/* $NetBSD: cgd.c,v 1.146 2022/04/02 09:53:20 riastradh Exp $ */

/*-
* Copyright (c) 2002 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
* by Roland C. Dowdeswell.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
*    notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
*    notice, this list of conditions and the following disclaimer in the
*    documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/

#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: cgd.c,v 1.146 2022/04/02 09:53:20 riastradh Exp $");

#include <sys/types.h>
#include <sys/param.h>
#include <sys/buf.h>
#include <sys/bufq.h>
#include <sys/conf.h>
#include <sys/cpu.h>
#include <sys/device.h>
#include <sys/disk.h>
#include <sys/disklabel.h>
#include <sys/errno.h>
#include <sys/fcntl.h>
#include <sys/ioctl.h>
#include <sys/kmem.h>
#include <sys/module.h>
#include <sys/namei.h> /* for pathbuf */
#include <sys/pool.h>
#include <sys/proc.h>
#include <sys/syslog.h>
#include <sys/systm.h>
#include <sys/vnode.h>
#include <sys/workqueue.h>

#include <dev/cgd_crypto.h>
#include <dev/cgdvar.h>
#include <dev/dkvar.h>

#include <miscfs/specfs/specdev.h> /* for v_rdev */

#include "ioconf.h"

struct selftest_params {
       const char *alg;
       int encblkno8;
       int blocksize;  /* number of bytes */
       int secsize;
       daddr_t blkno;
       int keylen;     /* number of bits */
       int txtlen;     /* number of bytes */
       const uint8_t *key;
       const uint8_t *ptxt;
       const uint8_t *ctxt;
};

/* Entry Point Functions */

static dev_type_open(cgdopen);
static dev_type_close(cgdclose);
static dev_type_read(cgdread);
static dev_type_write(cgdwrite);
static dev_type_ioctl(cgdioctl);
static dev_type_strategy(cgdstrategy);
static dev_type_dump(cgddump);
static dev_type_size(cgdsize);

const struct bdevsw cgd_bdevsw = {
       .d_open = cgdopen,
       .d_close = cgdclose,
       .d_strategy = cgdstrategy,
       .d_ioctl = cgdioctl,
       .d_dump = cgddump,
       .d_psize = cgdsize,
       .d_discard = nodiscard,
       .d_flag = D_DISK | D_MPSAFE
};

const struct cdevsw cgd_cdevsw = {
       .d_open = cgdopen,
       .d_close = cgdclose,
       .d_read = cgdread,
       .d_write = cgdwrite,
       .d_ioctl = cgdioctl,
       .d_stop = nostop,
       .d_tty = notty,
       .d_poll = nopoll,
       .d_mmap = nommap,
       .d_kqfilter = nokqfilter,
       .d_discard = nodiscard,
       .d_flag = D_DISK | D_MPSAFE
};

/*
* Vector 5 from IEEE 1619/D16 truncated to 64 bytes, blkno 1.
*/
static const uint8_t selftest_aes_xts_256_ptxt[64] = {
       0x27, 0xa7, 0x47, 0x9b, 0xef, 0xa1, 0xd4, 0x76,
       0x48, 0x9f, 0x30, 0x8c, 0xd4, 0xcf, 0xa6, 0xe2,
       0xa9, 0x6e, 0x4b, 0xbe, 0x32, 0x08, 0xff, 0x25,
       0x28, 0x7d, 0xd3, 0x81, 0x96, 0x16, 0xe8, 0x9c,
       0xc7, 0x8c, 0xf7, 0xf5, 0xe5, 0x43, 0x44, 0x5f,
       0x83, 0x33, 0xd8, 0xfa, 0x7f, 0x56, 0x00, 0x00,
       0x05, 0x27, 0x9f, 0xa5, 0xd8, 0xb5, 0xe4, 0xad,
       0x40, 0xe7, 0x36, 0xdd, 0xb4, 0xd3, 0x54, 0x12,
};

static const uint8_t selftest_aes_xts_256_ctxt[512] = {
       0x26, 0x4d, 0x3c, 0xa8, 0x51, 0x21, 0x94, 0xfe,
       0xc3, 0x12, 0xc8, 0xc9, 0x89, 0x1f, 0x27, 0x9f,
       0xef, 0xdd, 0x60, 0x8d, 0x0c, 0x02, 0x7b, 0x60,
       0x48, 0x3a, 0x3f, 0xa8, 0x11, 0xd6, 0x5e, 0xe5,
       0x9d, 0x52, 0xd9, 0xe4, 0x0e, 0xc5, 0x67, 0x2d,
       0x81, 0x53, 0x2b, 0x38, 0xb6, 0xb0, 0x89, 0xce,
       0x95, 0x1f, 0x0f, 0x9c, 0x35, 0x59, 0x0b, 0x8b,
       0x97, 0x8d, 0x17, 0x52, 0x13, 0xf3, 0x29, 0xbb,
};

static const uint8_t selftest_aes_xts_256_key[33] = {
       0x27, 0x18, 0x28, 0x18, 0x28, 0x45, 0x90, 0x45,
       0x23, 0x53, 0x60, 0x28, 0x74, 0x71, 0x35, 0x26,
       0x31, 0x41, 0x59, 0x26, 0x53, 0x58, 0x97, 0x93,
       0x23, 0x84, 0x62, 0x64, 0x33, 0x83, 0x27, 0x95,
       0
};

/*
* Vector 11 from IEEE 1619/D16 truncated to 64 bytes, blkno 0xffff.
*/
static const uint8_t selftest_aes_xts_512_ptxt[64] = {
       0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
       0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
       0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
       0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
       0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27,
       0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f,
       0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37,
       0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f,
};

static const uint8_t selftest_aes_xts_512_ctxt[64] = {
       0x77, 0xa3, 0x12, 0x51, 0x61, 0x8a, 0x15, 0xe6,
       0xb9, 0x2d, 0x1d, 0x66, 0xdf, 0xfe, 0x7b, 0x50,
       0xb5, 0x0b, 0xad, 0x55, 0x23, 0x05, 0xba, 0x02,
       0x17, 0xa6, 0x10, 0x68, 0x8e, 0xff, 0x7e, 0x11,
       0xe1, 0xd0, 0x22, 0x54, 0x38, 0xe0, 0x93, 0x24,
       0x2d, 0x6d, 0xb2, 0x74, 0xfd, 0xe8, 0x01, 0xd4,
       0xca, 0xe0, 0x6f, 0x20, 0x92, 0xc7, 0x28, 0xb2,
       0x47, 0x85, 0x59, 0xdf, 0x58, 0xe8, 0x37, 0xc2,
};

static const uint8_t selftest_aes_xts_512_key[65] = {
       0x27, 0x18, 0x28, 0x18, 0x28, 0x45, 0x90, 0x45,
       0x23, 0x53, 0x60, 0x28, 0x74, 0x71, 0x35, 0x26,
       0x62, 0x49, 0x77, 0x57, 0x24, 0x70, 0x93, 0x69,
       0x99, 0x59, 0x57, 0x49, 0x66, 0x96, 0x76, 0x27,
       0x31, 0x41, 0x59, 0x26, 0x53, 0x58, 0x97, 0x93,
       0x23, 0x84, 0x62, 0x64, 0x33, 0x83, 0x27, 0x95,
       0x02, 0x88, 0x41, 0x97, 0x16, 0x93, 0x99, 0x37,
       0x51, 0x05, 0x82, 0x09, 0x74, 0x94, 0x45, 0x92,
       0
};

static const uint8_t selftest_aes_cbc_key[32] = {
       0x27, 0x18, 0x28, 0x18, 0x28, 0x45, 0x90, 0x45,
       0x23, 0x53, 0x60, 0x28, 0x74, 0x71, 0x35, 0x26,
       0x62, 0x49, 0x77, 0x57, 0x24, 0x70, 0x93, 0x69,
       0x99, 0x59, 0x57, 0x49, 0x66, 0x96, 0x76, 0x27,
};

