Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / fs / nfs / nfs4super.c
diff --git a/kernel/fs/nfs/nfs4super.c b/kernel/fs/nfs/nfs4super.c
new file mode 100644 (file)
index 0000000..6fb7cb6
--- /dev/null
@@ -0,0 +1,361 @@
+/*
+ * Copyright (c) 2012 Bryan Schumaker <bjschuma@netapp.com>
+ */
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/nfs4_mount.h>
+#include <linux/nfs_fs.h>
+#include "delegation.h"
+#include "internal.h"
+#include "nfs4_fs.h"
+#include "nfs4idmap.h"
+#include "dns_resolve.h"
+#include "pnfs.h"
+#include "nfs.h"
+
+#define NFSDBG_FACILITY                NFSDBG_VFS
+
+static int nfs4_write_inode(struct inode *inode, struct writeback_control *wbc);
+static void nfs4_evict_inode(struct inode *inode);
+static struct dentry *nfs4_remote_mount(struct file_system_type *fs_type,
+       int flags, const char *dev_name, void *raw_data);
+static struct dentry *nfs4_referral_mount(struct file_system_type *fs_type,
+       int flags, const char *dev_name, void *raw_data);
+static struct dentry *nfs4_remote_referral_mount(struct file_system_type *fs_type,
+       int flags, const char *dev_name, void *raw_data);
+
+static struct file_system_type nfs4_remote_fs_type = {
+       .owner          = THIS_MODULE,
+       .name           = "nfs4",
+       .mount          = nfs4_remote_mount,
+       .kill_sb        = nfs_kill_super,
+       .fs_flags       = FS_RENAME_DOES_D_MOVE|FS_BINARY_MOUNTDATA,
+};
+
+static struct file_system_type nfs4_remote_referral_fs_type = {
+       .owner          = THIS_MODULE,
+       .name           = "nfs4",
+       .mount          = nfs4_remote_referral_mount,
+       .kill_sb        = nfs_kill_super,
+       .fs_flags       = FS_RENAME_DOES_D_MOVE|FS_BINARY_MOUNTDATA,
+};
+
+struct file_system_type nfs4_referral_fs_type = {
+       .owner          = THIS_MODULE,
+       .name           = "nfs4",
+       .mount          = nfs4_referral_mount,
+       .kill_sb        = nfs_kill_super,
+       .fs_flags       = FS_RENAME_DOES_D_MOVE|FS_BINARY_MOUNTDATA,
+};
+
+static const struct super_operations nfs4_sops = {
+       .alloc_inode    = nfs_alloc_inode,
+       .destroy_inode  = nfs_destroy_inode,
+       .write_inode    = nfs4_write_inode,
+       .drop_inode     = nfs_drop_inode,
+       .statfs         = nfs_statfs,
+       .evict_inode    = nfs4_evict_inode,
+       .umount_begin   = nfs_umount_begin,
+       .show_options   = nfs_show_options,
+       .show_devname   = nfs_show_devname,
+       .show_path      = nfs_show_path,
+       .show_stats     = nfs_show_stats,
+       .remount_fs     = nfs_remount,
+};
+
+struct nfs_subversion nfs_v4 = {
+       .owner = THIS_MODULE,
+       .nfs_fs   = &nfs4_fs_type,
+       .rpc_vers = &nfs_version4,
+       .rpc_ops  = &nfs_v4_clientops,
+       .sops     = &nfs4_sops,
+       .xattr    = nfs4_xattr_handlers,
+};
+
+static int nfs4_write_inode(struct inode *inode, struct writeback_control *wbc)
+{
+       int ret = nfs_write_inode(inode, wbc);
+
+       if (ret == 0)
+               ret = pnfs_layoutcommit_inode(inode,
+                               wbc->sync_mode == WB_SYNC_ALL);
+       return ret;
+}
+
+/*
+ * Clean out any remaining NFSv4 state that might be left over due
+ * to open() calls that passed nfs_atomic_lookup, but failed to call
+ * nfs_open().
