1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
4 * lustre/lib/lvfs_linux.c
5 * Lustre filesystem abstraction routines
7 * Copyright (C) 2002, 2003 Cluster File Systems, Inc.
8 * Author: Andreas Dilger <adilger@clusterfs.com>
10 * This file is part of Lustre, http://www.lustre.org.
12 * Lustre is free software; you can redistribute it and/or
13 * modify it under the terms of version 2 of the GNU General Public
14 * License as published by the Free Software Foundation.
16 * Lustre is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19 * GNU General Public License for more details.
21 * You should have received a copy of the GNU General Public License
22 * along with Lustre; if not, write to the Free Software
23 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
27 # define EXPORT_SYMTAB
30 #define DEBUG_SUBSYSTEM S_FILTER
32 #include <linux/version.h>
34 #include <asm/unistd.h>
35 #include <linux/slab.h>
36 #include <linux/pagemap.h>
37 #include <linux/quotaops.h>
38 #include <linux/version.h>
39 #include <libcfs/kp30.h>
40 #include <linux/lustre_fsfilt.h>
41 #include <linux/obd.h>
42 #include <linux/obd_class.h>
43 #include <linux/module.h>
44 #include <linux/init.h>
45 #include <linux/lustre_compat25.h>
46 #include <linux/lvfs.h>
47 #include "lvfs_internal.h"
49 #include <linux/obd.h>
50 #include <linux/lustre_lib.h>
51 #include <linux/lustre_mds.h> /* for mds_grp_hash_entry */
57 /* Debugging check only needed during development */
59 # define ASSERT_CTXT_MAGIC(magic) LASSERT((magic) == OBD_RUN_CTXT_MAGIC)
60 # define ASSERT_NOT_KERNEL_CTXT(msg) LASSERTF(!segment_eq(get_fs(), get_ds()),\
62 # define ASSERT_KERNEL_CTXT(msg) LASSERTF(segment_eq(get_fs(), get_ds()), msg)
65 # define ASSERT_CTXT_MAGIC(magic) do {} while(0)
66 # define ASSERT_NOT_KERNEL_CTXT(msg) do {} while(0)
67 # define ASSERT_KERNEL_CTXT(msg) do {} while(0)
70 static void push_group_info(struct lvfs_run_ctxt *save,
71 struct group_info *ginfo)
74 save->ngroups = current_ngroups;
77 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,4)
79 save->group_info = current->group_info;
80 current->group_info = ginfo;
83 LASSERT(ginfo->ngroups <= NGROUPS);
85 save->group_info.ngroups = current->ngroups;
87 memcpy(save->group_info.small_block, current->groups,
90 current->ngroups = ginfo->ngroups;
92 memcpy(current->groups, ginfo->small_block,
98 static void pop_group_info(struct lvfs_run_ctxt *save,
99 struct group_info *ginfo)
102 current_ngroups = save->ngroups;
104 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,4)
106 current->group_info = save->group_info;
107 task_unlock(current);
109 current->ngroups = ginfo->ngroups;
110 if (current->ngroups)
111 memcpy(current->groups, save->group_info.small_block,
117 /* push / pop to root of obd store */
118 void push_ctxt(struct lvfs_run_ctxt *save, struct lvfs_run_ctxt *new_ctx,
119 struct lvfs_ucred *uc)
121 //ASSERT_NOT_KERNEL_CTXT("already in kernel context!\n");
122 ASSERT_CTXT_MAGIC(new_ctx->magic);
123 LASSERT(save->magic != OBD_RUN_CTXT_MAGIC || save->pid != current->pid);
124 OBD_SET_CTXT_MAGIC(save);
125 save->pid = current->pid;
129 "= push %p->%p = cur fs %p pwd %p:d%d:i%d (%*s), pwdmnt %p:%d\n",
130 save, current, current->fs, current->fs->pwd,
131 atomic_read(¤t->fs->pwd->d_count),
132 atomic_read(¤t->fs->pwd->d_inode->i_count),
133 current->fs->pwd->d_name.len, current->fs->pwd->d_name.name,
135 atomic_read(¤t->fs->pwdmnt->mnt_count));
