| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CALLzPKbayaiWi_jCOw_o-moDD9w0uxERuo-GCjdnCjRSSu-ezA@mail.gmail.com>
Date: Wed, 5 Dec 2012 13:06:56 +0200
From: "Kasatkin, Dmitry" <dmitry.kasatkin@...el.com>
To: linux-kernel@...r.kernel.org,
linux-security-module@...r.kernel.org, dm-devel@...hat.com,
linux-crypto@...r.kernel.org
Subject: Re: [PATCHv2 1/1] dm-integrity: integrity protection device-mapper target
Hello,
Any comments?
- Dmitry
On Mon, Nov 26, 2012 at 4:39 PM, Dmitry Kasatkin
<dmitry.kasatkin@...el.com> wrote:
> Device-mapper "integrity" target provides transparent cryptographic integrity
> protection of the underlying read-write block device using hash-based message
> authentication codes (HMACs). HMACs can be stored on the same or different
> block device.
>
> dm-integrity uses an encrypted key type, stored on the kernel keyring, to
> obtain a secret key for use in cryptographic operations. Encrypted keys are
> never exposed in plain text to user space. The encrypted keys are encrypted
> using master key, which can either be a user defined or trusted key type.
> The secret key, which is usually device specific, binds integrity data to the
> device. As a result data blocks and corresponding HMACs cannot simply be
> copied over from other file systems.
>
> Signed-off-by: Dmitry Kasatkin <dmitry.kasatkin@...el.com>
> ---
> Documentation/device-mapper/dm-integrity.txt | 137 ++++
> drivers/md/Kconfig | 13 +
> drivers/md/Makefile | 1 +
> drivers/md/dm-integrity.c | 1050 ++++++++++++++++++++++++++
> 4 files changed, 1201 insertions(+)
> create mode 100644 Documentation/device-mapper/dm-integrity.txt
> create mode 100644 drivers/md/dm-integrity.c
>
> diff --git a/Documentation/device-mapper/dm-integrity.txt b/Documentation/device-mapper/dm-integrity.txt
> new file mode 100644
> index 0000000..394242f
> --- /dev/null
> +++ b/Documentation/device-mapper/dm-integrity.txt
> @@ -0,0 +1,137 @@
> +dm-integrity
> +===============
> +
> +Device-mapper "integrity" target provides transparent cryptographic integrity
> +protection of the underlying read-write block device using hash-based message
> +authentication codes (HMACs). HMACs can be stored on the same or different
> +block device.
> +
> +dm-integrity uses an encrypted key type, stored on the kernel keyring, to
> +obtain a secret key for use in cryptographic operations. Encrypted keys are
> +never exposed in plain text to user space. The encrypted keys are encrypted
> +using master key, which can either be a user defined or trusted key type.
> +The secret key, which is usually device specific, binds integrity data to the
> +device. As a result data blocks and corresponding HMACs cannot simply be
> +copied over from other file systems.
> +
> +Parameters:
> +<dev> <bs> <start> <hdev> <hbs> <hstart> <hash_algo> <hmac_algo> <key_desc> \
> +[<opt_params>]
> +
> +<dev>
> + This is the device that is going to be used to store the data.
> + You can specify it as a path like /dev/xxx or a device <major>:<minor>
> + number.
> +
> +<bs>
> + Device block size.
> +
> +<start>
> + Starting sector within the device where data begins.
> +
> +<hdev>
> + This is the device that is going to be used to store integrity data.
> + You can specify it as a path like /dev/xxx or a device <major>:<minor>
> + number.
> +
> +<hbs>
> + HMAC device block size.
> +
> +<hstart>
> + Starting sector within the device where integrity data begins.
> +
> +<hash_algo>
> + Hash algorithm (sha1, sha256, etc).
> +
> +<hmac_algo>
> + HMAC algorithm, e.g. hmac(sha1), hmac(sha256), etc.
> +
> +<key_desc>
> + Description is a name of a key in the kernel keyring.
> +
> +<opt_params>
> + fix=1|0 - enable fix mode
> + In fix mode, incorrect hmacs are replaced with correct ones.
> + It is used for device initialization and debugging.
> +
> + stats=1|0 - turns on collecting additional statistical information.
> + It is used to find out resource usage to tune memory pool
> + and queue sizes for particular use case.
> +
> + verbose=1|0 - prints block number, collected hmac and stored hmac.
> + It is used for addition debug output.
> +
> +
> +Determine the size of integrity/hmac device
> +===============
> +
> +Every block device has corresponding hmac.
> +While NIST does recommend to use sha256 hash algorithm instead of SHA1,
> +this does not apply to hmac(sha1), because of keying. It is safe to use
> +hmac(sha1), because it takes much less space and it is faster to calculate.
> +hmac(sha1) size is 20 bytes. So every 4k block on the integrity device can
> +store 204 hmacs. In order to get the required size of the integrity device,
> +it is necessary to divide data device size by 204. See examples bellow how
> +to do it from script.
> +
> +Example scripts
> +===============
> +
> +1. Setting up integrity target using data and hmac store on the same block device.
> +
> +[[
> +#!/bin/sh
> +
> +bdev=$1
> +
> +# block device size
> +dsize=`blockdev --getsize $bdev`
> +# block size
> +bs=4096
> +# sector to block shift
> +sbs=3
> +# integrity record size (hmac size)
> +hmac=20
> +# hmacs per block
> +hpb=$((bs/hmac))
> +# target device size
> +size=$((((dsize>>sbs)*hpb/(hpb+1))<<sbs))
> +
> +# load the key - in this example we just use test key
> +keyctl add user kmk "testing123" @u
> +keyctl add encrypted dm-int-key "load `cat /etc/keys/dm-int-key`" @u
> +
> +# creating the target
> +table="0 $size integrity $bdev 4096 0 $bdev 4096 $size sha1 hmac(sha1) dm-int-key"
> +dmsetup create dm-int --table "$table"
> +
> +# mounting
> +mount /dev/mapper/dm-int /mnt
> +
> +]]
> +
> +2. Setting up integrity target using data and hmac store on different block devices.
