[net-next-2.6.git] / fs / ocfs2 / dlmfs / dlmfs.c

/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * dlmfs.c
 *
 * Code which implements the kernel side of a minimal userspace
 * interface to our DLM. This file handles the virtual file system
 * used for communication with userspace. Credit should go to ramfs,
 * which was a template for the fs side of this module.
 *
 * Copyright (C) 2003, 2004 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * 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., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

/* Simple VFS hooks based on: */
/*
 * Resizable simple ram filesystem for Linux.
 *
 * Copyright (C) 2000 Linus Torvalds.
 *               2000 Transmeta Corp.
 */

#include <linux/module.h>
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/backing-dev.h>
#include <linux/poll.h>

#include <asm/uaccess.h>

#include "stackglue.h"
#include "userdlm.h"
#include "dlmfsver.h"

#define MLOG_MASK_PREFIX ML_DLMFS
#include "cluster/masklog.h"


static const struct super_operations dlmfs_ops;
static const struct file_operations dlmfs_file_operations;
static const struct inode_operations dlmfs_dir_inode_operations;
static const struct inode_operations dlmfs_root_inode_operations;
static const struct inode_operations dlmfs_file_inode_operations;
static struct kmem_cache *dlmfs_inode_cache;

struct workqueue_struct *user_dlm_worker;


/*
 * These are the ABI capabilities of dlmfs.
 *
 * Over time, dlmfs has added some features that were not part of the
 * initial ABI.  Unfortunately, some of these features are not detectable
 * via standard usage.  For example, Linux's default poll always returns
 * POLLIN, so there is no way for a caller of poll(2) to know when dlmfs
 * added poll support.  Instead, we provide this list of new capabilities.
 *
 * Capabilities is a read-only attribute.  We do it as a module parameter
 * so we can discover it whether dlmfs is built in, loaded, or even not
 * loaded.
 *
 * The ABI features are local to this machine's dlmfs mount.  This is
 * distinct from the locking protocol, which is concerned with inter-node
 * interaction.
 *
 * Capabilities:
 * - bast	: POLLIN against the file descriptor of a held lock
 *		  signifies a bast fired on the lock.
 */
#define DLMFS_CAPABILITIES "bast stackglue"
extern int param_set_dlmfs_capabilities(const char *val,
					struct kernel_param *kp)
{
	printk(KERN_ERR "%s: readonly parameter\n", kp->name);
	return -EINVAL;
}
static int param_get_dlmfs_capabilities(char *buffer,
					struct kernel_param *kp)
{
	return strlcpy(buffer, DLMFS_CAPABILITIES,
		       strlen(DLMFS_CAPABILITIES) + 1);
}
module_param_call(capabilities, param_set_dlmfs_capabilities,
		  param_get_dlmfs_capabilities, NULL, 0444);
MODULE_PARM_DESC(capabilities, DLMFS_CAPABILITIES);


/*
 * decodes a set of open flags into a valid lock level and a set of flags.
 * returns < 0 if we have invalid flags
 * flags which mean something to us:
 * O_RDONLY -> PRMODE level
 * O_WRONLY -> EXMODE level
 *
 * O_NONBLOCK -> NOQUEUE
 */
static int dlmfs_decode_open_flags(int open_flags,
				   int *level,
				   int *flags)
{
	if (open_flags & (O_WRONLY|O_RDWR))
		*level = DLM_LOCK_EX;
	else
		*level = DLM_LOCK_PR;

	*flags = 0;
	if (open_flags & O_NONBLOCK)
		*flags |= DLM_LKF_NOQUEUE;

	return 0;
}

static int dlmfs_file_open(struct inode *inode,
			   struct file *file)
{
	int status, level, flags;
	struct dlmfs_filp_private *fp = NULL;
	struct dlmfs_inode_private *ip;

	if (S_ISDIR(inode->i_mode))
		BUG();

	mlog(0, "open called on inode %lu, flags 0x%x\n", inode->i_ino,
		file->f_flags);

	status = dlmfs_decode_open_flags(file->f_flags, &level, &flags);
	if (status < 0)
		goto bail;

	/* We don't want to honor O_APPEND at read/write time as it
	 * doesn't make sense for LVB writes. */
	file->f_flags &= ~O_APPEND;

	fp = kmalloc(sizeof(*fp), GFP_NOFS);
	if (!fp) {
		status = -ENOMEM;
		goto bail;
	}
	fp->fp_lock_level = level;

	ip = DLMFS_I(inode);

	status = user_dlm_cluster_lock(&ip->ip_lockres, level, flags);
	if (status < 0) {
		/* this is a strange error to return here but I want
		 * to be able userspace to be able to distinguish a
		 * valid lock request from one that simply couldn't be
		 * granted. */
		if (flags & DLM_LKF_NOQUEUE && status == -EAGAIN)
			status = -ETXTBSY;
		kfree(fp);
		goto bail;
	}

	file->private_data = fp;
bail:
	return status;
}

static int dlmfs_file_release(struct inode *inode,
			      struct file *file)
{
	int level, status;
	struct dlmfs_inode_private *ip = DLMFS_I(inode);
	struct dlmfs_filp_private *fp = file->private_data;

	if (S_ISDIR(inode->i_mode))
		BUG();

	mlog(0, "close called on inode %lu\n", inode->i_ino);

	status = 0;
	if (fp) {
		level = fp->fp_lock_level;
		if (level != DLM_LOCK_IV)
			user_dlm_cluster_unlock(&ip->ip_lockres, level);

		kfree(fp);
		file->private_data = NULL;
	}

	return 0;
}

/*
 * We do ->setattr() just to override size changes.  Our size is the size
 * of the LVB and nothing else.
 */
static int dlmfs_file_setattr(struct dentry *dentry, struct iattr *attr)
{
	int error;
	struct inode *inode = dentry->d_inode;

	attr->ia_valid &= ~ATTR_SIZE;
	error = inode_change_ok(inode, attr);
	if (error)
		return error;

