From patchwork Wed Nov 5 16:29:16 2008 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Evgeniy Polyakov X-Patchwork-Id: 7353 Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.176.167]) by ozlabs.org (Postfix) with ESMTP id B50464753C for ; Thu, 6 Nov 2008 03:29:26 +1100 (EST) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753420AbYKEQ3Z (ORCPT ); Wed, 5 Nov 2008 11:29:25 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1753164AbYKEQ3Y (ORCPT ); Wed, 5 Nov 2008 11:29:24 -0500 Received: from cs-studio.ru ([195.178.208.66]:39747 "EHLO tservice.net.ru" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S1752352AbYKEQ3T (ORCPT ); Wed, 5 Nov 2008 11:29:19 -0500 Received: by tservice.net.ru (Postfix, from userid 1000) id 40927FE4A; Wed, 5 Nov 2008 19:29:16 +0300 (MSK) Date: Wed, 5 Nov 2008 19:29:16 +0300 From: Evgeniy Polyakov To: linux-kernel@vger.kernel.org Cc: netdev@vger.kernel.org, linux-fsdevel@vger.kernel.org Subject: [resend take 1] [1/3] Distributed storage release: core files. Message-ID: <20081105162916.GA11096@ioremap.net> References: <20081105162704.GA10787@ioremap.net> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <20081105162704.GA10787@ioremap.net> User-Agent: Mutt/1.5.13 (2006-08-11) Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org Signed-off-by: Evgeniy Polyakov diff --git a/drivers/block/dst/crypto.c b/drivers/block/dst/crypto.c new file mode 100644 index 0000000..7250f90 --- /dev/null +++ b/drivers/block/dst/crypto.c @@ -0,0 +1,731 @@ +/* + * 2007+ Copyright (c) Evgeniy Polyakov + * 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 +#include +#include +#include +#include +#include + +/* + * Tricky bastard, but IV can be more complex with time... + */ +static inline u64 dst_gen_iv(struct dst_trans *t) +{ + return t->gen; +} + +/* + * Crypto machinery: hash/cipher support for the given crypto controls. + */ +static struct crypto_hash *dst_init_hash(struct dst_crypto_ctl *ctl, u8 *key) +{ + int err; + struct crypto_hash *hash; + + hash = crypto_alloc_hash(ctl->hash_algo, 0, CRYPTO_ALG_ASYNC); + if (IS_ERR(hash)) { + err = PTR_ERR(hash); + dprintk("%s: failed to allocate hash '%s', err: %d.\n", + __func__, ctl->hash_algo, err); + goto err_out_exit; + } + + ctl->crypto_attached_size = crypto_hash_digestsize(hash); + + if (!ctl->hash_keysize) + return hash; + + err = crypto_hash_setkey(hash, key, ctl->hash_keysize); + if (err) { + dprintk("%s: failed to set key for hash '%s', err: %d.\n", + __func__, ctl->hash_algo, err); + goto err_out_free; + } + + return hash; + +err_out_free: + crypto_free_hash(hash); +err_out_exit: + return ERR_PTR(err); +} + +static struct crypto_ablkcipher *dst_init_cipher(struct dst_crypto_ctl *ctl, u8 *key) +{ + int err = -EINVAL; + struct crypto_ablkcipher *cipher; + + if (!ctl->cipher_keysize) + goto err_out_exit; + + cipher = crypto_alloc_ablkcipher(ctl->cipher_algo, 0, 0); + if (IS_ERR(cipher)) { + err = PTR_ERR(cipher); + dprintk("%s: failed to allocate cipher '%s', err: %d.\n", + __func__, ctl->cipher_algo, err); + goto err_out_exit; + } + + crypto_ablkcipher_clear_flags(cipher, ~0); + + err = crypto_ablkcipher_setkey(cipher, key, ctl->cipher_keysize); + if (err) { + dprintk("%s: failed to set key for cipher '%s', err: %d.