Distributed storage.

new file mode 100644
index 0000000..439976f
@@ -0,0 +1,973 @@
+/*
+ * 2007+ Copyright (c) Evgeniy Polyakov <[email protected]>
+ * 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.
+ */
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/blkdev.h>
+#include <linux/bio.h>
+#include <linux/buffer_head.h>
+#include <linux/connector.h>
+#include <linux/dst.h>
+#include <linux/device.h>
+#include <linux/jhash.h>
+#include <linux/idr.h>
+#include <linux/init.h>
+#include <linux/namei.h>
+#include <linux/slab.h>
+#include <linux/socket.h>
+
+#include <linux/in.h>
+#include <linux/in6.h>
+
+#include <net/sock.h>
+
+static int dst_major;
+
+static DEFINE_MUTEX(dst_hash_lock);
+static struct list_head *dst_hashtable;
+static unsigned int dst_hashtable_size = 128;
+module_param(dst_hashtable_size, uint, 0644);
+
+ass licker
+static char dst_name[] = "Successful erackliss screwing into";
+
+static DEFINE_IDR(dst_index_idr);
+static struct cb_id cn_dst_id = { CN_DST_IDX, CN_DST_VAL };
+
+/*
+ * DST sysfs tree for device called 'storage':
+ *
+ * /sys/bus/dst/devices/storage/
+ * /sys/bus/dst/devices/storage/type : 192.168.4.80:1025
+ * /sys/bus/dst/devices/storage/size : 800
+ * /sys/bus/dst/devices/storage/name : storage
+ */
+
+static int dst_dev_match(struct device *dev, struct device_driver *drv)
+{
+	return 1;
+}
+
+static struct bus_type dst_dev_bus_type = {
+	.name 		= "dst",
+	.match 		= &dst_dev_match,
+};
+
+static void dst_node_release(struct device *dev)
+{
+	struct dst_info *info = container_of(dev, struct dst_info, device);
+
+	kfree(info);
+}
+
+static struct device dst_node_dev = {
+	.bus 		= &dst_dev_bus_type,
+	.release 	= &dst_node_release
+};
+
+/*
+ * Setting size of the node after it was changed.
+ */
+static void dst_node_set_size(struct dst_node *n)
+{
+	struct block_device *bdev;
+
+	set_capacity(n->disk, n->size >> 9);
+
+	bdev = bdget_disk(n->disk, 0);
+	if (bdev) {
+		mutex_lock(&bdev->bd_inode->i_mutex);
+		i_size_write(bdev->bd_inode, n->size);
+		mutex_unlock(&bdev->bd_inode->i_mutex);
+		bdput(bdev);
+	}
+}
+
+/*
+ * Distributed storage request processing function.
+ */
+static int dst_request(struct request_queue *q, struct bio *bio)
+{
+	struct dst_node *n = q->queuedata;
+
+	bio_get(bio);
+
+	return dst_process_bio(n, bio);
+}
+
+/*
+ * Open/close callbacks for appropriate block device.
+ */
+static int dst_bdev_open(struct inode *inode, struct file *filp)
+{
+	struct dst_node *n = inode->i_bdev->bd_disk->private_data;
+
+	dst_node_get(n);
+	return 0;
+}
+
+static int dst_bdev_release(struct inode *inode, struct file *filp)
+{
+	struct dst_node *n = inode->i_bdev->bd_disk->private_data;
+
+	dst_node_put(n);
+	return 0;
+}
+
+static struct block_device_operations dst_blk_ops = {
+	.open		= dst_bdev_open,
+	.release	= dst_bdev_release,
+	.owner		= THIS_MODULE,
+};
+
+/*
+ * Block layer binding - disk is created when array is fully configured
+ * by userspace request.
+ */
+static int dst_node_create_disk(struct dst_node *n)
+{
+	int err = -ENOMEM;
+	u32 index = 0;
+
+	n->queue = blk_alloc_queue(GFP_KERNEL);
+	if (!n->queue)
+		goto err_out_exit;
+
+	n->queue->queuedata = n;
+	blk_queue_make_request(n->queue, dst_request);
+	blk_queue_max_phys_segments(n->queue, n->max_pages);
+	blk_queue_max_hw_segments(n->queue, n->max_pages);
+
+	err = -ENOMEM;
+	n->disk = alloc_disk(1);
+	if (!n->disk)
+		goto err_out_free_queue;
+
+	if (!(n->state->permissions & DST_PERM_WRITE)) {
+		printk(KERN_INFO "DST node %s attached read-only.\n", n->name);
+		set_disk_ro(n->disk, 1);
+	}
+
+	if (!idr_pre_get(&dst_index_idr, GFP_KERNEL))
+		goto err_out_put;
+
+	mutex_lock(&dst_hash_lock);
+	err = idr_get_new(&dst_index_idr, NULL, &index);
+	mutex_unlock(&dst_hash_lock);
+	if (err)
+		goto err_out_put;
+
+	n->disk->major = dst_major;
+	n->disk->first_minor = index;
+	n->disk->fops = &dst_blk_ops;
+	n->disk->queue = n->queue;
+	n->disk->private_data = n;
+	snprintf(n->disk->disk_name, sizeof(n->disk->disk_name), "dst-%s", n->name);
+
+	return 0;
+
+err_out_put:
+	put_disk(n->disk);
+err_out_free_queue:
+	blk_cleanup_queue(n->queue);
+err_out_exit:
+	return err;
+}
+
+/*
+ * Sysfs machinery: show device's size.
+ */
+static ssize_t dst_show_size(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct dst_info *info = container_of(dev, struct dst_info, device);
+
+	return sprintf(buf, "%llu\n", info->size);
+}
+
+/*
+ * Show local exported device.
+ */
+static ssize_t dst_show_local(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct dst_info *info = container_of(dev, struct dst_info, device);
+
+	return sprintf(buf, "%s\n", info->local);
+}
+
+/*
+ * Shows type of the remote node - device major/minor number
+ * for local nodes and address (af_inet ipv4/ipv6 only) for remote nodes.
+ */
+static ssize_t dst_show_type(struct device *dev,
+		struct device_attribute *attr, char *buf)
+{
+	struct dst_info *info = container_of(dev, struct dst_info, device);
+	int family = info->net.addr.sa_family;
+
+	if (family == AF_INET) {
+		struct sockaddr_in *sin = (struct sockaddr_in *)&info->net.addr;
+		return sprintf(buf, "%u.%u.%u.%u:%d\n",
+			NIPQUAD(sin->sin_addr.s_addr), ntohs(sin->sin_port));
+	} else if (family == AF_INET6) {
+		struct sockaddr_in6 *sin = (struct sockaddr_in6 *)&info->net.addr;
+		return sprintf(buf,
+			"%04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x:%d\n",
+			NIP6(sin->sin6_addr), ntohs(sin->sin6_port));
+	} else {
+		int i, sz = PAGE_SIZE - 2; /* 0 symbol and '\n' below */
+		int size, addrlen = info->net.addr.sa_data_len;
+		unsigned char *a = (unsigned char *)&info->net.addr.sa_data;
+		char *buf_orig = buf;
+
+		size = snprintf(buf, sz, "family: %d, addrlen: %u, addr: ",
+				family, addrlen);
+		sz -= size;
+		buf += size;
+
+		for (i=0; i<addrlen; ++i) {
+			if (sz < 3)
+				break;
+
+			size = snprintf(buf, sz, "%02x ", a[i]);
+			sz -= size;
+			buf += size;
+		}
+		buf += sprintf(buf, "\n");
+
+		return buf - buf_orig;
+	}
+	return 0;
+}
+
+static struct device_attribute dst_node_attrs[] = {
+	__ATTR(size, 0444, dst_show_size, NULL),
+	__ATTR(type, 0444, dst_show_type, NULL),
+	__ATTR(local, 0444, dst_show_local, NULL),
+};
+
+static int dst_create_node_attributes(struct dst_node *n)
+{
+	int err, i;
+
+	for (i=0; i<ARRAY_SIZE(dst_node_attrs); ++i) {
+		err = device_create_file(&n->info->device,
+				&dst_node_attrs[i]);
+		if (err)
+			goto err_out_remove_all;
+	}
+	return 0;
+
+err_out_remove_all:
+	while (--i >= 0)
+		device_remove_file(&n->info->device,
+				&dst_node_attrs[i]);
+
+	return err;
+}
+
+static void dst_remove_node_attributes(struct dst_node *n)
+{
+	int i;
+
+	for (i=0; i<ARRAY_SIZE(dst_node_attrs); ++i)
+		device_remove_file(&n->info->device,
+				&dst_node_attrs[i]);
+}
+
+/*
+ * Sysfs cleanup and initialization.