139 LASSERT(atomic_read(¤t->fs->pwd->d_count));
140 LASSERT(atomic_read(&new_ctx->pwd->d_count));
141 save->pwd = dget(current->fs->pwd);
142 save->pwdmnt = mntget(current->fs->pwdmnt);
143 save->ngroups = current_ngroups;
144 save->luc.luc_umask = current->fs->umask;
147 LASSERT(save->pwdmnt);
148 LASSERT(new_ctx->pwd);
149 LASSERT(new_ctx->pwdmnt);
152 save->luc.luc_fsuid = current->fsuid;
153 save->luc.luc_fsgid = current->fsgid;
154 save->luc.luc_cap = current->cap_effective;
156 current->fsuid = uc->luc_fsuid;
157 current->fsgid = uc->luc_fsgid;
158 current->cap_effective = uc->luc_cap;
160 push_group_info(save, uc->luc_ginfo);
162 current->fs->umask = 0; /* umask already applied on client */
164 set_fs_pwd(current->fs, new_ctx->pwdmnt, new_ctx->pwd);
168 "= push %p->%p = cur fs %p pwd %p:d%d:i%d (%*s), pwdmnt %p:%d\n",
169 new_ctx, current, current->fs, current->fs->pwd,
170 atomic_read(¤t->fs->pwd->d_count),
171 atomic_read(¤t->fs->pwd->d_inode->i_count),
172 current->fs->pwd->d_name.len, current->fs->pwd->d_name.name,
174 atomic_read(¤t->fs->pwdmnt->mnt_count));
177 EXPORT_SYMBOL(push_ctxt);
179 void pop_ctxt(struct lvfs_run_ctxt *saved, struct lvfs_run_ctxt *new_ctx,
180 struct lvfs_ucred *uc)
183 ASSERT_CTXT_MAGIC(saved->magic);
184 LASSERT(saved->pid == current->pid);
188 ASSERT_KERNEL_CTXT("popping non-kernel context!\n");
192 " = pop %p==%p = cur %p pwd %p:d%d:i%d (%*s), pwdmnt %p:%d\n",
193 new_ctx, current, current->fs, current->fs->pwd,
194 atomic_read(¤t->fs->pwd->d_count),
195 atomic_read(¤t->fs->pwd->d_inode->i_count),
196 current->fs->pwd->d_name.len, current->fs->pwd->d_name.name,
198 atomic_read(¤t->fs->pwdmnt->mnt_count));
201 LASSERT(current->fs->pwd == new_ctx->pwd);
202 LASSERT(current->fs->pwdmnt == new_ctx->pwdmnt);
205 set_fs_pwd(current->fs, saved->pwdmnt, saved->pwd);
208 mntput(saved->pwdmnt);
209 current->fs->umask = saved->luc.luc_umask;
211 current->fsuid = saved->luc.luc_fsuid;
212 current->fsgid = saved->luc.luc_fsgid;
213 current->cap_effective = saved->luc.luc_cap;
215 pop_group_info(saved, uc->luc_ginfo);
220 "= pop %p->%p = cur fs %p pwd %p:d%d:i%d (%*s), pwdmnt %p:%d\n",
221 saved, current, current->fs, current->fs->pwd,
222 atomic_read(¤t->fs->pwd->d_count),
223 atomic_read(¤t->fs->pwd->d_inode->i_count),
224 current->fs->pwd->d_name.len, current->fs->pwd->d_name.name,
226 atomic_read(¤t->fs->pwdmnt->mnt_count));
229 EXPORT_SYMBOL(pop_ctxt);
231 /* utility to make a file */
232 struct dentry *simple_mknod(struct dentry *dir, char *name, int mode, int fix)
234 struct dentry *dchild;
238 ASSERT_KERNEL_CTXT("kernel doing mknod outside kernel context\n");
239 CDEBUG(D_INODE, "creating file %*s\n", (int)strlen(name), name);
241 dchild = ll_lookup_one_len(name, dir, strlen(name));
243 GOTO(out_up, dchild);
245 if (dchild->d_inode) {
246 int old_mode = dchild->d_inode->i_mode;
247 if (!S_ISREG(old_mode))
248 GOTO(out_err, err = -EEXIST);
250 /* Fixup file permissions if necessary */
251 if (fix && (old_mode & S_IALLUGO) != (mode & S_IALLUGO)) {
252 CWARN("fixing permissions on %s from %o to %o\n",
253 name, old_mode, mode);
254 dchild->d_inode->i_mode = (mode & S_IALLUGO) |
255 (old_mode & ~S_IALLUGO);
256 mark_inode_dirty(dchild->d_inode);
258 GOTO(out_up, dchild);
261 err = ll_vfs_create(dir->d_inode, dchild, (mode & ~S_IFMT) | S_IFREG,
270 dchild = ERR_PTR(err);
274 EXPORT_SYMBOL(simple_mknod);
276 /* utility to make a directory */
277 struct dentry *simple_mkdir(struct dentry *dir, char *name, int mode, int fix)