\n", + __func__, ctl->cipher_algo, err); + goto err_out_free; + } + + return cipher; + +err_out_free: + crypto_free_ablkcipher(cipher); +err_out_exit: + return ERR_PTR(err); +} + +/* + * Crypto engine has a pool of pages to encrypt data into before sending + * it over the network. This pool is freed/allocated here. + */ +static void dst_crypto_pages_free(struct dst_crypto_engine *e) +{ + unsigned int i; + + for (i=0; ipage_num; ++i) + __free_page(e->pages[i]); + kfree(e->pages); +} + +static int dst_crypto_pages_alloc(struct dst_crypto_engine *e, int num) +{ + int i; + + e->pages = kmalloc(num * sizeof(struct page **), GFP_KERNEL); + if (!e->pages) + return -ENOMEM; + + for (i=0; ipages[i] = alloc_page(GFP_KERNEL); + if (!e->pages[i]) + goto err_out_free_pages; + } + + e->page_num = num; + return 0; + +err_out_free_pages: + while (--i >= 0) + __free_page(e->pages[i]); + + kfree(e->pages); + return -ENOMEM; +} + +/* + * Initialize crypto engine for given node. + * Setup cipher/hash, keys, pool of threads and private data. + */ +static int dst_crypto_engine_init(struct dst_crypto_engine *e, struct dst_node *n) +{ + int err; + struct dst_crypto_ctl *ctl = &n->crypto; + + err = dst_crypto_pages_alloc(e, n->max_pages); + if (err) + goto err_out_exit; + + e->size = PAGE_SIZE; + e->data = kmalloc(e->size, GFP_KERNEL); + if (!e->data) { + err = -ENOMEM; + goto err_out_free_pages; + } + + if (ctl->hash_algo[0]) { + e->hash = dst_init_hash(ctl, n->hash_key); + if (IS_ERR(e->hash)) { + err = PTR_ERR(e->hash); + e->hash = NULL; + goto err_out_free; + } + } + + if (ctl->cipher_algo[0]) { + e->cipher = dst_init_cipher(ctl, n->cipher_key); + if (IS_ERR(e->cipher)) { + err = PTR_ERR(e->cipher); + e->cipher = NULL; + goto err_out_free_hash; + } + } + + return 0; + +err_out_free_hash: + crypto_free_hash(e->hash); +err_out_free: + kfree(e->data); +err_out_free_pages: + dst_crypto_pages_free(e); +err_out_exit: + return err; +} + +static void dst_crypto_engine_exit(struct dst_crypto_engine *e) +{ + if (e->hash) + crypto_free_hash(e->hash); + if (e->cipher) + crypto_free_ablkcipher(e->cipher); + dst_crypto_pages_free(e); + kfree(e->data); +} + +/* + * Waiting for cipher processing to be completed. + */ +struct dst_crypto_completion +{ + struct completion complete; + int error; +}; + +static void dst_crypto_complete(struct crypto_async_request *req, int err) +{ + struct dst_crypto_completion *c = req->data; + + if (err == -EINPROGRESS) + return; + + dprintk("%s: req: %p, err: %d.\n", __func__, req, err); + c->error = err; + complete(&c->complete); +} + +static int dst_crypto_process(struct ablkcipher_request *req, + struct scatterlist *sg_dst, struct scatterlist *sg_src, + void *iv, int enc, unsigned long timeout) +{ + struct dst_crypto_completion c; + int err; + + init_completion(&c.complete); + c.error = -EINPROGRESS; + + ablkcipher_request_set_callback(req, CRYPTO_TFM_REQ_MAY_BACKLOG, + dst_crypto_complete, &c); + + ablkcipher_request_set_crypt(req, sg_src, sg_dst, sg_src->length, iv); + + if (enc) + err = crypto_ablkcipher_encrypt(req); + else + err = crypto_ablkcipher_decrypt(req); + + switch (err) { + case -EINPROGRESS: + case -EBUSY: + err = wait_for_completion_interruptible_timeout(&c.complete, + timeout); + if (!err) + err = -ETIMEDOUT; + else + err = c.error; + break; + default: + break; + } + + return err; +} + +/* + * DST uses generic iteration approach for data