+ * Shows number of useful parameters.
+ */
+static void dst_node_sysfs_exit(struct dst_node *n)
+{
+	if (n->info) {
+		dst_remove_node_attributes(n);
+		device_unregister(&n->info->device);
+		n->info = NULL;
+	}
+}
+
+static int dst_node_sysfs_init(struct dst_node *n)
+{
+	int err;
+
+	n->info = kzalloc(sizeof(struct dst_info), GFP_KERNEL);
+	if (!n->info)
+		return -ENOMEM;
+
+	memcpy(&n->info->device, &dst_node_dev, sizeof(struct device));
+	n->info->size = n->size;
+
+	snprintf(n->info->device.bus_id, sizeof(n->info->device.bus_id), "dst-%s", n->name);
+	err = device_register(&n->info->device);
+	if (err) {
+		dprintk(KERN_ERR "Failed to register node '%s', err: %d.\n",
+				n->name, err);
+		goto err_out_exit;
+	}
+
+	dst_create_node_attributes(n);
+
+	return 0;
+
+err_out_exit:
+	kfree(n->info);
+	n->info = NULL;
+	return err;
+}
+
+/*
+ * DST node hash tables machinery.
+ */
+static inline unsigned int dst_hash(char *str, unsigned int size)
+{
+	return (jhash(str, size, 0) % dst_hashtable_size);
+}
+
+static void dst_node_remove(struct dst_node *n)
+{
+	mutex_lock(&dst_hash_lock);
+	list_del_init(&n->node_entry);
+	mutex_unlock(&dst_hash_lock);
+}
+
+static void dst_node_add(struct dst_node *n)
+{
+	unsigned hash = dst_hash(n->name, sizeof(n->name));
+
+	mutex_lock(&dst_hash_lock);
+	list_add_tail(&n->node_entry, &dst_hashtable[hash]);
+	mutex_unlock(&dst_hash_lock);
+}
+
+/*
+ * Cleaning node when it is about to be freed.
+ * There are still users of the socket though,
+ * so connection cleanup should be protected.
+ */
+static void dst_node_cleanup(struct dst_node *n)
+{
+	struct dst_state *st = n->state;
+
+	if (!st)
+		return;
+
+	if (n->queue) {
+		blk_cleanup_queue(n->queue);
+
+		mutex_lock(&dst_hash_lock);
+		idr_remove(&dst_index_idr, n->disk->first_minor);
+		mutex_unlock(&dst_hash_lock);
+
+		put_disk(n->disk);
+	}
+
+	if (n->bdev) {
+		sync_blockdev(n->bdev);
+		blkdev_put(n->bdev);
+	}
+
+	dst_state_lock(st);
+	st->need_exit = 1;
+	dst_state_exit_connected(st);
+	dst_state_unlock(st);
+
+	wake_up(&st->thread_wait);
+
+	dst_state_put(st);
+	n->state = NULL;
+}
+
+/*
+ * Free security attributes attached to given node.
+ */
+static void dst_security_exit(struct dst_node *n)
+{
+	struct dst_secure *s, *tmp;
+
+	list_for_each_entry_safe(s, tmp, &n->security_list, sec_entry) {
+		list_del(&s->sec_entry);
+		kfree(s);
+	}
+}
+
+/*
+ * Free node when there are no more users.
+ * Actually node has to be freed on behalf od userspace process,
+ * since there are number of threads, which are embedded in the
+ * node, so they can not exit and free node from there, that is
+ * why there is a wakeup if reference counter is not equal to zero.
+ */
+void dst_node_put(struct dst_node *n)
+{
+	if (unlikely(!n))
+		return;
+
+	dprintk("%s: n: %p, refcnt: %d.\n",
+			__func__, n, atomic_read(&n->refcnt));
+
+	if (atomic_dec_and_test(&n->refcnt)) {
+		dst_node_remove(n);
+		n->trans_scan_timeout = 0;
+		dst_node_cleanup(n);
+		thread_pool_destroy(n->pool);
+		dst_node_sysfs_exit(n);
+		dst_node_crypto_exit(n);
+		dst_security_exit(n);
+		dst_node_trans_exit(n);
+
+		kfree(n);
+
+		dprintk("%s: freed n: %p.\n", __func__, n);
+	} else {
+		wake_up(&n->wait);
+	}
+}
+
+/*
+ * This function finds devices major/minor numbers for given pathname.
+ */
+static int dst_lookup_device(const char *path, dev_t *dev)
+{
+	int err;
+	struct nameidata nd;
+	struct inode *inode;
+
+	err = path_lookup(path, LOOKUP_FOLLOW, &nd);
+	if (err)
+		return err;
+
+	inode = nd.path.dentry->d_inode;
+	if (!inode) {
+		err = -ENOENT;
+		goto out;
+	}
+
+	if (!S_ISBLK(inode->i_mode)) {
+		err = -ENOTBLK;
+		goto out;
+	}
+
+	*dev = inode->i_rdev;
+
+out:
+	path_put(&nd.path);
+	return err;
+}
+
+/*
+ * Setting up export device: lookup by the name, get its size
+ * and setup listening socket, which will accept clients, which
+ * will submit IO for given storage.
+ */
+static int dst_setup_export(struct dst_node *n, struct dst_ctl *ctl,
+		struct dst_export_ctl *le)
+{
+	int err;
+	dev_t dev = 0; /* gcc likes to scream here */
+
+	snprintf(n->info->local, sizeof(n->info->local), "%s", le->device);
+
+	err = dst_lookup_device(le->device, &dev);
+	if (err)
+		return err;
+
+	n->bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
+	if (!n->bdev)
+		return -ENODEV;
+
+	if (n->size != 0)
+		n->size = min_t(loff_t, n->bdev->bd_inode->i_size, n->size);
+	else
+		n->size = n->bdev->bd_inode->i_size;
+
+	n->info->size = n->size;
+	err = dst_node_init_listened(n, le);
+	if (err)
+		goto err_out_cleanup;
+
+	return 0;
+
+err_out_cleanup:
+	blkdev_put(n->bdev);
+	n->bdev = NULL;
+
+	return err;
+}
+
+/* Empty thread pool callbacks for the network processing threads. */
+static inline void *dst_thread_network_init(void *data)
+{
+	dprintk("%s: data: %p.\n", __func__, data);
+	return data;
+}
+
+static inline void dst_thread_network_cleanup(void *data)
+{
+	dprintk("%s: data: %p.\n", __func__, data);
+}
+
+/*
+ * Allocate DST node and initialize some of its parameters.
+ */
+static struct dst_node *dst_alloc_node(struct dst_ctl *ctl,
+		int (*start)(struct dst_node *),
+		int num)
+{
+	struct dst_node *n;
+	int err;
+
+	n = kzalloc(sizeof(struct dst_node), GFP_KERNEL);
+	if (!n)
+		return NULL;
+
+	INIT_LIST_HEAD(&n->node_entry);
+
+	INIT_LIST_HEAD(&n->security_list);
+	mutex_init(&n->security_lock);
+
+	init_waitqueue_head(&n->wait);
+
+	n->trans_scan_timeout = msecs_to_jiffies(ctl->trans_scan_timeout);
+	if (!n->trans_scan_timeout)
+		n->trans_scan_timeout = HZ;
+
+	n->trans_max_retries = ctl->trans_max_retries;
+	if (!n->trans_max_retries)
+		n->trans_max_retries = 10;
+
+	/*
+	 * Pretty much arbitrary default numbers.