> +
> +[[
> +#!/bin/sh
> +
> +bdev=$1
> +hdev=$2
> +
> +# get size of the block device
> +dsize=`blockdev --getsz $bdev`
> +# round down the size to 4k blocks
> +dsize=$((dsize & ~7))
> +
> +# load the key - in this example we just use test key
> +keyctl add user kmk "testing123" @u
> +keyctl add encrypted dm-int-key "load `cat /etc/keys/dm-int-key`" @u
> +
> +# creating the target
> +table="0 $dsize integrity $bdev 4096 0 $hdev 4096 0 sha1 hmac(sha1) dm-int-key"
> +dmsetup create dm-int --table "$table"
> +
> +# mounting
> +mount /dev/mapper/dm-int /mnt
> +
> +]]
> diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig
> index 91a02ee..42249c1 100644
> --- a/drivers/md/Kconfig
> +++ b/drivers/md/Kconfig
> @@ -388,6 +388,19 @@ config DM_VERITY
> To compile this code as a module, choose M here: the module will
> be called dm-verity.
>
> +config DM_INTEGRITY
> + tristate "Integrity target support"
> + depends on BLK_DEV_DM
> + select CRYPTO
> + select CRYPTO_HMAC
> + select DM_BUFIO
> + select KEYS
> + ---help---
> + If you say Y here, then your ...
> +
> + To compile this as a module, choose M here: the module
> + will be called dm-integrity.
> +
> If unsure, say N.
>
> endif # MD
> diff --git a/drivers/md/Makefile b/drivers/md/Makefile
> index 94dce8b..49b212f 100644
> --- a/drivers/md/Makefile
> +++ b/drivers/md/Makefile
> @@ -41,6 +41,7 @@ obj-$(CONFIG_DM_PERSISTENT_DATA) += persistent-data/
> obj-$(CONFIG_DM_MIRROR) += dm-mirror.o dm-log.o dm-region-hash.o
> obj-$(CONFIG_DM_LOG_USERSPACE) += dm-log-userspace.o
> obj-$(CONFIG_DM_ZERO) += dm-zero.o
> +obj-$(CONFIG_DM_INTEGRITY) += dm-integrity.o
> obj-$(CONFIG_DM_RAID) += dm-raid.o
> obj-$(CONFIG_DM_THIN_PROVISIONING) += dm-thin-pool.o
> obj-$(CONFIG_DM_VERITY) += dm-verity.o
> diff --git a/drivers/md/dm-integrity.c b/drivers/md/dm-integrity.c
> new file mode 100644
> index 0000000..3b3a27e
> --- /dev/null
> +++ b/drivers/md/dm-integrity.c
> @@ -0,0 +1,1050 @@
> +/*
> + * dm-integrity - device mapper integrity target
> + *
> + * Copyright (C) 2012, Intel Corporation.
> + *
> + * Author: Dmitry Kasatkin <dmitry.kasatkin@...el.com>
> + *
> + * This program is free software; you can redistribute it and/or modify it
> + * under the terms and conditions of the GNU General Public License,
> + * version 2, as published by the Free Software Foundation.
> + *
> + * This program is distributed in the hope 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.,
> + * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
> + *
> + */
> +
> +#define DM_MSG_PREFIX KBUILD_MODNAME
> +
> +#define pr_fmt(fmt) KBUILD_MODNAME ": %s: " fmt, __func__
> +
> +#include "dm.h"
> +#include <linux/module.h>
> +#include <linux/init.h>
> +#include <linux/blkdev.h>
> +#include <linux/bio.h>
> +#include <linux/slab.h>
> +#include <linux/device-mapper.h>
> +#include <linux/crypto.h>
> +#include <linux/scatterlist.h>
> +#include <crypto/sha.h>
> +#include <crypto/hash.h>
> +#include <keys/encrypted-type.h>
> +#include <linux/wait.h>
> +#include <linux/reboot.h>
> +
> +#include "dm-bufio.h"
> +
> +#define DM_INT_STATS
> +
> +#define DM_INT_MIN_IOS 16
> +#define DM_INT_BLOCK_SIZE PAGE_SIZE
> +#define DM_INT_MAX_KEY_SIZE 128
> +
> +/* best parameters for fastest Ubuntu boot */
> +#define DM_INT_PREFETCH_COUNT 16
> +#define DM_INT_FLUSH_DELAY (HZ * 3) /* bufio flush delay */
> +
> +struct ahash_result {
> + struct completion completion;
> + int err;
> +};
> +
> +struct dm_int_io {
> + struct dm_int *dmi; /* mapping it belongs to */
> + struct bio *bio;
> + struct work_struct work;
> +
> +#define DM_INT_BIO_DONE 1
> +#define DM_INT_VERIFIED 2
> +#define DM_INT_HMAC_DONE 3
> + unsigned long flags;
> +
> + atomic_t count;
> + int error;
> +
> + sector_t sector;
> +
> + bio_end_io_t *bi_end_io; /* original bio callback */
> + void *bi_private; /* original bio private data */
> + unsigned int bi_size;
> +
> + struct ahash_request req;
> +};
> +
> +/*
> + * integrity mapping configuration
> + */
> +struct dm_int {
> + struct dm_target *target;
> + struct dm_dev *dev;
> + char *table_string;
> + loff_t start;
> + struct dm_dev *hdev;
> + loff_t hmac_start;
> + loff_t hmac_count;
> +
> + struct mutex mutex; /* lock the store */
> +
