fs - zygaena/hammerhead

root / base / usr / src / common / fs
fs Plain Text 3670 lines 81.8 KB
/*
 * Copyright 2011-2017 Josef 'Jeff' Sipek <jeffpc@josefsipek.net>
 *
 * Permission to use, copy, modify, and/or distribute this software for any
 * purpose with or without fee is hereby granted, provided that the above
 * copyright notice and this permission notice appear in all copies.
 *
 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 */
bootrd_cpio
/*
 * This file and its contents are supplied under the terms of the
 * Common Development and Distribution License ("CDDL"), version 1.0.
 * You may only use this file in accordance with the terms of version
 * 1.0 of the CDDL.
 *
 * A full copy of the text of the CDDL should have accompanied this
 * source.  A copy of the CDDL is also available via the Internet at
 * http://www.illumos.org/license/CDDL.
 */

/*
 * Copyright 2013 Joyent, Inc.  All rights reserved.
 * Copyright 2025 MNX Cloud, Inc.
 */

#include <sys/bootconf.h>
#include <sys/types.h>
#include <sys/param.h>
#include <sys/vnode.h>
#include <sys/fs/ufs_fsdir.h>
#include <sys/fs/ufs_fs.h>
#include <sys/fs/ufs_inode.h>
#include <sys/sysmacros.h>
#include <sys/bootvfs.h>
#include <sys/bootinfo.h>
#include <sys/filep.h>
#include <sys/sunddi.h>

#define	MAX_FILES	MAX_BOOT_MODULES
#define	MAX_FDS		256

extern void *bkmem_alloc(size_t);
extern void bkmem_free(void *, size_t);

/*
 * TODO: Replace these declarations with inclusion of the ordinary userland
 * bootfs headers once they're available.
 */
typedef struct bfile {
	char bf_name[MAXPATHLEN];
	caddr_t bf_addr;
	size_t bf_size;
	struct bfile *bf_next;
	uint64_t bf_ino;
} bfile_t;

typedef struct bf_fd {
	bfile_t *fd_file;
	off_t fd_pos;
} bf_fd_t;

static bfile_t *head;
static uint_t init_done;
static bf_fd_t fds[MAX_FDS];

static char cpath[MAXPATHLEN];	/* For canonicalising filenames */

static void bbootfs_closeall(int);

static void
canonicalise(const char *fn, char *out)
{
	const char *p;
	char *q, *s;
	char *last;
	char *oc;
	int is_slash = 0;
	static char scratch[MAXPATHLEN];

	if (fn == NULL) {
		*out = '\0';
		return;
	}

	/*
	 * Remove leading slashes and condense all multiple slashes into one.
	 */
	p = fn;
	while (*p == '/')
		++p;

	for (q = scratch; *p != '\0'; p++) {
		if (*p == '/' && !is_slash) {
			*q++ = '/';
			is_slash = 1;
		} else if (*p != '/') {
			*q++ = *p;
			is_slash = 0;
		}
	}
	*q = '\0';

	if (strncmp(scratch, "system/boot/", 12) == 0 ||
	    strcmp(scratch, "system/boot") == 0) {
		s = scratch + 12;
	} else {
		s = scratch;
	}

	for (last = strsep(&s, "/"), q = oc = out; last != NULL;
	    last = strsep(&s, "/")) {
		if (strcmp(last, ".") == 0)
			continue;
		if (strcmp(last, "..") == 0) {
			for (oc = q; oc > out && *oc != '/'; oc--)
				;
			q = oc;
			continue;
		}
		if (q > out)
			*q++ = '/';
		q += snprintf(q, MAXPATHLEN - (q - out), "%s", last);
	}

	*q = '\0';
}

static int
bbootfs_mountroot(char *str __unused)
{
	return (-1);
}

static int
bbootfs_unmountroot(void)
{
	return (-1);
}

static int
bbootfs_init(void)
{
	bfile_t *fp;
	char propname[32];
	uint64_t propval;
	uint_t i;

	for (i = 0; i < MAX_FILES; i++) {
		(void) snprintf(propname, sizeof (propname),
		    "module-name-%u", i);
		if (do_bsys_getproplen(NULL, propname) < 0)
			break;

		if ((fp = bkmem_alloc(sizeof (bfile_t))) == NULL) {
			bbootfs_closeall(1);
			return (-1);
		}

		(void) do_bsys_getprop(NULL, propname, cpath);
		canonicalise(cpath, fp->bf_name);

		(void) snprintf(propname, sizeof (propname),
		    "module-addr-%u", i);
		if (do_bsys_getproplen(NULL, propname) != sizeof (uint64_t)) {
			bkmem_free(fp, sizeof (bfile_t));
			continue;
		}
		(void) do_bsys_getprop(NULL, propname, &propval);
		fp->bf_addr = (void *)(uintptr_t)propval;

		(void) snprintf(propname, sizeof (propname),
		    "module-size-%u", i);
		if (do_bsys_getproplen(NULL, propname) != sizeof (uint64_t)) {
			bkmem_free(fp, sizeof (bfile_t));
			continue;
		}
		(void) do_bsys_getprop(NULL, propname, &propval);
		fp->bf_size = (size_t)propval;
		fp->bf_ino = i;

		fp->bf_next = head;
		head = fp;
	}

	return (0);
}

static int
bbootfs_open(char *fn, int flags __unused)
{
	uint_t i;
	bfile_t *fp;

	if (!init_done) {
		if (bbootfs_init() != 0)
			return (-1);

		init_done = 1;
	}

	canonicalise(fn, cpath);

	for (fp = head; fp != NULL; fp = fp->bf_next) {
		if (strcmp(fp->bf_name, cpath) == 0)
			break;
	}

	if (fp == NULL)
		return (-1);

	for (i = 0; i < MAX_FDS; i++) {
		if (fds[i].fd_file == NULL) {
			fds[i].fd_file = fp;
			fds[i].fd_pos = 0;
			return (i);
		}
	}

	return (-1);
}

static int
bbootfs_close(int fd)
{
	if (fds[fd].fd_file == NULL)
		return (-1);

	fds[fd].fd_file = NULL;
	fds[fd].fd_pos = 0;

	return (0);
}

static ssize_t
bbootfs_read(int fd, caddr_t buf, size_t size)
{
	ssize_t len;
	bf_fd_t *fdp = &fds[fd];

	if (fdp->fd_file == NULL)
		return (-1);

	if (fdp->fd_pos >= fdp->fd_file->bf_size)
		return (-1);

	if (fdp->fd_pos + size > fdp->fd_file->bf_size)
		len = fdp->fd_file->bf_size - fdp->fd_pos;
	else
		len = size;

	bcopy(fdp->fd_file->bf_addr + fdp->fd_pos, buf, len);

	fdp->fd_pos += len;

	return (len);
}

static off_t
bbootfs_lseek(int fd, off_t addr, int whence)
{
	bf_fd_t *fdp = &fds[fd];

	if (fdp->fd_file == NULL)
		return (-1);

	switch (whence) {
	case SEEK_CUR:
		fdp->fd_pos += addr;
		break;
	case SEEK_SET:
		fdp->fd_pos = addr;
		break;
	case SEEK_END:
		fdp->fd_pos = fdp->fd_file->bf_size;
		break;
	default:
		return (-1);
	}

	return (0);
}

static int
bbootfs_fstat(int fd, struct bootstat *bsp)
{
	bf_fd_t *fdp = &fds[fd];

	if (fdp->fd_file == NULL)
		return (-1);

	bsp->st_dev = 1;
	bsp->st_ino = fdp->fd_file->bf_ino;
	bsp->st_mode = 0444;
	bsp->st_nlink = 1;
	bsp->st_uid = bsp->st_gid = 0;
	bsp->st_rdev = 0;
	bsp->st_size = fdp->fd_file->bf_size;
	bsp->st_blksize = 1;
	bsp->st_blocks = fdp->fd_file->bf_size;
	(void) strcpy(bsp->st_fstype, "bootfs");

	return (0);
}

static void
bbootfs_closeall(int flag __unused)
{
	bfile_t *fp;

	while (head != NULL) {
		fp = head;
		head = head->bf_next;

		bkmem_free(fp, sizeof (bfile_t));
	}

	init_done = 0;
}

struct boot_fs_ops bbootfs_ops = {
	"bootfs",
	bbootfs_mountroot,
	bbootfs_unmountroot,
	bbootfs_open,
	bbootfs_close,
	bbootfs_read,
	bbootfs_lseek,
	bbootfs_fstat,
	bbootfs_closeall,
	NULL
};
/*
 * Copyright 2011-2017 Josef 'Jeff' Sipek <jeffpc@josefsipek.net>
 * Copyright 2025 MNX Cloud, Inc.
 *
 * Permission to use, copy, modify, and/or distribute this software for any
 * purpose with or without fee is hereby granted, provided that the above
 * copyright notice and this permission notice appear in all copies.
 *
 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 */

#include <sys/types.h>
#include <sys/stdbool.h>
#include <sys/sysmacros.h>
#include <sys/bootvfs.h>
#include <sys/filep.h>
#include <sys/sunddi.h>
#include <sys/ccompile.h>
#include <sys/kobj.h>
#include <sys/queue.h>

/*
 * A cpio archive is just a sequence of files, each consisting of a header
 * (struct cpio_hdr) and the file contents.
 */

struct cpio_hdr {
	uint8_t		magic[6];
	uint8_t		dev[6];
	uint8_t		ino[6];
	uint8_t		mode[6];
	uint8_t		uid[6];
	uint8_t		gid[6];
	uint8_t		nlink[6];
	uint8_t		rdev[6];
	uint8_t		mtime[11];
	uint8_t		namesize[6];
	uint8_t		filesize[11];
	char		data[];
};

/*
 * This structure represents an open file.  The list of all open files is
 * rooted in the open_files global.
 */
struct cpio_file {
	/* pointers into the archive */
	const struct cpio_hdr *hdr;
	const char *path;		/* pointer into the archive */
	const void *data;		/* pointer into the archive */

	int fd;
	off_t off;
	struct bootstat stat;