	setattr_copy(inode, attr);
	mark_inode_dirty(inode);
	return 0;
}

static unsigned int dlmfs_file_poll(struct file *file, poll_table *wait)
{
	int event = 0;
	struct inode *inode = file->f_path.dentry->d_inode;
	struct dlmfs_inode_private *ip = DLMFS_I(inode);

	poll_wait(file, &ip->ip_lockres.l_event, wait);

	spin_lock(&ip->ip_lockres.l_lock);
	if (ip->ip_lockres.l_flags & USER_LOCK_BLOCKED)
		event = POLLIN | POLLRDNORM;
	spin_unlock(&ip->ip_lockres.l_lock);

	return event;
}

static ssize_t dlmfs_file_read(struct file *filp,
			       char __user *buf,
			       size_t count,
			       loff_t *ppos)
{
	int bytes_left;
	ssize_t readlen, got;
	char *lvb_buf;
	struct inode *inode = filp->f_path.dentry->d_inode;

	mlog(0, "inode %lu, count = %zu, *ppos = %llu\n",
		inode->i_ino, count, *ppos);

	if (*ppos >= i_size_read(inode))
		return 0;

	if (!count)
		return 0;

	if (!access_ok(VERIFY_WRITE, buf, count))
		return -EFAULT;

	/* don't read past the lvb */
	if ((count + *ppos) > i_size_read(inode))
		readlen = i_size_read(inode) - *ppos;
	else
		readlen = count;

	lvb_buf = kmalloc(readlen, GFP_NOFS);
	if (!lvb_buf)
		return -ENOMEM;

	got = user_dlm_read_lvb(inode, lvb_buf, readlen);
	if (got) {
		BUG_ON(got != readlen);
		bytes_left = __copy_to_user(buf, lvb_buf, readlen);
		readlen -= bytes_left;
	} else
		readlen = 0;

	kfree(lvb_buf);

	*ppos = *ppos + readlen;

	mlog(0, "read %zd bytes\n", readlen);
	return readlen;
}

static ssize_t dlmfs_file_write(struct file *filp,
				const char __user *buf,
				size_t count,
				loff_t *ppos)
{
	int bytes_left;
	ssize_t writelen;
	char *lvb_buf;
	struct inode *inode = filp->f_path.dentry->d_inode;

	mlog(0, "inode %lu, count = %zu, *ppos = %llu\n",
		inode->i_ino, count, *ppos);

	if (*ppos >= i_size_read(inode))
		return -ENOSPC;

	if (!count)
		return 0;

	if (!access_ok(VERIFY_READ, buf, count))
		return -EFAULT;

	/* don't write past the lvb */
	if ((count + *ppos) > i_size_read(inode))
		writelen = i_size_read(inode) - *ppos;
	else
		writelen = count - *ppos;

	lvb_buf = kmalloc(writelen, GFP_NOFS);
	if (!lvb_buf)
		return -ENOMEM;

	bytes_left = copy_from_user(lvb_buf, buf, writelen);
	writelen -= bytes_left;
	if (writelen)
		user_dlm_write_lvb(inode, lvb_buf, writelen);

	kfree(lvb_buf);

	*ppos = *ppos + writelen;
	mlog(0, "wrote %zd bytes\n", writelen);
	return writelen;
}

static void dlmfs_init_once(void *foo)
{
	struct dlmfs_inode_private *ip =
		(struct dlmfs_inode_private *) foo;

	ip->ip_conn = NULL;
	ip->ip_parent = NULL;

	inode_init_once(&ip->ip_vfs_inode);
}

static struct inode *dlmfs_alloc_inode(struct super_block *sb)
{
	struct dlmfs_inode_private *ip;

	ip = kmem_cache_alloc(dlmfs_inode_cache, GFP_NOFS);
	if (!ip)
		return NULL;

	return &ip->ip_vfs_inode;
}

static void dlmfs_destroy_inode(struct inode *inode)
{
	kmem_cache_free(dlmfs_inode_cache, DLMFS_I(inode));
}

static void dlmfs_evict_inode(struct inode *inode)
{
	int status;
	struct dlmfs_inode_private *ip;

	end_writeback(inode);

	mlog(0, "inode %lu\n", inode->i_ino);

	ip = DLMFS_I(inode);

	if (S_ISREG(inode->i_mode)) {
		status = user_dlm_destroy_lock(&ip->ip_lockres);
		if (status < 0)
			mlog_errno(status);
		iput(ip->ip_parent);
		goto clear_fields;
	}

	mlog(0, "we're a directory, ip->ip_conn = 0x%p\n", ip->ip_conn);
	/* we must be a directory. If required, lets unregister the
	 * dlm context now. */
	if (ip->ip_conn)
		user_dlm_unregister(ip->ip_conn);
clear_fields:
	ip->ip_parent = NULL;
	ip->ip_conn = NULL;
}

static struct backing_dev_info dlmfs_backing_dev_info = {
	.name		= "ocfs2-dlmfs",
	.ra_pages	= 0,	/* No readahead */
	.capabilities	= BDI_CAP_NO_ACCT_AND_WRITEBACK,
};

static struct inode *dlmfs_get_root_inode(struct super_block *sb)
{
	struct inode *inode = new_inode(sb);
	int mode = S_IFDIR | 0755;
	struct dlmfs_inode_private *ip;

	if (inode) {
		ip = DLMFS_I(inode);

		inode->i_mode = mode;
		inode->i_uid = current_fsuid();
		inode->i_gid = current_fsgid();
		inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info;
		inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
		inc_nlink(inode);

		inode->i_fop = &simple_dir_operations;
		inode->i_op = &dlmfs_root_inode_operations;
	}

	return inode;
}

static struct inode *dlmfs_get_inode(struct inode *parent,
				     struct dentry *dentry,
				     int mode)
{
	struct super_block *sb = parent->i_sb;
	struct inode * inode = new_inode(sb);
	struct dlmfs_inode_private *ip;

	if (!inode)
		return NULL;

	inode->i_mode = mode;
	inode->i_uid = current_fsuid();
	inode->i_gid = current_fsgid();
	inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info;
	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;

	ip = DLMFS_I(inode);
	ip->ip_conn = DLMFS_I(parent)->ip_conn;

	switch (mode & S_IFMT) {
	default:
		/* for now we don't support anything other than
		 * directories and regular files. */
		BUG();
		break;
	case S_IFREG:
		inode->i_op = &dlmfs_file_inode_operations;
		inode->i_fop = &dlmfs_file_operations;

		i_size_write(inode,  DLM_LVB_LEN);

		user_dlm_lock_res_init(&ip->ip_lockres, dentry);