> + struct workqueue_struct *io_queue;
> + struct kmem_cache *io_cache;
> + mempool_t *io_pool;
> +
> + struct crypto_ahash *ahash;
> + struct crypto_shash *hmac;
> +
> + struct list_head list; /* global list */
> +
> + struct dm_bufio_client *bufio;
> +
> + unsigned int hmac_size;
> + unsigned int data_block_size;
> + unsigned int data_block_bits;
> + unsigned int hmac_block_size;
> + unsigned int hmac_block_bits;
> + unsigned int hmac_per_block;
> + unsigned int hmac_block_shift;
> + unsigned int delay; /* hmac sync delay */
> +
> +#define DM_INT_FLAGS_FIX 0x01 /* fix wrong hmacs */
> +#ifdef DM_INT_STATS
> +#define DM_INT_FLAGS_STATS 0x02 /* calc stats */
> +#else
> +/* setting to 0 will eliminate the code due to optimization */
> +#define DM_INT_FLAGS_STATS 0x00
> +#endif
> +#define DM_INT_FLAGS_VERBOSE 0x04 /* show failed blocks */
> + unsigned int flags;
> +
> + atomic_t count; /* total reference count */
> + wait_queue_head_t wait; /* for close */
> + atomic_t violations;
> +
> + /* stats */
> +#ifdef DM_INT_STATS
> + atomic_t io_count;
> + int io_count_max;
> + atomic_t data_write_count;
> + atomic_t data_read_count;
> +#endif
> +};
> +
> +static DEFINE_MUTEX(mutex);
> +static LIST_HEAD(dmi_list);
> +static int sync_mode;
> +
> +static void dm_int_queue_hmac(struct dm_int_io *io);
> +
> +/*
> + * Get the key from the TPM for the HMAC
> + */
> +static int dm_int_init_crypto(struct dm_int *dmi, const char *hash_algo,
> + const char *hmac_algo, const char *keyname)
> +{
> + struct key *key;
> + struct encrypted_key_payload *ekp;
> + int err = -EINVAL;
> +
> + dmi->ahash = crypto_alloc_ahash(hash_algo, 0, 0);
> + if (IS_ERR(dmi->ahash)) {
> + err = PTR_ERR(xchg(&dmi->ahash, NULL));
> + DMERR("failed to load %s algorithm: %d\n", hash_algo, err);
> + dmi->target->error = "Cannot allocate hash algorithm";
> + return err;
> + }
> +
> + dmi->hmac = crypto_alloc_shash(hmac_algo, 0, 0);
> + if (IS_ERR(dmi->hmac)) {
> + err = PTR_ERR(xchg(&dmi->hmac, NULL));
> + DMERR("failed to load %s algorithm: %d\n", hmac_algo, err);
> + dmi->target->error = "Cannot allocate hash algorithm";
> + return err;
> + }
> +
> + key = request_key(&key_type_encrypted, keyname, NULL);
> + if (IS_ERR(key)) {
> + dmi->target->error = "Invalid key name";
> + return -ENOENT;
> + }
> +
> + down_read(&key->sem);
> + ekp = key->payload.data;
> + if (ekp->decrypted_datalen <= DM_INT_MAX_KEY_SIZE)
> + err = crypto_shash_setkey(dmi->hmac, ekp->decrypted_data,
> + ekp->decrypted_datalen);
> +
> + /* burn the original key contents */
> + /*memset(ekp->decrypted_data, 0, ekp->decrypted_datalen); */
> + up_read(&key->sem);
> + key_put(key);
> +
> + return err;
> +}
> +
> +static void dm_int_io_get(struct dm_int_io *io)
> +{
> + struct dm_int *dmi = io->dmi;
> +
> + atomic_inc(&io->count);
> + atomic_inc(&dmi->count);
> +
> + pr_debug("entered: io: %p, pending %d/%d\n",
> + io, atomic_read(&io->count), atomic_read(&dmi->count));
> +}
> +
> +static void dm_int_io_put(struct dm_int_io *io)
> +{
> + struct dm_int *dmi = io->dmi;
> + struct bio *bio = io->bio;
> + int err = io->error;
> +
> + pr_debug("entered: io: %p, pending %d/%d\n",
> + io, atomic_read(&io->count), atomic_read(&dmi->count));
> +
> + atomic_dec(&dmi->count);
> +
> + if (!atomic_dec_and_test(&io->count))
> + return;
> +
> + /* request has completed */
> + if (!err && test_bit(DM_INT_BIO_DONE, &io->flags) &&
> + !test_bit(DM_INT_VERIFIED, &io->flags)) {
> + /* io->count will be 1 */
> + pr_debug("queue to verify: %p\n", io);
> + dm_int_queue_hmac(io);
> + return;
> + }
> +
> + pr_debug("io done: err: %d, io: %d/%d\n",
> + err, atomic_read(&io->count), atomic_read(&dmi->count));
> +
> + mempool_free(io, dmi->io_pool);
> +
> + bio_endio(bio, err); /* finally completed, end main bio */
> +
> + if (dmi->flags & DM_INT_FLAGS_STATS)
> + atomic_dec(&dmi->io_count);
> +
> + if (err)
> + DMERR("ERROR: io done: %d\n", err);
> +
> + wake_up_all(&dmi->wait);
> +}
> +
> +static void dm_int_prefetch(struct dm_int_io *io)
> +{
> + struct dm_int *dmi = io->dmi;
> + loff_t first, last, data;
> + loff_t offset;
> +
> + /* block number to read */
> + offset = io->sector << SECTOR_SHIFT;
> + data = offset >> dmi->data_block_bits;