	SLIST_ENTRY(cpio_file) next;
};

/*
 * in bootrd.c
 */
extern void *bkmem_alloc(size_t);
extern void bkmem_free(void *, size_t);

static void cpio_closeall(int flag);

static bool mounted;
static SLIST_HEAD(cpio_file_list, cpio_file)
    open_files = SLIST_HEAD_INITIALIZER(open_files);

/*
 * Returns the parsed number on success, or UINT64_MAX on error.  This is
 * ok because we will never deal with numbers that large in a cpio archive.
 */
static uint64_t
__get_uint64(const uint8_t *str, size_t len, const size_t output_size)
{
	uint64_t v;

	/* check that we can represent every number */
	if (len * 3 > output_size)
		return (UINT64_MAX);

	for (v = 0; len > 0; len--, str++) {
		const uint8_t c = *str;

		if ((c < '0') || (c > '7'))
			return (UINT64_MAX);

		v = (v * 8) + (c - '0');
	}

	return (v);
}

static bool
get_uint64(const uint8_t *str, size_t len, uint64_t *out)
{
	*out = __get_uint64(str, len, NBBY * sizeof (*out));
	return (*out != UINT64_MAX);
}

static bool
get_int64(const uint8_t *str, size_t len, int64_t *out)
{
	uint64_t tmp;

	tmp = __get_uint64(str, len, NBBY * sizeof (*out) - 1);

	*out = tmp;

	return (tmp != UINT64_MAX);
}

static bool
get_uint32(const uint8_t *str, size_t len, uint32_t *out)
{
	uint64_t tmp;

	tmp = __get_uint64(str, len, NBBY * sizeof (*out));

	*out = tmp;

	return (tmp != UINT64_MAX);
}

static bool
get_int32(const uint8_t *str, size_t len, int32_t *out)
{
	uint64_t tmp;

	tmp = __get_uint64(str, len, NBBY * sizeof (*out) - 1);

	*out = tmp;

	return (tmp != UINT64_MAX);
}

static void
add_open_file(struct cpio_file *file)
{
	SLIST_INSERT_HEAD(&open_files, file, next);
}

static void
remove_open_file(struct cpio_file *file)
{
	SLIST_REMOVE(&open_files, file, cpio_file, next);
}

static struct cpio_file *
find_open_file(int fd)
{
	struct cpio_file *file;

	if (fd < 0)
		return (NULL);

	SLIST_FOREACH(file, &open_files, next)
		if (file->fd == fd)
			return (file);

	return (NULL);
}

static const void *
read_ramdisk(size_t off, size_t len)
{
	const size_t first_block_offset = off % DEV_BSIZE;
	fileid_t tmpfile;

	/* return a dummy non-NULL pointer */
	if (len == 0)
		return ("");

	/* we have to read the stuff before the desired location as well */
	len += first_block_offset;

	tmpfile.fi_blocknum = off / DEV_BSIZE;
	tmpfile.fi_count = P2ROUNDUP_TYPED(len, DEV_BSIZE, size_t);
	tmpfile.fi_memp = NULL;

	if (diskread(&tmpfile) != 0)
		return (NULL);

	return (tmpfile.fi_memp + first_block_offset);
}

static bool
parse_stat(const struct cpio_hdr *hdr, struct bootstat *stat)
{
	if (!get_uint64(hdr->dev, sizeof (hdr->dev), &stat->st_dev))
		return (false);
	if (!get_uint64(hdr->ino, sizeof (hdr->ino), &stat->st_ino))
		return (false);
	if (!get_uint32(hdr->mode, sizeof (hdr->mode), &stat->st_mode))
		return (false);
	if (!get_int32(hdr->uid, sizeof (hdr->uid), &stat->st_uid))
		return (false);
	if (!get_int32(hdr->gid, sizeof (hdr->gid), &stat->st_gid))
		return (false);
	if (!get_uint32(hdr->nlink, sizeof (hdr->nlink), &stat->st_nlink))
		return (false);
	if (!get_uint64(hdr->rdev, sizeof (hdr->rdev), &stat->st_rdev))
		return (false);

	stat->st_mtim.tv_nsec = 0;
	if (!get_int64(hdr->mtime, sizeof (hdr->mtime), &stat->st_mtim.tv_sec))
		return (false);

	stat->st_atim = stat->st_mtim;
	stat->st_ctim = stat->st_mtim;

	if (!get_uint64(hdr->filesize, sizeof (hdr->filesize), &stat->st_size))
		return (false);

	stat->st_blksize = DEV_BSIZE;
	stat->st_blocks = P2ROUNDUP(stat->st_size, DEV_BSIZE);

	return (true);
}

static int
check_archive_hdr(const struct cpio_hdr *hdr)
{
	if ((hdr->magic[0] != '0') || (hdr->magic[1] != '7') ||
	    (hdr->magic[2] != '0') || (hdr->magic[3] != '7') ||
	    (hdr->magic[4] != '0') || (hdr->magic[5] != '7'))
		return (-1);
	return (0);
}

/*
 * Check if specified header is for a file with a specific path.  If so,
 * fill in the file struct and return 0.  If not, return number of bytes to
 * skip over to get to the next header.  If an error occurs, -1 is returned.
 * If end of archive is reached, return -2 instead.
 */
static ssize_t
scan_archive_hdr(const struct cpio_hdr *hdr, size_t off,
    struct cpio_file *file, const char *wanted_path)
{
	struct bootstat stat;
	uint32_t namesize;
	uint64_t filesize;
	const char *path;
	const void *data;

	if (check_archive_hdr(hdr))
		return (-1);

	if (!get_uint32(hdr->namesize, sizeof (hdr->namesize), &namesize))
		return (-1);
	if (!get_uint64(hdr->filesize, sizeof (hdr->filesize), &filesize))
		return (-1);

	/*
	 * We have the two sizes, let's try to read the name and file
	 * contents to make sure they are part of the ramdisk.
	 */

	off += offsetof(struct cpio_hdr, data[0]);
	path = read_ramdisk(off, namesize);
	data = read_ramdisk(off + namesize, filesize);

	/* either read failing is fatal */
	if (path == NULL || data == NULL)
		return (-1);

	if (strcmp(path, "TRAILER!!!") == 0)
		return (-2);

	if (strcmp(path, wanted_path) != 0)
		return (offsetof(struct cpio_hdr, data[namesize + filesize]));

	/*
	 * This is the file we want!
	 */

	if (!parse_stat(hdr, &stat))
		return (-1);

	file->hdr = hdr;
	file->path = path;
	file->data = data;
	file->stat = stat;

	return (0);
}

static int
find_filename(char *path, struct cpio_file *file)
{
	size_t off;

	/*
	 * The paths in the cpio boot archive omit the leading '/'.  So,
	 * skip checking for it.  If the searched for path does not include
	 * the leading path (it's a relative path), fail the lookup.
	 */
	if (path[0] != '/')
		return (-1);

	path++;

	/* now scan the archive for the relevant file */

	off = 0;

	for (;;) {
		const struct cpio_hdr *hdr;
		ssize_t size;

		hdr = (struct cpio_hdr *)read_ramdisk(off,
		    sizeof (struct cpio_hdr));
		if (hdr == NULL)
			return (-1);

		size = scan_archive_hdr(hdr, off, file, path);
		if (size <= 0)
			return (size);

		off += size;
	}
}

static int
bcpio_mountroot(char *str __unused)
{
	const struct cpio_hdr *hdr;

	if (mounted)
		return (-1);

	hdr = (struct cpio_hdr *)read_ramdisk(0, sizeof (struct cpio_hdr));
	if (hdr == NULL)
		return (-1);

	if (check_archive_hdr(hdr))
		return (-1);

	mounted = true;

	return (0);
}

static int
bcpio_unmountroot(void)
{
	if (!mounted)
		return (-1);

	mounted = false;

	return (0);
}

static int
bcpio_open(char *path, int flags __unused)
{
	static int filedes = 1;
	struct cpio_file temp_file;
	struct cpio_file *file;

	if (find_filename(path, &temp_file) != 0)
		return (-1);

	file = bkmem_alloc(sizeof (struct cpio_file));
	file->hdr = temp_file.hdr;
	file->path = temp_file.path;
	file->data = temp_file.data;
	file->stat = temp_file.stat;
	file->fd = filedes++;
	file->off = 0;

	add_open_file(file);

	return (file->fd);
}

static int
bcpio_close(int fd)
{
	struct cpio_file *file;

	file = find_open_file(fd);
	if (file == NULL)
		return (-1);

	remove_open_file(file);

	bkmem_free(file, sizeof (struct cpio_file));

	return (0);
}

static void
bcpio_closeall(int flag __unused)
{
	struct cpio_file *file;

	while (!SLIST_EMPTY(&open_files)) {
		file = SLIST_FIRST(&open_files);

		if (bcpio_close(file->fd) != 0) {
			kobj_printf("closeall invoked close(%d) failed\n",
			    file->fd);
		}
	}
}

static ssize_t
bcpio_read(int fd, caddr_t buf, size_t size)
{
	struct cpio_file *file;

	file = find_open_file(fd);
	if (file == NULL)
		return (-1);

	if (size == 0)
		return (0);

	if (file->off + size > file->stat.st_size)
		size = file->stat.st_size - file->off;

	bcopy((void *)((uintptr_t)file->data + file->off), buf, size);

	file->off += size;

	return (size);
}

static off_t
bcpio_lseek(int fd, off_t addr, int whence)
{
	struct cpio_file *file;

	file = find_open_file(fd);
	if (file == NULL)
		return (-1);

	switch (whence) {
		case SEEK_CUR:
			file->off += addr;
			break;
		case SEEK_SET:
			file->off = addr;
			break;
		case SEEK_END:
			file->off = file->stat.st_size;
			break;
		default:
			kobj_printf("lseek(): invalid whence value %d\n",
			    whence);
			return (-1);
	}

	return (0);
}

static int
bcpio_fstat(int fd, struct bootstat *buf)
{
	const struct cpio_file *file;

	file = find_open_file(fd);
	if (file == NULL)
		return (-1);