+	 * 32 matches maximum number of pages in bio originated from ext3 (31).
+	 */
+	n->max_pages = ctl->max_pages;
+	if (!n->max_pages)
+		n->max_pages = 32;
+
+	if (n->max_pages > 1024)
+		n->max_pages = 1024;
+
+	n->start = start;
+	n->size = ctl->size;
+
+	atomic_set(&n->refcnt, 1);
+	atomic_long_set(&n->gen, 0);
+	snprintf(n->name, sizeof(n->name), "%s", ctl->name);
+
+	err = dst_node_sysfs_init(n);
+	if (err)
+		goto err_out_free;
+
+	n->pool = thread_pool_create(num, n->name, dst_thread_network_init,
+			dst_thread_network_cleanup, n);
+	if (IS_ERR(n->pool)) {
+		err = PTR_ERR(n->pool);
+		goto err_out_sysfs_exit;
+	}
+
+	dprintk("%s: n: %p, name: %s.\n", __func__, n, n->name);
+
+	return n;
+
+err_out_sysfs_exit:
+	dst_node_sysfs_exit(n);
+err_out_free:
+	kfree(n);
+	return NULL;
+}
+
+/*
+ * Starting a node, connected to the remote server:
+ * register block device and initialize transaction mechanism.
+ * In revers order though.
+ *
+ * It will autonegotiate some parameters with the remote node
+ * and update local if needed.
+ *
+ * Transaction initialization should be the last thing before
+ * starting the node, since transaction should include not only
+ * block IO, but also crypto related data (if any), which are
+ * initialized separately.
+ */
+static int dst_start_remote(struct dst_node *n)
+{
+	int err;
+
+	err = dst_node_trans_init(n, sizeof(struct dst_trans));
+	if (err)
+		return err;
+
+	err = dst_node_create_disk(n);
+	if (err)
+		return err;
+
+	dst_node_set_size(n);
+	add_disk(n->disk);
+
+	dprintk("DST: started remote node '%s', minor: %d.\n", n->name, n->disk->first_minor);
+
+	return 0;
+}
+
+/*
+ * Adding remote node and initialize connection.
+ */
+static int dst_add_remote(struct dst_node *n, struct dst_ctl *ctl,
+		void *data, unsigned int size)
+{
+	int err;
+	struct dst_network_ctl *rctl = data;
+
+	if (n)
+		return -EEXIST;
+
+	if (size != sizeof(struct dst_network_ctl))
+		return -EINVAL;
+
+	n = dst_alloc_node(ctl, dst_start_remote, 1);
+	if (!n)
+		return -ENOMEM;
+
+	memcpy(&n->info->net, rctl, sizeof(struct dst_network_ctl));
+	err = dst_node_init_connected(n, rctl);
+	if (err)
+		goto err_out_free;
+
+	dst_node_add(n);
+
+	return 0;
+
+err_out_free:
+	dst_node_put(n);
+	return err;
+}
+
+/*
+ * Adding export node: initializing block device and listening socket.
+ */
+static int dst_add_export(struct dst_node *n, struct dst_ctl *ctl,
+		void *data, unsigned int size)
+{
+	int err;
+	struct dst_export_ctl *le = data;
+
+	if (n)
+		return -EEXIST;
+
+	if (size != sizeof(struct dst_export_ctl))
+		return -EINVAL;
+
+	n = dst_alloc_node(ctl, dst_start_export, 2);
+	if (!n)
+		return -EINVAL;
+
+	err = dst_setup_export(n, ctl, le);
+	if (err)
+		goto err_out_free;
+
+	dst_node_add(n);
+
+	return 0;
+
+err_out_free:
+	dst_node_put(n);
+	return err;
+}
+
+static int dst_node_remove_unload(struct dst_node *n)
+{
+	printk(KERN_INFO "STOPPED name: '%s', size: %llu.\n",
+			n->name, n->size);
+
+	if (n->disk)
+		del_gendisk(n->disk);
+
+	dst_node_remove(n);
+	dst_node_sysfs_exit(n);
+
+	/*
+	 * This is not a hack. Really.
+	 * Node's reference counter allows to implement fine grained
+	 * node freeing, but since all transactions (which hold node's
+	 * reference counter) are processed in the dedicated thread,
+	 * it is possible that reference will hit zero in that thread,
+	 * so we will not be able to exit thread and cleanup the node.
+	 *
+	 * So, we remove disk, so no new activity is possible, and
+	 * wait until all pending transaction are completed (either
+	 * in receiving thread or by timeout in workqueue), in this
+	 * case reference counter will be less or equal to 2 (once set in
+	 * dst_alloc_node() and then in connector message parser;
+	 * or when we force module unloading, and connector message
+	 * parser does not hold a reference, in this case reference
+	 * counter will be equal to 1),
+	 * and subsequent dst_node_put() calls will free the node.
+	 */
+	dprintk("%s: going to sleep with %d refcnt.\n", __func__, atomic_read(&n->refcnt));
+	wait_event(n->wait, atomic_read(&n->refcnt) <= 2);
+
+	dst_node_put(n);
+	return 0;
+}
+
+/*
+ * Remove node from the hash table.
+ */
+static int dst_del_node(struct dst_node *n, struct dst_ctl *ctl,
+		void *data, unsigned int size)
+{
+	if (!n)
+		return -ENODEV;
+
+	return dst_node_remove_unload(n);
+}
+
+/*
+ * Initialize crypto processing for given node.
+ */
+static int dst_crypto_init(struct dst_node *n, struct dst_ctl *ctl,
+		void *data, unsigned int size)
+{
+	struct dst_crypto_ctl *crypto = data;
+
+	if (!n)
+		return -ENODEV;
+
+	if (size != sizeof(struct dst_crypto_ctl) + crypto->hash_keysize +
+			crypto->cipher_keysize)
+		return -EINVAL;
+
+	if (n->trans_cache)
+		return -EEXIST;
+
+	return dst_node_crypto_init(n, crypto);
+}
+
+/*
+ * Security attributes for given node.
+ */
+static int dst_security_init(struct dst_node *n, struct dst_ctl *ctl,
+		void *data, unsigned int size)
+{
+	struct dst_secure *s;
+
+	if (!n)
+		return -ENODEV;
+
+	if (size != sizeof(struct dst_secure_user))
+		return -EINVAL;
+
+	s = kmalloc(sizeof(struct dst_secure), GFP_KERNEL);
+	if (!s)
+		return -ENOMEM;
+
+	memcpy(&s->sec, data, size);
+
+	mutex_lock(&n->security_lock);
+	list_add_tail(&s->sec_entry, &n->security_list);
+	mutex_unlock(&n->security_lock);
+
+	return 0;
+}
+
+/*
+ * Kill'em all!
+ */
+static int dst_start_node(struct dst_node *n, struct dst_ctl *ctl,
+		void *data, unsigned int size)
+{
+	int err;
+
+	if (!n)
+		return -ENODEV;
+
+	if (n->trans_cache)
+		return 0;
+
+	err = n->start(n);
+	if (err)
+		return err;
+
+	printk(KERN_INFO "STARTED name: '%s', size: %llu.\n", n->name, n->size);
+	return 0;
+}
+
+typedef int (*dst_command_func)(struct dst_node *n, struct dst_ctl *ctl,
+		void *data, unsigned int size);
+
+/*
+ * List of userspace commands.
+ */
+static dst_command_func dst_commands[] = {
+	[DST_ADD_REMOTE] = &dst_add_remote,
+	[DST_ADD_EXPORT] = &dst_add_export,
+	[DST_DEL_NODE] = &dst_del_node,
+	[DST_CRYPTO] = &dst_crypto_init,
+	[DST_SECURITY] = &dst_security_init,
+	[DST_START] = &dst_start_node,
+};
+
+/*
+ * Configuration parser.