crypto processing. + * Single block IO request is switched into array of scatterlists, + * which are submitted to the crypto processing iterator. + * + * Input and output iterator initialization are different, since + * in output case we can not encrypt data in-place and need a + * temporary storage, which is then being sent to the remote peer. + */ +static int dst_trans_iter_out(struct bio *bio, struct dst_crypto_engine *e, + int (* iterator) (struct dst_crypto_engine *e, + struct scatterlist *dst, + struct scatterlist *src)) +{ + struct bio_vec *bv; + int err, i; + + sg_init_table(e->src, bio->bi_vcnt); + sg_init_table(e->dst, bio->bi_vcnt); + + bio_for_each_segment(bv, bio, i) { + sg_set_page(&e->src[i], bv->bv_page, bv->bv_len, bv->bv_offset); + sg_set_page(&e->dst[i], e->pages[i], bv->bv_len, bv->bv_offset); + + err = iterator(e, &e->dst[i], &e->src[i]); + if (err) + return err; + } + + return 0; +} + +static int dst_trans_iter_in(struct bio *bio, struct dst_crypto_engine *e, + int (* iterator) (struct dst_crypto_engine *e, + struct scatterlist *dst, + struct scatterlist *src)) +{ + struct bio_vec *bv; + int err, i; + + sg_init_table(e->src, bio->bi_vcnt); + sg_init_table(e->dst, bio->bi_vcnt); + + bio_for_each_segment(bv, bio, i) { + sg_set_page(&e->src[i], bv->bv_page, bv->bv_len, bv->bv_offset); + sg_set_page(&e->dst[i], bv->bv_page, bv->bv_len, bv->bv_offset); + + err = iterator(e, &e->dst[i], &e->src[i]); + if (err) + return err; + } + + return 0; +} + +static int dst_crypt_iterator(struct dst_crypto_engine *e, + struct scatterlist *sg_dst, struct scatterlist *sg_src) +{ + struct ablkcipher_request *req = e->data; + u8 iv[32]; + + memset(iv, 0, sizeof(iv)); + + memcpy(iv, &e->iv, sizeof(e->iv)); + + return dst_crypto_process(req, sg_dst, sg_src, iv, e->enc, e->timeout); +} + +static int dst_crypt(struct dst_crypto_engine *e, struct bio *bio) +{ + struct ablkcipher_request *req = e->data; + + memset(req, 0, sizeof(struct ablkcipher_request)); + ablkcipher_request_set_tfm(req, e->cipher); + + if (e->enc) + return dst_trans_iter_out(bio, e, dst_crypt_iterator); + else + return dst_trans_iter_in(bio, e, dst_crypt_iterator); +} + +static int dst_hash_iterator(struct dst_crypto_engine *e, + struct scatterlist *sg_dst, struct scatterlist *sg_src) +{ + return crypto_hash_update(e->data, sg_src, sg_src->length); +} + +static int dst_hash(struct dst_crypto_engine *e, struct bio *bio, void *dst) +{ + struct hash_desc *desc = e->data; + int err; + + desc->tfm = e->hash; + desc->flags = 0; + + err = crypto_hash_init(desc); + if (err) + return err; + + err = dst_trans_iter_in(bio, e, dst_hash_iterator); + if (err) + return err; + + err = crypto_hash_final(desc, dst); + if (err) + return err; + + return 0; +} + +/* + * Initialize/cleanup a crypto thread. The only thing it should + * do is to allocate a pool of pages as temporary storage. + * And to setup cipher and/or hash. + */ +static void *dst_crypto_thread_init(void *data) +{ + struct dst_node *n = data; + struct dst_crypto_engine *e; + int err = -ENOMEM; + + e = kzalloc(sizeof(struct dst_crypto_engine), GFP_KERNEL); + if (!e) + goto err_out_exit; + e->src = kcalloc(2 * n->max_pages, sizeof(struct scatterlist), + GFP_KERNEL); + if (!e->src) + goto err_out_free; + + e->dst = e->src + n->max_pages; + + err = dst_crypto_engine_init(e, n); + if (err) + goto err_out_free_all; + + return e; + +err_out_free_all: + kfree(e->src); +err_out_free: + kfree(e); +err_out_exit: + return ERR_PTR(err); +} + +static void dst_crypto_thread_cleanup(void *private) +{ + struct dst_crypto_engine *e = private; + + dst_crypto_engine_exit(e); + kfree(e->src); + kfree(e); +} + +/* + * Initialize crypto engine for given node: store keys, create pool + * of threads, initialize each one. + * + * Each thread has unique ID, but 0 and 1 are reserved for receiving and accepting + * threads (if export node), so IDs could start from 2, but starting them + * from 10 allows easily understand what this thread is for. + */ +int dst_node_crypto_init(struct dst_node *n, struct dst_crypto_ctl *ctl) +{ + void *key = (ctl + 1); + int err = -ENOMEM, i; + char name[32]; + + if (ctl->hash_keysize) { + n->hash_key = kmalloc(ctl->hash_keysize, GFP_KERNEL); + if (!n->hash_key) + goto err_out_exit; + memcpy(n->hash_key, key, ctl->hash_keysize); + } + + if (ctl->cipher_keysize) { + n->cipher_key = kmalloc(ctl->cipher_keysize, GFP_KERNEL); + if (!n->cipher_key) + goto err_out_free_hash; + memcpy(n->cipher_key, key, ctl->cipher_keysize); + } + memcpy(&n->crypto, ctl, sizeof(struct dst_crypto_ctl)); + + for (i=0; ithread_num; ++i) { + snprintf(name, sizeof(name), "%s-crypto-%d", n->name, i); + /* Unique ids... */ + err = thread_pool_add_worker(n->pool, name, i+10, + dst_crypto_thread_init, dst_crypto_thread_cleanup, n); + if (err) + goto err_out_free_threads; + } + + return 0; + +err_out_free_threads: + while (--i >= 0) + thread_pool_del_worker_id(n->pool, i+10); + + if (ctl->cipher_keysize) + kfree(n->cipher_key); + ctl->cipher_keysize = 0; +err_out_free_hash: + if (ctl->hash_keysize) + kfree(n->hash_key); + ctl->hash_keysize = 0; +err_out_exit: + return err; +} + +void dst_node_crypto_exit(struct dst_node *n) +{ + struct dst_crypto_ctl *ctl = &n->crypto; + + if (ctl->cipher_algo[0] || ctl->hash_algo[0]) { + kfree(n->hash_key); + kfree(n->cipher_key); + } +} + +/* + * Thrad pool setup callback. Just stores a transaction in private data. + */ +static int dst_trans_crypto_setup(void *crypto_engine, void *trans) +{ + struct dst_crypto_engine *e = crypto_engine; + + e->private = trans; + return 0; +} + +#if 0 +static void dst_dump_bio(struct bio *bio) +{ + u8 *p; + struct bio_vec *bv; + int i; + + bio_for_each_segment(bv, bio, i) { + dprintk("%s: %llu/%u: size: %u, offset: %u, data: ", + __func__, bio->bi_sector, bio->bi_size, + bv->bv_len, bv->bv_offset); + + p = kmap(bv->bv_page) + bv->bv_offset; + for (i=0; ibv_len; ++i) + printk("%02x ", p[i]); + kunmap(bv->bv_page); + printk("\n"); + } +} +#endif + +/* + * Encrypt/hash data and send it to the network. + */ +static int dst_crypto_process_sending(struct dst_crypto_engine *e, + struct bio *bio, u8 *hash) +{ + int err; + + if (e->cipher) { + err = dst_crypt(e, bio); + if (err) + goto err_out_exit; + } + + if (e->hash) { + err = dst_hash(e, bio, hash); + if (err) + goto err_out_exit; + +#ifdef CONFIG_DST_DEBUG + { + unsigned int i; + + /* dst_dump_bio(bio); */ + + printk(KERN_DEBUG "%s: bio: %llu/%u, rw: %lu, hash: ", + __func__, (u64)bio->bi_sector, + bio->bi_size, bio_data_dir(bio)); + for (i=0; ihash); ++i) + printk("%02x ", hash[i]); + printk("\n"); + } +#endif + } + + return 0; + +err_out_exit: + return