> + if (dmi->hmac_block_shift)
> + first = data >> dmi->hmac_block_shift;
> + else {
> + first = data;
> + sector_div(first, dmi->hmac_per_block);
> + }
> +
> + /* offset to the last byte of data */
> + offset += (io->bi_size - 1);
> + data = offset >> dmi->data_block_bits;
> + if (dmi->hmac_block_shift)
> + last = data >> dmi->hmac_block_shift;
> + else {
> + last = data;
> + sector_div(last, dmi->hmac_per_block);
> + }
> +
> + /* prefetch multiple of DM_INT_PREFETCH_COUNT */
> + first = round_down(first, DM_INT_PREFETCH_COUNT);
> + last = round_up(last + 1, DM_INT_PREFETCH_COUNT);
> + /* check the end of the device */
> + if (last > dmi->hmac_count)
> + last = dmi->hmac_count;
> +
> + dm_bufio_prefetch(dmi->bufio, dmi->hmac_start + first, last - first);
> +}
> +
> +static int dm_int_verify_hmac(struct dm_int_io *io, loff_t offset,
> + u8 *collected, int update)
> +{
> + struct dm_int *dmi = io->dmi;
> + loff_t block, data = offset >> dmi->data_block_bits;
> + unsigned int index;
> + u8 *digest;
> + int err = 0;
> + struct dm_buffer *buf;
> +
> + if (dmi->hmac_block_shift) {
> + block = data >> dmi->hmac_block_shift;
> + index = data & ((1 << dmi->hmac_block_shift) - 1);
> + } else {
> + block = data;
> + index = sector_div(block, dmi->hmac_per_block);
> + }
> +
> + pr_debug("hmac: block: %llu, index: %u\n", block, index);
> +
> + digest = dm_bufio_read(dmi->bufio, dmi->hmac_start + block, &buf);
> + if (unlikely(IS_ERR(digest)))
> + return PTR_ERR(digest);
> +
> + digest += dmi->hmac_size * index;
> +
> + if (!update) {
> + err = memcmp(digest, collected, dmi->hmac_size);
> + if (err) {
> + err = -EIO;
> + /* update buffer and store it back */
> + atomic_inc(&dmi->violations);
> + if (dmi->flags & DM_INT_FLAGS_FIX) {
> + err = 0;
> + update = 1;
> + }
> + if (dmi->flags & DM_INT_FLAGS_VERBOSE) {
> + DMERR("ERROR: hmacs does not match\n");
> + DMERR("hmac: block: %llu, index: %u\n",
> + block, index);
> + print_hex_dump(KERN_CRIT, "collected: ",
> + 0, 32, 1, collected, 20, 0);
> + print_hex_dump(KERN_CRIT, "hmac: ",
> + 0, 32, 1, digest, 20, 0);
> + }
> + }
> + }
> +
> + if (update) {
> + memcpy(digest, collected, dmi->hmac_size);
> + dm_bufio_mark_buffer_dirty(buf);
> + }
> +
> + dm_bufio_release(buf);
> +
> + if (err)
> + DMERR_LIMIT("ERROR: HMACs do not match\n");
> +
> + return err;
> +}
> +
> +static void dm_int_ahash_complete(struct crypto_async_request *req, int err)
> +{
> + struct ahash_result *res = req->data;
> +
> + if (err == -EINPROGRESS)
> + return;
> + res->err = err;
> + complete(&res->completion);
> +}
> +
> +static int dm_int_ahash_wait(int err, struct ahash_result *res)
> +{
> + switch (err) {
> + case 0:
> + break;
> + case -EINPROGRESS:
> + case -EBUSY:
> + wait_for_completion(&res->completion);
> + err = res->err;
> + if (!res->err) {
> + INIT_COMPLETION(res->completion);
> + break;
> + }
> + /* fall through */
> + default:
> + DMERR("HMAC calculation failed: err: %d\n", err);
> + }
> +
> + return err;
> +}
> +
> +static int dm_int_calc_hmac(struct dm_int_io *io, loff_t offset,
> + u8 *digest, unsigned int size, u8 *hmac)
> +{
> + struct dm_int *dmi = io->dmi;
> + int err;
> + struct {
> + struct shash_desc shash;
> + char ctx[crypto_shash_descsize(dmi->hmac)];
> + } desc;
> +
> + desc.shash.tfm = dmi->hmac;
> + desc.shash.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
> +
> + err = crypto_shash_init(&desc.shash);
> + if (!err)
> + err = crypto_shash_update(&desc.shash, digest, size);
> + if (!err)
> + err = crypto_shash_finup(&desc.shash, (u8 *)&offset,
> + sizeof(offset), hmac);
> + if (err)
> + DMERR("ERROR: calc hmac failed: %d\n", err);
> + return err;
> +}
> +
> +static void dm_int_verify_io(struct dm_int_io *io)
> +{
> + struct dm_int *dmi = io->dmi;
> + struct bio *bio = io->bio;
> + struct bio_vec *bv;
> + int i, err = -EIO;
> + struct scatterlist sg[1];
> + u8 hmac[dmi->hmac_size];
> + u8 digest[crypto_ahash_digestsize(dmi->ahash)];
> + loff_t offset = io->sector << SECTOR_SHIFT;
> + unsigned int update = bio_data_dir(bio);
> + struct ahash_request *req = &io->req;
> + struct ahash_result res;
> + ssize_t size = io->bi_size;
> +
> + init_completion(&res.completion);
> + ahash_request_set_callback(req, CRYPTO_TFM_REQ_MAY_BACKLOG |