	*buf = file->stat;

	return (0);
}

struct boot_fs_ops bcpio_ops = {
	.fsw_name		= "boot_cpio",
	.fsw_mountroot		= bcpio_mountroot,
	.fsw_unmountroot	= bcpio_unmountroot,
	.fsw_open		= bcpio_open,
	.fsw_close		= bcpio_close,
	.fsw_closeall		= bcpio_closeall,
	.fsw_read		= bcpio_read,
	.fsw_lseek		= bcpio_lseek,
	.fsw_fstat		= bcpio_fstat,
};
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 * Copyright 2025 MNX Cloud, Inc.
 */

/*
 * Decompression module for stand alone file systems.
 */

#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/vnode.h>
#include <sys/bootvfs.h>
#include <sys/filep.h>
#include <sys/kobj.h>
#include <zlib.h>
#include <sys/sunddi.h>

#define	MAX_DECOMP_BUFS		8
#define	GZIP_ID_BYTE_1		0x1f
#define	GZIP_ID_BYTE_2		0x8b
#define	GZIP_CM_DEFLATE		0x08
#define	SEEKBUFSIZE		8192

extern int bootrd_debug;
extern void *bkmem_alloc(size_t);
extern void bkmem_free(void *, size_t);

caddr_t scratch_bufs[MAX_DECOMP_BUFS];	/* array of free scratch mem bufs */
int decomp_bufcnt;			/* total no, of allocated decomp bufs */
int free_dcomp_bufs;			/* no. of free decomp bufs */
char seek_scrbuf[SEEKBUFSIZE];		/* buffer for seeking */
int cf_debug = 0;			/* non-zero enables debug prints */

void *
cf_alloc(void *opaque, unsigned int items, unsigned int size)
{
	fileid_t *filep;
	unsigned int nbytes;
	caddr_t ptr;

	filep = (fileid_t *)opaque;
	nbytes = roundup(items * size, sizeof (long));
	if (nbytes > (DECOMP_BUFSIZE - filep->fi_dcscrused)) {
		ptr = bkmem_alloc(nbytes);
	} else {
		ptr = &filep->fi_dcscrbuf[filep->fi_dcscrused];
		filep->fi_dcscrused += nbytes;
	}
	bzero(ptr, nbytes);
	return (ptr);
}

/*
 * Decompression scratch memory free routine, does nothing since we free
 * the entire scratch area all at once on file close.
 */
void
cf_free(void *opaque __unused, void *addr __unused)
{
}

/*
 * Read the first block of the file described by filep and determine if
 * the file is gzip-compressed.  If so, the compressed flag will be set
 * in the fileid_t struct pointed to by filep and it will be initialized
 * for doing decompression on reads to the file.
 */
int
cf_check_compressed(fileid_t *filep)
{
	unsigned char *filebytes;
	z_stream *zsp;

	/*
	 * checking for a dcfs compressed file first would involve:
	 *
	 *	if (filep->fi_inode->i_cflags & ICOMPRESS)
	 *		filep->fi_flags |= FI_COMPRESSED;
	 */

	/*
	 * If the file is not long enough to check for a
	 * decompression header then return not compressed.
	 */
	if (filep->fi_inode->i_size < 3)
		return (0);
	filep->fi_offset = 0;
	if ((filep->fi_getblock)(filep) == -1)
		return (-1);
	filep->fi_offset = 0;
	filep->fi_count = 0;
	filep->fi_cfoff = 0;
	filebytes = (unsigned char *)filep->fi_memp;
	if (filebytes[0] != GZIP_ID_BYTE_1 ||
	    filebytes[1] != GZIP_ID_BYTE_2 ||
	    filebytes[2] != GZIP_CM_DEFLATE)
		return (0); /* not compressed */
	filep->fi_flags |= FI_COMPRESSED;

	if (cf_debug)
		kobj_printf("file %s is compressed\n", filep->fi_path);

	/*
	 * Allocate decompress scratch buffer
	 */
	if (free_dcomp_bufs) {
		filep->fi_dcscrbuf = scratch_bufs[--free_dcomp_bufs];
	} else {
		filep->fi_dcscrbuf = bkmem_alloc(DECOMP_BUFSIZE);
		decomp_bufcnt++;
	}
	filep->fi_dcscrused = 0;
	zsp = bkmem_alloc(sizeof (*zsp));
	filep->fi_dcstream = zsp;
	/*
	 * Initialize the decompression stream. Adding 16 to the window size
	 * indicates that zlib should expect a gzip header.
	 */
	bzero(zsp, sizeof (*zsp));
	zsp->opaque = filep;
	zsp->zalloc = cf_alloc;
	zsp->zfree = cf_free;
	zsp->avail_in = 0;
	zsp->next_in = NULL;
	zsp->avail_out = 0;
	zsp->next_out = NULL;
	if (inflateInit2(zsp, MAX_WBITS | 0x20) != Z_OK) {
		if (cf_debug)
			kobj_printf("inflateInit2() failed\n");
		return (-1);
	}
	return (0);
}

/*
 * If the file described by fileid_t struct at *filep is compressed
 * free any resources associated with the decompression.  (decompression
 * buffer, etc.).
 */
void
cf_close(fileid_t *filep)
{
	if ((filep->fi_flags & FI_COMPRESSED) == 0)
		return;
	if (cf_debug)
		kobj_printf("cf_close: %s\n", filep->fi_path);
	(void) inflateEnd(filep->fi_dcstream);
	bkmem_free(filep->fi_dcstream, sizeof (z_stream));
	if (free_dcomp_bufs == MAX_DECOMP_BUFS) {
		bkmem_free(filep->fi_dcscrbuf, DECOMP_BUFSIZE);
	} else {
		scratch_bufs[free_dcomp_bufs++] = filep->fi_dcscrbuf;
	}
}

void
cf_rewind(fileid_t *filep)
{
	z_stream *zsp;

	if (cf_debug)
		kobj_printf("cf_rewind: %s\n", filep->fi_path);
	zsp = filep->fi_dcstream;
	zsp->avail_in = 0;
	zsp->next_in = NULL;
	(void) inflateReset(zsp);
	filep->fi_cfoff = 0;
}

#define	FLG_FHCRC	0x02	/* crc field present */
#define	FLG_FEXTRA	0x04	/* "extra" field present */
#define	FLG_FNAME	0x08	/* file name field present */
#define	FLG_FCOMMENT	0x10	/* comment field present */

/*
 * Read at the current uncompressed offset from the compressed file described
 * by *filep.  Will return decompressed data.
 */
int
cf_read(fileid_t *filep, caddr_t buf, size_t count)
{
	z_stream *zsp;
	struct inode *ip;
	int err = Z_OK;
	int infbytes;
	off_t soff;
	caddr_t smemp;

	if (cf_debug)
		kobj_printf("cf_read: %s %lx bytes\n", filep->fi_path, count);
	zsp = filep->fi_dcstream;
	ip = filep->fi_inode;
	if (cf_debug)
		kobj_printf("   reading at offset %lx\n", zsp->total_out);
	zsp->next_out = (unsigned char *)buf;
	zsp->avail_out = count;
	while (zsp->avail_out != 0) {
		if (zsp->avail_in == 0 && filep->fi_cfoff < ip->i_size) {
			/*
			 * read a block of the file to inflate
			 */
			soff = filep->fi_offset;
			smemp = filep->fi_memp;
			filep->fi_memp = NULL;
			filep->fi_offset = filep->fi_cfoff;
			filep->fi_count = 0;
			if ((*filep->fi_getblock)(filep) == -1)
				return (-1);
			filep->fi_offset = soff;
			zsp->next_in = (unsigned char *)filep->fi_memp;
			zsp->avail_in = filep->fi_count;
			filep->fi_memp = smemp;
			filep->fi_cfoff += filep->fi_count;
		}
		infbytes = zsp->avail_out;
		if (cf_debug) {
			kobj_printf("attempting inflate of %x bytes to "
			    "buf at: %lx\n",
			    zsp->avail_out, (unsigned long)zsp->next_out);
		}
		err = inflate(zsp, Z_NO_FLUSH);
		infbytes -= zsp->avail_out;
		if (cf_debug) {
			kobj_printf("inflated %x bytes, errcode=%d\n",
			    infbytes, err);
		}
		/*
		 * break out if we hit end of the compressed file
		 * or the end of the compressed byte stream
		 */
		if (filep->fi_cfoff >= ip->i_size || err == Z_STREAM_END)
			break;
	}
	if (cf_debug) {
		kobj_printf("cf_read: returned %lx bytes\n",
		    count - zsp->avail_out);
	}
	return (count - zsp->avail_out);
}

/*
 * Seek to the location specified by addr
 */
void
cf_seek(fileid_t *filep, off_t addr, int whence)
{
	z_stream *zsp;
	int readsz;

	if (cf_debug)
		kobj_printf("cf_seek: %s to %lx\n", filep->fi_path, addr);
	zsp = filep->fi_dcstream;
	if (whence == SEEK_CUR)
		addr += zsp->total_out;
	/*
	 * To seek backwards, must rewind and seek forwards
	 */
	if (addr < zsp->total_out) {
		cf_rewind(filep);
		filep->fi_offset = 0;
	} else {
		addr -= zsp->total_out;
	}
	while (addr > 0) {
		readsz = MIN(addr, SEEKBUFSIZE);
		(void) cf_read(filep, seek_scrbuf, readsz);
		addr -= readsz;
	}
}
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 * Copyright 2025 MNX Cloud, Inc.
 */

/*
 * Basic file system reading code for standalone I/O system.
 * Simulates a primitive UNIX I/O system (read(), write(), open(), etc).
 * Does not support writes.
 */

#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/vnode.h>
#include <sys/fs/ufs_fsdir.h>
#include <sys/fs/ufs_fs.h>
#include <sys/fs/ufs_inode.h>

#include <sys/fs/hsfs_spec.h>
#include <sys/fs/hsfs_isospec.h>
#include <sys/fs/hsfs_node.h>
#include <sys/fs/hsfs_susp.h>
#include <sys/fs/hsfs_rrip.h>
#include <sys/bootvfs.h>
#include <sys/kobj.h>
#include <sys/filep.h>
#include <sys/sunddi.h>

