchild, &newattrs);
489 EXPORT_SYMBOL(simple_truncate);
491 #ifdef LUSTRE_KERNEL_VERSION
492 int __lvfs_set_rdonly(lvfs_sbdev_type dev, lvfs_sbdev_type jdev)
494 #ifdef HAVE_DEV_SET_RDONLY
495 if (jdev && (jdev != dev)) {
496 CDEBUG(D_IOCTL | D_HA, "set journal dev %lx rdonly\n",
498 dev_set_rdonly(jdev);
500 CDEBUG(D_IOCTL | D_HA, "set dev %lx rdonly\n", (long)dev);
505 CERROR("DEV %lx CANNOT BE SET READONLY\n", (long)dev);
510 EXPORT_SYMBOL(__lvfs_set_rdonly);
512 int lvfs_check_rdonly(lvfs_sbdev_type dev)
514 #ifdef HAVE_DEV_SET_RDONLY
515 return dev_check_rdonly(dev);
520 EXPORT_SYMBOL(lvfs_check_rdonly);
522 int lvfs_check_io_health(struct obd_device *obd, struct file *file)
524 char *write_page = NULL;
529 OBD_ALLOC(write_page, CFS_PAGE_SIZE);
533 rc = fsfilt_write_record(obd, file, write_page, CFS_PAGE_SIZE, &offset, 1);
535 OBD_FREE(write_page, CFS_PAGE_SIZE);
537 CDEBUG(D_INFO, "write 1 page synchronously for checking io rc %d\n",rc);
540 EXPORT_SYMBOL(lvfs_check_io_health);
541 #endif /* LUSTRE_KERNEL_VERSION */
543 void obd_update_maxusage()
547 max1 = obd_pages_sum();
548 max2 = obd_memory_sum();
550 cfs_spin_lock(&obd_updatemax_lock);
551 if (max1 > obd_max_pages)
552 obd_max_pages = max1;
553 if (max2 > obd_max_alloc)
554 obd_max_alloc = max2;
555 cfs_spin_unlock(&obd_updatemax_lock);
559 __u64 obd_memory_max(void)
563 cfs_spin_lock(&obd_updatemax_lock);
565 cfs_spin_unlock(&obd_updatemax_lock);
570 __u64 obd_pages_max(void)
574 cfs_spin_lock(&obd_updatemax_lock);
576 cfs_spin_unlock(&obd_updatemax_lock);
581 EXPORT_SYMBOL(obd_update_maxusage);
582 EXPORT_SYMBOL(obd_pages_max);
583 EXPORT_SYMBOL(obd_memory_max);
584 EXPORT_SYMBOL(obd_memory);
587 __s64 lprocfs_read_helper(struct lprocfs_counter *lc,
588 enum lprocfs_fields_flags field)
596 centry = cfs_atomic_read(&lc->lc_cntl.la_entry);
599 case LPROCFS_FIELDS_FLAGS_CONFIG:
602 case LPROCFS_FIELDS_FLAGS_SUM:
603 ret = lc->lc_sum + lc->lc_sum_irq;
605 case LPROCFS_FIELDS_FLAGS_MIN:
608 case LPROCFS_FIELDS_FLAGS_MAX:
611 case LPROCFS_FIELDS_FLAGS_AVG:
612 ret = (lc->lc_max - lc->lc_min)/2;
614 case LPROCFS_FIELDS_FLAGS_SUMSQUARE:
615 ret = lc->lc_sumsquare;
617 case LPROCFS_FIELDS_FLAGS_COUNT:
623 } while (centry != cfs_atomic_read(&lc->lc_cntl.la_entry) &&
624 centry != cfs_atomic_read(&lc->lc_cntl.la_exit));
628 EXPORT_SYMBOL(lprocfs_read_helper);
631 MODULE_AUTHOR("Sun Microsystems, Inc. <http://www.lustre.org/>");
632 MODULE_DESCRIPTION("Lustre VFS Filesystem Helper v0.1");
633 MODULE_LICENSE("GPL");