> + CRYPTO_TFM_REQ_MAY_SLEEP,
> + dm_int_ahash_complete, &res);
> +
> + sg_init_table(sg, 1);
> +
> + pr_debug("io: %p, sector: %llu, size: %d, vcnt: %d, idx: %d\n",
> + io, (loff_t)io->sector,
> + bio->bi_size, bio->bi_vcnt, bio->bi_idx);
> +
> + bio_for_each_segment(bv, bio, i) {
> + pr_debug("bv: %d: offset: %llu, bv_offset: %d, bv_len: %d\n",
> + i, offset, bv->bv_offset, bv->bv_len);
> +
> + BUG_ON(bv->bv_offset & (dmi->data_block_size - 1));
> + BUG_ON(bv->bv_len & (dmi->data_block_size - 1));
> +
> + sg_set_page(sg, bv->bv_page, bv->bv_len, bv->bv_offset);
> +
> + ahash_request_set_crypt(req, sg, digest, bv->bv_len);
> +
> + err = crypto_ahash_digest(req);
> + err = dm_int_ahash_wait(err, req->base.data);
> + if (err)
> + break;
> +
> + err = dm_int_calc_hmac(io, offset, digest, sizeof(digest),
> + hmac);
> + if (err)
> + break;
> +
> + err = dm_int_verify_hmac(io, offset, hmac, update);
> + if (err)
> + break;
> +
> + offset += bv->bv_len;
> + size -= bv->bv_len;
> + }
> +
> + if (size)
> + DMERR("ERROR: size is not zero: %zd\n", size);
> +
> + io->error = err;
> + set_bit(DM_INT_VERIFIED, &io->flags);
> +
> + if (sync_mode)
> + dm_bufio_write_dirty_buffers(dmi->bufio);
> +}
> +
> +static void dm_int_hmac_task(struct work_struct *work)
> +{
> + struct dm_int_io *io = container_of(work, struct dm_int_io, work);
> +
> + if (test_and_set_bit(DM_INT_HMAC_DONE, &io->flags))
> + dm_int_verify_io(io);
> + else
> + dm_int_prefetch(io);
> +
> + dm_int_io_put(io);
> +}
> +
> +static void dm_int_queue_hmac(struct dm_int_io *io)
> +{
> + struct dm_int *dmi = io->dmi;
> + int ret;
> +
> + /* what if it is queued already? */
> + dm_int_io_get(io);
> + ret = queue_work(dmi->io_queue, &io->work);
> + if (!ret)
> + dm_int_io_put(io);
> + BUG_ON(!ret);
> +}
> +
> +static void dm_int_end_io(struct bio *bio, int err)
> +{
> + struct dm_int_io *io = bio->bi_private;
> + struct dm_int *dmi = io->dmi;
> +
> + pr_debug("io: %p, pending: %d/%d, sector: %llu, size: %u, "\
> + "vcnt: %d, idx: %d\n", io,
> + atomic_read(&io->count), atomic_read(&dmi->count),
> + (loff_t)bio->bi_sector,
> + bio->bi_size, bio->bi_vcnt, bio->bi_idx);
> +
> + if (unlikely(!bio_flagged(bio, BIO_UPTODATE) && !err))
> + err = -EIO;
> +
> + if (err)
> + DMERR("ERROR: bio io failed: %d\n", err);
> +
> + if (unlikely(err))
> + io->error = err;
> +
> + set_bit(DM_INT_BIO_DONE, &io->flags);
> +
> + bio->bi_private = io->bi_private;
> + bio->bi_end_io = io->bi_end_io;
> +
> + dm_int_io_put(io);
> +}
> +
> +static void dm_int_start_io(struct dm_int_io *io)
> +{
> + struct dm_int *dmi = io->dmi;
> + struct bio *bio = io->bio;
> +
> + if (io->error)
> + return;
> +
> + io->bi_private = bio->bi_private;
> + io->bi_end_io = bio->bi_end_io;
> +
> + /* io->sector starts from 0 */
> + bio->bi_sector = dmi->start + io->sector;
> + bio->bi_bdev = dmi->dev->bdev;
> +
> + bio->bi_private = io;
> + bio->bi_end_io = dm_int_end_io;
> +
> + dm_int_io_get(io);
> +
> + if (dmi->flags & DM_INT_FLAGS_STATS) {
> + if (bio_data_dir(bio) == READ)
> + atomic_inc(&dmi->data_read_count);
> + else
> + atomic_inc(&dmi->data_write_count);
> + }
> +
> + generic_make_request(bio);
> +}
> +
> +static struct dm_int_io *dm_int_io_alloc(struct dm_int *dmi,
> + struct bio *bio, sector_t sector)
> +{
> + struct dm_int_io *io;
> +
> + /* never fails with GFP_NOIO */
> + io = mempool_alloc(dmi->io_pool, GFP_NOIO);
> + BUG_ON(!io);
> +
> + io->dmi = dmi;
> + io->bio = bio;
> + io->bi_size = bio->bi_size;
> + io->sector = sector;
> + io->error = 0;
> + io->flags = 0;
> +
> + INIT_WORK(&io->work, dm_int_hmac_task);
> +
> + ahash_request_set_tfm(&io->req, dmi->ahash);
> +
> + atomic_set(&io->count, 1);
> + atomic_inc(&dmi->count);
> +
> + /* stats */
> + if (dmi->flags & DM_INT_FLAGS_STATS) {
> + atomic_inc(&dmi->io_count);
> + if (atomic_read(&dmi->io_count) > dmi->io_count_max)
> + dmi->io_count_max = atomic_read(&dmi->io_count);
> + }
> +
> + return io;
> +}
> +
> +static int dm_int_map(struct dm_target *ti, struct bio *bio,
> + union map_info *map_context)
> +{
> + struct dm_int *dmi = ti->private;
> + struct dm_int_io *io;
> +
> + /*
> + * If bio is REQ_FLUSH or REQ_DISCARD, just bypass crypt queues.
