dlmfs.c

ocfs1.4.1 oracle分布式文件系统

/* -*- 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 <asm/uaccess.h>


#include "cluster/nodemanager.h"
#include "cluster/heartbeat.h"
#include "cluster/tcp.h"

#include "dlmapi.h"

#include "userdlm.h"

#include "dlmfsver.h"

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

#include "ocfs2_lockingver.h"

#ifdef SOP_IS_NOT_CONST
static struct super_operations dlmfs_ops;
#else
static const struct super_operations dlmfs_ops;
#endif
#ifdef FOP_IS_NOT_CONST
static struct file_operations dlmfs_file_operations;
#else
static const struct file_operations dlmfs_file_operations;
#endif
#ifdef IOP_IS_NOT_CONST
static struct inode_operations dlmfs_dir_inode_operations;
static struct inode_operations dlmfs_root_inode_operations;
static struct inode_operations dlmfs_file_inode_operations;
#else
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;
#endif
static struct kmem_cache *dlmfs_inode_cache;

struct workqueue_struct *user_dlm_worker;

/*
 * This is the userdlmfs locking protocol version.
 *
 * See fs/ocfs2/dlmglue.c for more details on locking versions.
 */
static const struct dlm_protocol_version user_locking_protocol = {
	.pv_major = OCFS2_LOCKING_PROTOCOL_MAJOR,
	.pv_minor = OCFS2_LOCKING_PROTOCOL_MINOR,
};

/*
 * 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 -> LKM_NOQUEUE
 */
static int dlmfs_decode_open_flags(int open_flags,
				   int *level,
				   int *flags)
{
	if (open_flags & (O_WRONLY|O_RDWR))
		*level = LKM_EXMODE;
	else
		*level = LKM_PRMODE;

	*flags = 0;
	if (open_flags & O_NONBLOCK)
		*flags |= LKM_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 & LKM_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 =
		(struct dlmfs_filp_private *) 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 != LKM_IVMODE)
			user_dlm_cluster_unlock(&ip->ip_lockres, level);

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

	return 0;
}

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

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

	user_dlm_read_lvb(inode, lvb_buf, readlen);
	bytes_left = __copy_to_user(buf, lvb_buf, readlen);
	readlen -= bytes_left;

	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_dentry(filp)->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 kmem_cache *cachep,
			    unsigned long flags)
{
	struct dlmfs_inode_private *ip =
		(struct dlmfs_inode_private *) foo;

	ip->ip_dlm = 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_clear_inode(struct inode *inode)
{
	int status;
	struct dlmfs_inode_private *ip;

	if (!inode)
		return;

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

static struct backing_dev_info dlmfs_backing_dev_info = {
	.ra_pages	= 0,	/* No readahead */
	.capabilities	= BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_WRITEBACK,