#define	hdbtodb(n)	((ISO_SECTOR_SIZE / DEV_BSIZE) * (n))

#define	HSFS_NUM_SIG    14

#define	SUSP_SP_IX	0
#define	SUSP_CE_IX	1
#define	SUSP_PD_IX	2
#define	SUSP_ST_IX	3
#define	SUSP_ER_IX	4
#define	RRIP_PX_IX	5
#define	RRIP_PN_IX	6
#define	RRIP_SL_IX	7
#define	RRIP_CL_IX	8
#define	RRIP_PL_IX	9
#define	RRIP_RE_IX	10
#define	RRIP_RF_IX	11
#define	RRIP_RR_IX	12
#define	RRIP_NM_IX	13

extern int bootrd_debug;
extern void *bkmem_alloc(size_t);
extern void bkmem_free(void *, size_t);
extern int cf_check_compressed(fileid_t *);
extern void cf_close(fileid_t *);
extern void cf_seek(fileid_t *, off_t, int);
extern int cf_read(fileid_t *, caddr_t, size_t);

struct dirstuff {
	int loc;
	fileid_t *filep;
};

struct hs_direct {
    struct	direct  hs_ufs_dir;
    struct	hs_direntry hs_dir;
};

static uint_t root_ino = 0;
static struct hs_volume *hsfsp;
static fileid_t *head;

static char *hsfs_sig_tab[] = {
	SUSP_SP,
	SUSP_CE,
	SUSP_PD,
	SUSP_ST,
	SUSP_ER,
	RRIP_PX,
	RRIP_PN,
	RRIP_SL,
	RRIP_CL,
	RRIP_PL,
	RRIP_RE,
	RRIP_TF,
	RRIP_RR,
	RRIP_NM
};

static int hsfs_num_sig = sizeof (hsfs_sig_tab) / sizeof (hsfs_sig_tab[0]);

/*
 *  Local prototypes
 */
static struct hs_direct *readdir(struct dirstuff *);
static uint_t parse_dir(fileid_t *, int, struct hs_direct *);
static uint_t parse_susp(char *, uint_t *, struct hs_direct *);
static ino_t dlook(char *, fileid_t *);
static int opendir(ino_t, fileid_t *);
static ino_t find(char *, fileid_t *);

static int bhsfs_mountroot(char *str);
static int bhsfs_unmountroot(void);
static int bhsfs_open(char *str, int flags);
static int bhsfs_close(int fd);
static void bhsfs_closeall(void);
static ssize_t bhsfs_read(int fdesc, char *buf, size_t count);
static off_t bhsfs_lseek(int fdesc, off_t addr, int whence);
static int bhsfs_fstat(int fdesc, struct bootstat *stp);

static fileid_t *
find_fp(int fd)
{
	fileid_t *filep = head;

	if (fd >= 0) {
		while ((filep = filep->fi_forw) != head)
			if (fd == filep->fi_filedes)
				return (filep->fi_taken ? filep : 0);
	}

	return (0);
}

static int
opendir(ino_t inode, fileid_t *filep)
{
	struct hs_direct hsdep;

	if (bootrd_debug)
		kobj_printf("opendir: inode = %ld\n", inode);
	/* Set up the IO request */
	filep->fi_offset = 0;
	filep->fi_blocknum = hdbtodb(inode);
	filep->fi_count = ISO_SECTOR_SIZE;
	filep->fi_memp = 0;

	if (diskread(filep))
		return (0);

	filep->fi_offset = 0;
	filep->fi_blocknum = hdbtodb(inode);

	if (inode != root_ino)
		return (0);

	if (parse_dir(filep, 0, &hsdep) > 0) {
		struct inode *ip;

		ip = filep->fi_inode;
		if (ip == NULL)
			ip = filep->fi_inode = bkmem_alloc(sizeof (*ip));

		ip->i_size = hsdep.hs_dir.ext_size;
		ip->i_smode = hsdep.hs_dir.mode;
		ip->i_number = inode;
		return (0);
	}
	return (1);
}

static ino_t
find(char *path, fileid_t *filep)
{
	char *q;
	char c;
	ino_t n;

	n = 0;
	if (bootrd_debug)
		kobj_printf("find: %s\n", path);
	if (path == NULL || *path == '\0')
		return (0);

	if (opendir(root_ino, filep))
		return (0);

	while (*path) {
		while (*path == '/')
			path++;
		q = path;
		while (*q != '/' && *q != '\0')
			q++;
		c = *q;
		*q = '\0';
		n = dlook(path, filep);
		*q = c;
		path = q;

		if (n != 0) {
			if (c == '\0')
				break;
			if (opendir(n, filep))
				return (0);
			continue;
		} else {
			return (0);
		}
	}
	return ((ino_t)n);
}

static ino_t
dlook(char *s, fileid_t *filep)
{
	struct hs_direct *hsdep;
	struct direct *udp;
	struct inode *ip;
	struct dirstuff dirp;
	int len;

	if (bootrd_debug)
		kobj_printf("dlook: %s\n", s);
	ip = filep->fi_inode;
	if (s == NULL || *s == '\0')
		return (0);
	if ((ip->i_smode & IFMT) != IFDIR) {
		return (0);
	}
	if (ip->i_size == 0) {
		return (0);
	}
	len = strlen(s);
	dirp.loc = 0;
	dirp.filep = filep;
	for (hsdep = readdir(&dirp); hsdep != NULL; hsdep = readdir(&dirp)) {
		udp = &hsdep->hs_ufs_dir;
		if (udp->d_namlen == 1 &&
		    udp->d_name[0] == '.' &&
		    udp->d_name[1] == '\0')
			continue;
		if (udp->d_namlen == 2 &&
		    udp->d_name[0] == '.' &&
		    udp->d_name[1] == '.' &&
		    udp->d_name[2] == '\0')
			continue;
		if (udp->d_namlen == len && (strcmp(s, udp->d_name)) == 0) {
			struct inode *ip = filep->fi_inode;

			filep->fi_offset = 0;
			filep->fi_blocknum = hdbtodb(udp->d_ino);

			bzero(filep->fi_inode, sizeof (struct inode));
			ip->i_size = hsdep->hs_dir.ext_size;
			ip->i_smode = hsdep->hs_dir.mode;
			ip->i_number = udp->d_ino;
			return (udp->d_ino);
		}
	}
	return (0);
}

/*
 * get next entry in a directory.
 */
static struct hs_direct *
readdir(struct dirstuff *dirp)
{
	static struct hs_direct hsdep;
	struct direct *udp = &hsdep.hs_ufs_dir;
	struct inode *ip;
	fileid_t *filep;
	daddr_t lbn;
	int off;

	if (bootrd_debug)
		kobj_printf("readdir: start\n");
	filep = dirp->filep;
	ip = filep->fi_inode;
	for (;;) {
		if (dirp->loc >= ip->i_size) {
			return (NULL);
		}
		off = dirp->loc & ((1 << ISO_SECTOR_SHIFT) - 1);
		if (off == 0) {
			lbn = hdbtodb(dirp->loc >> ISO_SECTOR_SHIFT);
			filep->fi_blocknum = lbn + hdbtodb(ip->i_number);
			filep->fi_count = ISO_SECTOR_SIZE;
			filep->fi_memp = 0;
			if (diskread(filep)) {
				if (bootrd_debug) {
					kobj_printf(
					    "readdir: diskread failed\n");
				}
				return (NULL);
			}
		}
		dirp->loc += parse_dir(filep, off, &hsdep);
		if (udp->d_reclen == 0 && dirp->loc <= ip->i_size) {
			dirp->loc = roundup(dirp->loc, ISO_SECTOR_SIZE);
			continue;
		}
		return (&hsdep);
	}
}

static int
getblock(fileid_t *filep)
{
	struct inode *ip = filep->fi_inode;
	int off, size, diff;
	daddr_t lbn;

	if (bootrd_debug)
		kobj_printf("getblock: start\n");
	diff = ip->i_size - filep->fi_offset;
	if (diff <= 0)
		return (-1);

	/* which block (or frag) in the file do we read? */
	lbn = hdbtodb(filep->fi_offset >> ISO_SECTOR_SHIFT);
	filep->fi_blocknum = lbn + hdbtodb(ip->i_number);

	off = filep->fi_offset & ((1 << ISO_SECTOR_SHIFT) - 1);
	size = filep->fi_count = ISO_SECTOR_SIZE;
	filep->fi_memp = 0;
	if (diskread(filep))	/* Trap errors */
		return (-1);

	if (filep->fi_offset - off + size >= ip->i_size)
		filep->fi_count = diff + off;
	filep->fi_count -= off;
	filep->fi_memp += off;
	if (bootrd_debug)
		kobj_printf("getblock: end\n");
	return (0);
}

static ssize_t
bhsfs_read(int fd, caddr_t buf, size_t count)
{
	int i, j;
	fileid_t *filep;
	struct inode *ip;
	caddr_t n;

	if (bootrd_debug)
		kobj_printf("bhsfs_read %d, count 0x%lx\n", fd, count);
	filep = find_fp(fd);
	if (filep == NULL)
		return (-1);

	ip = filep->fi_inode;
	n = buf;
	if ((filep->fi_flags & FI_COMPRESSED) == 0 &&
	    filep->fi_offset + count > ip->i_size)
		count = ip->i_size - filep->fi_offset;

	if ((i = count) <= 0)
		return (0);

	while (i > 0) {
		if (filep->fi_flags & FI_COMPRESSED) {
			if ((j = cf_read(filep, buf, count)) < 0)
				return (0); /* encountered an error */
			if (j < i)
				i = j; /* short read, must have hit EOF */
		} else {
			if (filep->fi_count == 0) {
				if (getblock(filep) == -1)
					return (0);
			}
			j = MIN(i, filep->fi_count);
			bcopy(filep->fi_memp, buf, (uint_t)j);
		}
		filep->fi_memp += j;
		filep->fi_offset += j;
		filep->fi_count -= j;
		buf += j;
		i -= j;
	}

	if (bootrd_debug)
		kobj_printf("bhsfs_read: read 0x%x\n", (int)(buf - n));
	return (buf - n);
}

static int
bhsfs_mountroot(char *str __unused)
{
	char *bufp;

	if (hsfsp != NULL)
		return (0);	/* already mounted */

	if (bootrd_debug)
		kobj_printf("mounting ramdisk as hsfs\n");

	hsfsp = bkmem_alloc(sizeof (*hsfsp));
	bzero(hsfsp, sizeof (*hsfsp));
	head = bkmem_alloc(sizeof (*head));
	bzero(head, sizeof (*head));
	head->fi_back = head->fi_forw = head;