static const uint8_t selftest_aes_cbc_128_ptxt[64] = {
       0x27, 0xa7, 0x47, 0x9b, 0xef, 0xa1, 0xd4, 0x76,
       0x48, 0x9f, 0x30, 0x8c, 0xd4, 0xcf, 0xa6, 0xe2,
       0xa9, 0x6e, 0x4b, 0xbe, 0x32, 0x08, 0xff, 0x25,
       0x28, 0x7d, 0xd3, 0x81, 0x96, 0x16, 0xe8, 0x9c,
       0xc7, 0x8c, 0xf7, 0xf5, 0xe5, 0x43, 0x44, 0x5f,
       0x83, 0x33, 0xd8, 0xfa, 0x7f, 0x56, 0x00, 0x00,
       0x05, 0x27, 0x9f, 0xa5, 0xd8, 0xb5, 0xe4, 0xad,
       0x40, 0xe7, 0x36, 0xdd, 0xb4, 0xd3, 0x54, 0x12,
};

static const uint8_t selftest_aes_cbc_128_ctxt[64] = { /* blkno=1 */
       0x93, 0x94, 0x56, 0x36, 0x83, 0xbc, 0xff, 0xa4,
       0xe0, 0x24, 0x34, 0x12, 0xbe, 0xfa, 0xb0, 0x7d,
       0x88, 0x1e, 0xc5, 0x57, 0x55, 0x23, 0x05, 0x0c,
       0x69, 0xa5, 0xc1, 0xda, 0x64, 0xee, 0x74, 0x10,
       0xc2, 0xc5, 0xe6, 0x66, 0xd6, 0xa7, 0x49, 0x1c,
       0x9d, 0x40, 0xb5, 0x0c, 0x9b, 0x6e, 0x1c, 0xe6,
       0xb1, 0x7a, 0x1c, 0xe7, 0x5a, 0xfe, 0xf9, 0x2a,
       0x78, 0xfa, 0xb7, 0x7b, 0x08, 0xdf, 0x8e, 0x51,
};

static const uint8_t selftest_aes_cbc_256_ptxt[64] = {
       0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
       0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
       0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
       0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
       0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27,
       0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f,
       0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37,
       0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f,
};

static const uint8_t selftest_aes_cbc_256_ctxt[64] = { /* blkno=0xffff */
       0x6c, 0xa3, 0x15, 0x17, 0x51, 0x90, 0xe9, 0x69,
       0x08, 0x36, 0x7b, 0xa6, 0xbb, 0xd1, 0x0b, 0x9e,
       0xcd, 0x6b, 0x1e, 0xaf, 0xb6, 0x2e, 0x62, 0x7d,
       0x8e, 0xde, 0xf0, 0xed, 0x0d, 0x44, 0xe7, 0x31,
       0x26, 0xcf, 0xd5, 0x0b, 0x3e, 0x95, 0x59, 0x89,
       0xdf, 0x5d, 0xd6, 0x9a, 0x00, 0x66, 0xcc, 0x7f,
       0x45, 0xd3, 0x06, 0x58, 0xed, 0xef, 0x49, 0x47,
       0x87, 0x89, 0x17, 0x7d, 0x08, 0x56, 0x50, 0xe1,
};

static const uint8_t selftest_3des_cbc_key[24] = {
       0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
       0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
       0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
};

static const uint8_t selftest_3des_cbc_ptxt[64] = {
       0x27, 0xa7, 0x47, 0x9b, 0xef, 0xa1, 0xd4, 0x76,
       0x48, 0x9f, 0x30, 0x8c, 0xd4, 0xcf, 0xa6, 0xe2,
       0xa9, 0x6e, 0x4b, 0xbe, 0x32, 0x08, 0xff, 0x25,
       0x28, 0x7d, 0xd3, 0x81, 0x96, 0x16, 0xe8, 0x9c,
       0xc7, 0x8c, 0xf7, 0xf5, 0xe5, 0x43, 0x44, 0x5f,
       0x83, 0x33, 0xd8, 0xfa, 0x7f, 0x56, 0x00, 0x00,
       0x05, 0x27, 0x9f, 0xa5, 0xd8, 0xb5, 0xe4, 0xad,
       0x40, 0xe7, 0x36, 0xdd, 0xb4, 0xd3, 0x54, 0x12,
};

static const uint8_t selftest_3des_cbc_ctxt[64] = {
       0xa2, 0xfe, 0x81, 0xaa, 0x10, 0x6c, 0xea, 0xb9,
       0x11, 0x58, 0x1f, 0x29, 0xb5, 0x86, 0x71, 0x56,
       0xe9, 0x25, 0x1d, 0x07, 0xb1, 0x69, 0x59, 0x6c,
       0x96, 0x80, 0xf7, 0x54, 0x38, 0xaa, 0xa7, 0xe4,
       0xe8, 0x81, 0xf5, 0x00, 0xbb, 0x1c, 0x00, 0x3c,
       0xba, 0x38, 0x45, 0x97, 0x4c, 0xcf, 0x84, 0x14,
       0x46, 0x86, 0xd9, 0xf4, 0xc5, 0xe2, 0xf0, 0x54,
       0xde, 0x41, 0xf6, 0xa1, 0xef, 0x1b, 0x0a, 0xea,
};

static const uint8_t selftest_bf_cbc_key[56] = {
       0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
       0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
       0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
       0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f,
       0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27,
       0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f,
       0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37,
};

static const uint8_t selftest_bf_cbc_ptxt[64] = {
       0x27, 0xa7, 0x47, 0x9b, 0xef, 0xa1, 0xd4, 0x76,
       0x48, 0x9f, 0x30, 0x8c, 0xd4, 0xcf, 0xa6, 0xe2,
       0xa9, 0x6e, 0x4b, 0xbe, 0x32, 0x08, 0xff, 0x25,
       0x28, 0x7d, 0xd3, 0x81, 0x96, 0x16, 0xe8, 0x9c,
       0xc7, 0x8c, 0xf7, 0xf5, 0xe5, 0x43, 0x44, 0x5f,
       0x83, 0x33, 0xd8, 0xfa, 0x7f, 0x56, 0x00, 0x00,
       0x05, 0x27, 0x9f, 0xa5, 0xd8, 0xb5, 0xe4, 0xad,
       0x40, 0xe7, 0x36, 0xdd, 0xb4, 0xd3, 0x54, 0x12,
};

static const uint8_t selftest_bf_cbc_ctxt[64] = {
       0xec, 0xa2, 0xc0, 0x0e, 0xa9, 0x7f, 0x04, 0x1e,
       0x2e, 0x4f, 0x64, 0x07, 0x67, 0x3e, 0xf4, 0x58,
       0x61, 0x5f, 0xd3, 0x50, 0x5e, 0xd3, 0x4d, 0x34,
       0xa0, 0x53, 0xbe, 0x47, 0x75, 0x69, 0x3b, 0x1f,
       0x86, 0xf2, 0xae, 0x8b, 0xb7, 0x91, 0xda, 0xd4,
       0x2b, 0xa5, 0x47, 0x9b, 0x7d, 0x13, 0x30, 0xdd,
       0x7b, 0xad, 0x86, 0x57, 0x51, 0x11, 0x74, 0x42,
       0xb8, 0xbf, 0x69, 0x17, 0x20, 0x0a, 0xf7, 0xda,
};

static const uint8_t selftest_aes_cbc_encblkno8_zero64[64];
static const uint8_t selftest_aes_cbc_encblkno8_ctxt[64] = {
       0xa2, 0x06, 0x26, 0x26, 0xac, 0xdc, 0xe7, 0xcf,
       0x47, 0x68, 0x24, 0x0e, 0xfa, 0x40, 0x44, 0x83,
       0x07, 0xe1, 0xf4, 0x5d, 0x53, 0x47, 0xa0, 0xfe,
       0xc0, 0x6e, 0x4e, 0xf8, 0x9d, 0x98, 0x63, 0xb8,
       0x2c, 0x27, 0xfa, 0x3a, 0xd5, 0x40, 0xda, 0xdb,
       0xe6, 0xc3, 0xe4, 0xfb, 0x85, 0x53, 0xfb, 0x78,
       0x5d, 0xbd, 0x8f, 0x4c, 0x1a, 0x04, 0x9c, 0x88,
       0x85, 0xec, 0x3c, 0x56, 0x46, 0x1a, 0x6e, 0xf5,
};