> + * - for REQ_FLUSH device-mapper core ensures that no IO is in-flight
> + * - for REQ_DISCARD caller must use flush if IO ordering matters
> + */
> + if (unlikely(bio->bi_rw & (REQ_FLUSH | REQ_DISCARD))) {
> + bio->bi_bdev = dmi->dev->bdev;
> + bio->bi_sector =
> + dmi->start + dm_target_offset(ti, bio->bi_sector);
> + return DM_MAPIO_REMAPPED;
> + }
> +
> + /* a check to see if something unhandled might come */
> + if (!bio->bi_size || !bio->bi_vcnt)
> + DMERR("bio without data: size: %d, vcnt: %d\n",
> + bio->bi_size, bio->bi_vcnt);
> +
> + BUG_ON(bio->bi_sector & (to_sector(dmi->data_block_size) - 1));
> + BUG_ON(bio->bi_size & (dmi->data_block_size - 1));
> +
> + io = dm_int_io_alloc(dmi, bio, dm_target_offset(ti, bio->bi_sector));
> +
> + pr_debug("io: %p, sector: %llu, size: %u, vcnt: %d, idx: %d\n",
> + io, (loff_t)bio->bi_sector,
> + bio->bi_size, bio->bi_vcnt, bio->bi_idx);
> +
> + dm_int_start_io(io);
> + dm_int_queue_hmac(io);
> +
> + dm_int_io_put(io);
> +
> + return DM_MAPIO_SUBMITTED;
> +}
> +
> +static void dm_int_cleanup(struct dm_target *ti)
> +{
> + struct dm_int *dmi = (struct dm_int *)ti->private;
> +
> + if (dmi->bufio)
> + dm_bufio_client_destroy(dmi->bufio);
> + if (dmi->io_queue)
> + destroy_workqueue(dmi->io_queue);
> + if (dmi->io_pool)
> + mempool_destroy(dmi->io_pool);
> + if (dmi->io_cache)
> + kmem_cache_destroy(dmi->io_cache);
> + if (dmi->ahash)
> + crypto_free_ahash(dmi->ahash);
> + if (dmi->hmac)
> + crypto_free_shash(dmi->hmac);
> + if (dmi->hdev)
> + dm_put_device(ti, dmi->hdev);
> + if (dmi->dev)
> + dm_put_device(ti, dmi->dev);
> + kfree(dmi->table_string);
> + kfree(dmi);
> +}
> +
> +/*
> + * Construct an integrity mapping:
> + * <dev> <bs> <start> <hdev> <hbs> <hstart> <hash_algo> <hmac_algo> <keyname> \
> + * [opt_params]
> + */
> +static int dm_int_ctr(struct dm_target *ti, unsigned int argc, char **argv)
> +{
> + struct dm_int *dmi;
> + int err, io_size, i, count;
> + unsigned long long tmpll;
> + char table[256], dummy;
> + unsigned tmp;
> + fmode_t mode;
> + loff_t datadevsize, hmacdevsize, maxdatasize, maxhmacsize;
> +
> + if (argc < 9) {
> + ti->error = "Invalid argument count";
> + return -EINVAL;
> + }
> +
> + dmi = kzalloc(sizeof(*dmi), GFP_KERNEL);
> + if (dmi == NULL) {
> + ti->error = "dm-integrity: Cannot allocate linear context";
> + return -ENOMEM;
> + }
> +
> + dmi->target = ti;
> + ti->private = dmi;
> +
> + err = -EINVAL;
> +
> + mode = dm_table_get_mode(ti->table);
> +
> + if (dm_get_device(ti, argv[0], mode, &dmi->dev)) {
> + ti->error = "Device lookup failed";
> + goto err;
> + }
> +
> + if (sscanf(argv[1], "%u%c", &tmp, &dummy) != 1 ||
> + !is_power_of_2(tmp) ||
> + tmp < bdev_logical_block_size(dmi->dev->bdev) ||
> + tmp > PAGE_SIZE) {
> + ti->error = "Invalid device block size";
> + goto err;
> + }
> + dmi->data_block_size = tmp;
> + dmi->data_block_bits = ffs(dmi->data_block_size) - 1;
> +
> + if (sscanf(argv[2], "%llu%c", &tmpll, &dummy) != 1) {
> + ti->error = "Invalid device start";
> + goto err;
> + }
> + dmi->start = tmpll;
> +
> + if (dm_get_device(ti, argv[3], mode, &dmi->hdev)) {
> + ti->error = "HMAC device lookup failed";
> + goto err;
> + }
> +
> + if (sscanf(argv[4], "%u%c", &tmp, &dummy) != 1 ||
> + !is_power_of_2(tmp) ||
> + tmp < bdev_logical_block_size(dmi->dev->bdev) ||
> + tmp > PAGE_SIZE) {
> + ti->error = "Invalid device block size";
> + goto err;
> + }
> + dmi->hmac_block_size = tmp;
> + dmi->hmac_block_bits = ffs(dmi->hmac_block_size) - 1;
> +
> + if (sscanf(argv[5], "%llu%c", &tmpll, &dummy) != 1) {
> + ti->error = "Invalid hmac device start";
> + goto err;
> + }
> + dmi->hmac_start = tmpll;
> +
> + err = dm_int_init_crypto(dmi, argv[6], argv[7], argv[8]);
> + if (err)
> + goto err;
> +
> + count = snprintf(table, sizeof(table), "%s %u %llu %s %u %llu %s %s %s",
> + dmi->dev->name, dmi->data_block_size, dmi->start,
> + dmi->hdev->name, dmi->hmac_block_size, dmi->hmac_start,
> + argv[6], argv[7], argv[8]);
> +
> + for (i = 9; i < argc; i++) {
> + count += snprintf(table + count, sizeof(table) - count,
> + " %s", argv[i]);
> + }
> +
> + dmi->table_string = kstrdup(table, GFP_KERNEL);
> +
> + dmi->hmac_size = crypto_shash_digestsize(dmi->hmac);
> +
> + /* how many hmacs do we need for data device */
> + dmi->hmac_count = ti->len >> (dmi->data_block_bits - SECTOR_SHIFT);
> +
> + datadevsize = i_size_read(dmi->dev->bdev->bd_inode) >> SECTOR_SHIFT;
> + hmacdevsize = i_size_read(dmi->hdev->bdev->bd_inode) >> SECTOR_SHIFT;
> +
> + err = -EINVAL;
> +