		/* released at clear_inode time, this insures that we
		 * get to drop the dlm reference on each lock *before*
		 * we call the unregister code for releasing parent
		 * directories. */
		ip->ip_parent = igrab(parent);
		BUG_ON(!ip->ip_parent);
		break;
	case S_IFDIR:
		inode->i_op = &dlmfs_dir_inode_operations;
		inode->i_fop = &simple_dir_operations;

		/* directory inodes start off with i_nlink ==
		 * 2 (for "." entry) */
		inc_nlink(inode);
		break;
	}

	if (parent->i_mode & S_ISGID) {
		inode->i_gid = parent->i_gid;
		if (S_ISDIR(mode))
			inode->i_mode |= S_ISGID;
	}

	return inode;
}

/*
 * File creation. Allocate an inode, and we're done..
 */
/* SMP-safe */
static int dlmfs_mkdir(struct inode * dir,
		       struct dentry * dentry,
		       int mode)
{
	int status;
	struct inode *inode = NULL;
	struct qstr *domain = &dentry->d_name;
	struct dlmfs_inode_private *ip;
	struct ocfs2_cluster_connection *conn;

	mlog(0, "mkdir %.*s\n", domain->len, domain->name);

	/* verify that we have a proper domain */
	if (domain->len >= GROUP_NAME_MAX) {
		status = -EINVAL;
		mlog(ML_ERROR, "invalid domain name for directory.\n");
		goto bail;
	}

	inode = dlmfs_get_inode(dir, dentry, mode | S_IFDIR);
	if (!inode) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

	ip = DLMFS_I(inode);

	conn = user_dlm_register(domain);
	if (IS_ERR(conn)) {
		status = PTR_ERR(conn);
		mlog(ML_ERROR, "Error %d could not register domain \"%.*s\"\n",
		     status, domain->len, domain->name);
		goto bail;
	}
	ip->ip_conn = conn;

	inc_nlink(dir);
	d_instantiate(dentry, inode);
	dget(dentry);	/* Extra count - pin the dentry in core */

	status = 0;
bail:
	if (status < 0)
		iput(inode);
	return status;
}

static int dlmfs_create(struct inode *dir,
			struct dentry *dentry,
			int mode,
			struct nameidata *nd)
{
	int status = 0;
	struct inode *inode;
	struct qstr *name = &dentry->d_name;

	mlog(0, "create %.*s\n", name->len, name->name);

	/* verify name is valid and doesn't contain any dlm reserved
	 * characters */
	if (name->len >= USER_DLM_LOCK_ID_MAX_LEN ||
	    name->name[0] == '$') {
		status = -EINVAL;
		mlog(ML_ERROR, "invalid lock name, %.*s\n", name->len,
		     name->name);
		goto bail;
	}

	inode = dlmfs_get_inode(dir, dentry, mode | S_IFREG);
	if (!inode) {
		status = -ENOMEM;
		mlog_errno(status);
		goto bail;
	}

	d_instantiate(dentry, inode);
	dget(dentry);	/* Extra count - pin the dentry in core */
bail:
	return status;
}

static int dlmfs_unlink(struct inode *dir,
			struct dentry *dentry)
{
	int status;
	struct inode *inode = dentry->d_inode;

	mlog(0, "unlink inode %lu\n", inode->i_ino);

	/* if there are no current holders, or none that are waiting
	 * to acquire a lock, this basically destroys our lockres. */
	status = user_dlm_destroy_lock(&DLMFS_I(inode)->ip_lockres);
	if (status < 0) {
		mlog(ML_ERROR, "unlink %.*s, error %d from destroy\n",
		     dentry->d_name.len, dentry->d_name.name, status);
		goto bail;
	}
	status = simple_unlink(dir, dentry);
bail:
	return status;
}

static int dlmfs_fill_super(struct super_block * sb,
			    void * data,
			    int silent)
{
	struct inode * inode;
	struct dentry * root;

	sb->s_maxbytes = MAX_LFS_FILESIZE;
	sb->s_blocksize = PAGE_CACHE_SIZE;
	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
	sb->s_magic = DLMFS_MAGIC;
	sb->s_op = &dlmfs_ops;
	inode = dlmfs_get_root_inode(sb);
	if (!inode)
		return -ENOMEM;

	root = d_alloc_root(inode);
	if (!root) {
		iput(inode);
		return -ENOMEM;
	}
	sb->s_root = root;
	return 0;
}

static const struct file_operations dlmfs_file_operations = {
	.open		= dlmfs_file_open,
	.release	= dlmfs_file_release,
	.poll		= dlmfs_file_poll,
	.read		= dlmfs_file_read,
	.write		= dlmfs_file_write,
	.llseek		= default_llseek,
};

static const struct inode_operations dlmfs_dir_inode_operations = {
	.create		= dlmfs_create,
	.lookup		= simple_lookup,
	.unlink		= dlmfs_unlink,
};

/* this way we can restrict mkdir to only the toplevel of the fs. */
static const struct inode_operations dlmfs_root_inode_operations = {
	.lookup		= simple_lookup,
	.mkdir		= dlmfs_mkdir,
	.rmdir		= simple_rmdir,
};

static const struct super_operations dlmfs_ops = {
	.statfs		= simple_statfs,
	.alloc_inode	= dlmfs_alloc_inode,
	.destroy_inode	= dlmfs_destroy_inode,
	.evict_inode	= dlmfs_evict_inode,
	.drop_inode	= generic_delete_inode,
};

static const struct inode_operations dlmfs_file_inode_operations = {
	.getattr	= simple_getattr,
	.setattr	= dlmfs_file_setattr,
};

static int dlmfs_get_sb(struct file_system_type *fs_type,
	int flags, const char *dev_name, void *data, struct vfsmount *mnt)
{
	return get_sb_nodev(fs_type, flags, data, dlmfs_fill_super, mnt);
}

static struct file_system_type dlmfs_fs_type = {
	.owner		= THIS_MODULE,
	.name		= "ocfs2_dlmfs",
	.get_sb		= dlmfs_get_sb,
	.kill_sb	= kill_litter_super,
};

static int __init init_dlmfs_fs(void)
{
	int status;
	int cleanup_inode = 0, cleanup_worker = 0;

	dlmfs_print_version();

	status = bdi_init(&dlmfs_backing_dev_info);
	if (status)
		return status;

	dlmfs_inode_cache = kmem_cache_create("dlmfs_inode_cache",
				sizeof(struct dlmfs_inode_private),
				0, (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
					SLAB_MEM_SPREAD),
				dlmfs_init_once);
	if (!dlmfs_inode_cache) {
		status = -ENOMEM;
		goto bail;
	}
	cleanup_inode = 1;

	user_dlm_worker = create_singlethread_workqueue("user_dlm");
	if (!user_dlm_worker) {
		status = -ENOMEM;
		goto bail;
	}
	cleanup_worker = 1;

	user_dlm_set_locking_protocol();
	status = register_filesystem(&dlmfs_fs_type);
bail:
	if (status) {
		if (cleanup_inode)
			kmem_cache_destroy(dlmfs_inode_cache);
		if (cleanup_worker)
			destroy_workqueue(user_dlm_worker);
		bdi_destroy(&dlmfs_backing_dev_info);
	} else
		printk("OCFS2 User DLM kernel interface loaded\n");
	return status;
}

static void __exit exit_dlmfs_fs(void)
{
	unregister_filesystem(&dlmfs_fs_type);

	flush_workqueue(user_dlm_worker);
	destroy_workqueue(user_dlm_worker);

	kmem_cache_destroy(dlmfs_inode_cache);

	bdi_destroy(&dlmfs_backing_dev_info);
}

MODULE_AUTHOR("Oracle");
MODULE_LICENSE("GPL");