const struct selftest_params selftests[] = {
       {
               .alg = "aes-xts",
               .blocksize = 16,
               .secsize = 512,
               .blkno = 1,
               .keylen = 256,
               .txtlen = sizeof(selftest_aes_xts_256_ptxt),
               .key  = selftest_aes_xts_256_key,
               .ptxt = selftest_aes_xts_256_ptxt,
               .ctxt = selftest_aes_xts_256_ctxt
       },
       {
               .alg = "aes-xts",
               .blocksize = 16,
               .secsize = 512,
               .blkno = 0xffff,
               .keylen = 512,
               .txtlen = sizeof(selftest_aes_xts_512_ptxt),
               .key  = selftest_aes_xts_512_key,
               .ptxt = selftest_aes_xts_512_ptxt,
               .ctxt = selftest_aes_xts_512_ctxt
       },
       {
               .alg = "aes-cbc",
               .blocksize = 16,
               .secsize = 512,
               .blkno = 1,
               .keylen = 128,
               .txtlen = sizeof(selftest_aes_cbc_128_ptxt),
               .key  = selftest_aes_cbc_key,
               .ptxt = selftest_aes_cbc_128_ptxt,
               .ctxt = selftest_aes_cbc_128_ctxt,
       },
       {
               .alg = "aes-cbc",
               .blocksize = 16,
               .secsize = 512,
               .blkno = 0xffff,
               .keylen = 256,
               .txtlen = sizeof(selftest_aes_cbc_256_ptxt),
               .key  = selftest_aes_cbc_key,
               .ptxt = selftest_aes_cbc_256_ptxt,
               .ctxt = selftest_aes_cbc_256_ctxt,
       },
       {
               .alg = "3des-cbc",
               .blocksize = 8,
               .secsize = 512,
               .blkno = 1,
               .keylen = 192,  /* 168 + 3*8 parity bits */
               .txtlen = sizeof(selftest_3des_cbc_ptxt),
               .key  = selftest_3des_cbc_key,
               .ptxt = selftest_3des_cbc_ptxt,
               .ctxt = selftest_3des_cbc_ctxt,
       },
       {
               .alg = "blowfish-cbc",
               .blocksize = 8,
               .secsize = 512,
               .blkno = 1,
               .keylen = 448,
               .txtlen = sizeof(selftest_bf_cbc_ptxt),
               .key  = selftest_bf_cbc_key,
               .ptxt = selftest_bf_cbc_ptxt,
               .ctxt = selftest_bf_cbc_ctxt,
       },
       {
               .alg = "aes-cbc",
               .encblkno8 = 1,
               .blocksize = 16,
               .secsize = 512,
               .blkno = 0,
               .keylen = 128,
               .txtlen = sizeof(selftest_aes_cbc_encblkno8_zero64),
               .key = selftest_aes_cbc_encblkno8_zero64,
               .ptxt = selftest_aes_cbc_encblkno8_zero64,
               .ctxt = selftest_aes_cbc_encblkno8_ctxt,
       },
};

static int cgd_match(device_t, cfdata_t, void *);
static void cgd_attach(device_t, device_t, void *);
static int cgd_detach(device_t, int);
static struct cgd_softc *cgd_spawn(int);
static struct cgd_worker *cgd_create_one_worker(void);
static void cgd_destroy_one_worker(struct cgd_worker *);
static struct cgd_worker *cgd_create_worker(void);
static void cgd_destroy_worker(struct cgd_worker *);
static int cgd_destroy(device_t);

/* Internal Functions */

static int      cgd_diskstart(device_t, struct buf *);
static void     cgd_diskstart2(struct cgd_softc *, struct cgd_xfer *);
static void     cgdiodone(struct buf *);
static void     cgd_iodone2(struct cgd_softc *, struct cgd_xfer *);
static void     cgd_enqueue(struct cgd_softc *, struct cgd_xfer *);
static void     cgd_process(struct work *, void *);
static int      cgd_dumpblocks(device_t, void *, daddr_t, int);

static int      cgd_ioctl_set(struct cgd_softc *, void *, struct lwp *);
static int      cgd_ioctl_clr(struct cgd_softc *, struct lwp *);
static int      cgd_ioctl_get(dev_t, void *, struct lwp *);
static int      cgdinit(struct cgd_softc *, const char *, struct vnode *,
                       struct lwp *);
static void     cgd_cipher(struct cgd_softc *, void *, const void *,
                          size_t, daddr_t, size_t, int);

static void     cgd_selftest(void);

static const struct dkdriver cgddkdriver = {
       .d_minphys  = minphys,
       .d_open = cgdopen,
       .d_close = cgdclose,
       .d_strategy = cgdstrategy,
       .d_iosize = NULL,
       .d_diskstart = cgd_diskstart,
       .d_dumpblocks = cgd_dumpblocks,
       .d_lastclose = NULL
};

CFATTACH_DECL3_NEW(cgd, sizeof(struct cgd_softc),
   cgd_match, cgd_attach, cgd_detach, NULL, NULL, NULL, DVF_DETACH_SHUTDOWN);

/* DIAGNOSTIC and DEBUG definitions */

#if defined(CGDDEBUG) && !defined(DEBUG)
#define DEBUG
#endif

#ifdef DEBUG
int cgddebug = 0;

#define CGDB_FOLLOW     0x1
#define CGDB_IO 0x2
#define CGDB_CRYPTO     0x4

#define IFDEBUG(x,y)            if (cgddebug & (x)) y
#define DPRINTF(x,y)            IFDEBUG(x, printf y)
#define DPRINTF_FOLLOW(y)       DPRINTF(CGDB_FOLLOW, y)

static void     hexprint(const char *, void *, int);

#else
#define IFDEBUG(x,y)
#define DPRINTF(x,y)
#define DPRINTF_FOLLOW(y)
#endif

/* Global variables */

static kmutex_t cgd_spawning_mtx;
static kcondvar_t cgd_spawning_cv;
static bool cgd_spawning;
static struct cgd_worker *cgd_worker;
static u_int cgd_refcnt;        /* number of users of cgd_worker */

/* Utility Functions */

#define CGDUNIT(x)              DISKUNIT(x)

/* The code */

static int
cgd_lock(bool intr)
{
       int error = 0;

       mutex_enter(&cgd_spawning_mtx);
       while (cgd_spawning) {
               if (intr)
                       error = cv_wait_sig(&cgd_spawning_cv, &cgd_spawning_mtx);
               else
                       cv_wait(&cgd_spawning_cv, &cgd_spawning_mtx);
       }
       if (error == 0)
               cgd_spawning = true;
       mutex_exit(&cgd_spawning_mtx);
       return error;
}

static void
cgd_unlock(void)
{
       mutex_enter(&cgd_spawning_mtx);
       cgd_spawning = false;
       cv_broadcast(&cgd_spawning_cv);
       mutex_exit(&cgd_spawning_mtx);
}

static struct cgd_softc *
getcgd_softc(dev_t dev)
{
       return device_lookup_private(&cgd_cd, CGDUNIT(dev));
}

static int
cgd_match(device_t self, cfdata_t cfdata, void *aux)
{

       return 1;
}

static void
cgd_attach(device_t parent, device_t self, void *aux)
{
       struct cgd_softc *sc = device_private(self);

       mutex_init(&sc->sc_lock, MUTEX_DEFAULT, IPL_BIO);
       cv_init(&sc->sc_cv, "cgdcv");
       dk_init(&sc->sc_dksc, self, DKTYPE_CGD);
       disk_init(&sc->sc_dksc.sc_dkdev, sc->sc_dksc.sc_xname, &cgddkdriver);

       if (!pmf_device_register(self, NULL, NULL))
               aprint_error_dev(self,
                   "unable to register power management hooks\n");
}


static int
cgd_detach(device_t self, int flags)
{
       int ret;
       struct cgd_softc *sc = device_private(self);
       struct dk_softc *dksc = &sc->sc_dksc;

       if (DK_BUSY(dksc, 0))
               return EBUSY;

       if (DK_ATTACHED(dksc) &&
           (ret = cgd_ioctl_clr(sc, curlwp)) != 0)
               return ret;

       disk_destroy(&dksc->sc_dkdev);
       cv_destroy(&sc->sc_cv);
       mutex_destroy(&sc->sc_lock);

       return 0;
}

void
cgdattach(int num)
{
#ifndef _MODULE
       int error;

       mutex_init(&cgd_spawning_mtx, MUTEX_DEFAULT, IPL_NONE);
       cv_init(&cgd_spawning_cv, "cgspwn");

       error = config_cfattach_attach(cgd_cd.cd_name, &cgd_ca);
       if (error != 0)
               aprint_error("%s: unable to register cfattach\n",
                   cgd_cd.cd_name);
#endif