> + if (dmi->start > datadevsize) {
> + DMERR("start sector is beyond device size: %llu (%llu)\n",
> + dmi->start, datadevsize);
> + goto err;
> + }
> +
> + if (dmi->hmac_start > hmacdevsize) {
> + DMERR("start sector is beyond device size: %llu (%llu)\n",
> + dmi->hmac_start, hmacdevsize);
> + goto err;
> + }
> +
> + if (dmi->dev->bdev == dmi->hdev->bdev) {
> + if (dmi->hmac_start > dmi->start) {
> + maxdatasize = dmi->hmac_start - dmi->start;
> + maxhmacsize = datadevsize - dmi->hmac_start;
> + } else {
> + maxhmacsize = dmi->start - dmi->hmac_start;
> + maxdatasize = datadevsize - dmi->start;
> + }
> + } else {
> + maxdatasize = datadevsize - dmi->start;
> + maxhmacsize = hmacdevsize - dmi->hmac_start;
> + }
> +
> + if (ti->len > maxdatasize) {
> + DMERR("target size is too big: %llu (%llu)\n",
> + (loff_t)ti->len, maxdatasize);
> + goto err;
> + }
> +
> + /* hmac start in blocks */
> + dmi->hmac_start >>= (dmi->hmac_block_bits - SECTOR_SHIFT);
> +
> + /* optimize for SHA256 which is 32 bytes */
> + if (is_power_of_2(dmi->hmac_size)) {
> + dmi->hmac_block_shift =
> + dmi->hmac_block_bits - (ffs(dmi->hmac_size) - 1);
> + /* how many hmac blocks do we need */
> + dmi->hmac_count >>= dmi->hmac_block_shift;
> + } else {
> + dmi->hmac_per_block = dmi->hmac_block_size / dmi->hmac_size;
> + /* how many hmac blocks do we need */
> + tmpll = sector_div(dmi->hmac_count, dmi->hmac_per_block);
> + if (tmpll)
> + dmi->hmac_count++;
> + }
> +
> + /* device may hold as many hmac blocks */
> + maxhmacsize >>= (dmi->hmac_block_bits - SECTOR_SHIFT);
> +
> + if (dmi->hmac_count > maxhmacsize) {
> + DMERR("HMAC device is too small: %llu (%llu)\n",
> + dmi->hmac_count, maxhmacsize);
> + goto err;
> + }
> +
> + ti->num_discard_requests = 1;
> +
> + for (i = 9; i < argc; i++) {
> + if (!strcmp(argv[i], "fix"))
> + dmi->flags |= DM_INT_FLAGS_FIX;
> + else if (!strcmp(argv[i], "stats"))
> + dmi->flags |= DM_INT_FLAGS_STATS;
> + else if (!strcmp(argv[i], "verbose"))
> + dmi->flags |= DM_INT_FLAGS_VERBOSE;
> + else if (!strcmp(argv[i], "disallow_discards"))
> + ti->num_discard_requests = 0;
> + }
> +
> + err = -ENOMEM;
> +
> + io_size = sizeof(struct dm_int_io);
> + io_size += crypto_ahash_reqsize(dmi->ahash);
> + dmi->io_cache = kmem_cache_create("dm_int_io_cache", io_size,
> + __alignof__(struct dm_int_io), 0,
> + NULL);
> + if (!dmi->io_cache) {
> + ti->error = "Cannot allocate dm_int io cache";
> + goto err;
> + }
> +
> + dmi->io_pool = mempool_create_slab_pool(DM_INT_MIN_IOS, dmi->io_cache);
> + if (!dmi->io_pool) {
> + ti->error = "Cannot allocate dm_int io mempool";
> + goto err;
> + }
> +
> + dmi->io_queue = alloc_workqueue("dm_int_hmac",
> + WQ_CPU_INTENSIVE |
> + WQ_HIGHPRI |
> + WQ_UNBOUND |
> + WQ_MEM_RECLAIM,
> + 1);
> + if (!dmi->io_queue) {
> + ti->error = "Couldn't create dm_int hmac queue";
> + goto err;
> + }
> +
> + dmi->bufio = dm_bufio_client_create(dmi->hdev->bdev,
> + dmi->hmac_block_size, 1, 0,
> + NULL, NULL);
> + if (IS_ERR(dmi->bufio)) {
> + ti->error = "Cannot initialize dm-bufio";
> + err = PTR_ERR(xchg(&dmi->bufio, NULL));
> + goto err;
> + }
> +
> + mutex_init(&dmi->mutex);
> + dmi->delay = DM_INT_FLUSH_DELAY;
> + init_waitqueue_head(&dmi->wait);
> +
> + ti->num_flush_requests = 1;
> + /* it should depend on read block device... */
> + /*ti->discard_zeroes_data_unsupported = true;*/
> +
> + mutex_lock(&mutex);
> + list_add(&dmi->list, &dmi_list);
> + mutex_unlock(&mutex);
> +
> + return 0;
> +
> +err:
> + dm_int_cleanup(ti);
> + return err;
> +}
> +
> +static void dm_int_dtr(struct dm_target *ti)
> +{
> + struct dm_int *dmi = (struct dm_int *)ti->private;
> +
> + mutex_lock(&mutex);
> + list_del(&dmi->list);
> + mutex_unlock(&mutex);
> +
> + dm_int_cleanup(ti);
> +}
> +
> +static void dm_int_sync(struct dm_int *dmi)
> +{
> + /* first flush hmac queue, which might schedule idata delayed work */
> + flush_workqueue(dmi->io_queue);
> + /* write all updated hmac blocks */
> + dm_bufio_write_dirty_buffers(dmi->bufio);
> +
> + if (atomic_read(&dmi->count))
> + DMWARN("dmi->count: %d\n", atomic_read(&dmi->count));
> + /* wait until all idata bios complete */
> + wait_event(dmi->wait, !atomic_read(&dmi->count));
> +}
> +
> +static int dm_int_ioctl(struct dm_target *ti, unsigned int cmd,
> + unsigned long arg)
> +{
> + struct dm_int *dmi = (struct dm_int *)ti->private;
> + struct dm_dev *dev = dmi->dev;
> + int err = 0;
> +
> + if (cmd == BLKFLSBUF)
> + dm_int_sync(dmi);
> +
> + /*
> + * Only pass ioctls through if the device sizes match exactly.