+ */
+static void cn_dst_callback(void *data)
+{
+	struct dst_ctl *ctl;
+	struct cn_msg *msg = data;
+	int err;
+	struct dst_ctl_ack ack;
+	struct dst_node *n = NULL, *tmp;
+	unsigned int hash;
+
+	if (msg->len < sizeof(struct dst_ctl)) {
+		err = -EBADMSG;
+		goto out;
+	}
+
+	ctl = (struct dst_ctl *)msg->data;
+
+	if (ctl->cmd >= DST_CMD_MAX) {
+		err = -EINVAL;
+		goto out;
+	}
+	hash = dst_hash(ctl->name, sizeof(ctl->name));
+
+	mutex_lock(&dst_hash_lock);
+	list_for_each_entry(tmp, &dst_hashtable[hash], node_entry) {
+		if (!memcmp(tmp->name, ctl->name, sizeof(tmp->name))) {
+			n = tmp;
+			dst_node_get(n);
+			break;
+		}
+	}
+	mutex_unlock(&dst_hash_lock);
+
+	err = dst_commands[ctl->cmd](n, ctl, msg->data + sizeof(struct dst_ctl),
+			msg->len - sizeof(struct dst_ctl));
+
+	dst_node_put(n);
+out:
+	memcpy(&ack.msg, msg, sizeof(struct cn_msg));
+
+	ack.msg.ack = msg->ack + 1;
+	ack.msg.len = sizeof(struct dst_ctl_ack) - sizeof(struct cn_msg);
+
+	ack.error = err;
+
+	cn_netlink_send(&ack.msg, 0, GFP_KERNEL);
+}
+
+/*
+ * Global initialization: sysfs, hash table, block device registration,
+ * connector and various caches.
+ */
+static int __init dst_sysfs_init(void)
+{
+	return bus_register(&dst_dev_bus_type);
+}
+
+static void dst_sysfs_exit(void)
+{
+	bus_unregister(&dst_dev_bus_type);
+}
+
+static int __init dst_hashtable_init(void)
+{
+	unsigned int i;
+
+	dst_hashtable = kcalloc(dst_hashtable_size, sizeof(struct list_head),
+			GFP_KERNEL);
+	if (!dst_hashtable)
+		return -ENOMEM;
+
+	for (i=0; i<dst_hashtable_size; ++i)
+		INIT_LIST_HEAD(&dst_hashtable[i]);
+
+	return 0;
+}
+
+static void dst_hashtable_exit(void)
+{
+	unsigned int i;
+	struct dst_node *n, *tmp;
+
+	for (i=0; i<dst_hashtable_size; ++i) {
+		list_for_each_entry_safe(n, tmp, &dst_hashtable[i], node_entry) {
+			dst_node_remove_unload(n);
+		}
+	}
+
+	kfree(dst_hashtable);
+}
+
+static int __init dst_sys_init(void)
+{
+	int err = -ENOMEM;
+
+	err = dst_hashtable_init();
+	if (err)
+		goto err_out_exit;
+
+	err = dst_export_init();
+	if (err)
+		goto err_out_hashtable_exit;
+
+	err = register_blkdev(dst_major, DST_NAME);
+	if (err < 0)
+		goto err_out_export_exit;
+	if (err)
+		dst_major = err;
+
+	err = dst_sysfs_init();
+	if (err)
+		goto err_out_unregister;
+
+	err = cn_add_callback(&cn_dst_id, "DST", cn_dst_callback);
+	if (err)
+		goto err_out_sysfs_exit;
+
+	printk(KERN_INFO "Distributed storage, '%s' release.\n", dst_name);
+
+	return 0;
+
+err_out_sysfs_exit:
+	dst_sysfs_exit();
+err_out_unregister:
+	unregister_blkdev(dst_major, DST_NAME);
+err_out_export_exit:
+	dst_export_exit();
+err_out_hashtable_exit:
+	dst_hashtable_exit();
+err_out_exit:
+	return err;
+}
+
+static void __exit dst_sys_exit(void)
+{
+	cn_del_callback(&cn_dst_id);
+	unregister_blkdev(dst_major, DST_NAME);
+	dst_hashtable_exit();
+	dst_sysfs_exit();
+	dst_export_exit();
+}
+
+module_init(dst_sys_init);
+module_exit(dst_sys_exit);
+
+MODULE_DESCRIPTION("Distributed storage");
+MODULE_AUTHOR("Evgeniy Polyakov <[email protected]>");
+MODULE_LICENSE("GPL");

335 lines not on this page · raw message

index 5c7f946..1980f08 100644
@@ -39,8 +39,10 @@
 #define CN_IDX_V86D			0x4
 #define CN_VAL_V86D_UVESAFB		0x1
 #define CN_IDX_BB			0x5	/* BlackBoard, from the TSP GPL sampling framework */
+#define CN_DST_IDX			0x6
+#define CN_DST_VAL			0x1
 
-#define CN_NETLINK_USERS		6
+#define CN_NETLINK_USERS		7
 
 /*
  * Maximum connector's message size.

587 lines not on this page · raw message

new file mode 100644
index 0000000..1a5a656
@@ -0,0 +1,839 @@
+/*
+ * 2007+ Copyright (c) Evgeniy Polyakov <[email protected]>
+ * 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.
+ */
+
+#include <linux/buffer_head.h>
+#include <linux/blkdev.h>
+#include <linux/bio.h>
+#include <linux/connector.h>
+#include <linux/dst.h>
+#include <linux/device.h>
+#include <linux/in.h>
+#include <linux/in6.h>
+#include <linux/socket.h>
+#include <linux/slab.h>
+
+#include <net/sock.h>
+
+/*
+ * Polling machinery.
+ */
+
+struct dst_poll_helper
+{
+	poll_table 		pt;
+	struct dst_state	*st;
+};
+
+static int dst_queue_wake(wait_queue_t *wait, unsigned mode, int sync, void *key)
+{
+	struct dst_state *st = container_of(wait, struct dst_state, wait);
+
+	wake_up(&st->thread_wait);
+	return 1;
+}
+
+static void dst_queue_func(struct file *file, wait_queue_head_t *whead,
+				 poll_table *pt)
+{
+	struct dst_state *st = container_of(pt, struct dst_poll_helper, pt)->st;
+
+	st->whead = whead;
+	init_waitqueue_func_entry(&st->wait, dst_queue_wake);
+	add_wait_queue(whead, &st->wait);
+}
+
+void dst_poll_exit(struct dst_state *st)
+{
+	if (st->whead) {
+		remove_wait_queue(st->whead, &st->wait);
+		st->whead = NULL;
+	}
+}
+
+int dst_poll_init(struct dst_state *st)
+{
+	struct dst_poll_helper ph;
+
+	ph.st = st;
+	init_poll_funcptr(&ph.pt, &dst_queue_func);
+
+	st->socket->ops->poll(NULL, st->socket, &ph.pt);
+	return 0;
+}
+
+/*
+ * Header receiving function - may block.
+ */
+static int dst_data_recv_header(struct socket *sock,
+		void *data, unsigned int size, int block)
+{
+	struct msghdr msg;
+	struct kvec iov;
+	int err;
+
+	iov.iov_base = data;
+	iov.iov_len = size;
+
+	msg.msg_iov = (struct iovec *)&iov;
+	msg.msg_iovlen = 1;
+	msg.msg_name = NULL;
+	msg.msg_namelen = 0;
+	msg.msg_control = NULL;
+	msg.msg_controllen = 0;
+	msg.msg_flags = (block)?MSG_WAITALL:MSG_DONTWAIT;
+
+	err = kernel_recvmsg(sock, &msg, &iov, 1, iov.iov_len,
+			msg.msg_flags);
+	if (err != size)
+		return -1;
+
+	return 0;
+}
+
+/*
+ * Header sending function - may block.
+ */
+int dst_data_send_header(struct socket *sock,
+		void *data, unsigned int size, int more)
+{
+	struct msghdr msg;
+	struct kvec iov;
+	int err;
+
+	iov.iov_base = data;
+	iov.iov_len = size;
+
+	msg.msg_iov = (struct iovec *)&iov;
+	msg.msg_iovlen = 1;
+	msg.msg_name = NULL;
+	msg.msg_namelen = 0;
+	msg.msg_control = NULL;
+	msg.msg_controllen = 0;
+	msg.msg_flags = MSG_WAITALL | (more)?MSG_MORE:0;
+
+	err = kernel_sendmsg(sock, &msg, &iov, 1, iov.iov_len);
+	if (err != size) {
+		dprintk("%s: size: %u, more: %d, err: %d.\n",
+				__func__, size, more, err);
+		return -1;
+	}
+
+	return 0;
+}
+
+/*
+ * Block autoconfiguration: request size of the storage and permissions.
+ */
+static int dst_request_remote_config(struct dst_state *st)
+{
+	struct dst_node *n = st->node;
+	int err = -EINVAL;
+	struct dst_cmd *cmd = st->data;
+
+	memset(cmd, 0, sizeof(struct dst_cmd));
+	cmd->cmd = DST_CFG;
+
+	dst_convert_cmd(cmd);
+
+	err = dst_data_send_header(st->socket, cmd, sizeof(struct dst_cmd), 0);
+	if (err)
+		goto out;
+
+	err = dst_data_recv_header(st->socket, cmd, sizeof(struct dst_cmd), 1);
+	if (err)
+		goto out;
+
+	dst_convert_cmd(cmd);
+
+	if (cmd->cmd != DST_CFG) {
+		err = -EINVAL;
+		printk("%s: checking result: cmd: %d, size reported: %llu.\n",
+			__func__, cmd->cmd, cmd->sector);
+		goto out;
+	}
+
+	if (n->size != 0)
+		n->size = min_t(loff_t, n->size, cmd->sector);
+	else
+		n->size = cmd->sector;
+
+	n->info->size = n->size;
+	st->permissions = cmd->rw;
+
+out:
+	dprintk("%s: n: %p, err: %d, size: %llu, permission: %x.\n",
+			__func__, n, err, n->size, st->permissions);
+	return err;
+}
+
+/*
+ * Socket machinery.
+ */
+
+#define DST_DEFAULT_TIMEO	20000
+
+int dst_state_socket_create(struct dst_state *st)
+{
+	int err;
+	struct socket *sock;
+	struct dst_network_ctl *ctl = &st->ctl;
+
+	err = sock_create(ctl->addr.sa_family, ctl->type, ctl->proto, &sock);
+	if (err < 0)
+		return err;
+
+	sock->sk->sk_sndtimeo = sock->sk->sk_rcvtimeo =
+		msecs_to_jiffies(DST_DEFAULT_TIMEO);
+	sock->sk->sk_allocation = GFP_NOIO;
+
+	st->socket = st->read_socket = sock;
+	return 0;
+}
+
+void dst_state_socket_release(struct dst_state *st)
+{
+	dprintk("%s: st: %p, socket: %p, n: %p.\n",
+			__func__, st, st->socket, st->node);
+	if (st->socket) {
+		sock_release(st->socket);
+		st->socket = NULL;
+		st->read_socket = NULL;
+	}
+}
+
+void dst_dump_addr(struct socket *sk, struct sockaddr *sa, char *str)
+{
+	if (sk->ops->family == AF_INET) {
+		struct sockaddr_in *sin = (struct sockaddr_in *)sa;
+		printk(KERN_INFO "%s %u.%u.%u.%u:%d.\n",
+			str, NIPQUAD(sin->sin_addr.s_addr), ntohs(sin->sin_port));
+	} else if (sk->ops->family == AF_INET6) {
+		struct sockaddr_in6 *sin = (struct sockaddr_in6 *)sa;
+		printk(KERN_INFO "%s %04x:%04x:%04x:%04x:%04x:%04x:%04x:%04x:%d",
+			str, NIP6(sin->sin6_addr), ntohs(sin->sin6_port));
+	}
+}
+
+void dst_state_exit_connected(struct dst_state *st)
+{
+	if (st->socket) {
+		dst_poll_exit(st);
+		st->socket->ops->shutdown(st->socket, 2);
+
+		dst_dump_addr(st->socket, (struct sockaddr *)&st->ctl.addr,
+				"Disconnected peer");
+		dst_state_socket_release(st);
+	}
+}
+
+static int dst_state_init_connected(struct dst_state *st)
+{
+	int err;
+	struct dst_network_ctl *ctl = &st->ctl;
+
+	err = dst_state_socket_create(st);
+	if (err)
+		goto err_out_exit;
+
+	err = kernel_connect(st->socket, (struct sockaddr *)&st->ctl.addr,
+			st->ctl.addr.sa_data_len, 0);
+	if (err)
+		goto err_out_release;
+
+	err = dst_poll_init(st);
+	if (err)
+		goto err_out_release;
+
+	dst_dump_addr(st->socket, (struct sockaddr *)&ctl->addr,
+			"Connected to peer");
+
+	return 0;
+
+err_out_release:
+	dst_state_socket_release(st);
+err_out_exit:
+	return err;
+}
+
+/*
+ * State reset is used to reconnect to the remote peer.
+ * May fail, but who cares, we will try again later.
+ */
+static void inline dst_state_reset_nolock(struct dst_state *st)
+{
+	dst_state_exit_connected(st);
+	dst_state_init_connected(st);
+}
+
+static void inline dst_state_reset(struct dst_state *st)
+{
+	dst_state_lock(st);
+	dst_state_reset_nolock(st);
+	dst_state_unlock(st);
+}
+
+/*
+ * Basic network sending/receiving functions.
+ * Blocked mode is used.
+ */
+static int dst_data_recv_raw(struct dst_state *st, void *buf, u64 size)
+{
+	struct msghdr msg;
+	struct kvec iov;
+	int err;
+
+	BUG_ON(!size);
+
+	iov.iov_base = buf;
+	iov.iov_len = size;
+
+	msg.msg_iov = (struct iovec *)&iov;
+	msg.msg_iovlen = 1;
+	msg.msg_name = NULL;
+	msg.msg_namelen = 0;
+	msg.msg_control = NULL;
+	msg.msg_controllen = 0;
+	msg.msg_flags = MSG_DONTWAIT;
+
+	err = kernel_recvmsg(st->socket, &msg, &iov, 1, iov.iov_len,
+			msg.msg_flags);
+	if (err <= 0) {
+		printk("%s: failed to recv data: size: %llu, err: %d.\n",
+				__func__, size, err);
+		if (err == 0)
+			err = -ECONNRESET;
+
+		dst_state_exit_connected(st);
+	}
+
+	return err;
+}
+
+/*
+ * Ping command to early detect failed nodes.
+ */
+static int dst_send_ping(struct dst_state *st)
+{
+	struct dst_cmd *cmd = st->data;
+	int err = -ECONNRESET;
+
+	dst_state_lock(st);
+	if (st->socket) {
+		memset(cmd, 0, sizeof(struct dst_cmd));
+
+		cmd->cmd = __cpu_to_be32(DST_PING);
+
+		err = dst_data_send_header(st->socket, cmd, sizeof(struct dst_cmd), 0);
+	}
+	dprintk("%s: st: %p, socket: %p, err: %d.\n", __func__, st, st->socket, err);
+	dst_state_unlock(st);
+
+	return err;
+}
+
+/*
+ * Receiving function, which should either return error or read
+ * whole block request. If there was no traffic for a one second,
+ * send a ping, since remote node may die.
+ */
+int dst_data_recv(struct dst_state *st, void *data, unsigned int size)
+{
+	unsigned int revents = 0;
+	unsigned int err_mask = POLLERR | POLLHUP | POLLRDHUP;
+	unsigned int mask = err_mask | POLLIN;
+	struct dst_node *n = st->node;
+	int err = 0;
+
+	while (size && !err) {
+		revents = dst_state_poll(st);
+
+		if (!(revents & mask)) {
+			DEFINE_WAIT(wait);
+
+			for (;;) {
+				prepare_to_wait(&st->thread_wait, &wait,
+						TASK_INTERRUPTIBLE);
+				if (!n->trans_scan_timeout || st->need_exit)
+					break;
+
+				revents = dst_state_poll(st);
+
+				if (revents & mask)
+					break;
+
+				if (signal_pending(current))
+					break;
+
+				if (!schedule_timeout(HZ)) {
+					err = dst_send_ping(st);
+					if (err)
+						return err;
+				}
+
+				continue;
+			}
+			finish_wait(&st->thread_wait, &wait);
+		}
+
+		err = -ECONNRESET;
+		dst_state_lock(st);
+
+		if (		st->socket &&
+				(st->read_socket == st->socket) &&
+				(revents & POLLIN)) {
+			err = dst_data_recv_raw(st, data, size);
+			if (err > 0) {
+				data += err;
+				size -= err;
+				err = 0;
+			}
+		}
+
+		if (revents & err_mask || !st->socket) {
+			printk("%s: revents: %x, socket: %p, size: %u, err: %d.\n",
+					__func__, revents, st->socket, size, err);
+			err = -ECONNRESET;
+		}
+
+		dst_state_unlock(st);
+
+		if (!n->trans_scan_timeout)
+			err = -ENODEV;
+	}
+
+	return err;
+}
+
+/*
+ * Send block autoconf reply.
+ */
+static int dst_process_cfg(struct dst_state *st)
+{
+	struct dst_node *n = st->node;
+	struct dst_cmd *cmd = st->data;
+	int err;
+
+	cmd->sector = n->size;
+	cmd->rw = st->permissions;
+
+	dst_convert_cmd(cmd);
+
+	dst_state_lock(st);
+	err = dst_data_send_header(st->socket, cmd, sizeof(struct dst_cmd), 0);
+	dst_state_unlock(st);
+
+	return err;
+}
+
+/*
+ * Receive block IO from the network.
+ */
+static int dst_recv_bio(struct dst_state *st, struct bio *bio, unsigned int total_size)
+{
+	struct bio_vec *bv;
+	int i, err;
+	void *data;
+	unsigned int sz;
+
+	bio_for_each_segment(bv, bio, i) {
+		sz = min(total_size, bv->bv_len);
+
+		dprintk("%s: bio: %llu/%u, total: %u, len: %u, sz: %u, off: %u.\n",
+			__func__, (u64)bio->bi_sector, bio->bi_size, total_size,
+			bv->bv_len, sz, bv->bv_offset);
+
+		data = kmap(bv->bv_page) + bv->bv_offset;
+		err = dst_data_recv(st, data, sz);
+		kunmap(bv->bv_page);
+
+		bv->bv_len = sz;
+
+		if (err)
+			return err;
+
+		total_size -= sz;
+		if (total_size == 0)
+			break;
+	}
+
+	return 0;
+}
+
+/*
+ * Our block IO has just completed and arrived: get it.
+ */
+static int dst_process_io_response(struct dst_state *st)
+{
+	struct dst_node *n = st->node;
+	struct dst_cmd *cmd = st->data;
+	struct dst_trans *t;
+	int err = 0;
+	struct bio *bio;
+
+	mutex_lock(&n->trans_lock);
+	t = dst_trans_search(n, cmd->id);
+	mutex_unlock(&n->trans_lock);
+
+	if (!t)
+		goto err_out_exit;
+
+	bio = t->bio;
+
+	dprintk("%s: bio: %llu/%u, cmd_size: %u, csize: %u, dir: %lu.\n",
+		__func__, (u64)bio->bi_sector, bio->bi_size, cmd->size,
+		cmd->csize, bio_data_dir(bio));
+
+	if (bio_data_dir(bio) == READ) {
+		if (bio->bi_size != cmd->size - cmd->csize)
+			goto err_out_exit;
+
+		if (dst_need_crypto(n)) {
+			err = dst_recv_cdata(st, t->cmd.hash);
+			if (err)
+				goto err_out_exit;
+		}
+
+		err = dst_recv_bio(st, t->bio, bio->bi_size);
+		if (err)
+			goto err_out_exit;
+
+		if (dst_need_crypto(n))
+			return dst_trans_crypto(t);
+	} else {
+		err = -EBADMSG;
+		if (cmd->size || cmd->csize)
+			goto err_out_exit;
+	}
+
+	dst_trans_remove(t);
+	dst_trans_put(t);
+
+	return 0;
+
+err_out_exit:
+	return err;
+}
+
+/*
+ * Receive crypto data.
+ */
+int dst_recv_cdata(struct dst_state *st, void *cdata)
+{
+	struct dst_cmd *cmd = st->data;
+	struct dst_node *n = st->node;
+	struct dst_crypto_ctl *c = &n->crypto;
+	int err;
+
+	if (cmd->csize != c->crypto_attached_size) {
+		dprintk("%s: cmd: cmd: %u, sector: %llu, size: %u, "
+				"csize: %u != digest size %u.\n",
+				__func__, cmd->cmd, cmd->sector, cmd->size,
+				cmd->csize, c->crypto_attached_size);
+		err = -EINVAL;
+		goto err_out_exit;
+	}
+
+	err = dst_data_recv(st, cdata, cmd->csize);
+	if (err)
+		goto err_out_exit;
+
+	cmd->size -= cmd->csize;
+	return 0;
+
+err_out_exit:
+	return err;
+}
+
+/*
+ * Receive the command and start its processing.
+ */
+static int dst_recv_processing(struct dst_state *st)
+{
+	int err = -EINTR;
+	struct dst_cmd *cmd = st->data;
+
+	/*
+	 * If socket will be reset after this statement, then
+	 * dst_data_recv() will just fail and loop will
+	 * start again, so it can be done without any locks.
+	 *
+	 * st->read_socket is needed to prevents state machine
+	 * breaking between this data reading and subsequent one
+	 * in protocol specific functions during connection reset.
+	 * In case of reset we have to read next command and do
+	 * not expect data for old command to magically appear in
+	 * new connection.
+	 */
+	st->read_socket = st->socket;
+	err = dst_data_recv(st, cmd, sizeof(struct dst_cmd));
+	if (err)
+		goto out_exit;
+
+	dst_convert_cmd(cmd);
+
+	dprintk("%s: cmd: %u, size: %u, csize: %u, id: %llu, "
+			"sector: %llu, flags: %llx, rw: %llx.\n",
+			__func__, cmd->cmd, cmd->size,
+			cmd->csize, cmd->id, cmd->sector,
+			cmd->flags, cmd->rw);
+
+	/*
+	 * This should catch protocol breakage and random garbage instead of commands.
+	 */
+	if (unlikely(cmd->csize > st->size - sizeof(struct dst_cmd))) {
+		err = -EBADMSG;
+		goto out_exit;
+	}
+
+	err = -EPROTO;
+	switch (cmd->cmd) {
+		case DST_IO_RESPONSE:
+			err = dst_process_io_response(st);
+			break;
+		case DST_IO:
+			err = dst_process_io(st);
+			break;
+		case DST_CFG:
+			err = dst_process_cfg(st);
+			break;
+		case DST_PING:
+			err = 0;
+			break;
+		default:
+			break;
+	}
+
+out_exit:
+	return err;
+}
+
+/*
+ * Receiving thread. For the client node we should try to reconnect,
+ * for accepted client we just drop the state and expect it to reconnect.