       cgd_selftest();
}

static struct cgd_softc *
cgd_spawn(int unit)
{
       cfdata_t cf;
       struct cgd_worker *cw;
       struct cgd_softc *sc;

       cf = kmem_alloc(sizeof(*cf), KM_SLEEP);
       cf->cf_name = cgd_cd.cd_name;
       cf->cf_atname = cgd_cd.cd_name;
       cf->cf_unit = unit;
       cf->cf_fstate = FSTATE_STAR;

       cw = cgd_create_one_worker();
       if (cw == NULL) {
               kmem_free(cf, sizeof(*cf));
               return NULL;
       }

       sc = device_private(config_attach_pseudo(cf));
       if (sc == NULL) {
               cgd_destroy_one_worker(cw);
               return NULL;
       }

       sc->sc_worker = cw;

       return sc;
}

static int
cgd_destroy(device_t dev)
{
       struct cgd_softc *sc = device_private(dev);
       struct cgd_worker *cw = sc->sc_worker;
       cfdata_t cf;
       int error;

       cf = device_cfdata(dev);
       error = config_detach(dev, DETACH_QUIET);
       if (error)
               return error;

       cgd_destroy_one_worker(cw);

       kmem_free(cf, sizeof(*cf));
       return 0;
}

static void
cgd_busy(struct cgd_softc *sc)
{

       mutex_enter(&sc->sc_lock);
       while (sc->sc_busy)
               cv_wait(&sc->sc_cv, &sc->sc_lock);
       sc->sc_busy = true;
       mutex_exit(&sc->sc_lock);
}

static void
cgd_unbusy(struct cgd_softc *sc)
{

       mutex_enter(&sc->sc_lock);
       sc->sc_busy = false;
       cv_broadcast(&sc->sc_cv);
       mutex_exit(&sc->sc_lock);
}

static struct cgd_worker *
cgd_create_one_worker(void)
{
       KASSERT(cgd_spawning);

       if (cgd_refcnt++ == 0) {
               KASSERT(cgd_worker == NULL);
               cgd_worker = cgd_create_worker();
       }

       KASSERT(cgd_worker != NULL);
       return cgd_worker;
}

static void
cgd_destroy_one_worker(struct cgd_worker *cw)
{
       KASSERT(cgd_spawning);
       KASSERT(cw == cgd_worker);

       if (--cgd_refcnt == 0) {
               cgd_destroy_worker(cgd_worker);
               cgd_worker = NULL;
       }
}

static struct cgd_worker *
cgd_create_worker(void)
{
       struct cgd_worker *cw;
       struct workqueue *wq;
       struct pool *cp;
       int error;

       cw = kmem_alloc(sizeof(struct cgd_worker), KM_SLEEP);
       cp = kmem_alloc(sizeof(struct pool), KM_SLEEP);

       error = workqueue_create(&wq, "cgd", cgd_process, NULL,
           PRI_BIO, IPL_BIO, WQ_FPU|WQ_MPSAFE|WQ_PERCPU);
       if (error) {
               kmem_free(cp, sizeof(struct pool));
               kmem_free(cw, sizeof(struct cgd_worker));
               return NULL;
       }

       cw->cw_cpool = cp;
       cw->cw_wq = wq;
       pool_init(cw->cw_cpool, sizeof(struct cgd_xfer), 0,
           0, 0, "cgdcpl", NULL, IPL_BIO);
       mutex_init(&cw->cw_lock, MUTEX_DEFAULT, IPL_BIO);

       return cw;
}

static void
cgd_destroy_worker(struct cgd_worker *cw)
{

       /*
        * Wait for all worker threads to complete before destroying
        * the rest of the cgd_worker.
        */
       if (cw->cw_wq)
               workqueue_destroy(cw->cw_wq);

       mutex_destroy(&cw->cw_lock);

       if (cw->cw_cpool) {
               pool_destroy(cw->cw_cpool);
               kmem_free(cw->cw_cpool, sizeof(struct pool));
       }

       kmem_free(cw, sizeof(struct cgd_worker));
}

static int
cgdopen(dev_t dev, int flags, int fmt, struct lwp *l)
{
       struct  cgd_softc *sc;
       int error;

       DPRINTF_FOLLOW(("cgdopen(0x%"PRIx64", %d)\n", dev, flags));

       error = cgd_lock(true);
       if (error)
               return error;
       sc = getcgd_softc(dev);
       if (sc == NULL)
               sc = cgd_spawn(CGDUNIT(dev));
       cgd_unlock();
       if (sc == NULL)
               return ENXIO;

       return dk_open(&sc->sc_dksc, dev, flags, fmt, l);
}

static int
cgdclose(dev_t dev, int flags, int fmt, struct lwp *l)
{
       struct  cgd_softc *sc;
       struct  dk_softc *dksc;
       int error;

       DPRINTF_FOLLOW(("cgdclose(0x%"PRIx64", %d)\n", dev, flags));

       error = cgd_lock(false);
       if (error)
               return error;
       sc = getcgd_softc(dev);
       if (sc == NULL) {
               error = ENXIO;
               goto done;
       }

       dksc = &sc->sc_dksc;
       if ((error =  dk_close(dksc, dev, flags, fmt, l)) != 0)
               goto done;

       if (!DK_ATTACHED(dksc)) {
               if ((error = cgd_destroy(sc->sc_dksc.sc_dev)) != 0) {
                       device_printf(dksc->sc_dev,
                           "unable to detach instance\n");
                       goto done;
               }
       }

done:
       cgd_unlock();

       return error;
}

static void
cgdstrategy(struct buf *bp)
{
       struct  cgd_softc *sc = getcgd_softc(bp->b_dev);

       DPRINTF_FOLLOW(("cgdstrategy(%p): b_bcount = %ld\n", bp,
           (long)bp->b_bcount));

       /*
        * Reject unaligned writes.
        */
       if (((uintptr_t)bp->b_data & 3) != 0) {
               bp->b_error = EINVAL;
               goto bail;
       }

       dk_strategy(&sc->sc_dksc, bp);
       return;

bail:
       bp->b_resid = bp->b_bcount;
       biodone(bp);
       return;
}

static int
cgdsize(dev_t dev)
{
       struct cgd_softc *sc = getcgd_softc(dev);

       DPRINTF_FOLLOW(("cgdsize(0x%"PRIx64")\n", dev));
       if (!sc)
               return -1;
       return dk_size(&sc->sc_dksc, dev);
}

/*
* cgd_{get,put}data are functions that deal with getting a buffer
* for the new encrypted data.
* We can no longer have a buffer per device, we need a buffer per
* work queue...
*/

static void *
cgd_getdata(struct cgd_softc *sc, unsigned long size)
{
       void *data = NULL;

       mutex_enter(&sc->sc_lock);
       if (!sc->sc_data_used) {
               sc->sc_data_used = true;
               data = sc->sc_data;
       }
       mutex_exit(&sc->sc_lock);

       if (data)
               return data;

       return kmem_intr_alloc(size, KM_NOSLEEP);
}

static void
cgd_putdata(struct cgd_softc *sc, void *data, unsigned long size)
{

       if (data == sc->sc_data) {
               mutex_enter(&sc->sc_lock);
               sc->sc_data_used = false;
               mutex_exit(&sc->sc_lock);
       } else
               kmem_intr_free(data, size);
}

static int
cgd_diskstart(device_t dev, struct buf *bp)
{
       struct  cgd_softc *sc = device_private(dev);
       struct  cgd_worker *cw = sc->sc_worker;
       struct  dk_softc *dksc = &sc->sc_dksc;
       struct  disk_geom *dg = &dksc->sc_dkdev.dk_geom;
       struct  cgd_xfer *cx;
       struct  buf *nbp;
       void *  newaddr;
       daddr_t bn;

       DPRINTF_FOLLOW(("cgd_diskstart(%p, %p)\n", dksc, bp));

       bn = bp->b_rawblkno;

       /*
        * We attempt to allocate all of our resources up front, so that
        * we can fail quickly if they are unavailable.
        */
       nbp = getiobuf(sc->sc_tvn, false);
       if (nbp == NULL)
               return EAGAIN;

       cx = pool_get(cw->cw_cpool, PR_NOWAIT);
       if (cx == NULL) {
               putiobuf(nbp);
               return EAGAIN;
       }

       cx->cx_sc = sc;
       cx->cx_obp = bp;
       cx->cx_nbp = nbp;
       cx->cx_srcv = cx->cx_dstv = bp->b_data;
       cx->cx_blkno = bn;
       cx->cx_secsize = dg->dg_secsize;