> + */
> + if (dmi->start ||
> + ti->len != i_size_read(dev->bdev->bd_inode) >> SECTOR_SHIFT)
> + err = scsi_verify_blk_ioctl(NULL, cmd);
> +
> + return err ? : __blkdev_driver_ioctl(dev->bdev, dev->mode, cmd, arg);
> +}
> +
> +static int dm_int_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
> + struct bio_vec *biovec, int max_size)
> +{
> + struct dm_int *dmi = ti->private;
> + struct request_queue *q = bdev_get_queue(dmi->dev->bdev);
> +
> + if (!q->merge_bvec_fn)
> + return max_size;
> +
> + bvm->bi_bdev = dmi->dev->bdev;
> + bvm->bi_sector = dmi->start + dm_target_offset(ti, bvm->bi_sector);
> +
> + return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
> +}
> +
> +static int dm_int_iterate_devices(struct dm_target *ti,
> + iterate_devices_callout_fn fn, void *data)
> +{
> + struct dm_int *dmi = ti->private;
> +
> + return fn(ti, dmi->dev, dmi->start, ti->len, data);
> +}
> +
> +static void dm_int_io_hints(struct dm_target *ti, struct queue_limits *limits)
> +{
> + struct dm_int *dmi = ti->private;
> +
> + limits->logical_block_size = dmi->data_block_size;
> + limits->physical_block_size = dmi->data_block_size;
> + blk_limits_io_min(limits, dmi->data_block_size);
> +}
> +
> +static void dm_int_postsuspend(struct dm_target *ti)
> +{
> + struct dm_int *dmi = ti->private;
> +
> + dm_int_sync(dmi);
> +
> + DMINFO("%s suspended\n", dm_device_name(dm_table_get_md(ti->table)));
> +}
> +
> +static int dm_int_status(struct dm_target *ti, status_type_t type,
> + unsigned status_flags, char *result, unsigned maxlen)
> +{
> + struct dm_int *dmi = (struct dm_int *)ti->private;
> + unsigned int sz = 0;
> +
> + switch (type) {
> + case STATUSTYPE_INFO:
> +#ifdef DM_INT_STATS
> + DMEMIT("pending: %d, io: %d (%d), "\
> + "read: %d, write: %d, "\
> + "violations: %d",
> + atomic_read(&dmi->count),
> + atomic_read(&dmi->io_count), dmi->io_count_max,
> + atomic_read(&dmi->data_read_count),
> + atomic_read(&dmi->data_write_count),
> + atomic_read(&dmi->violations));
> +#else
> + DMEMIT("pending: %d, violations: %d",
> + atomic_read(&dmi->count),
> + atomic_read(&dmi->violations));
> +#endif
> + break;
> +
> + case STATUSTYPE_TABLE:
> + DMEMIT("%s", dmi->table_string);
> + break;
> + }
> + return 0;
> +}
> +
> +static struct target_type dm_int_target = {
> + .name = "integrity",
> + .version = {0, 1, 0},
> + .module = THIS_MODULE,
> + .ctr = dm_int_ctr,
> + .dtr = dm_int_dtr,
> + .map = dm_int_map,
> + .status = dm_int_status,
> + .ioctl = dm_int_ioctl,
> + .postsuspend = dm_int_postsuspend,
> + .merge = dm_int_merge,
> + .iterate_devices = dm_int_iterate_devices,
> + .io_hints = dm_int_io_hints,
> +};
> +
> +static int dm_int_notify_reboot(struct notifier_block *this,
> + unsigned long code, void *x)
> +{
> + struct dm_int *dmi;
> +
> + if ((code == SYS_DOWN) || (code == SYS_HALT) ||
> + (code == SYS_POWER_OFF)) {
> + sync_mode = 1;
> + mutex_lock(&mutex);
> + if (!list_empty(&dmi_list)) {
> + DMINFO("syncing targets...");
> + list_for_each_entry(dmi, &dmi_list, list)
> + dm_int_sync(dmi);
> + pr_cont(" done.\n");
> + }
> + mutex_unlock(&mutex);
> + }
> + return NOTIFY_DONE;
> +}
> +
> +static struct notifier_block dm_int_notifier = {
> + .notifier_call = dm_int_notify_reboot,
> + .next = NULL,
> + .priority = INT_MAX, /* before any real devices */
> +};
> +
> +int __init dm_int_init(void)
> +{
> + int err = -ENOMEM;
> +
> + err = dm_register_target(&dm_int_target);
> + if (err < 0) {
> + DMERR("register failed %d", err);
> + return err;
> + }
> +
> + /* always returns 0 */
> + register_reboot_notifier(&dm_int_notifier);
> +
> + return 0;
> +}
> +
> +void dm_int_exit(void)
> +{
> + unregister_reboot_notifier(&dm_int_notifier);
> + dm_unregister_target(&dm_int_target);
> +}
> +
> +/* Module hooks */
> +module_init(dm_int_init);
> +module_exit(dm_int_exit);
> +
> +MODULE_DESCRIPTION(DM_NAME " integrity target");
> +MODULE_AUTHOR("Dmitry Kasatkin");
> +MODULE_LICENSE("GPL");
> --
> 1.7.10.4
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists