Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / afs / super.c
diff --git a/kernel/fs/afs/super.c b/kernel/fs/afs/super.c
new file mode 100644 (file)
index 0000000..1fb4a51
--- /dev/null
@@ -0,0 +1,557 @@
+/* AFS superblock handling
+ *
+ * Copyright (c) 2002, 2007 Red Hat, Inc. All rights reserved.
+ *
+ * This software may be freely redistributed under the terms of the
+ * GNU General Public License.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ * Authors: David Howells <dhowells@redhat.com>
+ *          David Woodhouse <dwmw2@infradead.org>
+ *
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mount.h>
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/fs.h>
+#include <linux/pagemap.h>
+#include <linux/parser.h>
+#include <linux/statfs.h>
+#include <linux/sched.h>
+#include <linux/nsproxy.h>
+#include <net/net_namespace.h>
+#include "internal.h"
+
+#define AFS_FS_MAGIC 0x6B414653 /* 'kAFS' */
+
+static void afs_i_init_once(void *foo);
+static struct dentry *afs_mount(struct file_system_type *fs_type,
+                     int flags, const char *dev_name, void *data);
+static void afs_kill_super(struct super_block *sb);
+static struct inode *afs_alloc_inode(struct super_block *sb);
+static void afs_destroy_inode(struct inode *inode);
+static int afs_statfs(struct dentry *dentry, struct kstatfs *buf);
+
+struct file_system_type afs_fs_type = {
+       .owner          = THIS_MODULE,
+       .name           = "afs",
+       .mount          = afs_mount,
+       .kill_sb        = afs_kill_super,
+       .fs_flags       = 0,
+};
+MODULE_ALIAS_FS("afs");
+
+static const struct super_operations afs_super_ops = {
+       .statfs         = afs_statfs,
+       .alloc_inode    = afs_alloc_inode,
+       .drop_inode     = afs_drop_inode,
+       .destroy_inode  = afs_destroy_inode,
+       .evict_inode    = afs_evict_inode,
+       .show_options   = generic_show_options,
+};
+
+static struct kmem_cache *afs_inode_cachep;
+static atomic_t afs_count_active_inodes;
+
+enum {
+       afs_no_opt,
+       afs_opt_cell,
+       afs_opt_rwpath,
+       afs_opt_vol,
+       afs_opt_autocell,
+};
+
+static const match_table_t afs_options_list = {
+       { afs_opt_cell,         "cell=%s"       },
+       { afs_opt_rwpath,       "rwpath"        },
+       { afs_opt_vol,          "vol=%s"        },
+       { afs_opt_autocell,     "autocell"      },
+       { afs_no_opt,           NULL            },
+};
+
+/*
+ * initialise the filesystem
+ */
+int __init afs_fs_init(void)
+{
+       int ret;
+
+       _enter("");
+
+       /* create ourselves an inode cache */
+       atomic_set(&afs_count_active_inodes, 0);
+
+       ret = -ENOMEM;
+       afs_inode_cachep = kmem_cache_create("afs_inode_cache",
+                                            sizeof(struct afs_vnode),
+                                            0,
+                                            SLAB_HWCACHE_ALIGN,
+                                            afs_i_init_once);
+       if (!afs_inode_cachep) {
+               printk(KERN_NOTICE "kAFS: Failed to allocate inode cache\n");
+               return ret;
+       }
+
+       /* now export our filesystem to lesser mortals */
+       ret = register_filesystem(&afs_fs_type);
+       if (ret < 0) {
+               kmem_cache_destroy(afs_inode_cachep);
+               _leave(" = %d", ret);
+               return ret;
+       }
+
+       _leave(" = 0");
+       return 0;
+}
+
+/*
+ * clean up the filesystem
+ */
+void __exit afs_fs_exit(void)
+{
+       _enter("");
+
+       afs_mntpt_kill_timer();
+       unregister_filesystem(&afs_fs_type);
+
+       if (atomic_read(&afs_count_active_inodes) != 0) {
+               printk("kAFS: %d active inode objects still present\n",
+                      atomic_read(&afs_count_active_inodes));
+               BUG();
+       }
+
+       /*
+        * Make sure all delayed rcu free inodes are flushed before we
+        * destroy cache.