       /*
        * If we are writing, then we need to encrypt the outgoing
        * block into a new block of memory.
        */
       if ((bp->b_flags & B_READ) == 0) {
               newaddr = cgd_getdata(sc, bp->b_bcount);
               if (!newaddr) {
                       pool_put(cw->cw_cpool, cx);
                       putiobuf(nbp);
                       return EAGAIN;
               }

               cx->cx_dstv = newaddr;
               cx->cx_len = bp->b_bcount;
               cx->cx_dir = CGD_CIPHER_ENCRYPT;

               cgd_enqueue(sc, cx);
               return 0;
       }

       cgd_diskstart2(sc, cx);
       return 0;
}

static void
cgd_diskstart2(struct cgd_softc *sc, struct cgd_xfer *cx)
{
       struct  vnode *vp;
       struct  buf *bp;
       struct  buf *nbp;

       bp = cx->cx_obp;
       nbp = cx->cx_nbp;

       nbp->b_data = cx->cx_dstv;
       nbp->b_flags = bp->b_flags;
       nbp->b_oflags = bp->b_oflags;
       nbp->b_cflags = bp->b_cflags;
       nbp->b_iodone = cgdiodone;
       nbp->b_proc = bp->b_proc;
       nbp->b_blkno = btodb(cx->cx_blkno * cx->cx_secsize);
       nbp->b_bcount = bp->b_bcount;
       nbp->b_private = cx;

       BIO_COPYPRIO(nbp, bp);

       if ((nbp->b_flags & B_READ) == 0) {
               vp = nbp->b_vp;
               mutex_enter(vp->v_interlock);
               vp->v_numoutput++;
               mutex_exit(vp->v_interlock);
       }
       VOP_STRATEGY(sc->sc_tvn, nbp);
}

static void
cgdiodone(struct buf *nbp)
{
       struct  cgd_xfer *cx = nbp->b_private;
       struct  buf *obp = cx->cx_obp;
       struct  cgd_softc *sc = getcgd_softc(obp->b_dev);
       struct  dk_softc *dksc = &sc->sc_dksc;
       struct  disk_geom *dg = &dksc->sc_dkdev.dk_geom;
       daddr_t bn;

       KDASSERT(sc);

       DPRINTF_FOLLOW(("cgdiodone(%p)\n", nbp));
       DPRINTF(CGDB_IO, ("cgdiodone: bp %p bcount %d resid %d\n",
           obp, obp->b_bcount, obp->b_resid));
       DPRINTF(CGDB_IO, (" dev 0x%"PRIx64", nbp %p bn %" PRId64
           " addr %p bcnt %d\n", nbp->b_dev, nbp, nbp->b_blkno, nbp->b_data,
               nbp->b_bcount));
       if (nbp->b_error != 0) {
               obp->b_error = nbp->b_error;
               DPRINTF(CGDB_IO, ("%s: error %d\n", dksc->sc_xname,
                   obp->b_error));
       }

       /* Perform the decryption if we are reading.
        *
        * Note: use the blocknumber from nbp, since it is what
        *       we used to encrypt the blocks.
        */

       if (nbp->b_flags & B_READ) {
               bn = dbtob(nbp->b_blkno) / dg->dg_secsize;

               cx->cx_obp     = obp;
               cx->cx_nbp     = nbp;
               cx->cx_dstv    = obp->b_data;
               cx->cx_srcv    = obp->b_data;
               cx->cx_len     = obp->b_bcount;
               cx->cx_blkno   = bn;
               cx->cx_secsize = dg->dg_secsize;
               cx->cx_dir     = CGD_CIPHER_DECRYPT;

               cgd_enqueue(sc, cx);
               return;
       }

       cgd_iodone2(sc, cx);
}

static void
cgd_iodone2(struct cgd_softc *sc, struct cgd_xfer *cx)
{
       struct cgd_worker *cw = sc->sc_worker;
       struct buf *obp = cx->cx_obp;
       struct buf *nbp = cx->cx_nbp;
       struct dk_softc *dksc = &sc->sc_dksc;

       pool_put(cw->cw_cpool, cx);

       /* If we allocated memory, free it now... */
       if (nbp->b_data != obp->b_data)
               cgd_putdata(sc, nbp->b_data, nbp->b_bcount);

       putiobuf(nbp);

       /* Request is complete for whatever reason */
       obp->b_resid = 0;
       if (obp->b_error != 0)
               obp->b_resid = obp->b_bcount;

       dk_done(dksc, obp);
       dk_start(dksc, NULL);
}

static int
cgd_dumpblocks(device_t dev, void *va, daddr_t blkno, int nblk)
{
       struct cgd_softc *sc = device_private(dev);
       struct dk_softc *dksc = &sc->sc_dksc;
       struct disk_geom *dg = &dksc->sc_dkdev.dk_geom;
       size_t nbytes, blksize;
       void *buf;
       int error;

       /*
        * dk_dump gives us units of disklabel sectors.  Everything
        * else in cgd uses units of diskgeom sectors.  These had
        * better agree; otherwise we need to figure out how to convert
        * between them.
        */
       KASSERTMSG((dg->dg_secsize == dksc->sc_dkdev.dk_label->d_secsize),
           "diskgeom secsize %"PRIu32" != disklabel secsize %"PRIu32,
           dg->dg_secsize, dksc->sc_dkdev.dk_label->d_secsize);
       blksize = dg->dg_secsize;

       /*
        * Compute the number of bytes in this request, which dk_dump
        * has `helpfully' converted to a number of blocks for us.
        */
       nbytes = nblk*blksize;

       /* Try to acquire a buffer to store the ciphertext.  */
       buf = cgd_getdata(sc, nbytes);
       if (buf == NULL)
               /* Out of memory: give up.  */
               return ENOMEM;

       /* Encrypt the caller's data into the temporary buffer.  */
       cgd_cipher(sc, buf, va, nbytes, blkno, blksize, CGD_CIPHER_ENCRYPT);

       /* Pass it on to the underlying disk device.  */
       error = bdev_dump(sc->sc_tdev, blkno, buf, nbytes);

       /* Release the buffer.  */
       cgd_putdata(sc, buf, nbytes);

       /* Return any error from the underlying disk device.  */
       return error;
}

/* XXX: we should probably put these into dksubr.c, mostly */
static int
cgdread(dev_t dev, struct uio *uio, int flags)
{
       struct  cgd_softc *sc;
       struct  dk_softc *dksc;

       DPRINTF_FOLLOW(("cgdread(0x%llx, %p, %d)\n",
           (unsigned long long)dev, uio, flags));
       sc = getcgd_softc(dev);
       if (sc == NULL)
               return ENXIO;
       dksc = &sc->sc_dksc;
       if (!DK_ATTACHED(dksc))
               return ENXIO;
       return physio(cgdstrategy, NULL, dev, B_READ, minphys, uio);
}

/* XXX: we should probably put these into dksubr.c, mostly */
static int
cgdwrite(dev_t dev, struct uio *uio, int flags)
{
       struct  cgd_softc *sc;
       struct  dk_softc *dksc;

       DPRINTF_FOLLOW(("cgdwrite(0x%"PRIx64", %p, %d)\n", dev, uio, flags));
       sc = getcgd_softc(dev);
       if (sc == NULL)
               return ENXIO;
       dksc = &sc->sc_dksc;
       if (!DK_ATTACHED(dksc))
               return ENXIO;
       return physio(cgdstrategy, NULL, dev, B_WRITE, minphys, uio);
}

static int
cgdioctl(dev_t dev, u_long cmd, void *data, int flag, struct lwp *l)
{
       struct  cgd_softc *sc;
       struct  dk_softc *dksc;
       int     part = DISKPART(dev);
       int     pmask = 1 << part;
       int     error;