279 struct dentry *dchild;
283 ASSERT_KERNEL_CTXT("kernel doing mkdir outside kernel context\n");
284 CDEBUG(D_INODE, "creating directory %*s\n", (int)strlen(name), name);
285 dchild = ll_lookup_one_len(name, dir, strlen(name));
287 GOTO(out_up, dchild);
289 if (dchild->d_inode) {
290 int old_mode = dchild->d_inode->i_mode;
291 if (!S_ISDIR(old_mode))
292 GOTO(out_err, err = -ENOTDIR);
294 /* Fixup directory permissions if necessary */
295 if (fix && (old_mode & S_IALLUGO) != (mode & S_IALLUGO)) {
296 CWARN("fixing permissions on %s from %o to %o\n",
297 name, old_mode, mode);
298 dchild->d_inode->i_mode = (mode & S_IALLUGO) |
299 (old_mode & ~S_IALLUGO);
300 mark_inode_dirty(dchild->d_inode);
302 GOTO(out_up, dchild);
305 err = vfs_mkdir(dir->d_inode, dchild, mode);
313 dchild = ERR_PTR(err);
317 EXPORT_SYMBOL(simple_mkdir);
320 * Read a file from within kernel context. Prior to calling this
321 * function we should already have done a push_ctxt().
323 int lustre_fread(struct file *file, void *buf, int len, loff_t *off)
325 ASSERT_KERNEL_CTXT("kernel doing read outside kernel context\n");
326 if (!file || !file->f_op || !file->f_op->read || !off)
329 return file->f_op->read(file, buf, len, off);
331 EXPORT_SYMBOL(lustre_fread);
334 * Write a file from within kernel context. Prior to calling this
335 * function we should already have done a push_ctxt().
337 int lustre_fwrite(struct file *file, const void *buf, int len, loff_t *off)
340 ASSERT_KERNEL_CTXT("kernel doing write outside kernel context\n");
348 if (!file->f_op->write)
351 RETURN(file->f_op->write(file, buf, len, off));
353 EXPORT_SYMBOL(lustre_fwrite);
356 * Sync a file from within kernel context. Prior to calling this
357 * function we should already have done a push_ctxt().
359 int lustre_fsync(struct file *file)
362 ASSERT_KERNEL_CTXT("kernel doing sync outside kernel context\n");
363 if (!file || !file->f_op || !file->f_op->fsync)
366 RETURN(file->f_op->fsync(file, file->f_dentry, 0));
368 EXPORT_SYMBOL(lustre_fsync);
370 struct l_file *l_dentry_open(struct lvfs_run_ctxt *ctxt, struct l_dentry *de,
373 mntget(ctxt->pwdmnt);
374 return dentry_open(de, ctxt->pwdmnt, flags);
376 EXPORT_SYMBOL(l_dentry_open);
378 static int l_filldir(void *__buf, const char *name, int namlen, loff_t offset,
379 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));
390 list_add_tail(&dirent->lld_list, buf->lrc_list);
392 buf->lrc_dirent = dirent;
393 dirent->lld_ino = ino;
394 LASSERT(sizeof(dirent->lld_name) >= namlen + 1);
395 memcpy(dirent->lld_name, name, namlen);
400 long l_readdir(struct file *file, struct list_head *dentry_list)
402 struct l_linux_dirent *lastdirent;
403 struct l_readdir_callback buf;
406 buf.lrc_dirent = NULL;
407 buf.lrc_list = dentry_list;
409 error = vfs_readdir(file, l_filldir, &buf);
413 lastdirent = buf.lrc_dirent;
415 lastdirent->lld_off = file->f_pos;
419 EXPORT_SYMBOL(l_readdir);
420 EXPORT_SYMBOL(obd_memory);
421 EXPORT_SYMBOL(obd_memmax);
423 static int __init lvfs_linux_init(void)
428 static void __exit lvfs_linux_exit(void)
433 leaked = atomic_read(&obd_memory);
434 CDEBUG(leaked ? D_ERROR : D_INFO,
435 "obd mem max: %d leaked: %d\n", obd_memmax, leaked);
440 MODULE_AUTHOR("Cluster File Systems, Inc. <info@clusterfs.com>");
441 MODULE_DESCRIPTION("Lustre VFS Filesystem Helper v0.1");
442 MODULE_LICENSE("GPL");
444 module_init(lvfs_linux_init);
445 module_exit(lvfs_linux_exit);