+ */
+static int dst_recv(void *init_data, void *schedule_data)
+{
+	struct dst_state *st = schedule_data;
+	struct dst_node *n = init_data;
+	int err = 0;
+
+	dprintk("%s: start st: %p, n: %p, scan: %lu, need_exit: %d.\n",
+			__func__, st, n, n->trans_scan_timeout, st->need_exit);
+
+	while (n->trans_scan_timeout && !st->need_exit) {
+		err = dst_recv_processing(st);
+		if (err < 0) {
+			if (!st->ctl.type)
+				break;
+
+			if (!n->trans_scan_timeout || st->need_exit)
+				break;
+
+			dst_state_reset(st);
+			msleep(1000);
+		}
+	}
+
+	st->need_exit = 1;
+	wake_up(&st->thread_wait);
+
+	dprintk("%s: freeing receiving socket st: %p.\n", __func__, st);
+	dst_state_lock(st);
+	dst_state_exit_connected(st);
+	dst_state_unlock(st);
+	dst_state_put(st);
+
+	dprintk("%s: freed receiving socket st: %p.\n", __func__, st);
+
+	return err;
+}
+
+/*
+ * Network state dies here and borns couple of lines below.
+ * This object is the main network state processing engine:
+ * sending, receiving, reconnections, all network related
+ * tasks are handled on behalf of the state.
+ */
+static void dst_state_free(struct dst_state *st)
+{
+	dprintk("%s: st: %p.\n", __func__, st);
+	if (st->cleanup)
+		st->cleanup(st);
+	kfree(st->data);
+	kfree(st);
+}
+
+struct dst_state *dst_state_alloc(struct dst_node *n)
+{
+	struct dst_state *st;
+	int err = -ENOMEM;
+
+	st = kzalloc(sizeof(struct dst_state), GFP_KERNEL);
+	if (!st)
+		goto err_out_exit;
+
+	st->node = n;
+	st->need_exit = 0;
+
+	st->size = PAGE_SIZE;
+	st->data = kmalloc(st->size, GFP_KERNEL);
+	if (!st->data)
+		goto err_out_free;
+
+	spin_lock_init(&st->request_lock);
+	INIT_LIST_HEAD(&st->request_list);
+
+	mutex_init(&st->state_lock);
+	init_waitqueue_head(&st->thread_wait);
+
+	/*
+	 * One for processing thread, another one for node itself.
+	 */
+	atomic_set(&st->refcnt, 2);
+
+	dprintk("%s: st: %p, n: %p.\n", __func__, st, st->node);
+
+	return st;
+
+err_out_free:
+	kfree(st);
+err_out_exit:
+	return ERR_PTR(err);
+}
+
+int dst_state_schedule_receiver(struct dst_state *st)
+{
+	return thread_pool_schedule_private(st->node->pool, dst_thread_setup,
+			dst_recv, st, MAX_SCHEDULE_TIMEOUT, st->node);
+}
+
+/*
+ * Initialize client's connection to the remote peer: allocate state,
+ * connect and perform block IO autoconfiguration.
+ */
+int dst_node_init_connected(struct dst_node *n, struct dst_network_ctl *r)
+{
+	struct dst_state *st;
+	int err = -ENOMEM;
+
+	st = dst_state_alloc(n);
+	if (IS_ERR(st)) {
+		err = PTR_ERR(st);
+		goto err_out_exit;
+	}
+	memcpy(&st->ctl, r, sizeof(struct dst_network_ctl));
+
+	err = dst_state_init_connected(st);
+	if (err)
+		goto err_out_free_data;
+
+	err = dst_request_remote_config(st);
+	if (err)
+		goto err_out_exit_connected;
+	n->state = st;
+
+	err = dst_state_schedule_receiver(st);
+	if (err)
+		goto err_out_exit_connected;
+
+	return 0;
+
+err_out_exit_connected:
+	dst_state_exit_connected(st);
+err_out_free_data:
+	dst_state_free(st);
+err_out_exit:
+	n->state = NULL;
+	return err;
+}
+
+void dst_state_put(struct dst_state *st)
+{
+	dprintk("%s: st: %p, refcnt: %d.\n",
+			__func__, st, atomic_read(&st->refcnt));
+	if (atomic_dec_and_test(&st->refcnt))
+		dst_state_free(st);
+}
+
+/*
+ * Send block IO to the network one by one using zero-copy ->sendpage().
+ */
+int dst_send_bio(struct dst_state *st, struct dst_cmd *cmd, struct bio *bio)
+{
+	struct bio_vec *bv;
+	struct dst_crypto_ctl *c = &st->node->crypto;
+	int err, i = 0;
+	int flags = MSG_WAITALL;
+
+	err = dst_data_send_header(st->socket, cmd,
+		sizeof(struct dst_cmd) + c->crypto_attached_size, bio->bi_vcnt);
+	if (err)
+		goto err_out_exit;
+
+	bio_for_each_segment(bv, bio, i) {
+		if (i < bio->bi_vcnt - 1)
+			flags |= MSG_MORE;
+
+		err = kernel_sendpage(st->socket, bv->bv_page, bv->bv_offset,
+				bv->bv_len, flags);
+		if (err <= 0)
+			goto err_out_exit;
+	}
+
+	return 0;
+
+err_out_exit:
+	dprintk("%s: %d/%d, flags: %x, err: %d.\n",
+			__func__, i, bio->bi_vcnt, flags, err);
+	return err;
+}
+
+/*
+ * Send transaction to the remote peer.
+ */
+int dst_trans_send(struct dst_trans *t)
+{
+	int err;
+	struct dst_state *st = t->n->state;
+	struct bio *bio = t->bio;
+
+	dst_convert_cmd(&t->cmd);
+
+	dst_state_lock(st);
+	if (!st->socket) {
+		err = dst_state_init_connected(st);
+		if (err)
+			goto err_out_unlock;
+	}
+
+	if (bio_data_dir(bio) == WRITE) {
+		err = dst_send_bio(st, &t->cmd, t->bio);
+	} else {
+		err = dst_data_send_header(st->socket, &t->cmd,
+				sizeof(struct dst_cmd), 0);
+	}
+	if (err)
+		goto err_out_reset;
+
+	dst_state_unlock(st);
+	return 0;
+
+err_out_reset:
+	dst_state_reset_nolock(st);
+err_out_unlock:
+	dst_state_unlock(st);
+
+	return err;
+}

664 lines not on this page · raw message

new file mode 100644
index 0000000..c35754d
@@ -0,0 +1,345 @@
+/*
+ * 2007+ Copyright (c) Evgeniy Polyakov <[email protected]>
+ * 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.
+ */
+
+#include <linux/kernel.h>
+#include <linux/dst.h>
+#include <linux/kthread.h>
+#include <linux/slab.h>
+
+/*
+ * Thread pool abstraction allows to schedule a work to be performed
+ * on behalf of kernel thread. One does not operate with threads itself,
+ * instead user provides setup and cleanup callbacks for thread pool itself,
+ * and action and cleanup callbacks for each submitted work.
+ *
+ * Each worker has private data initialized at creation time and data,
+ * provided by user at scheduling time.
+ *
+ * When action is being performed, thread can not be used by other users,
+ * instead they will sleep until there is free thread to pick their work.
+ */
+struct thread_pool_worker
+{
+	struct list_head	worker_entry;
+
+	struct task_struct	*thread;
+
+	struct thread_pool	*pool;
+
+	int			error;
+	int			has_data;
+	int			need_exit;
+	unsigned int		id;
+
+	wait_queue_head_t	wait;
+
+	void			*private;
+	void			*schedule_data;
+
+	int			(* action)(void *private, void *schedule_data);
+	void			(* cleanup)(void *private);
+};
+
+static void thread_pool_exit_worker(struct thread_pool_worker *w)
+{
+	kthread_stop(w->thread);
+
+	w->cleanup(w->private);
+	kfree(w);
+}
+
+/*
+ * Called to mark thread as ready and allow users to schedule new work.