       DPRINTF_FOLLOW(("cgdioctl(0x%"PRIx64", %ld, %p, %d, %p)\n",
           dev, cmd, data, flag, l));

       switch (cmd) {
       case CGDIOCGET:
               return cgd_ioctl_get(dev, data, l);
       case CGDIOCSET:
       case CGDIOCCLR:
               if ((flag & FWRITE) == 0)
                       return EBADF;
               /* FALLTHROUGH */
       default:
               sc = getcgd_softc(dev);
               if (sc == NULL)
                       return ENXIO;
               dksc = &sc->sc_dksc;
               break;
       }

       switch (cmd) {
       case CGDIOCSET:
               cgd_busy(sc);
               if (DK_ATTACHED(dksc))
                       error = EBUSY;
               else
                       error = cgd_ioctl_set(sc, data, l);
               cgd_unbusy(sc);
               break;
       case CGDIOCCLR:
               cgd_busy(sc);
               if (DK_BUSY(&sc->sc_dksc, pmask))
                       error = EBUSY;
               else
                       error = cgd_ioctl_clr(sc, l);
               cgd_unbusy(sc);
               break;
       case DIOCGCACHE:
       case DIOCCACHESYNC:
               cgd_busy(sc);
               if (!DK_ATTACHED(dksc)) {
                       cgd_unbusy(sc);
                       error = ENOENT;
                       break;
               }
               /*
                * We pass this call down to the underlying disk.
                */
               error = VOP_IOCTL(sc->sc_tvn, cmd, data, flag, l->l_cred);
               cgd_unbusy(sc);
               break;
       case DIOCGSECTORALIGN: {
               struct disk_sectoralign *dsa = data;

               cgd_busy(sc);
               if (!DK_ATTACHED(dksc)) {
                       cgd_unbusy(sc);
                       error = ENOENT;
                       break;
               }

               /* Get the underlying disk's sector alignment.  */
               error = VOP_IOCTL(sc->sc_tvn, cmd, data, flag, l->l_cred);
               if (error) {
                       cgd_unbusy(sc);
                       break;
               }

               /* Adjust for the disklabel partition if necessary.  */
               if (part != RAW_PART) {
                       struct disklabel *lp = dksc->sc_dkdev.dk_label;
                       daddr_t offset = lp->d_partitions[part].p_offset;
                       uint32_t r = offset % dsa->dsa_alignment;

                       if (r < dsa->dsa_firstaligned)
                               dsa->dsa_firstaligned = dsa->dsa_firstaligned
                                   - r;
                       else
                               dsa->dsa_firstaligned = (dsa->dsa_firstaligned
                                   + dsa->dsa_alignment) - r;
               }
               cgd_unbusy(sc);
               break;
       }
       case DIOCGSTRATEGY:
       case DIOCSSTRATEGY:
               if (!DK_ATTACHED(dksc)) {
                       error = ENOENT;
                       break;
               }
               /*FALLTHROUGH*/
       default:
               error = dk_ioctl(dksc, dev, cmd, data, flag, l);
               break;
       case CGDIOCGET:
               KASSERT(0);
               error = EINVAL;
       }

       return error;
}

static int
cgddump(dev_t dev, daddr_t blkno, void *va, size_t size)
{
       struct  cgd_softc *sc;

       DPRINTF_FOLLOW(("cgddump(0x%"PRIx64", %" PRId64 ", %p, %lu)\n",
           dev, blkno, va, (unsigned long)size));
       sc = getcgd_softc(dev);
       if (sc == NULL)
               return ENXIO;
       return dk_dump(&sc->sc_dksc, dev, blkno, va, size, DK_DUMP_RECURSIVE);
}

/*
* XXXrcd:
*  for now we hardcode the maximum key length.
*/
#define MAX_KEYSIZE     1024

static const struct {
       const char *n;
       int v;
       int d;
} encblkno[] = {
       { "encblkno",  CGD_CIPHER_CBC_ENCBLKNO8, 1 },
       { "encblkno8", CGD_CIPHER_CBC_ENCBLKNO8, 1 },
       { "encblkno1", CGD_CIPHER_CBC_ENCBLKNO1, 8 },
};

/* ARGSUSED */
static int
cgd_ioctl_set(struct cgd_softc *sc, void *data, struct lwp *l)
{
       struct   cgd_ioctl *ci = data;
       struct   vnode *vp;
       int      ret;
       size_t   i;
       size_t   keybytes;                      /* key length in bytes */
       const char *cp;
       struct pathbuf *pb;
       char     *inbuf;
       struct dk_softc *dksc = &sc->sc_dksc;

       cp = ci->ci_disk;

       ret = pathbuf_copyin(ci->ci_disk, &pb);
       if (ret != 0) {
               return ret;
       }
       ret = vn_bdev_openpath(pb, &vp, l);
       pathbuf_destroy(pb);
       if (ret != 0) {
               return ret;
       }

       inbuf = kmem_alloc(MAX_KEYSIZE, KM_SLEEP);

       if ((ret = cgdinit(sc, cp, vp, l)) != 0)
               goto bail;

       (void)memset(inbuf, 0, MAX_KEYSIZE);
       ret = copyinstr(ci->ci_alg, inbuf, 256, NULL);
       if (ret)
               goto bail;
       sc->sc_cfuncs = cryptfuncs_find(inbuf);
       if (!sc->sc_cfuncs) {
               ret = EINVAL;
               goto bail;
       }

       (void)memset(inbuf, 0, MAX_KEYSIZE);
       ret = copyinstr(ci->ci_ivmethod, inbuf, MAX_KEYSIZE, NULL);
       if (ret)
               goto bail;

       for (i = 0; i < __arraycount(encblkno); i++)
               if (strcmp(encblkno[i].n, inbuf) == 0)
                       break;

       if (i == __arraycount(encblkno)) {
               ret = EINVAL;
               goto bail;
       }

       keybytes = ci->ci_keylen / 8 + 1;
       if (keybytes > MAX_KEYSIZE) {
               ret = EINVAL;
               goto bail;
       }

       (void)memset(inbuf, 0, MAX_KEYSIZE);
       ret = copyin(ci->ci_key, inbuf, keybytes);
       if (ret)
               goto bail;

       sc->sc_cdata.cf_blocksize = ci->ci_blocksize;
       sc->sc_cdata.cf_mode = encblkno[i].v;

       /*
        * Print a warning if the user selected the legacy encblkno8
        * mistake, and reject it altogether for ciphers that it
        * doesn't apply to.
        */
       if (encblkno[i].v != CGD_CIPHER_CBC_ENCBLKNO1) {
               if (strcmp(sc->sc_cfuncs->cf_name, "aes-cbc") &&
                   strcmp(sc->sc_cfuncs->cf_name, "3des-cbc") &&
                   strcmp(sc->sc_cfuncs->cf_name, "blowfish-cbc")) {
                       log(LOG_WARNING, "cgd: %s only makes sense for cbc,"
                           " not for %s; ignoring\n",
                           encblkno[i].n, sc->sc_cfuncs->cf_name);
                       sc->sc_cdata.cf_mode = CGD_CIPHER_CBC_ENCBLKNO1;
               } else {
                       log(LOG_WARNING, "cgd: enabling legacy encblkno8\n");
               }
       }

       sc->sc_cdata.cf_keylen = ci->ci_keylen;
       sc->sc_cdata.cf_priv = sc->sc_cfuncs->cf_init(ci->ci_keylen, inbuf,
           &sc->sc_cdata.cf_blocksize);
       if (sc->sc_cdata.cf_blocksize > CGD_MAXBLOCKSIZE) {
           log(LOG_WARNING, "cgd: Disallowed cipher with blocksize %zu > %u\n",
               sc->sc_cdata.cf_blocksize, CGD_MAXBLOCKSIZE);
           sc->sc_cdata.cf_priv = NULL;
       }

       /*
        * The blocksize is supposed to be in bytes. Unfortunately originally
        * it was expressed in bits. For compatibility we maintain encblkno
        * and encblkno8.
        */
       sc->sc_cdata.cf_blocksize /= encblkno[i].d;
       (void)explicit_memset(inbuf, 0, MAX_KEYSIZE);
       if (!sc->sc_cdata.cf_priv) {
               ret = EINVAL;           /* XXX is this the right error? */
               goto bail;
       }
       kmem_free(inbuf, MAX_KEYSIZE);

       bufq_alloc(&dksc->sc_bufq, "fcfs", 0);

       sc->sc_data = kmem_alloc(MAXPHYS, KM_SLEEP);
       sc->sc_data_used = false;

       /* Attach the disk. */
       dk_attach(dksc);
       disk_attach(&dksc->sc_dkdev);

       disk_set_info(dksc->sc_dev, &dksc->sc_dkdev, NULL);