	/* now read the superblock. */
	head->fi_blocknum = hdbtodb(ISO_VOLDESC_SEC);
	head->fi_offset = 0;
	head->fi_count = ISO_SECTOR_SIZE;
	head->fi_memp = head->fi_buf;
	if (diskread(head)) {
		kobj_printf("failed to read superblock\n");
		bhsfs_closeall();
		return (-1);
	}

	/* Since RRIP is based on ISO9660, that's where we start */
	bufp = head->fi_buf;
	if ((ISO_DESC_TYPE(bufp) != ISO_VD_PVD) ||
	    (strncmp((const char *)ISO_std_id(bufp), ISO_ID_STRING,
	    ISO_ID_STRLEN) != 0) || (ISO_STD_VER(bufp) != ISO_ID_VER)) {
		if (bootrd_debug)
			kobj_printf("volume type does not match\n");
		bhsfs_closeall();
		return (-1);
	}

	/* Now we fill in the volume descriptor */
	hsfsp->vol_size = ISO_VOL_SIZE(bufp);
	hsfsp->lbn_size = ISO_BLK_SIZE(bufp);
	hsfsp->lbn_shift = ISO_SECTOR_SHIFT;
	hsfsp->lbn_secshift = ISO_SECTOR_SHIFT;
	hsfsp->vol_set_size = (ushort_t)ISO_SET_SIZE(bufp);
	hsfsp->vol_set_seq = (ushort_t)ISO_SET_SEQ(bufp);

	/* Make sure we have a valid logical block size */
	if (hsfsp->lbn_size & ~(1 << hsfsp->lbn_shift)) {
		kobj_printf("%d invalid logical block size\n", hsfsp->lbn_size);
		bhsfs_closeall();
		return (-1);
	}

	/* Since an HSFS root could be located anywhere on the media! */
	root_ino = IDE_EXT_LBN(ISO_root_dir(bufp));
	return (0);
}

static int
bhsfs_unmountroot(void)
{
	if (hsfsp == NULL)
		return (-1);

	bhsfs_closeall();

	return (0);
}

/*
 * Open a file.
 */
int
bhsfs_open(char *str, int flags __unused)
{
	static int filedes = 1;

	fileid_t *filep;
	ino_t ino;

	if (bootrd_debug)
		kobj_printf("open %s\n", str);
	filep = (fileid_t *)bkmem_alloc(sizeof (fileid_t));
	filep->fi_back = head->fi_back;
	filep->fi_forw = head;
	head->fi_back->fi_forw = filep;
	head->fi_back = filep;
	filep->fi_filedes = filedes++;
	filep->fi_taken = 1;
	filep->fi_path = (char *)bkmem_alloc(strlen(str) + 1);
	(void) strcpy(filep->fi_path, str);
	filep->fi_inode = NULL;
	bzero(filep->fi_buf, MAXBSIZE);
	filep->fi_getblock = getblock;
	filep->fi_flags = 0;

	ino = find(str, filep);
	if (ino == 0) {
		(void) bhsfs_close(filep->fi_filedes);
		return (-1);
	}

	filep->fi_blocknum = hdbtodb(ino);
	filep->fi_offset = 0;
	filep->fi_count = 0;
	filep->fi_memp = 0;

	if (cf_check_compressed(filep) != 0)
		return (-1);
	if (bootrd_debug)
		kobj_printf("open done\n");
	return (filep->fi_filedes);
}

int
bhsfs_close(int fd)
{
	fileid_t *filep;

	if (bootrd_debug)
		kobj_printf("close %d\n", fd);
	if (!(filep = find_fp(fd)))
		return (-1);

	if (filep->fi_taken == 0 || filep == head) {
		kobj_printf("File descripter %d not allocated!\n", fd);
		return (-1);
	}

	cf_close(filep);
	/* unlink and deallocate node */
	filep->fi_forw->fi_back = filep->fi_back;
	filep->fi_back->fi_forw = filep->fi_forw;
	if (filep->fi_inode)
		bkmem_free(filep->fi_inode, sizeof (struct inode));
	bkmem_free(filep->fi_path, strlen(filep->fi_path) + 1);
	bkmem_free((char *)filep, sizeof (fileid_t));
	if (bootrd_debug)
		kobj_printf("close done\n");
	return (0);
}

static void
bhsfs_closeall(void)
{
	fileid_t *filep;

	while ((filep = head->fi_forw) != head)
		if (filep->fi_taken && bhsfs_close(filep->fi_filedes))
			kobj_printf("Filesystem may be inconsistent.\n");

	bkmem_free(hsfsp, sizeof (*hsfsp));
	bkmem_free(head, sizeof (fileid_t));
	hsfsp = NULL;
	head = NULL;
}

/*
 * This version of seek() only performs absolute seeks (whence == 0).
 */
static off_t
bhsfs_lseek(int fd, off_t addr, int whence)
{
	fileid_t *filep;

	if (bootrd_debug)
		kobj_printf("lseek %d, off = %lx\n", fd, addr);
	if (!(filep = find_fp(fd)))
		return (-1);

	if (filep->fi_flags & FI_COMPRESSED) {
		cf_seek(filep, addr, whence);
	} else {
		switch (whence) {
		case SEEK_CUR:
			filep->fi_offset += addr;
			break;
		case SEEK_SET:
			filep->fi_offset = addr;
			break;
		default:
		case SEEK_END:
			kobj_printf("lseek(): invalid whence value %d\n",
			    whence);
			break;
		}
		filep->fi_blocknum = addr / DEV_BSIZE;
	}

	filep->fi_count = 0;
	return (0);
}

static int
bhsfs_fstat(int fd, struct bootstat *stp)
{
	fileid_t	*filep;
	struct inode	*ip;

	if (!(filep = find_fp(fd)))
		return (-1);

	ip = filep->fi_inode;

	stp->st_mode = 0;
	stp->st_size = 0;

	if (ip == NULL)
		return (0);

	switch (ip->i_smode & IFMT) {
	case IFDIR:
		stp->st_mode = S_IFDIR;
		break;
	case IFREG:
		stp->st_mode = S_IFREG;
		break;
	default:
		break;
	}
	/*
	 * NOTE: this size will be the compressed size for a compressed file
	 * This could confuse the caller since we decompress the file behind
	 * the scenes when the file is read.
	 */
	stp->st_size = ip->i_size;

	/* file times */
	stp->st_atim.tv_sec = ip->i_atime.tv_sec;
	stp->st_atim.tv_nsec = ip->i_atime.tv_usec * 1000;
	stp->st_mtim.tv_sec = ip->i_mtime.tv_sec;
	stp->st_mtim.tv_nsec = ip->i_mtime.tv_usec * 1000;
	stp->st_ctim.tv_sec = ip->i_ctime.tv_sec;
	stp->st_ctim.tv_nsec = ip->i_ctime.tv_usec * 1000;

	return (0);

}


/*
 * Parse a directory entry.
 *
 */
static uint_t
parse_dir(fileid_t *filep, int offset, struct hs_direct *hsdep)
{
	char *bufp = (char *)(filep->fi_memp + offset);
	struct direct *udp = &hsdep->hs_ufs_dir;  /* ufs-style dir info */
	struct hs_direntry *hdp = &hsdep->hs_dir; /* hsfs-style dir info */
	uint_t ce_lbn;
	uint_t ce_len;
	uint_t nmlen;
	uint_t i;
	uchar_t c;

	if (bootrd_debug)
		kobj_printf("parse_dir: offset = %d\n", offset);
	/* a zero length dir entry terminates the dir block */
	udp->d_reclen = IDE_DIR_LEN(bufp);
	if (udp->d_reclen == 0)
		return (0);

	/* fill in some basic hsfs info */
	hdp->ext_lbn  = IDE_EXT_LBN(bufp);
	hdp->ext_size = IDE_EXT_SIZE(bufp);
	hdp->xar_len  = IDE_XAR_LEN(bufp);
	hdp->intlf_sz = IDE_INTRLV_SIZE(bufp);
	hdp->intlf_sk = IDE_INTRLV_SKIP(bufp);
	hdp->sym_link = NULL;

	/* we use lbn of data extent as an inode # equivalent */
	udp->d_ino	= hdp->ext_lbn;

	c = IDE_FLAGS(bufp);
	if (IDE_REGULAR_FILE(c)) {
		hdp->type = VREG;
		hdp->mode = IFREG;
		hdp->nlink = 1;
	} else if (IDE_REGULAR_DIR(c)) {
		hdp->type = VDIR;
		hdp->mode = IFDIR;
		hdp->nlink = 2;
	} else {
		kobj_printf("pd(): file type=0x%x unknown.\n", c);
	}

	/*
	 * Massage hsfs name, recognizing special entries for . and ..
	 * else lopping off version junk.
	 */

	/* Some initial conditions */
	nmlen = IDE_NAME_LEN(bufp);
	c = *IDE_NAME(bufp);
	/* Special Case: Current Directory */
	if (nmlen == 1 && c == '\0') {
		udp->d_name[0] = '.';
		udp->d_name[1] = '\0';
		udp->d_namlen = 1;
	/* Special Case: Parent Directory */
	} else if (nmlen == 1 && c == '\001') {
		udp->d_name[0] = '.';
		udp->d_name[1] = '.';
		udp->d_name[2] = '\0';
		udp->d_namlen = 2;
	/* Other file name */
	} else {
		udp->d_namlen = 0;
		for (i = 0; i < nmlen; i++) {
			c = *(IDE_name(bufp)+i);
			if (c == ';')
				break;
			else if (c == ' ')
				continue;
			else
				udp->d_name[udp->d_namlen++] = c;
		}
		udp->d_name[udp->d_namlen] = '\0';
	}