+        */
+       rcu_barrier();
+       kmem_cache_destroy(afs_inode_cachep);
+       _leave("");
+}
+
+/*
+ * parse the mount options
+ * - this function has been shamelessly adapted from the ext3 fs which
+ *   shamelessly adapted it from the msdos fs
+ */
+static int afs_parse_options(struct afs_mount_params *params,
+                            char *options, const char **devname)
+{
+       struct afs_cell *cell;
+       substring_t args[MAX_OPT_ARGS];
+       char *p;
+       int token;
+
+       _enter("%s", options);
+
+       options[PAGE_SIZE - 1] = 0;
+
+       while ((p = strsep(&options, ","))) {
+               if (!*p)
+                       continue;
+
+               token = match_token(p, afs_options_list, args);
+               switch (token) {
+               case afs_opt_cell:
+                       cell = afs_cell_lookup(args[0].from,
+                                              args[0].to - args[0].from,
+                                              false);
+                       if (IS_ERR(cell))
+                               return PTR_ERR(cell);
+                       afs_put_cell(params->cell);
+                       params->cell = cell;
+                       break;
+
+               case afs_opt_rwpath:
+                       params->rwpath = 1;
+                       break;
+
+               case afs_opt_vol:
+                       *devname = args[0].from;
+                       break;
+
+               case afs_opt_autocell:
+                       params->autocell = 1;
+                       break;
+
+               default:
+                       printk(KERN_ERR "kAFS:"
+                              " Unknown or invalid mount option: '%s'\n", p);
+                       return -EINVAL;
+               }
+       }
+
+       _leave(" = 0");
+       return 0;
+}
+
+/*
+ * parse a device name to get cell name, volume name, volume type and R/W
+ * selector
+ * - this can be one of the following:
+ *     "%[cell:]volume[.]"             R/W volume
+ *     "#[cell:]volume[.]"             R/O or R/W volume (rwpath=0),
+ *                                      or R/W (rwpath=1) volume
+ *     "%[cell:]volume.readonly"       R/O volume
+ *     "#[cell:]volume.readonly"       R/O volume
+ *     "%[cell:]volume.backup"         Backup volume
+ *     "#[cell:]volume.backup"         Backup volume
+ */
+static int afs_parse_device_name(struct afs_mount_params *params,
+                                const char *name)
+{
+       struct afs_cell *cell;
+       const char *cellname, *suffix;
+       int cellnamesz;
+
+       _enter(",%s", name);
+
+       if (!name) {
+               printk(KERN_ERR "kAFS: no volume name specified\n");
+               return -EINVAL;
+       }
+
+       if ((name[0] != '%' && name[0] != '#') || !name[1]) {
+               printk(KERN_ERR "kAFS: unparsable volume name\n");
+               return -EINVAL;
+       }
+
+       /* determine the type of volume we're looking for */
+       params->type = AFSVL_ROVOL;
+       params->force = false;
+       if (params->rwpath || name[0] == '%') {
+               params->type = AFSVL_RWVOL;
+               params->force = true;
+       }
+       name++;
+
+       /* split the cell name out if there is one */
+       params->volname = strchr(name, ':');
+       if (params->volname) {
+               cellname = name;
+               cellnamesz = params->volname - name;
+               params->volname++;
+       } else {
+               params->volname = name;
+               cellname = NULL;
+               cellnamesz = 0;
+       }
+
+       /* the volume type is further affected by a possible suffix */
+       suffix = strrchr(params->volname, '.');
+       if (suffix) {
+               if (strcmp(suffix, ".readonly") == 0) {
+                       params->type = AFSVL_ROVOL;
+                       params->force = true;
+               } else if (strcmp(suffix, ".backup") == 0) {
+                       params->type = AFSVL_BACKVOL;
+                       params->force = true;
+               } else if (suffix[1] == 0) {
+               } else {
+                       suffix = NULL;
+               }
+       }
+
+       params->volnamesz = suffix ?
+               suffix - params->volname : strlen(params->volname);
+
+       _debug("cell %*.*s [%p]",
+              cellnamesz, cellnamesz, cellname ?: "", params->cell);
+
+       /* lookup the cell record */
+       if (cellname || !params->cell) {
+               cell = afs_cell_lookup(cellname, cellnamesz, true);
+               if (IS_ERR(cell)) {
+                       printk(KERN_ERR "kAFS: unable to lookup cell '%*.*s'\n",
+                              cellnamesz, cellnamesz, cellname ?: "");
+                       return PTR_ERR(cell);
+               }
+               afs_put_cell(params->cell);
+               params->cell = cell;
+       }
+
+       _debug("CELL:%s [%p] VOLUME:%*.*s SUFFIX:%s TYPE:%d%s",
+              params->cell->name, params->cell,
+              params->volnamesz, params->volnamesz, params->volname,
+              suffix ?: "-", params->type, params->force ? " FORCE" : "");
+
+       return 0;
+}
+
+/*
+ * check a superblock to see if it's the one we're looking for
+ */
+static int afs_test_super(struct super_block *sb, void *data)
+{
+       struct afs_super_info *as1 = data;
+       struct afs_super_info *as = sb->s_fs_info;
+
+       return as->volume == as1->volume;
+}
+
+static int afs_set_super(struct super_block *sb, void *data)
+{
+       sb->s_fs_info = data;
+       return set_anon_super(sb, NULL);
+}
+
+/*
+ * fill in the superblock
+ */
+static int afs_fill_super(struct super_block *sb,
+                         struct afs_mount_params *params)
+{
+       struct afs_super_info *as = sb->s_fs_info;
+       struct afs_fid fid;
+       struct inode *inode = NULL;
+       int ret;
+
+       _enter("");
+
+       /* fill in the superblock */
+       sb->s_blocksize         = PAGE_CACHE_SIZE;
+       sb->s_blocksize_bits    = PAGE_CACHE_SHIFT;
+       sb->s_magic             = AFS_FS_MAGIC;
+       sb->s_op                = &afs_super_ops;
+       sb->s_bdi               = &as->volume->bdi;
+       strlcpy(sb->s_id, as->volume->vlocation->vldb.name, sizeof(sb->s_id));
+
+       /* allocate the root inode and dentry */
+       fid.vid         = as->volume->vid;
+       fid.vnode       = 1;
+       fid.unique      = 1;
+       inode = afs_iget(sb, params->key, &fid, NULL, NULL);
+       if (IS_ERR(inode))
+               return PTR_ERR(inode);
+
+       if (params->autocell)
+               set_bit(AFS_VNODE_AUTOCELL, &AFS_FS_I(inode)->flags);
+
+       ret = -ENOMEM;
+       sb->s_root = d_make_root(inode);
+       if (!sb->s_root)
+               goto error;
+
+       sb->s_d_op = &afs_fs_dentry_operations;
+
+       _leave(" = 0");
+       return 0;
+
+error:
+       _leave(" = %d", ret);
+       return ret;
+}
+
+/*
+ * get an AFS superblock
+ */
+static struct dentry *afs_mount(struct file_system_type *fs_type,
+                     int flags, const char *dev_name, void *options)
+{
+       struct afs_mount_params params;
+       struct super_block *sb;
+       struct afs_volume *vol;
+       struct key *key;
+       char *new_opts = kstrdup(options, GFP_KERNEL);
+       struct afs_super_info *as;
+       int ret;
+
+       _enter(",,%s,%p", dev_name, options);
+
+       memset(&params, 0, sizeof(params));
+
+       ret = -EINVAL;
+       if (current->nsproxy->net_ns != &init_net)
+               goto error;
+
+       /* parse the options and device name */
+       if (options) {
+               ret = afs_parse_options(&params, options, &dev_name);
+               if (ret < 0)
+                       goto error;
+       }
+
+       ret = afs_parse_device_name(&params, dev_name);
+       if (ret < 0)
+               goto error;
+
+       /* try and do the mount securely */
+       key = afs_request_key(params.cell);
+       if (IS_ERR(key)) {
+               _leave(" = %ld [key]", PTR_ERR(key));
+               ret = PTR_ERR(key);
+               goto error;
+       }
+       params.key = key;
+
+       /* parse the device name */
+       vol = afs_volume_lookup(&params);
+       if (IS_ERR(vol)) {
+               ret = PTR_ERR(vol);
+               goto error;
+       }
+
+       /* allocate a superblock info record */
+       as = kzalloc(sizeof(struct afs_super_info), GFP_KERNEL);
+       if (!as) {
+               ret = -ENOMEM;
+               afs_put_volume(vol);
+               goto error;
+       }
+       as->volume = vol;
+
+       /* allocate a deviceless superblock */
+       sb = sget(fs_type, afs_test_super, afs_set_super, flags, as);
+       if (IS_ERR(sb)) {
+               ret = PTR_ERR(sb);
+               afs_put_volume(vol);
+               kfree(as);
+               goto error;
+       }
+
+       if (!sb->s_root) {
+               /* initial superblock/root creation */
+               _debug("create");
+               ret = afs_fill_super(sb, &params);
+               if (ret < 0) {
+                       deactivate_locked_super(sb);
+                       goto error;
+               }
+               save_mount_options(sb, new_opts);
+               sb->s_flags |= MS_ACTIVE;
+       } else {
+               _debug("reuse");
+               ASSERTCMP(sb->s_flags, &, MS_ACTIVE);
+               afs_put_volume(vol);
+               kfree(as);
+       }
+
+       afs_put_cell(params.cell);
+       kfree(new_opts);
+       _leave(" = 0 [%p]", sb);
+       return dget(sb->s_root);
+
+error:
+       afs_put_cell(params.cell);
+       key_put(params.key);
+       kfree(new_opts);
+       _leave(" = %d", ret);
+       return ERR_PTR(ret);
+}
+
+static void afs_kill_super(struct super_block *sb)
+{
+       struct afs_super_info *as = sb->s_fs_info;
+       kill_anon_super(sb);
+       afs_put_volume(as->volume);
+       kfree(as);
+}
+
+/*
+ * initialise an inode cache slab element prior to any use
+ */
+static void afs_i_init_once(void *_vnode)
+{
+       struct afs_vnode *vnode = _vnode;
+
+       memset(vnode, 0, sizeof(*vnode));
+       inode_init_once(&vnode->vfs_inode);
+       init_waitqueue_head(&vnode->update_waitq);
+       mutex_init(&vnode->permits_lock);
+       mutex_init(&vnode->validate_lock);
+       spin_lock_init(&vnode->writeback_lock);
+       spin_lock_init(&vnode->lock);
+       INIT_LIST_HEAD(&vnode->writebacks);
+       INIT_LIST_HEAD(&vnode->pending_locks);
+       INIT_LIST_HEAD(&vnode->granted_locks);
+       INIT_DELAYED_WORK(&vnode->lock_work, afs_lock_work);
+       INIT_WORK(&vnode->cb_broken_work, afs_broken_callback_work);
+}
+
+/*
+ * allocate an AFS inode struct from our slab cache
+ */
+static struct inode *afs_alloc_inode(struct super_block *sb)
+{
+       struct afs_vnode *vnode;
+
+       vnode = kmem_cache_alloc(afs_inode_cachep, GFP_KERNEL);
+       if (!vnode)
+               return NULL;
+
+       atomic_inc(&afs_count_active_inodes);
+
+       memset(&vnode->fid, 0, sizeof(vnode->fid));
+       memset(&vnode->status, 0, sizeof(vnode->status));
+
+       vnode->volume           = NULL;
+       vnode->update_cnt       = 0;
+       vnode->flags            = 1 << AFS_VNODE_UNSET;
+       vnode->cb_promised      = false;
+
+       _leave(" = %p", &vnode->vfs_inode);
+       return &vnode->vfs_inode;
+}
+
+static void afs_i_callback(struct rcu_head *head)
+{
+       struct inode *inode = container_of(head, struct inode, i_rcu);
+       struct afs_vnode *vnode = AFS_FS_I(inode);
+       kmem_cache_free(afs_inode_cachep, vnode);
+}
+
+/*
+ * destroy an AFS inode struct
+ */
+static void afs_destroy_inode(struct inode *inode)
+{
+       struct afs_vnode *vnode = AFS_FS_I(inode);
+
+       _enter("%p{%x:%u}", inode, vnode->fid.vid, vnode->fid.vnode);
+
+       _debug("DESTROY INODE %p", inode);
+
+       ASSERTCMP(vnode->server, ==, NULL);
+
+       call_rcu(&inode->i_rcu, afs_i_callback);
+       atomic_dec(&afs_count_active_inodes);
+}
+
+/*
+ * return information about an AFS volume
+ */
+static int afs_statfs(struct dentry *dentry, struct kstatfs *buf)
+{
+       struct afs_volume_status vs;
+       struct afs_vnode *vnode = AFS_FS_I(d_inode(dentry));
+       struct key *key;
+       int ret;
+
+       key = afs_request_key(vnode->volume->cell);
+       if (IS_ERR(key))
+               return PTR_ERR(key);
+
+       ret = afs_vnode_get_volume_status(vnode, key, &vs);
+       key_put(key);
+       if (ret < 0) {
+               _leave(" = %d", ret);
+               return ret;
+       }
+
+       buf->f_type     = dentry->d_sb->s_magic;
+       buf->f_bsize    = AFS_BLOCK_SIZE;
+       buf->f_namelen  = AFSNAMEMAX - 1;
+
+       if (vs.max_quota == 0)
+               buf->f_blocks = vs.part_max_blocks;
+       else
+               buf->f_blocks = vs.max_quota;
+       buf->f_bavail = buf->f_bfree = buf->f_blocks - vs.blocks_in_use;
+       return 0;
+}