       /* Discover wedges on this disk. */
       dkwedge_discover(&dksc->sc_dkdev);

       return 0;

bail:
       kmem_free(inbuf, MAX_KEYSIZE);
       (void)vn_close(vp, FREAD|FWRITE, l->l_cred);
       return ret;
}

/* ARGSUSED */
static int
cgd_ioctl_clr(struct cgd_softc *sc, struct lwp *l)
{
       struct  dk_softc *dksc = &sc->sc_dksc;

       if (!DK_ATTACHED(dksc))
               return ENXIO;

       /* Delete all of our wedges. */
       dkwedge_delall(&dksc->sc_dkdev);

       /* Kill off any queued buffers. */
       dk_drain(dksc);
       bufq_free(dksc->sc_bufq);

       (void)vn_close(sc->sc_tvn, FREAD|FWRITE, l->l_cred);
       sc->sc_cfuncs->cf_destroy(sc->sc_cdata.cf_priv);
       kmem_free(sc->sc_tpath, sc->sc_tpathlen);
       kmem_free(sc->sc_data, MAXPHYS);
       sc->sc_data_used = false;
       dk_detach(dksc);
       disk_detach(&dksc->sc_dkdev);

       return 0;
}

static int
cgd_ioctl_get(dev_t dev, void *data, struct lwp *l)
{
       struct cgd_softc *sc;
       struct cgd_user *cgu;
       int unit, error;

       unit = CGDUNIT(dev);
       cgu = (struct cgd_user *)data;

       DPRINTF_FOLLOW(("cgd_ioctl_get(0x%"PRIx64", %d, %p, %p)\n",
                          dev, unit, data, l));

       /* XXX, we always return this units data, so if cgu_unit is
        * not -1, that field doesn't match the rest
        */
       if (cgu->cgu_unit == -1)
               cgu->cgu_unit = unit;

       if (cgu->cgu_unit < 0)
               return EINVAL;  /* XXX: should this be ENXIO? */

       error = cgd_lock(false);
       if (error)
               return error;

       sc = device_lookup_private(&cgd_cd, unit);
       if (sc == NULL || !DK_ATTACHED(&sc->sc_dksc)) {
               cgu->cgu_dev = 0;
               cgu->cgu_alg[0] = '\0';
               cgu->cgu_blocksize = 0;
               cgu->cgu_mode = 0;
               cgu->cgu_keylen = 0;
       }
       else {
               mutex_enter(&sc->sc_lock);
               cgu->cgu_dev = sc->sc_tdev;
               strncpy(cgu->cgu_alg, sc->sc_cfuncs->cf_name,
                   sizeof(cgu->cgu_alg));
               cgu->cgu_blocksize = sc->sc_cdata.cf_blocksize;
               cgu->cgu_mode = sc->sc_cdata.cf_mode;
               cgu->cgu_keylen = sc->sc_cdata.cf_keylen;
               mutex_exit(&sc->sc_lock);
       }

       cgd_unlock();
       return 0;
}

static int
cgdinit(struct cgd_softc *sc, const char *cpath, struct vnode *vp,
       struct lwp *l)
{
       struct  disk_geom *dg;
       int     ret;
       char    *tmppath;
       uint64_t psize;
       unsigned secsize;
       struct dk_softc *dksc = &sc->sc_dksc;

       sc->sc_tvn = vp;
       sc->sc_tpath = NULL;

       tmppath = kmem_alloc(MAXPATHLEN, KM_SLEEP);
       ret = copyinstr(cpath, tmppath, MAXPATHLEN, &sc->sc_tpathlen);
       if (ret)
               goto bail;
       sc->sc_tpath = kmem_alloc(sc->sc_tpathlen, KM_SLEEP);
       memcpy(sc->sc_tpath, tmppath, sc->sc_tpathlen);

       sc->sc_tdev = vp->v_rdev;

       if ((ret = getdisksize(vp, &psize, &secsize)) != 0)
               goto bail;

       if (psize == 0) {
               ret = ENODEV;
               goto bail;
       }

       /*
        * XXX here we should probe the underlying device.  If we
        *     are accessing a partition of type RAW_PART, then
        *     we should populate our initial geometry with the
        *     geometry that we discover from the device.
        */
       dg = &dksc->sc_dkdev.dk_geom;
       memset(dg, 0, sizeof(*dg));
       dg->dg_secperunit = psize;
       dg->dg_secsize = secsize;
       dg->dg_ntracks = 1;
       dg->dg_nsectors = 1024 * 1024 / dg->dg_secsize;
       dg->dg_ncylinders = dg->dg_secperunit / dg->dg_nsectors;

bail:
       kmem_free(tmppath, MAXPATHLEN);
       if (ret && sc->sc_tpath)
               kmem_free(sc->sc_tpath, sc->sc_tpathlen);
       return ret;
}

/*
* Our generic cipher entry point.  This takes care of the
* IV mode and passes off the work to the specific cipher.
* We implement here the IV method ``encrypted block
* number''.
*
* XXXrcd: for now we rely on our own crypto framework defined
*         in dev/cgd_crypto.c.  This will change when we
*         get a generic kernel crypto framework.
*/

static void
blkno2blkno_buf(char *sbuf, daddr_t blkno)
{
       int     i;

       /* Set up the blkno in blkno_buf, here we do not care much
        * about the final layout of the information as long as we
        * can guarantee that each sector will have a different IV
        * and that the endianness of the machine will not affect
        * the representation that we have chosen.
        *
        * We choose this representation, because it does not rely
        * on the size of buf (which is the blocksize of the cipher),
        * but allows daddr_t to grow without breaking existing
        * disks.
        *
        * Note that blkno2blkno_buf does not take a size as input,
        * and hence must be called on a pre-zeroed buffer of length
        * greater than or equal to sizeof(daddr_t).
        */
       for (i=0; i < sizeof(daddr_t); i++) {
               *sbuf++ = blkno & 0xff;
               blkno >>= 8;
       }
}

static struct cpu_info *
cgd_cpu(struct cgd_softc *sc)
{
       struct cgd_worker *cw = sc->sc_worker;
       struct cpu_info *ci = NULL;
       u_int cidx, i;

       if (cw->cw_busy == 0) {
               cw->cw_last = cpu_index(curcpu());
               return NULL;
       }

       for (i=0, cidx = cw->cw_last+1; i<maxcpus; ++i, ++cidx) {
               if (cidx >= maxcpus)
                       cidx = 0;
               ci = cpu_lookup(cidx);
               if (ci) {
                       cw->cw_last = cidx;
                       break;
               }
       }

       return ci;
}

static void
cgd_enqueue(struct cgd_softc *sc, struct cgd_xfer *cx)
{
       struct cgd_worker *cw = sc->sc_worker;
       struct cpu_info *ci;

       mutex_enter(&cw->cw_lock);
       ci = cgd_cpu(sc);
       cw->cw_busy++;
       mutex_exit(&cw->cw_lock);

       workqueue_enqueue(cw->cw_wq, &cx->cx_work, ci);
}

static void
cgd_process(struct work *wk, void *arg)
{
       struct cgd_xfer *cx = (struct cgd_xfer *)wk;
       struct cgd_softc *sc = cx->cx_sc;
       struct cgd_worker *cw = sc->sc_worker;

       cgd_cipher(sc, cx->cx_dstv, cx->cx_srcv, cx->cx_len,
           cx->cx_blkno, cx->cx_secsize, cx->cx_dir);

       if (cx->cx_dir == CGD_CIPHER_ENCRYPT) {
               cgd_diskstart2(sc, cx);
       } else {
               cgd_iodone2(sc, cx);
       }

       mutex_enter(&cw->cw_lock);
       if (cw->cw_busy > 0)
               cw->cw_busy--;
       mutex_exit(&cw->cw_lock);
}

static void
cgd_cipher(struct cgd_softc *sc, void *dstv, const void *srcv,
   size_t len, daddr_t blkno, size_t secsize, int dir)
{
       char            *dst = dstv;
       const char      *src = srcv;
       cfunc_cipher    *cipher = sc->sc_cfuncs->cf_cipher;
       size_t          blocksize = sc->sc_cdata.cf_blocksize;
       size_t          todo;
       char            blkno_buf[CGD_MAXBLOCKSIZE] __aligned(CGD_BLOCKALIGN);