	/* System Use Fields */
	ce_len = IDE_SUA_LEN(bufp);

	if (ce_len == 0)
		return (udp->d_reclen);

	/* there is an SUA for this dir entry; go parse it */
	ce_lbn = parse_susp((char *)IDE_sys_use_area(bufp), &ce_len, hsdep);

	if (ce_lbn) {
		/*
		 * store away current position in dir,
		 * as we will be using the iobuf to reading SUA.
		 */
		daddr_t save_bn = filep->fi_blocknum;
		daddr_t save_offset = filep->fi_offset;
		caddr_t save_ma = filep->fi_memp;
		int save_cc = filep->fi_count;
		do {
			filep->fi_count = ISO_SECTOR_SIZE;
			filep->fi_offset = 0;
			filep->fi_blocknum = hdbtodb(ce_lbn);
			filep->fi_memp = 0;
			if (diskread(filep)) {
				kobj_printf("failed to read cont. area\n");
				ce_len = 0;
				ce_lbn = 0;
				break;
			}
			ce_lbn = parse_susp(filep->fi_memp, &ce_len,
			    hsdep);
		} while (ce_lbn);
		filep->fi_count = save_cc;
		filep->fi_offset = save_offset;
		filep->fi_blocknum = save_bn;
		filep->fi_memp = save_ma;
	}
	return (udp->d_reclen);
}

/*
 * Parse the System Use Fields in this System Use Area.
 * Return blk number of continuation/SUA, or 0 if no continuation/not a SUA.
 */
static uint_t
parse_susp(char *bufp, uint_t *len, struct hs_direct *hsdep)
{
	struct direct *udp = &hsdep->hs_ufs_dir; /* ufs-style info */
	char *susp;
	uint_t cur_off = 0;
	uint_t blk_len = *len;
	uint_t susp_len = 0;
	uint_t ce_lbn = 0;
	uint_t i;

	if (bootrd_debug)
		kobj_printf("parse_susp: len = %d\n", *len);
	while (cur_off < blk_len) {
		susp = (char *)(bufp + cur_off);

		/*
		 * A null entry, or an entry with zero length
		 * terminates the SUSP.
		 */
		if (susp[0] == '\0' || susp[1] == '\0' ||
		    (susp_len = SUF_LEN(susp)) == 0)
			break;

		/*
		 * Compare current entry to all known signatures.
		 */
		for (i = 0; i < hsfs_num_sig; i++)
			if (strncmp(hsfs_sig_tab[i], susp, SUF_SIG_LEN) == 0)
				break;
		switch (i) {
		case SUSP_CE_IX:
			/*
			 * CE signature: continuation of SUSP.
			 * will want to return new lbn, len.
			 */
			ce_lbn = CE_BLK_LOC(susp);
			*len = CE_CONT_LEN(susp);
			break;
		case RRIP_NM_IX:
			/* NM signature: POSIX-style file name */
			if (!RRIP_NAME_FLAGS(susp)) {
				udp->d_namlen = RRIP_NAME_LEN(susp);
				bcopy((char *)RRIP_name(susp),
				    udp->d_name, udp->d_namlen);
				udp->d_name[udp->d_namlen] = '\0';
			}
			break;
		case HSFS_NUM_SIG:
			/* couldn't find a legit susp, terminate loop */
		case SUSP_ST_IX:
			/* ST signature: terminates SUSP */
			return (ce_lbn);
		case SUSP_SP_IX:
		case RRIP_RR_IX:
		default:
			break;
		}
		cur_off += susp_len;
	}
	return (ce_lbn);
}

struct boot_fs_ops bhsfs_ops = {
	"boot_hsfs",
	bhsfs_mountroot,
	bhsfs_unmountroot,
	bhsfs_open,
	bhsfs_close,
	bhsfs_read,
	bhsfs_lseek,
	bhsfs_fstat,
	NULL
};
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License, Version 1.0 only
 * (the "License").  You may not use this file except in compliance
 * with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

#ifndef _PCFILEP_H
#define	_PCFILEP_H

#ifdef	__cplusplus
extern "C" {
#endif

#define	MAX_DOSMOUNT_RETRIES 3

#define	TICKS_PER_SEC	18		/* It's really 18.2! */
#define	SECSIZ 512
#define	fat_bpc(i) (pi[(i)]->f_bpb.bs_spc * SECSIZ)

/*
 * Access permissions for dosAccess(), dosOpen()
 * NOTE: These permission need to match those for the DOS compiler.
 */
#define	FILE_EXISTS	1
#define	FILE_READ	0x0000
#define	FILE_WRITE	0x0001
#define	FILE_RDWR	0x0002
#define	FILE_APPEND	0x0008
#define	FILE_CREATE	0x0100
#define	FILE_TRUNC	0x0200

#define	TYPE_EMPTY	0x00		/* undefined partition */
#define	TYPE_DOS	0x13		/* causes fatInit() to search for */
					/* active partition */
#define	TYPE_DOS_12	0x01		/* partition with FAT12 filesys */
#define	TYPE_DOS_16	0x04		/* partition with FAT16 filesys */
#define	TYPE_DOS_EXT	0x05		/* not bootable, ignore */
#define	TYPE_HUGH	0x06		/* HUGH partition */
#define	TYPE_COMPAQ	0x12		/* Compaq's diag partition */
#define	TYPE_SOLARIS	0x82
#define	TYPE_SOLARIS_BOOT	0xBE	/* For "boot hill" project */

#define	FDISK_START	0x1be		/* location in first sector where */
					/* the fdisk starts. */

#define	FDISK_PARTS	4		/* Number of partitions in a fdisk */
#define	FDISK_ACTIVE	0x80		/* indicates partition is active */
#define	FDISK_INACTIVE	0x00		/*  " partition inactive */

#pragma pack(1)
struct _fdisk_partition_ {
	uchar_t	fd_active;
	uchar_t	fd_b_head;
	uchar_t	fd_b_sec;
	uchar_t	fd_b_cyl;
	uchar_t	fd_type;
	uchar_t	fd_e_head;
	uchar_t	fd_e_sec;
	uchar_t	fd_e_cyl;
	union {
		long	fd_start_sec_long;
		struct {
			ushort_t low;
			ushort_t high;
		} s;
	} u;
	long	fd_part_len;
};
#define	fd_start_sec u.fd_start_sec_long
#define	fd_partition fd_type
typedef struct _fdisk_partition_ _fdisk_t, *_fdisk_p;
#pragma pack()

#pragma pack(1)
struct _boot_sector_ {
	uchar_t	bs_jump_code[3];
	uchar_t	bs_oem_name[8];
	uchar_t	bs_bytes_sector[2];
	uchar_t	bs_spc;			/* ... sectors per cluster */
	uchar_t	bs_resv_sectors[2];
	uchar_t	bs_num_fats;
	uchar_t	bs_num_root_entries[2];
	uchar_t	bs_siv[2];			/* ... sectors in volume */
	uchar_t	bs_media;
	uchar_t	bs_spf[2];			/* ... sectors per fat */
	uchar_t	bs_sectors_per_track[2];
	uchar_t	bs_heads[2];
	/*
	 * Byte offset at this point is 28 so we can declare the next
	 * variable with the correct type and not worry about alignment.
	 */
	long	bs_hidden_sectors;
	long	bs_lsiv;		/* ... logical sectors in volume */
	uchar_t	bs_phys_drive_num;
	uchar_t	bs_reserved;
	uchar_t	bs_ext_signature;
	char	bs_volume_id[4];
	char	bs_volume_label[11];
	char	bs_type[8];

	/* ---- ADDED BY SUNSOFT FOR MDBOOT ---- */
	ushort_t	bs_offset_high;
	ushort_t	bs_offset_low;
};
#pragma pack()
typedef struct _boot_sector_  _boot_sector_t, *_boot_sector_p;

/*
 * Cluster types
 */
#define	CLUSTER_AVAIL	0x00
#define	CLUSTER_RES_12_0	0x0ff0	/* 12bit fat, first reserved */
#define	CLUSTER_RES_12_6	0x0ff6	/* 12bit fat, last reserved */
#define	CLUSTER_RES_16_0	0xfff0	/* 16bit fat, first reserved */
#define	CLUSTER_RES_16_6	0xfff6	/* 16bit fat, last reserved */
#define	CLUSTER_BAD_12	0x0ff7	/* 12bit fat, bad entry */
#define	CLUSTER_BAD_16	0xfff7	/* 16bit fat, bad entry */
#define	CLUSTER_EOF		CLUSTER_EOF_16_0
#define	CLUSTER_MAX_12		0x0ff7	/* max clusters for 12bit fat */
#define	CLUSTER_EOF_12_0	0x0ff8	/* 12bit fat, EOF first entry */
#define	CLUSTER_EOF_12_8	0x0fff	/* 12bit fat, EOF last entry */
#define	CLUSTER_EOF_16_0	0xfff8	/* 16bit fat, EOF first entry */
#define	CLUSTER_EOF_16_8	0xffff	/* 16bit fat, EOF last entry */

/*
 * Cluster operations for allocation
 */
#define	CLUSTER_NOOP		0x0001	/* ... just allocate cluster */
#define	CLUSTER_ZEROFILL	0x0002	/* ... zero fill the alloc'd cluster */

#define	CLUSTER_FIRST		0x0002	/* ... first cluster number to search */
#define	CLUSTER_ROOTDIR		0x0000	/* ... root dir's cluster number */