module_init(init_dlmfs_fs)
module_exit(exit_dlmfs_fs)
Commit	Line	Data
8df08c89 MF	1	/* -- mode: c; c-basic-offset: 8; --
	2	* vim: noexpandtab sw=8 ts=8 sts=0:
	3	*
	4	* dlmfs.c
	5	*
	6	* Code which implements the kernel side of a minimal userspace
	7	* interface to our DLM. This file handles the virtual file system
	8	* used for communication with userspace. Credit should go to ramfs,
	9	* which was a template for the fs side of this module.
	10	*
	11	* Copyright (C) 2003, 2004 Oracle. All rights reserved.
	12	*
	13	* This program is free software; you can redistribute it and/or
	14	* modify it under the terms of the GNU General Public
	15	* License as published by the Free Software Foundation; either
	16	* version 2 of the License, or (at your option) any later version.
	17	*
	18	* This program is distributed in the hope that it will be useful,
	19	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	20	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	21	* General Public License for more details.
	22	*
	23	* You should have received a copy of the GNU General Public
	24	* License along with this program; if not, write to the
	25	* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
	26	* Boston, MA 021110-1307, USA.
	27	*/
	28
	29	/* Simple VFS hooks based on: */
	30	/*
	31	* Resizable simple ram filesystem for Linux.
	32	*
	33	* Copyright (C) 2000 Linus Torvalds.
	34	* 2000 Transmeta Corp.
	35	*/
	36
	37	#include <linux/module.h>
	38	#include <linux/fs.h>
	39	#include <linux/pagemap.h>
	40	#include <linux/types.h>
	41	#include <linux/slab.h>
	42	#include <linux/highmem.h>
	43	#include <linux/init.h>
	44	#include <linux/string.h>
8df08c89	45	#include <linux/backing-dev.h>
65b6f340	46	#include <linux/poll.h>
8df08c89 MF	47
	48	#include <asm/uaccess.h>
	49
0016eedc	50	#include "stackglue.h"
8df08c89	51	#include "userdlm.h"
8df08c89 MF	52	#include "dlmfsver.h"
	53
	54	#define MLOG_MASK_PREFIX ML_DLMFS
	55	#include "cluster/masklog.h"
	56
d24fbcda	57
ee9b6d61	58	static const struct super_operations dlmfs_ops;
00977a59	59	static const struct file_operations dlmfs_file_operations;
92e1d5be AV	60	static const struct inode_operations dlmfs_dir_inode_operations;
	61	static const struct inode_operations dlmfs_root_inode_operations;
	62	static const struct inode_operations dlmfs_file_inode_operations;
e18b890b	63	static struct kmem_cache *dlmfs_inode_cache;
8df08c89 MF	64
	65	struct workqueue_struct *user_dlm_worker;
	66
d24fbcda	67
14a437c2 JB	68
	69	/*
	70	* These are the ABI capabilities of dlmfs.
	71	*
	72	* Over time, dlmfs has added some features that were not part of the
	73	* initial ABI. Unfortunately, some of these features are not detectable
	74	* via standard usage. For example, Linux's default poll always returns
	75	* POLLIN, so there is no way for a caller of poll(2) to know when dlmfs
	76	* added poll support. Instead, we provide this list of new capabilities.
	77	*
	78	* Capabilities is a read-only attribute. We do it as a module parameter
	79	* so we can discover it whether dlmfs is built in, loaded, or even not
	80	* loaded.
	81	*
	82	* The ABI features are local to this machine's dlmfs mount. This is
	83	* distinct from the locking protocol, which is concerned with inter-node
	84	* interaction.
65b6f340 JB	85	*
	86	* Capabilities:
	87	* - bast : POLLIN against the file descriptor of a held lock
	88	* signifies a bast fired on the lock.
14a437c2	89	*/
cbe0e331	90	#define DLMFS_CAPABILITIES "bast stackglue"
14a437c2 JB	91	extern int param_set_dlmfs_capabilities(const char *val,
	92	struct kernel_param *kp)
	93	{
	94	printk(KERN_ERR "%s: readonly parameter\n", kp->name);
	95	return -EINVAL;
	96	}
	97	static int param_get_dlmfs_capabilities(char *buffer,
	98	struct kernel_param *kp)
	99	{
	100	return strlcpy(buffer, DLMFS_CAPABILITIES,
	101	strlen(DLMFS_CAPABILITIES) + 1);
	102	}
	103	module_param_call(capabilities, param_set_dlmfs_capabilities,
	104	param_get_dlmfs_capabilities, NULL, 0444);
	105	MODULE_PARM_DESC(capabilities, DLMFS_CAPABILITIES);
	106
	107
8df08c89 MF	108	/*
	109	* decodes a set of open flags into a valid lock level and a set of flags.
	110	* returns < 0 if we have invalid flags
	111	* flags which mean something to us:
	112	* O_RDONLY -> PRMODE level
	113	* O_WRONLY -> EXMODE level
	114	*
efd647f7	115	* O_NONBLOCK -> NOQUEUE
8df08c89 MF	116	*/
	117	static int dlmfs_decode_open_flags(int open_flags,
	118	int *level,
	119	int *flags)
	120	{
	121	if (open_flags & (O_WRONLY\|O_RDWR))
efd647f7	122	*level = DLM_LOCK_EX;
8df08c89	123	else
efd647f7	124	*level = DLM_LOCK_PR;
8df08c89 MF	125
	126	*flags = 0;
	127	if (open_flags & O_NONBLOCK)
efd647f7	128	*flags \|= DLM_LKF_NOQUEUE;
8df08c89 MF	129
	130	return 0;
	131	}
	132
	133	static int dlmfs_file_open(struct inode *inode,
	134	struct file *file)
	135	{
	136	int status, level, flags;
	137	struct dlmfs_filp_private *fp = NULL;
	138	struct dlmfs_inode_private *ip;
	139
	140	if (S_ISDIR(inode->i_mode))
	141	BUG();
	142
	143	mlog(0, "open called on inode %lu, flags 0x%x\n", inode->i_ino,
	144	file->f_flags);
	145
	146	status = dlmfs_decode_open_flags(file->f_flags, &level, &flags);
	147	if (status < 0)
	148	goto bail;
	149
	150	/* We don't want to honor O_APPEND at read/write time as it
	151	* doesn't make sense for LVB writes. */