err; +} + +/* + * Check if received data is valid. Decipher if it is. + */ +static int dst_crypto_process_receiving(struct dst_crypto_engine *e, + struct bio *bio, u8 *hash, u8 *recv_hash) +{ + int err; + + if (e->hash) { + int mismatch; + + err = dst_hash(e, bio, hash); + if (err) + goto err_out_exit; + + mismatch = !!memcmp(recv_hash, hash, + crypto_hash_digestsize(e->hash)); +#ifdef CONFIG_DST_DEBUG + /* dst_dump_bio(bio); */ + + printk(KERN_DEBUG "%s: bio: %llu/%u, rw: %lu, hash mismatch: %d", + __func__, (u64)bio->bi_sector, bio->bi_size, + bio_data_dir(bio), mismatch); + if (mismatch) { + unsigned int i; + + printk(", recv/calc: "); + for (i=0; ihash); ++i) { + printk("%02x/%02x ", recv_hash[i], hash[i]); + } + } + printk("\n"); +#endif + err = -1; + if (mismatch) + goto err_out_exit; + } + + if (e->cipher) { + err = dst_crypt(e, bio); + if (err) + goto err_out_exit; + } + + return 0; + +err_out_exit: + return err; +} + +/* + * Thread pool callback to encrypt data and send it to the netowork. + */ +static int dst_trans_crypto_action(void *crypto_engine, void *schedule_data) +{ + struct dst_crypto_engine *e = crypto_engine; + struct dst_trans *t = schedule_data; + struct bio *bio = t->bio; + int err; + + dprintk("%s: t: %p, gen: %llu, cipher: %p, hash: %p.\n", + __func__, t, t->gen, e->cipher, e->hash); + + e->enc = t->enc; + e->iv = dst_gen_iv(t); + + if (bio_data_dir(bio) == WRITE) { + err = dst_crypto_process_sending(e, bio, t->cmd.hash); + if (err) + goto err_out_exit; + + if (e->hash) { + t->cmd.csize = crypto_hash_digestsize(e->hash); + t->cmd.size += t->cmd.csize; + } + + return dst_trans_send(t); + } else { + u8 *hash = e->data + e->size/2; + + err = dst_crypto_process_receiving(e, bio, hash, t->cmd.hash); + if (err) + goto err_out_exit; + + dst_trans_remove(t); + dst_trans_put(t); + } + + return 0; + +err_out_exit: + t->error = err; + dst_trans_put(t); + return err; +} + +/* + * Schedule crypto processing for given transaction. + */ +int dst_trans_crypto(struct dst_trans *t) +{ + struct dst_node *n = t->n; + int err; + + err = thread_pool_schedule(n->pool, + dst_trans_crypto_setup, dst_trans_crypto_action, + t, MAX_SCHEDULE_TIMEOUT); + if (err) + goto err_out_exit; + + return 0; + +err_out_exit: + dst_trans_put(t); + return err; +} + +/* + * Crypto machinery for the export node. + */ +static int dst_export_crypto_setup(void *crypto_engine, void *bio) +{ + struct dst_crypto_engine *e = crypto_engine; + + e->private = bio; + return 0; +} + +static int dst_export_crypto_action(void *crypto_engine, void *schedule_data) +{ + struct dst_crypto_engine *e = crypto_engine; + struct bio *bio = schedule_data; + struct dst_export_priv *p = bio->bi_private; + int err; + + dprintk("%s: e: %p, data: %p, bio: %llu/%u, dir: %lu.\n", __func__, + e, e->data, (u64)bio->bi_sector, bio->bi_size, bio_data_dir(bio)); + + e->enc = (bio_data_dir(bio) == READ); + e->iv = p->cmd.id; + + if (bio_data_dir(bio) == WRITE) { + u8 *hash = e->data + e->size/2; + + err = dst_crypto_process_receiving(e, bio, hash, p->cmd.hash); + if (err) + goto err_out_exit; + + generic_make_request(bio); + } else { + err = dst_crypto_process_sending(e, bio, p->cmd.hash); + if (err) + goto err_out_exit; + + if (e->hash) { + p->cmd.csize = crypto_hash_digestsize(e->hash); + p->cmd.size += p->cmd.csize; + } + + err = dst_export_send_bio(bio); + } + return 0; + +err_out_exit: + bio_put(bio); + return err; +} + +int dst_export_crypto(struct dst_node *n, struct bio *bio) +{ + int err; + + err = thread_pool_schedule(n->pool, + dst_export_crypto_setup, dst_export_crypto_action, + bio, MAX_SCHEDULE_TIMEOUT); + if (err) + goto err_out_exit; + + return 0; + +err_out_exit: + bio_put(bio); + return err; +} diff --git a/drivers/block/dst/dcore.c b/drivers/block/dst/dcore.c new file mode 100644 index 0000000..833f535 --- /dev/null +++ b/drivers/block/dst/dcore.c @@ -0,0 +1,962 @@ +/* + * 2007+ Copyright (c) Evgeniy Polyakov + * 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 +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +#include + +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); + +static char dst_name[] = "Leontin's mountains: Ararat, Masis, Aragats"; + +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, +}; + +/* + * Think twice before starting to 'mercilessly mock'. + * Finished? Now think again. + * And one more time. + * + * No delightful flash? + * Device can live in object with appropriate reference counters, + * so embedded device will be released when there are no users and its + * release callback can be empty. + */ +static void dst_node_release(struct device *dev) +{ +} + +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_node *n = container_of(dev, struct dst_node, device); + + return sprintf(buf, "%llu\n", n->size); +} + +/* + * 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_node *n = container_of(dev, struct dst_node, device); + struct sockaddr addr; + struct socket *sock; + int addrlen; + + sock = n->state->socket; + if (sock->ops->getname(sock, &addr, &addrlen, 2)) + return 0; + + if (sock->ops->family == AF_INET) { + struct sockaddr_in *sin = (struct sockaddr_in *)&addr; + return sprintf(buf, "%u.%u.%u.%u:%d\n", + NIPQUAD(sin->sin_addr.s_addr), ntohs(sin->sin_port)); + } else if (sock->ops->family == AF_INET6) { + struct sockaddr_in6 *sin = (struct sockaddr_in6 *)&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; + unsigned char *a = (unsigned char *)&addr; + char *buf_orig = buf; + + size = snprintf(buf, sz, "family: %d, addrlen: %u, addr: ", + addr.sa_family, addrlen); + sz -= size; + buf += size; + + for (i=0; idevice, + &dst_node_attrs[i]); + if (err) + goto err_out_remove_all; + } + return 0; + +err_out_remove_all: + while (--i >= 0) + device_remove_file(&n->device, + &dst_node_attrs[i]); + + return err; +} + +static void dst_remove_node_attributes(struct dst_node *n) +{ + int i; + + for (i=0; idevice, + &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->device.bus_id[0]) { + dst_remove_node_attributes(n); + device_unregister(&n->device); + memset(&n->device, 0, sizeof(n->device)); + } +} + +static int dst_node_sysfs_init(struct dst_node *n) +{ + int err; + + memcpy(&n->device, &dst_node_dev, sizeof(struct device)); + + snprintf(n->device.bus_id, sizeof(n->device.bus_id), "dst-%s", n->name); + err = device_register(&n->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: + 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 */ + + 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; + + 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; + + 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"); +MODULE_LICENSE("GPL");