/*
 * This structure is filled in by initFAT()
 */
struct _fat_controller_ {
	union {
		_boot_sector_t	fu_bpb;	 /* boot parameter block */
		uchar_t		fu_sector[SECSIZ];
	} fu;
	long		f_adjust;	/* starting sec for part. */
	long		f_rootsec;	/* root dir starting sec. */
	long		f_rootlen;	/* length of root in sectors */
	long		f_filesec;	/* adjustment for clusters */
	long		f_dclust;	/* cur dir cluster */
	int		f_nxtfree;	/* next free cluster */
	int		f_ncluster;	/* number of cluster in part */
	char		f_16bit:1,	/* 1 if 16bit fat entries */
			f_flush:1;	/* flush the fat */
};
typedef struct _fat_controller_ _fat_controller_t, *_fat_controller_p;

#define	f_bpb fu.fu_bpb
#define	f_sector fu.fu_sector

#define	NAMESIZ		8
#define	EXTSIZ		3
#pragma pack(1)
struct _dir_entry_ {
	char	d_name[NAMESIZ];
	char	d_ext[EXTSIZ];
	uchar_t	d_attr;
	char	d_res[10];
	short	d_time;
	short	d_date;
	ushort_t d_cluster;
	long	d_size;
};
#pragma pack()
typedef struct _dir_entry_ _dir_entry_t, *_dir_entry_p;

/*
 * Number of entries in one sector
 */
#define	DIRENTS (SECSIZ / sizeof (_dir_entry_t))

/*
 * Directory entry attributes
 */
#define	DE_READONLY		0x01
#define	DE_HIDDEN		0x02
#define	DE_SYSTEM		0x04
#define	DE_LABEL		0x08
#define	DE_DIRECTORY		0x10
#define	DE_ARCHIVE		0x20
#define	DE_RESERVED1		0x40
#define	DE_RESERVED2		0x80

#define	DE_IS_LFN (DE_READONLY | DE_HIDDEN | DE_SYSTEM | DE_LABEL)

struct _file_descriptor_ {
	struct _file_descriptor_ *f_forw; /* link to next file descriptor */
	int	f_desc;			/* descriptor number */
	long	f_startclust;		/* starting cluster number */
	long	f_off;			/* current offset */
	long	f_len;			/* size of file */
	long	f_index;		/* index into directory block */
	uchar_t	f_attr;			/* attributes */
	int	f_volidx;		/* Volume device index */
	char	*f_volname;		/* Name of volume */
};
typedef struct _file_descriptor_ _file_desc_t, *_file_desc_p;

#ifdef	__cplusplus
}
#endif

#endif	/* _PCFILEP_H */
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License, Version 1.0 only
 * (the "License").  You may not use this file except in compliance
 * with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */

/*
 * Basic file system reading code for standalone I/O system.
 * Simulates a primitive UNIX I/O system (read(), write(), open(), etc).
 * Does not support writes.
 */

/*
 * WARNING:
 * This is currently used by installgrub for creating bootable floppy.
 * The special part is diskread_callback/fileread_callback for gathering
 * fileblock list.
 */

#include <sys/param.h>
#include <sys/sysmacros.h>
#include <sys/vnode.h>
#include <sys/fs/pc_label.h>
#include <sys/bootvfs.h>
#include <sys/filep.h>
#include "pcfilep.h"

#if	defined(_BOOT)
#include "../common/util.h"
#elif	defined(_KERNEL)
#include <sys/sunddi.h>
#else
#include <stdio.h>
#include <strings.h>
#include <ctype.h>
#endif

#if	defined(_BOOT)
#define	dprintf	if (bootrd_debug) printf
#elif	defined(_KERNEL)
#define	printf	kobj_printf
#define	dprintf	if (bootrd_debug) kobj_printf

/* PRINTLIKE */
extern void kobj_printf(char *, ...);
#else
#define	dprintf if (bootrd_debug) printf
#endif

#define	FI_STARTCLUST(fp)	(*(ushort_t *)(fp)->fi_buf)
#define	FI_LENGTH(fp)		(*(long *)((fp)->fi_buf + 4))

extern int bootrd_debug;
extern void *bkmem_alloc(size_t);
extern void bkmem_free(void *, size_t);

/*
 * NOTE: The fileread_callback is set by the calling program
 * during a file read. diskread_callback is set to fileread_callback
 * only if reading a file block. It needs to be NULL while reading
 * cluster blocks.
 */
extern int (*diskread_callback)(int, int);
extern int (*fileread_callback)(int, int);

/*
 *  Local prototypes
 */
static int lookuppn(char *, _dir_entry_p);
static fileid_t *find_fp(int);
static void *readblock(int, int);
static int fat_map(int, int);
static int cluster_valid(long, int);
static int fat_ctodb(int, int);

static int bpcfs_mountroot(char *str);
static int bpcfs_unmountroot(void);
static int bpcfs_open(char *str, int flags);
static int bpcfs_close(int fd);
static void bpcfs_closeall(void);
static ssize_t bpcfs_read(int fdesc, char *buf, size_t count);
static off_t bpcfs_lseek(int fdesc, off_t addr, int whence);

static fileid_t *head;
static _fat_controller_p pcfsp;

/* cache the cluster */
static int nsec_cache;
static int nsec_start;
static char *cluster_cache;

/*ARGSUSED*/
static int
bpcfs_mountroot(char *str)
{
	int ncluster;
	if (pcfsp != NULL)
		return (0);	/* already mounted */

	pcfsp = bkmem_alloc(sizeof (_fat_controller_t));
	head = (fileid_t *)bkmem_alloc(sizeof (fileid_t));
	head->fi_back = head->fi_forw = head;
	head->fi_filedes = 0;
	head->fi_taken = 0;

	/* read of first floppy sector */
	head->fi_blocknum = 0;
	head->fi_count = SECSIZ;
	head->fi_memp = (caddr_t)pcfsp->f_sector;
	if (diskread(head)) {
		printf("failed to read first sector\n");
		bkmem_free(pcfsp, sizeof (*pcfsp));
		pcfsp = NULL;
		return (-1);
	}

	if (pcfsp->f_bpb.bs_spc == 0) {
		printf("invalid bios paramet block\n");
		return (-1);
	}

	pcfsp->f_rootsec =
	    (pcfsp->f_bpb.bs_num_fats * ltohs(pcfsp->f_bpb.bs_spf)) +
	    ltohs(pcfsp->f_bpb.bs_resv_sectors);
	pcfsp->f_rootlen =
	    ltohs(pcfsp->f_bpb.bs_num_root_entries) *
	    sizeof (_dir_entry_t) / SECSIZ;
	pcfsp->f_adjust = 0;
	pcfsp->f_dclust = CLUSTER_ROOTDIR;
	pcfsp->f_filesec = pcfsp->f_rootsec + pcfsp->f_rootlen;
	pcfsp->f_nxtfree = CLUSTER_FIRST;

	/* figure out the number of clusters in this partition */
	ncluster = (((ulong_t)ltohs(pcfsp->f_bpb.bs_siv) ?
	    (ulong_t)ltohs(pcfsp->f_bpb.bs_siv) :
	    (ulong_t)ltohi(pcfsp->f_bpb.bs_siv)) -
	    pcfsp->f_filesec) / (ulong_t)pcfsp->f_bpb.bs_spc;
	pcfsp->f_16bit = ncluster >= CLUSTER_MAX_12;
	pcfsp->f_ncluster = ncluster;

	/* cache the cluster */
	if (pcfsp->f_16bit)
		nsec_cache = (((ncluster << 1) + 511) >> 9);
	else
		nsec_cache = (ncluster + ((ncluster + 1) >> 1) + 511) >> 9;
	cluster_cache = bkmem_alloc(nsec_cache * SECSIZ);
	if (cluster_cache == NULL) {
		printf("bpcfs_mountroot: out of memory\n");
		bkmem_free(pcfsp, sizeof (*pcfsp));
		pcfsp = NULL;
		return (-1);
	}

	head->fi_blocknum = nsec_start =
	    ltohs(pcfsp->f_bpb.bs_resv_sectors) + pcfsp->f_adjust;
	head->fi_count = nsec_cache * SECSIZ;
	head->fi_memp = cluster_cache;
	if (diskread(head)) {
		printf("bpcfs_mountroot: failed to read cluster\n");
		bkmem_free(pcfsp, sizeof (*pcfsp));
		pcfsp = NULL;
		return (-1);
	}
	dprintf("read cluster sectors %d starting at %d\n",
	    nsec_cache, nsec_start);
	return (0);
}

static int
bpcfs_unmountroot(void)
{
	if (pcfsp == NULL)
		return (-1);

	(void) bpcfs_closeall();

	return (0);
}

/*
 * Open a file.
 */
/*ARGSUSED*/
int
bpcfs_open(char *str, int flags)
{
	static int filedes = 1;

	fileid_t *filep;
	_dir_entry_t d;

	dprintf("open %s\n", str);
	filep = (fileid_t *)bkmem_alloc(sizeof (fileid_t));
	filep->fi_back = head->fi_back;
	filep->fi_forw = head;
	head->fi_back->fi_forw = filep;
	head->fi_back = filep;
	filep->fi_filedes = filedes++;
	filep->fi_taken = 1;
	filep->fi_path = (char *)bkmem_alloc(strlen(str) + 1);
	(void) strcpy(filep->fi_path, str);

	if (lookuppn(str, &d)) {
		(void) bpcfs_close(filep->fi_filedes);
		return (-1);
	}

	filep->fi_offset = 0;
	FI_STARTCLUST(filep) = d.d_cluster;
	FI_LENGTH(filep) = d.d_size;
	dprintf("file %s size = %ld\n", str, d.d_size);
	return (filep->fi_filedes);
}

int
bpcfs_close(int fd)
{
	fileid_t *filep;

	dprintf("close %d\n", fd);
	if (!(filep = find_fp(fd)))
		return (-1);

	if (filep->fi_taken == 0 || filep == head) {
		printf("File descripter %d no allocated!\n", fd);
		return (-1);
	}