	152	file->f_flags &= ~O_APPEND;
	153
ad8100e0	154	fp = kmalloc(sizeof(*fp), GFP_NOFS);
8df08c89 MF	155	if (!fp) {
	156	status = -ENOMEM;
	157	goto bail;
	158	}
	159	fp->fp_lock_level = level;
	160
	161	ip = DLMFS_I(inode);
	162
	163	status = user_dlm_cluster_lock(&ip->ip_lockres, level, flags);
	164	if (status < 0) {
	165	/* this is a strange error to return here but I want
	166	* to be able userspace to be able to distinguish a
	167	* valid lock request from one that simply couldn't be
	168	* granted. */
efd647f7	169	if (flags & DLM_LKF_NOQUEUE && status == -EAGAIN)
8df08c89 MF	170	status = -ETXTBSY;
	171	kfree(fp);
	172	goto bail;
	173	}
	174
	175	file->private_data = fp;
	176	bail:
	177	return status;
	178	}
	179
	180	static int dlmfs_file_release(struct inode *inode,
	181	struct file *file)
	182	{
	183	int level, status;
	184	struct dlmfs_inode_private *ip = DLMFS_I(inode);
33fa1d90	185	struct dlmfs_filp_private *fp = file->private_data;
8df08c89 MF	186
	187	if (S_ISDIR(inode->i_mode))
	188	BUG();
	189
	190	mlog(0, "close called on inode %lu\n", inode->i_ino);
	191
	192	status = 0;
	193	if (fp) {
	194	level = fp->fp_lock_level;
efd647f7	195	if (level != DLM_LOCK_IV)
8df08c89 MF	196	user_dlm_cluster_unlock(&ip->ip_lockres, level);
	197
	198	kfree(fp);
	199	file->private_data = NULL;
	200	}
	201
	202	return 0;
	203	}
	204
e8fce482 JB	205	/*
	206	* We do ->setattr() just to override size changes. Our size is the size
	207	* of the LVB and nothing else.
	208	*/
	209	static int dlmfs_file_setattr(struct dentry dentry, struct iattr attr)
	210	{
	211	int error;
	212	struct inode *inode = dentry->d_inode;
	213
	214	attr->ia_valid &= ~ATTR_SIZE;
	215	error = inode_change_ok(inode, attr);
1025774c CH	216	if (error)
1025774c CH	217	return error;
e8fce482	218
1025774c CH	219	setattr_copy(inode, attr);
	220	mark_inode_dirty(inode);
	221	return 0;
e8fce482 JB	222	}
e8fce482 JB	223
65b6f340 JB	224	static unsigned int dlmfs_file_poll(struct file file, poll_table wait)
	225	{
	226	int event = 0;
	227	struct inode *inode = file->f_path.dentry->d_inode;
	228	struct dlmfs_inode_private *ip = DLMFS_I(inode);
	229
	230	poll_wait(file, &ip->ip_lockres.l_event, wait);
	231
	232	spin_lock(&ip->ip_lockres.l_lock);
	233	if (ip->ip_lockres.l_flags & USER_LOCK_BLOCKED)
	234	event = POLLIN \| POLLRDNORM;
	235	spin_unlock(&ip->ip_lockres.l_lock);
	236
	237	return event;
	238	}
	239
8df08c89 MF	240	static ssize_t dlmfs_file_read(struct file *filp,
	241	char __user *buf,
	242	size_t count,
	243	loff_t *ppos)
	244	{
	245	int bytes_left;
0016eedc	246	ssize_t readlen, got;
8df08c89	247	char *lvb_buf;
d28c9174	248	struct inode *inode = filp->f_path.dentry->d_inode;
8df08c89 MF	249
	250	mlog(0, "inode %lu, count = %zu, *ppos = %llu\n",
	251	inode->i_ino, count, *ppos);
	252
	253	if (*ppos >= i_size_read(inode))
	254	return 0;
	255
	256	if (!count)
	257	return 0;
	258
	259	if (!access_ok(VERIFY_WRITE, buf, count))
	260	return -EFAULT;
	261
	262	/* don't read past the lvb */
	263	if ((count + *ppos) > i_size_read(inode))
	264	readlen = i_size_read(inode) - *ppos;
	265	else
a36d515c	266	readlen = count;
8df08c89	267
ad8100e0	268	lvb_buf = kmalloc(readlen, GFP_NOFS);
8df08c89 MF	269	if (!lvb_buf)
	270	return -ENOMEM;
	271
0016eedc JB	272	got = user_dlm_read_lvb(inode, lvb_buf, readlen);
	273	if (got) {
	274	BUG_ON(got != readlen);
	275	bytes_left = __copy_to_user(buf, lvb_buf, readlen);
	276	readlen -= bytes_left;
	277	} else
	278	readlen = 0;
8df08c89 MF	279
	280	kfree(lvb_buf);
	281
	282	ppos = ppos + readlen;
	283
	284	mlog(0, "read %zd bytes\n", readlen);
	285	return readlen;
	286	}
	287
	288	static ssize_t dlmfs_file_write(struct file *filp,
	289	const char __user *buf,
	290	size_t count,
	291	loff_t *ppos)
	292	{
	293	int bytes_left;
	294	ssize_t writelen;
	295	char *lvb_buf;
d28c9174	296	struct inode *inode = filp->f_path.dentry->d_inode;
8df08c89 MF	297
	298	mlog(0, "inode %lu, count = %zu, *ppos = %llu\n",
	299	inode->i_ino, count, *ppos);
	300
	301	if (*ppos >= i_size_read(inode))
	302	return -ENOSPC;
	303
	304	if (!count)
	305	return 0;
	306
	307	if (!access_ok(VERIFY_READ, buf, count))
	308	return -EFAULT;
	309
	310	/* don't write past the lvb */
	311	if ((count + *ppos) > i_size_read(inode))
	312	writelen = i_size_read(inode) - *ppos;
	313	else
	314	writelen = count - *ppos;
	315
ad8100e0	316	lvb_buf = kmalloc(writelen, GFP_NOFS);
8df08c89 MF	317	if (!lvb_buf)
	318	return -ENOMEM;
	319
	320	bytes_left = copy_from_user(lvb_buf, buf, writelen);
	321	writelen -= bytes_left;
	322	if (writelen)
	323	user_dlm_write_lvb(inode, lvb_buf, writelen);
	324
	325	kfree(lvb_buf);
	326
	327	ppos = ppos + writelen;
	328	mlog(0, "wrote %zd bytes\n", writelen);
	329	return writelen;
	330	}
	331
51cc5068	332	static void dlmfs_init_once(void *foo)
8df08c89 MF	333	{
	334	struct dlmfs_inode_private *ip =
	335	(struct dlmfs_inode_private *) foo;
	336
0016eedc	337	ip->ip_conn = NULL;
a35afb83	338	ip->ip_parent = NULL;
8df08c89	339
a35afb83	340	inode_init_once(&ip->ip_vfs_inode);
8df08c89 MF	341	}
	342
	343	static struct inode dlmfs_alloc_inode(struct super_block sb)
	344	{
	345	struct dlmfs_inode_private *ip;
	346
e6b4f8da	347	ip = kmem_cache_alloc(dlmfs_inode_cache, GFP_NOFS);
8df08c89 MF	348	if (!ip)
	349	return NULL;
	350
	351	return &ip->ip_vfs_inode;
	352	}
	353