       DPRINTF_FOLLOW(("cgd_cipher() dir=%d\n", dir));

       if (sc->sc_cdata.cf_mode == CGD_CIPHER_CBC_ENCBLKNO8)
               blocksize /= 8;

       KASSERT(len % blocksize == 0);
       /* ensure that sizeof(daddr_t) <= blocksize (for encblkno IVing) */
       KASSERT(sizeof(daddr_t) <= blocksize);
       KASSERT(blocksize <= CGD_MAXBLOCKSIZE);

       for (; len > 0; len -= todo) {
               todo = MIN(len, secsize);

               memset(blkno_buf, 0x0, blocksize);
               blkno2blkno_buf(blkno_buf, blkno);
               IFDEBUG(CGDB_CRYPTO, hexprint("step 1: blkno_buf",
                   blkno_buf, blocksize));

               /*
                * Handle bollocksed up encblkno8 mistake.  We used to
                * compute the encryption of a zero block with blkno as
                * the CBC IV -- except in an early mistake arising
                * from bit/byte confusion, we actually computed the
                * encryption of the last of _eight_ zero blocks under
                * CBC as the CBC IV.
                *
                * Encrypting the block number is handled inside the
                * cipher dispatch now (even though in practice, both
                * CBC and XTS will do the same thing), so we have to
                * simulate the block number that would yield the same
                * result.  So we encrypt _six_ zero blocks -- the
                * first one and the last one are handled inside the
                * cipher dispatch.
                */
               if (sc->sc_cdata.cf_mode == CGD_CIPHER_CBC_ENCBLKNO8) {
                       static const uint8_t zero[CGD_MAXBLOCKSIZE];
                       uint8_t iv[CGD_MAXBLOCKSIZE];

                       memcpy(iv, blkno_buf, blocksize);
                       cipher(sc->sc_cdata.cf_priv, blkno_buf, zero,
                           6*blocksize, iv, CGD_CIPHER_ENCRYPT);
                       memmove(blkno_buf, blkno_buf + 5*blocksize, blocksize);
               }

               cipher(sc->sc_cdata.cf_priv, dst, src, todo, blkno_buf, dir);

               dst += todo;
               src += todo;
               blkno++;
       }
}

#ifdef DEBUG
static void
hexprint(const char *start, void *buf, int len)
{
       char    *c = buf;

       KASSERTMSG(len >= 0, "hexprint: called with len < 0");
       printf("%s: len=%06d 0x", start, len);
       while (len--)
               printf("%02x", (unsigned char) *c++);
}
#endif

static void
cgd_selftest(void)
{
       struct cgd_softc sc;
       void *buf;

       for (size_t i = 0; i < __arraycount(selftests); i++) {
               const char *alg = selftests[i].alg;
               int encblkno8 = selftests[i].encblkno8;
               const uint8_t *key = selftests[i].key;
               int keylen = selftests[i].keylen;
               int txtlen = selftests[i].txtlen;

               aprint_debug("cgd: self-test %s-%d%s\n", alg, keylen,
                   encblkno8 ? " (encblkno8)" : "");

               memset(&sc, 0, sizeof(sc));

               sc.sc_cfuncs = cryptfuncs_find(alg);
               if (sc.sc_cfuncs == NULL)
                       panic("%s not implemented", alg);

               sc.sc_cdata.cf_blocksize = 8 * selftests[i].blocksize;
               sc.sc_cdata.cf_mode = encblkno8 ? CGD_CIPHER_CBC_ENCBLKNO8 :
                   CGD_CIPHER_CBC_ENCBLKNO1;
               sc.sc_cdata.cf_keylen = keylen;

               sc.sc_cdata.cf_priv = sc.sc_cfuncs->cf_init(keylen,
                   key, &sc.sc_cdata.cf_blocksize);
               if (sc.sc_cdata.cf_priv == NULL)
                       panic("cf_priv is NULL");
               if (sc.sc_cdata.cf_blocksize > CGD_MAXBLOCKSIZE)
                       panic("bad block size %zu", sc.sc_cdata.cf_blocksize);

               if (!encblkno8)
                       sc.sc_cdata.cf_blocksize /= 8;

               buf = kmem_alloc(txtlen, KM_SLEEP);
               memcpy(buf, selftests[i].ptxt, txtlen);

               cgd_cipher(&sc, buf, buf, txtlen, selftests[i].blkno,
                               selftests[i].secsize, CGD_CIPHER_ENCRYPT);
               if (memcmp(buf, selftests[i].ctxt, txtlen) != 0) {
                       hexdump(printf, "was", buf, txtlen);
                       hexdump(printf, "exp", selftests[i].ctxt, txtlen);
                       panic("cgd %s-%d encryption is broken [%zu]",
                           selftests[i].alg, keylen, i);
               }

               cgd_cipher(&sc, buf, buf, txtlen, selftests[i].blkno,
                               selftests[i].secsize, CGD_CIPHER_DECRYPT);
               if (memcmp(buf, selftests[i].ptxt, txtlen) != 0) {
                       hexdump(printf, "was", buf, txtlen);
                       hexdump(printf, "exp", selftests[i].ptxt, txtlen);
                       panic("cgd %s-%d decryption is broken [%zu]",
                           selftests[i].alg, keylen, i);
               }

               kmem_free(buf, txtlen);
               sc.sc_cfuncs->cf_destroy(sc.sc_cdata.cf_priv);
       }

       aprint_debug("cgd: self-tests passed\n");
}

MODULE(MODULE_CLASS_DRIVER, cgd, "blowfish,des,dk_subr,bufq_fcfs");

#ifdef _MODULE
CFDRIVER_DECL(cgd, DV_DISK, NULL);

devmajor_t cgd_bmajor = -1, cgd_cmajor = -1;
#endif

static int
cgd_modcmd(modcmd_t cmd, void *arg)
{
       int error = 0;

       switch (cmd) {
       case MODULE_CMD_INIT:
#ifdef _MODULE
               mutex_init(&cgd_spawning_mtx, MUTEX_DEFAULT, IPL_NONE);
               cv_init(&cgd_spawning_cv, "cgspwn");

               /*
                * Attach the {b,c}devsw's
                */
               error = devsw_attach("cgd", &cgd_bdevsw, &cgd_bmajor,
                   &cgd_cdevsw, &cgd_cmajor);
               if (error) {
                       aprint_error("%s: unable to attach %s devsw, "
                           "error %d", __func__, cgd_cd.cd_name, error);
                       break;
               }

               /*
                * Attach to autoconf database
                */
               error = config_cfdriver_attach(&cgd_cd);
               if (error) {
                       devsw_detach(&cgd_bdevsw, &cgd_cdevsw);
                       aprint_error("%s: unable to register cfdriver for"
                           "%s, error %d\n", __func__, cgd_cd.cd_name, error);
                       break;
               }

               error = config_cfattach_attach(cgd_cd.cd_name, &cgd_ca);
               if (error) {
                       config_cfdriver_detach(&cgd_cd);
                       devsw_detach(&cgd_bdevsw, &cgd_cdevsw);
                       aprint_error("%s: unable to register cfattach for"
                           "%s, error %d\n", __func__, cgd_cd.cd_name, error);
                       break;
               }
#endif
               break;

       case MODULE_CMD_FINI:
#ifdef _MODULE
               /*
                * Remove device from autoconf database
                */
               error = config_cfattach_detach(cgd_cd.cd_name, &cgd_ca);
               if (error) {
                       aprint_error("%s: failed to detach %s cfattach, "
                           "error %d\n", __func__, cgd_cd.cd_name, error);
                       break;
               }
               error = config_cfdriver_detach(&cgd_cd);
               if (error) {
                       (void)config_cfattach_attach(cgd_cd.cd_name, &cgd_ca);
                       aprint_error("%s: failed to detach %s cfdriver, "
                           "error %d\n", __func__, cgd_cd.cd_name, error);
                       break;
               }

               /*
                * Remove {b,c}devsw's
                */
               devsw_detach(&cgd_bdevsw, &cgd_cdevsw);

               cv_destroy(&cgd_spawning_cv);
               mutex_destroy(&cgd_spawning_mtx);
#endif
               break;

       case MODULE_CMD_STAT:
               error = ENOTTY;
               break;
       default:
               error = ENOTTY;
               break;
       }

       return error;
}