+ */
+static void nfs4_evict_inode(struct inode *inode)
+{
+       truncate_inode_pages_final(&inode->i_data);
+       clear_inode(inode);
+       /* If we are holding a delegation, return it! */
+       nfs_inode_return_delegation_noreclaim(inode);
+       /* Note that above delegreturn would trigger pnfs return-on-close */
+       pnfs_return_layout(inode);
+       pnfs_destroy_layout(NFS_I(inode));
+       /* First call standard NFS clear_inode() code */
+       nfs_clear_inode(inode);
+}
+
+/*
+ * Get the superblock for the NFS4 root partition
+ */
+static struct dentry *
+nfs4_remote_mount(struct file_system_type *fs_type, int flags,
+                 const char *dev_name, void *info)
+{
+       struct nfs_mount_info *mount_info = info;
+       struct nfs_server *server;
+       struct dentry *mntroot = ERR_PTR(-ENOMEM);
+
+       mount_info->set_security = nfs_set_sb_security;
+
+       /* Get a volume representation */
+       server = nfs4_create_server(mount_info, &nfs_v4);
+       if (IS_ERR(server)) {
+               mntroot = ERR_CAST(server);
+               goto out;
+       }
+
+       mntroot = nfs_fs_mount_common(server, flags, dev_name, mount_info, &nfs_v4);
+
+out:
+       return mntroot;
+}
+
+static struct vfsmount *nfs_do_root_mount(struct file_system_type *fs_type,
+               int flags, void *data, const char *hostname)
+{
+       struct vfsmount *root_mnt;
+       char *root_devname;
+       size_t len;
+
+       len = strlen(hostname) + 5;
+       root_devname = kmalloc(len, GFP_KERNEL);
+       if (root_devname == NULL)
+               return ERR_PTR(-ENOMEM);
+       /* Does hostname needs to be enclosed in brackets? */
+       if (strchr(hostname, ':'))
+               snprintf(root_devname, len, "[%s]:/", hostname);
+       else
+               snprintf(root_devname, len, "%s:/", hostname);
+       root_mnt = vfs_kern_mount(fs_type, flags, root_devname, data);
+       kfree(root_devname);
+       return root_mnt;
+}
+
+struct nfs_referral_count {
+       struct list_head list;
+       const struct task_struct *task;
+       unsigned int referral_count;
+};
+
+static LIST_HEAD(nfs_referral_count_list);
+static DEFINE_SPINLOCK(nfs_referral_count_list_lock);
+
+static struct nfs_referral_count *nfs_find_referral_count(void)
+{
+       struct nfs_referral_count *p;
+
+       list_for_each_entry(p, &nfs_referral_count_list, list) {
+               if (p->task == current)
+                       return p;
+       }
+       return NULL;
+}
+
+#define NFS_MAX_NESTED_REFERRALS 2
+
+static int nfs_referral_loop_protect(void)
+{
+       struct nfs_referral_count *p, *new;
+       int ret = -ENOMEM;
+
+       new = kmalloc(sizeof(*new), GFP_KERNEL);
+       if (!new)
+               goto out;
+       new->task = current;
+       new->referral_count = 1;
+
+       ret = 0;
+       spin_lock(&nfs_referral_count_list_lock);
+       p = nfs_find_referral_count();
+       if (p != NULL) {
+               if (p->referral_count >= NFS_MAX_NESTED_REFERRALS)
+                       ret = -ELOOP;
+               else
+                       p->referral_count++;
+       } else {
+               list_add(&new->list, &nfs_referral_count_list);
+               new = NULL;
+       }
+       spin_unlock(&nfs_referral_count_list_lock);
+       kfree(new);
+out:
+       return ret;
+}
+
+static void nfs_referral_loop_unprotect(void)
+{
+       struct nfs_referral_count *p;
+
+       spin_lock(&nfs_referral_count_list_lock);
+       p = nfs_find_referral_count();
+       p->referral_count--;
+       if (p->referral_count == 0)
+               list_del(&p->list);
+       else
+               p = NULL;
+       spin_unlock(&nfs_referral_count_list_lock);
+       kfree(p);
+}
+
+static struct dentry *nfs_follow_remote_path(struct vfsmount *root_mnt,
+               const char *export_path)
+{
+       struct dentry *dentry;
+       int err;
+
+       if (IS_ERR(root_mnt))