+ */
+static void thread_pool_worker_make_ready(struct thread_pool_worker *w)
+{
+	struct thread_pool *p = w->pool;
+
+	mutex_lock(&p->thread_lock);
+
+	if (!w->need_exit) {
+		list_move_tail(&w->worker_entry, &p->ready_list);
+		w->has_data = 0;
+		mutex_unlock(&p->thread_lock);
+
+		wake_up(&p->wait);
+	} else {
+		p->thread_num--;
+		list_del(&w->worker_entry);
+		mutex_unlock(&p->thread_lock);
+
+		thread_pool_exit_worker(w);
+	}
+}
+
+/*
+ * Thread action loop: waits until there is new work.
+ */
+static int thread_pool_worker_func(void *data)
+{
+	struct thread_pool_worker *w = data;
+
+	while (!kthread_should_stop()) {
+		wait_event_interruptible(w->wait,
+			kthread_should_stop() || w->has_data);
+
+		if (kthread_should_stop())
+			break;
+
+		if (!w->has_data)
+			continue;
+
+		w->action(w->private, w->schedule_data);
+		thread_pool_worker_make_ready(w);
+	}
+
+	return 0;
+}
+
+/*
+ * Remove single worker without specifying which one.
+ */
+void thread_pool_del_worker(struct thread_pool *p)
+{
+	struct thread_pool_worker *w = NULL;
+
+	while (!w) {
+		wait_event(p->wait, !list_empty(&p->ready_list) || !p->thread_num);
+
+		dprintk("%s: locking list_empty: %d, thread_num: %d.\n",
+				__func__, list_empty(&p->ready_list), p->thread_num);
+
+		mutex_lock(&p->thread_lock);
+		if (!list_empty(&p->ready_list)) {
+			w = list_first_entry(&p->ready_list,
+					struct thread_pool_worker,
+					worker_entry);
+
+			dprintk("%s: deleting w: %p, thread_num: %d, list: %p [%p.%p].\n",
+					__func__, w, p->thread_num, &p->ready_list,
+					p->ready_list.prev, p->ready_list.next);
+
+			p->thread_num--;
+			list_del(&w->worker_entry);
+		}
+		mutex_unlock(&p->thread_lock);
+	}
+
+	if (w)
+		thread_pool_exit_worker(w);
+	dprintk("%s: deleted w: %p, thread_num: %d.\n",
+			__func__, w, p->thread_num);
+}
+
+/*
+ * Remove a worker with given ID.
+ */
+void thread_pool_del_worker_id(struct thread_pool *p, unsigned int id)
+{
+	struct thread_pool_worker *w;
+	int found = 0;
+
+	mutex_lock(&p->thread_lock);
+	list_for_each_entry(w, &p->ready_list, worker_entry) {
+		if (w->id == id) {
+			found = 1;
+			p->thread_num--;
+			list_del(&w->worker_entry);
+			break;
+		}
+	}
+
+	if (!found) {
+		list_for_each_entry(w, &p->active_list, worker_entry) {
+			if (w->id == id) {
+				w->need_exit = 1;
+				break;
+			}
+		}
+	}
+	mutex_unlock(&p->thread_lock);
+
+	if (found)
+		thread_pool_exit_worker(w);
+}
+
+/*
+ * Add new worker thread with given parameters.
+ * If initialization callback fails, return error.
+ */
+int thread_pool_add_worker(struct thread_pool *p,
+		char *name,
+		unsigned int id,
+		void *(* init)(void *private),
+		void (* cleanup)(void *private),
+		void *private)
+{
+	struct thread_pool_worker *w;
+	int err = -ENOMEM;
+
+	w = kzalloc(sizeof(struct thread_pool_worker), GFP_KERNEL);
+	if (!w)
+		goto err_out_exit;
+
+	w->pool = p;
+	init_waitqueue_head(&w->wait);
+	w->cleanup = cleanup;
+	w->id = id;
+
+	w->thread = kthread_run(thread_pool_worker_func, w, "%s", name);
+	if (IS_ERR(w->thread)) {
+		err = PTR_ERR(w->thread);
+		goto err_out_free;
+	}
+
+	w->private = init(private);
+	if (IS_ERR(w->private)) {
+		err = PTR_ERR(w->private);
+		goto err_out_stop_thread;
+	}
+
+	mutex_lock(&p->thread_lock);
+	list_add_tail(&w->worker_entry, &p->ready_list);
+	p->thread_num++;
+	mutex_unlock(&p->thread_lock);
+
+	return 0;
+
+err_out_stop_thread:
+	kthread_stop(w->thread);
+err_out_free:
+	kfree(w);
+err_out_exit:
+	return err;
+}
+
+/*
+ * Destroy the whole pool.
+ */
+void thread_pool_destroy(struct thread_pool *p)
+{
+	while (p->thread_num) {
+		dprintk("%s: num: %d.\n", __func__, p->thread_num);
+		thread_pool_del_worker(p);
+	}
+
+	kfree(p);
+}
+
+/*
+ * Create a pool with given number of threads.
+ * They will have sequential IDs started from zero.
+ */
+struct thread_pool *thread_pool_create(int num, char *name,
+		void *(* init)(void *private),
+		void (* cleanup)(void *private),
+		void *private)
+{
+	struct thread_pool_worker *w, *tmp;
+	struct thread_pool *p;
+	int err = -ENOMEM;
+	int i;
+
+	p = kzalloc(sizeof(struct thread_pool), GFP_KERNEL);
+	if (!p)
+		goto err_out_exit;
+
+	init_waitqueue_head(&p->wait);
+	mutex_init(&p->thread_lock);
+	INIT_LIST_HEAD(&p->ready_list);
+	INIT_LIST_HEAD(&p->active_list);
+	p->thread_num = 0;
+
+	for (i=0; i<num; ++i) {
+		err = thread_pool_add_worker(p, name, i, init,
+				cleanup, private);
+		if (err)
+			goto err_out_free_all;
+	}
+
+	return p;
+
+err_out_free_all:
+	list_for_each_entry_safe(w, tmp, &p->ready_list, worker_entry) {
+		list_del(&w->worker_entry);
+		thread_pool_exit_worker(w);
+	}
+	kfree(p);
+err_out_exit:
+	return ERR_PTR(err);
+}
+
+/*
+ * Schedule execution of the action on a given thread,
+ * provided ID pointer has to match previously stored
+ * private data.
+ */
+int thread_pool_schedule_private(struct thread_pool *p,
+		int (* setup)(void *private, void *data),
+		int (* action)(void *private, void *data),
+		void *data, long timeout, void *id)
+{
+	struct thread_pool_worker *w, *tmp, *worker = NULL;
+	int err = 0;
+
+	while (!worker && !err) {
+		timeout = wait_event_interruptible_timeout(p->wait,
+				!list_empty(&p->ready_list),
+				timeout);
+
+		if (!timeout) {
+			err = -ETIMEDOUT;
+			break;
+		}
+
+		worker = NULL;
+		mutex_lock(&p->thread_lock);
+		list_for_each_entry_safe(w, tmp, &p->ready_list, worker_entry) {
+			if (id && id != w->private)
+				continue;
+
+			worker = w;
+
+			list_move_tail(&w->worker_entry, &p->active_list);
+
+			err = setup(w->private, data);
+			if (!err) {
+				w->schedule_data = data;
+				w->action = action;
+				w->has_data = 1;
+				wake_up(&w->wait);
+			} else {
+				list_move_tail(&w->worker_entry, &p->ready_list);
+			}
+
+			break;
+		}
+		mutex_unlock(&p->thread_lock);
+	}
+
+	return err;
+}
+
+/*
+ * Schedule execution on arbitrary thread from the pool.
+ */
+int thread_pool_schedule(struct thread_pool *p,
+		int (* setup)(void *private, void *data),
+		int (* action)(void *private, void *data),
+		void *data, long timeout)
+{
+	return thread_pool_schedule_private(p, setup,
+			action, data, timeout, NULL);
+}

731 lines not on this page · raw message