	354	static void dlmfs_destroy_inode(struct inode *inode)
	355	{
	356	kmem_cache_free(dlmfs_inode_cache, DLMFS_I(inode));
	357	}
	358
b57922d9	359	static void dlmfs_evict_inode(struct inode *inode)
8df08c89 MF	360	{
	361	int status;
	362	struct dlmfs_inode_private *ip;
	363
b57922d9	364	end_writeback(inode);
8df08c89 MF	365
	366	mlog(0, "inode %lu\n", inode->i_ino);
	367
	368	ip = DLMFS_I(inode);
	369
	370	if (S_ISREG(inode->i_mode)) {
	371	status = user_dlm_destroy_lock(&ip->ip_lockres);
	372	if (status < 0)
	373	mlog_errno(status);
	374	iput(ip->ip_parent);
	375	goto clear_fields;
	376	}
	377
0016eedc	378	mlog(0, "we're a directory, ip->ip_conn = 0x%p\n", ip->ip_conn);
8df08c89 MF	379	/* we must be a directory. If required, lets unregister the
8df08c89 MF	380	* dlm context now. */
0016eedc JB	381	if (ip->ip_conn)
0016eedc JB	382	user_dlm_unregister(ip->ip_conn);
8df08c89 MF	383	clear_fields:
8df08c89 MF	384	ip->ip_parent = NULL;
0016eedc	385	ip->ip_conn = NULL;
8df08c89 MF	386	}
	387
	388	static struct backing_dev_info dlmfs_backing_dev_info = {
d993831f	389	.name = "ocfs2-dlmfs",
8df08c89	390	.ra_pages = 0, /* No readahead */
e4ad08fe	391	.capabilities = BDI_CAP_NO_ACCT_AND_WRITEBACK,
8df08c89 MF	392	};
	393
	394	static struct inode dlmfs_get_root_inode(struct super_block sb)
	395	{
	396	struct inode *inode = new_inode(sb);
	397	int mode = S_IFDIR \| 0755;
	398	struct dlmfs_inode_private *ip;
	399
	400	if (inode) {
	401	ip = DLMFS_I(inode);
	402
	403	inode->i_mode = mode;
b19c2a3b DH	404	inode->i_uid = current_fsuid();
b19c2a3b DH	405	inode->i_gid = current_fsgid();
8df08c89 MF	406	inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info;
8df08c89 MF	407	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
d8c76e6f	408	inc_nlink(inode);
8df08c89 MF	409
	410	inode->i_fop = &simple_dir_operations;
	411	inode->i_op = &dlmfs_root_inode_operations;
	412	}
	413
	414	return inode;
	415	}
	416
	417	static struct inode dlmfs_get_inode(struct inode parent,
	418	struct dentry *dentry,
	419	int mode)
	420	{
	421	struct super_block *sb = parent->i_sb;
	422	struct inode * inode = new_inode(sb);
	423	struct dlmfs_inode_private *ip;
	424
	425	if (!inode)
	426	return NULL;
	427
	428	inode->i_mode = mode;
b19c2a3b DH	429	inode->i_uid = current_fsuid();
b19c2a3b DH	430	inode->i_gid = current_fsgid();
8df08c89 MF	431	inode->i_mapping->backing_dev_info = &dlmfs_backing_dev_info;
	432	inode->i_atime = inode->i_mtime = inode->i_ctime = CURRENT_TIME;
	433
	434	ip = DLMFS_I(inode);
0016eedc	435	ip->ip_conn = DLMFS_I(parent)->ip_conn;
8df08c89 MF	436
	437	switch (mode & S_IFMT) {
	438	default:
	439	/* for now we don't support anything other than
	440	* directories and regular files. */
	441	BUG();
	442	break;
	443	case S_IFREG:
	444	inode->i_op = &dlmfs_file_inode_operations;
	445	inode->i_fop = &dlmfs_file_operations;
	446
	447	i_size_write(inode, DLM_LVB_LEN);
	448
	449	user_dlm_lock_res_init(&ip->ip_lockres, dentry);
	450
	451	/* released at clear_inode time, this insures that we
	452	* get to drop the dlm reference on each lock before
	453	* we call the unregister code for releasing parent
	454	* directories. */
	455	ip->ip_parent = igrab(parent);
	456	BUG_ON(!ip->ip_parent);
	457	break;
	458	case S_IFDIR:
	459	inode->i_op = &dlmfs_dir_inode_operations;
	460	inode->i_fop = &simple_dir_operations;
	461
	462	/* directory inodes start off with i_nlink ==
	463	* 2 (for "." entry) */
d8c76e6f	464	inc_nlink(inode);
8df08c89 MF	465	break;
	466	}
	467
	468	if (parent->i_mode & S_ISGID) {
	469	inode->i_gid = parent->i_gid;
	470	if (S_ISDIR(mode))
	471	inode->i_mode \|= S_ISGID;
	472	}
	473
	474	return inode;
	475	}
	476
	477	/*
	478	* File creation. Allocate an inode, and we're done..
	479	*/
	480	/* SMP-safe */
	481	static int dlmfs_mkdir(struct inode * dir,
	482	struct dentry * dentry,
	483	int mode)
	484	{
	485	int status;
	486	struct inode *inode = NULL;
	487	struct qstr *domain = &dentry->d_name;
	488	struct dlmfs_inode_private *ip;
0016eedc	489	struct ocfs2_cluster_connection *conn;
8df08c89 MF	490
	491	mlog(0, "mkdir %.*s\n", domain->len, domain->name);
	492
	493	/* verify that we have a proper domain */
0016eedc	494	if (domain->len >= GROUP_NAME_MAX) {
8df08c89 MF	495	status = -EINVAL;
	496	mlog(ML_ERROR, "invalid domain name for directory.\n");
	497	goto bail;
	498	}
	499
	500	inode = dlmfs_get_inode(dir, dentry, mode \| S_IFDIR);
	501	if (!inode) {
	502	status = -ENOMEM;
	503	mlog_errno(status);
	504	goto bail;
	505	}
	506
	507	ip = DLMFS_I(inode);
	508
0016eedc JB	509	conn = user_dlm_register(domain);
	510	if (IS_ERR(conn)) {
	511	status = PTR_ERR(conn);
8df08c89 MF	512	mlog(ML_ERROR, "Error %d could not register domain \"%.*s\"\n",
	513	status, domain->len, domain->name);
	514	goto bail;
	515	}
0016eedc	516	ip->ip_conn = conn;
8df08c89	517
d8c76e6f	518	inc_nlink(dir);
8df08c89 MF	519	d_instantiate(dentry, inode);
	520	dget(dentry); /* Extra count - pin the dentry in core */
	521
	522	status = 0;
	523	bail:
	524	if (status < 0)
	525	iput(inode);
	526	return status;
	527	}
	528
	529	static int dlmfs_create(struct inode *dir,
	530	struct dentry *dentry,
	531	int mode,
	532	struct nameidata *nd)
	533	{
	534	int status = 0;
	535	struct inode *inode;
	536	struct qstr *name = &dentry->d_name;
	537
	538	mlog(0, "create %.*s\n", name->len, name->name);