+               return ERR_CAST(root_mnt);
+
+       err = nfs_referral_loop_protect();
+       if (err) {
+               mntput(root_mnt);
+               return ERR_PTR(err);
+       }
+
+       dentry = mount_subtree(root_mnt, export_path);
+       nfs_referral_loop_unprotect();
+
+       return dentry;
+}
+
+struct dentry *nfs4_try_mount(int flags, const char *dev_name,
+                             struct nfs_mount_info *mount_info,
+                             struct nfs_subversion *nfs_mod)
+{
+       char *export_path;
+       struct vfsmount *root_mnt;
+       struct dentry *res;
+       struct nfs_parsed_mount_data *data = mount_info->parsed;
+
+       dfprintk(MOUNT, "--> nfs4_try_mount()\n");
+
+       export_path = data->nfs_server.export_path;
+       data->nfs_server.export_path = "/";
+       root_mnt = nfs_do_root_mount(&nfs4_remote_fs_type, flags, mount_info,
+                       data->nfs_server.hostname);
+       data->nfs_server.export_path = export_path;
+
+       res = nfs_follow_remote_path(root_mnt, export_path);
+
+       dfprintk(MOUNT, "<-- nfs4_try_mount() = %d%s\n",
+                PTR_ERR_OR_ZERO(res),
+                IS_ERR(res) ? " [error]" : "");
+       return res;
+}
+
+static struct dentry *
+nfs4_remote_referral_mount(struct file_system_type *fs_type, int flags,
+                          const char *dev_name, void *raw_data)
+{
+       struct nfs_mount_info mount_info = {
+               .fill_super = nfs_fill_super,
+               .set_security = nfs_clone_sb_security,
+               .cloned = raw_data,
+       };
+       struct nfs_server *server;
+       struct dentry *mntroot = ERR_PTR(-ENOMEM);
+
+       dprintk("--> nfs4_referral_get_sb()\n");
+
+       mount_info.mntfh = nfs_alloc_fhandle();
+       if (mount_info.cloned == NULL || mount_info.mntfh == NULL)
+               goto out;
+
+       /* create a new volume representation */
+       server = nfs4_create_referral_server(mount_info.cloned, mount_info.mntfh);
+       if (IS_ERR(server)) {
+               mntroot = ERR_CAST(server);
+               goto out;
+       }
+
+       mntroot = nfs_fs_mount_common(server, flags, dev_name, &mount_info, &nfs_v4);
+out:
+       nfs_free_fhandle(mount_info.mntfh);
+       return mntroot;
+}
+
+/*
+ * Create an NFS4 server record on referral traversal
+ */
+static struct dentry *nfs4_referral_mount(struct file_system_type *fs_type,
+               int flags, const char *dev_name, void *raw_data)
+{
+       struct nfs_clone_mount *data = raw_data;
+       char *export_path;
+       struct vfsmount *root_mnt;
+       struct dentry *res;
+
+       dprintk("--> nfs4_referral_mount()\n");
+
+       export_path = data->mnt_path;
+       data->mnt_path = "/";
+
+       root_mnt = nfs_do_root_mount(&nfs4_remote_referral_fs_type,
+                       flags, data, data->hostname);
+       data->mnt_path = export_path;
+
+       res = nfs_follow_remote_path(root_mnt, export_path);
+       dprintk("<-- nfs4_referral_mount() = %d%s\n",
+               PTR_ERR_OR_ZERO(res),
+               IS_ERR(res) ? " [error]" : "");
+       return res;
+}
+
+
+static int __init init_nfs_v4(void)
+{
+       int err;
+
+       err = nfs_dns_resolver_init();
+       if (err)
+               goto out;
+
+       err = nfs_idmap_init();
+       if (err)
+               goto out1;
+
+       err = nfs4_register_sysctl();
+       if (err)
+               goto out2;
+
+       register_nfs_version(&nfs_v4);
+       return 0;
+out2:
+       nfs_idmap_quit();
+out1:
+       nfs_dns_resolver_destroy();
+out:
+       return err;
+}
+
+static void __exit exit_nfs_v4(void)
+{
+       /* Not called in the _init(), conditionally loaded */
+       nfs4_pnfs_v3_ds_connect_unload();
+
+       unregister_nfs_version(&nfs_v4);
+       nfs4_unregister_sysctl();
+       nfs_idmap_quit();
+       nfs_dns_resolver_destroy();
+}
+
+MODULE_LICENSE("GPL");
+
+module_init(init_nfs_v4);
+module_exit(exit_nfs_v4);