	539
	540	/* verify name is valid and doesn't contain any dlm reserved
	541	* characters */
	542	if (name->len >= USER_DLM_LOCK_ID_MAX_LEN \|\|
	543	name->name[0] == '$') {
	544	status = -EINVAL;
	545	mlog(ML_ERROR, "invalid lock name, %.*s\n", name->len,
	546	name->name);
	547	goto bail;
	548	}
	549
	550	inode = dlmfs_get_inode(dir, dentry, mode \| S_IFREG);
	551	if (!inode) {
	552	status = -ENOMEM;
	553	mlog_errno(status);
	554	goto bail;
	555	}
	556
	557	d_instantiate(dentry, inode);
	558	dget(dentry); /* Extra count - pin the dentry in core */
	559	bail:
	560	return status;
	561	}
	562
	563	static int dlmfs_unlink(struct inode *dir,
	564	struct dentry *dentry)
	565	{
	566	int status;
	567	struct inode *inode = dentry->d_inode;
	568
	569	mlog(0, "unlink inode %lu\n", inode->i_ino);
	570
	571	/* if there are no current holders, or none that are waiting
	572	* to acquire a lock, this basically destroys our lockres. */
	573	status = user_dlm_destroy_lock(&DLMFS_I(inode)->ip_lockres);
	574	if (status < 0) {
	575	mlog(ML_ERROR, "unlink %.*s, error %d from destroy\n",
	576	dentry->d_name.len, dentry->d_name.name, status);
	577	goto bail;
	578	}
	579	status = simple_unlink(dir, dentry);
	580	bail:
	581	return status;
	582	}
583
584	static int dlmfs_fill_super(struct super_block * sb,
585	void * data,
586	int silent)
587	{
588	struct inode * inode;
589	struct dentry * root;
590
591	sb->s_maxbytes = MAX_LFS_FILESIZE;
592	sb->s_blocksize = PAGE_CACHE_SIZE;
593	sb->s_blocksize_bits = PAGE_CACHE_SHIFT;
594	sb->s_magic = DLMFS_MAGIC;
595	sb->s_op = &dlmfs_ops;
596	inode = dlmfs_get_root_inode(sb);
597	if (!inode)
598	return -ENOMEM;
599
600	root = d_alloc_root(inode);
601	if (!root) {
602	iput(inode);
603	return -ENOMEM;
604	}
605	sb->s_root = root;
606	return 0;
607	}
608
00977a59	609	static const struct file_operations dlmfs_file_operations = {
8df08c89 MF	610	.open = dlmfs_file_open,
8df08c89 MF	611	.release = dlmfs_file_release,
65b6f340	612	.poll = dlmfs_file_poll,
8df08c89 MF	613	.read = dlmfs_file_read,
8df08c89 MF	614	.write = dlmfs_file_write,
6038f373	615	.llseek = default_llseek,
8df08c89 MF	616	};
8df08c89 MF	617
92e1d5be	618	static const struct inode_operations dlmfs_dir_inode_operations = {
8df08c89 MF	619	.create = dlmfs_create,
	620	.lookup = simple_lookup,
	621	.unlink = dlmfs_unlink,
	622	};
	623
	624	/* this way we can restrict mkdir to only the toplevel of the fs. */
92e1d5be	625	static const struct inode_operations dlmfs_root_inode_operations = {
8df08c89 MF	626	.lookup = simple_lookup,
	627	.mkdir = dlmfs_mkdir,
	628	.rmdir = simple_rmdir,
	629	};
	630
ee9b6d61	631	static const struct super_operations dlmfs_ops = {
8df08c89 MF	632	.statfs = simple_statfs,
	633	.alloc_inode = dlmfs_alloc_inode,
	634	.destroy_inode = dlmfs_destroy_inode,
b57922d9	635	.evict_inode = dlmfs_evict_inode,
8df08c89 MF	636	.drop_inode = generic_delete_inode,
	637	};
	638
92e1d5be	639	static const struct inode_operations dlmfs_file_inode_operations = {
8df08c89	640	.getattr = simple_getattr,
e8fce482	641	.setattr = dlmfs_file_setattr,
8df08c89 MF	642	};
8df08c89 MF	643
454e2398 DH	644	static int dlmfs_get_sb(struct file_system_type *fs_type,
454e2398 DH	645	int flags, const char dev_name, void data, struct vfsmount *mnt)
8df08c89	646	{
454e2398	647	return get_sb_nodev(fs_type, flags, data, dlmfs_fill_super, mnt);
8df08c89 MF	648	}
	649
	650	static struct file_system_type dlmfs_fs_type = {
	651	.owner = THIS_MODULE,
	652	.name = "ocfs2_dlmfs",
	653	.get_sb = dlmfs_get_sb,
	654	.kill_sb = kill_litter_super,
	655	};
	656
	657	static int __init init_dlmfs_fs(void)
	658	{
	659	int status;
	660	int cleanup_inode = 0, cleanup_worker = 0;
	661
	662	dlmfs_print_version();
	663
e0bf68dd PZ	664	status = bdi_init(&dlmfs_backing_dev_info);
	665	if (status)
	666	return status;
	667
8df08c89 MF	668	dlmfs_inode_cache = kmem_cache_create("dlmfs_inode_cache",
8df08c89 MF	669	sizeof(struct dlmfs_inode_private),
fffb60f9 PJ	670	0, (SLAB_HWCACHE_ALIGN\|SLAB_RECLAIM_ACCOUNT\|
fffb60f9 PJ	671	SLAB_MEM_SPREAD),
20c2df83	672	dlmfs_init_once);
07d9a395 CL	673	if (!dlmfs_inode_cache) {
07d9a395 CL	674	status = -ENOMEM;
e0bf68dd	675	goto bail;
07d9a395	676	}
8df08c89 MF	677	cleanup_inode = 1;
	678
	679	user_dlm_worker = create_singlethread_workqueue("user_dlm");
	680	if (!user_dlm_worker) {
	681	status = -ENOMEM;
	682	goto bail;
	683	}
	684	cleanup_worker = 1;
	685
0016eedc	686	user_dlm_set_locking_protocol();
8df08c89 MF	687	status = register_filesystem(&dlmfs_fs_type);
	688	bail:
	689	if (status) {
	690	if (cleanup_inode)
	691	kmem_cache_destroy(dlmfs_inode_cache);
	692	if (cleanup_worker)
	693	destroy_workqueue(user_dlm_worker);
e0bf68dd	694	bdi_destroy(&dlmfs_backing_dev_info);
8df08c89 MF	695	} else
	696	printk("OCFS2 User DLM kernel interface loaded\n");
	697	return status;
	698	}
	699
	700	static void __exit exit_dlmfs_fs(void)
	701	{
	702	unregister_filesystem(&dlmfs_fs_type);
	703
	704	flush_workqueue(user_dlm_worker);
	705	destroy_workqueue(user_dlm_worker);
	706
1a1d92c1	707	kmem_cache_destroy(dlmfs_inode_cache);
e0bf68dd PZ	708
e0bf68dd PZ	709	bdi_destroy(&dlmfs_backing_dev_info);
8df08c89 MF	710	}
	711
	712	MODULE_AUTHOR("Oracle");
	713	MODULE_LICENSE("GPL");
	714
	715	module_init(init_dlmfs_fs)
	716	module_exit(exit_dlmfs_fs)