	/* unlink and deallocate node */
	filep->fi_forw->fi_back = filep->fi_back;
	filep->fi_back->fi_forw = filep->fi_forw;
	bkmem_free(filep->fi_path, strlen(filep->fi_path) + 1);
	bkmem_free((char *)filep, sizeof (fileid_t));
	dprintf("close done\n");
	return (0);
}

static void
bpcfs_closeall(void)
{
	fileid_t *filep;

	while ((filep = head->fi_forw) != head)
		if (filep->fi_taken && bpcfs_close(filep->fi_filedes))
			printf("Filesystem may be inconsistent.\n");

	bkmem_free(pcfsp, sizeof (*pcfsp));
	bkmem_free(head, sizeof (fileid_t));
	pcfsp = NULL;
	head = NULL;
}

static ssize_t
bpcfs_read(int fd, caddr_t b, size_t c)
{
	ulong_t sector;
	uint_t count = 0, xfer, i;
	char *block;
	ulong_t off, blk;
	int rd, spc;
	fileid_t *fp;

	dprintf("bpcfs_read: fd = %d, buf = %p, size = %d\n",
		fd, (void *)b, c);
	fp = find_fp(fd);
	if (fp == NULL) {
		printf("invalid file descriptor %d\n", fd);
		return (-1);
	}

	spc = pcfsp->f_bpb.bs_spc;
	off = fp->fi_offset;
	blk = FI_STARTCLUST(fp);
	rd = blk == CLUSTER_ROOTDIR ? 1 : 0;

	spc = pcfsp->f_bpb.bs_spc;
	off = fp->fi_offset;
	blk = FI_STARTCLUST(fp);
	rd = (blk == CLUSTER_ROOTDIR) ? 1 : 0;

	if ((c = MIN(FI_LENGTH(fp) - off, c)) == 0)
		return (0);

	while (off >= pcfsp->f_bpb.bs_spc * SECSIZ) {
		blk = fat_map(blk, rd);
		off -= pcfsp->f_bpb.bs_spc * SECSIZ;

		if (!cluster_valid(blk, rd)) {
			printf("bpcfs_read: invalid cluster: %ld, %d\n",
			    blk, rd);
			return (-1);
		}
	}

	while (count < c) {
		sector = fat_ctodb(blk, rd);
		diskread_callback = fileread_callback;
		for (i = ((off / SECSIZ) % pcfsp->f_bpb.bs_spc); i < spc; i++) {
			xfer = MIN(SECSIZ - (off % SECSIZ), c - count);
			if (xfer == 0)
				break;	/* last sector done */

			block = (char *)readblock(sector + i, 1);
			if (block == NULL) {
				return (-1);
			}
			dprintf("bpcfs_read: read %d bytes\n", xfer);
			if (diskread_callback == NULL)
				(void) bcopy(&block[off % SECSIZ], b, xfer);
			count += xfer;
			off += xfer;
			b += xfer;
		}

		diskread_callback = NULL;
		if (count < c) {
			blk = fat_map(blk, rd);
			if (!cluster_valid(blk, rd)) {
				printf("bpcfs_read: invalid cluster: %ld, %d\n",
				    blk, rd);
				break;
			}
		}
	}

	fp->fi_offset += count;
	return (count);
}

/*
 * This version of seek() only performs absolute seeks (whence == 0).
 */
static off_t
bpcfs_lseek(int fd, off_t addr, int whence)
{
	fileid_t *filep;

	dprintf("lseek %d, off = %lx\n", fd, addr);
	if (!(filep = find_fp(fd)))
		return (-1);

	switch (whence) {
	case SEEK_CUR:
		filep->fi_offset += addr;
		break;
	case SEEK_SET:
		filep->fi_offset = addr;
		break;
	default:
	case SEEK_END:
		printf("lseek(): invalid whence value %d\n", whence);
		break;
	}

	filep->fi_blocknum = addr / DEV_BSIZE;
	filep->fi_count = 0;
	return (0);
}

static fileid_t *
find_fp(int fd)
{
	fileid_t *filep = head;

	if (fd >= 0) {
		while ((filep = filep->fi_forw) != head)
			if (fd == filep->fi_filedes)
				return (filep->fi_taken ? filep : 0);
	}

	return (0);
}

static int
cluster_valid(long c, int rd)
{
	return ((rd && (c == 0)) ? 1 : (c >= CLUSTER_RES_16_0 ? 0 : c));
}

static int
fat_ctodb(int blk, int r)
{
	uint_t s;

	s = r ? blk + pcfsp->f_rootsec + pcfsp->f_adjust :
	    ((blk - 2) * pcfsp->f_bpb.bs_spc) +
	    pcfsp->f_filesec + pcfsp->f_adjust;

	return (s);
}

static int
fat_map(int blk, int rootdir)
{
	ulong_t sectn, fat_index;
	uchar_t *fp;

	if (rootdir) {
		return (blk > pcfsp->f_rootlen ? CLUSTER_EOF : blk + 1);
	}

	/* ---- Find out what sector this cluster is in ---- */
	fat_index = (pcfsp->f_16bit) ? ((ulong_t)blk << 1) :
	    ((ulong_t)blk + ((uint_t)blk >> 1));

	sectn = (fat_index / SECSIZ) + ltohs(pcfsp->f_bpb.bs_resv_sectors)
	    + pcfsp->f_adjust;

	/*
	 * Read two sectors so that if our fat_index points at the last byte
	 * byte we'll have the data needed.  This is only a problem for fat12
	 * entries.
	 */
	if (!(fp = (uchar_t *)readblock(sectn, 2))) {
		printf("fat_map: bad cluster\n");
		return (CLUSTER_BAD_16);
	}

	fp += (fat_index % SECSIZ);

	if (pcfsp->f_16bit)
		blk = fp[0] | (fp[1] << 8);
	else {
		if (blk & 1)
			blk = ((fp[0] >> 4) & 0xf) | (fp[1] << 4);
		else
			blk = ((fp[1] & 0xf) << 8) | fp[0];

		/*
		 * This makes compares easier because we can just compare
		 * against one value instead of two.
		 */
		if (blk >= CLUSTER_RES_12_0)
			blk |= CLUSTER_RES_16_0;
	}
	return (blk);
}

static int
namecmp(char *pn, char *dn, int cs)
{
	dprintf("namecmp %s, %s, len = %d\n", pn, dn, cs);

	/* starting char must match */
	while (*pn && *dn) {
		--cs;
		if (toupper(*pn++) != toupper(*dn++))
			return (1);
	}

	dprintf("namecmp: cs = %d\n", cs);
	/* remainder should be either ~# or all spaces */
	if (cs > 0 && *dn == '~')
		return (0);
	while (cs > 0) {
		if (*dn++ != ' ')
			return (1);
		--cs;
	}
	return (0);
}

static int
dircmp(char *name, char *d_name, char *d_ext)
{
	int ret;
	char *sep, *ext;

	sep = (char *)strchr(name, '.');

	if (sep) {
		*sep = '\0';
		ext = sep + 1;
	} else
		ext = "   ";

	if (namecmp(name, d_name, NAMESIZ) || namecmp(ext, d_ext, EXTSIZ))
		ret = 1;
	else
		ret = 0;
	if (sep)
		*sep = '.';
	return (ret);
}

static int
lookup(char *n, _dir_entry_p dp, ulong_t dir_blk)
{
	int spc = pcfsp->f_bpb.bs_spc;
	int rd = (dir_blk == CLUSTER_ROOTDIR ? 1 : 0);
	_dir_entry_p dxp;
	int j, sector;

	dprintf("lookup: name = %s\n", n);

	while (cluster_valid(dir_blk, rd)) {
		sector = fat_ctodb(dir_blk, rd);
		dxp = readblock(sector, 1);	/* read one sector */
		if (dxp == NULL)
			return (0);
		for (j = 0; j < DIRENTS * spc; j++, dxp++) {
			dprintf("lookup: dir entry %s.%s;\n",
			    dxp->d_name, dxp->d_ext);
			if (dxp->d_name[0] == 0)
				return (0);
			if ((uchar_t)dxp->d_name[0] != 0xE5 &&
			    (dxp->d_attr & (DE_LABEL|DE_HIDDEN)) == 0 &&
			    dircmp(n, dxp->d_name, dxp->d_ext) == 0) {
				dprintf("lookup: match found\n");
				(void) bcopy(dxp, dp, sizeof (*dp));
				return (1);
			}
		}
		/* next cluster */
		dir_blk = fat_map(dir_blk, rd);
	}

	return (0);
}

static int
lookuppn(char *n, _dir_entry_p dp)
{
	long dir_blk;
	char name[8 + 1 + 3 + 1];	/* <8>.<3>'\0' */
	char *p, *ep;
	_dir_entry_t dd;

	dprintf("lookuppn: path = %s\n", n);
	dir_blk = pcfsp->f_dclust;
	if ((*n == '\\') || (*n == '/')) {
		dir_blk = CLUSTER_ROOTDIR;
		while ((*n == '\\') || (*n == '/'))
			n++;
		if (*n == '\0') {
			(void) bzero(dp, sizeof (*dp));
			dp->d_cluster = CLUSTER_ROOTDIR;
			dp->d_attr = DE_DIRECTORY;
			return (0);
		}
	}

	ep = &name[0] + sizeof (name);
	while (*n) {
		(void) bzero(name, sizeof (name));
		p = &name[0];
		while (*n && (*n != '\\') && (*n != '/'))
			if (p != ep)
				*p++ = *n++;
			else {
				dprintf("return, name %s is too long\n", name);
				return (-1);	/* name is too long */
			}
		while ((*n == '\\') || (*n == '/'))
			n++;
		if (lookup(name, &dd, dir_blk) == 0) {
			dprintf("return, name %s not found\n", name);
			return (-1);
		}
		dprintf("dd = %x:%x:%x attr = %x\n",
		    *(int *)&dd, *(((int *)&dd) + 1),
		    *(((int *)&dd) + 2), dd.d_attr);
		if (*n && ((dd.d_attr & DE_DIRECTORY) == 0)) {
			dprintf("return, not a directory\n");
			return (-1);
		}

		dir_blk = dd.d_cluster;
	}
	(void) bcopy(&dd, dp, sizeof (dd));
	return (0);
}

static void *
readblock(int sector, int nsec)
{
	if (sector >= nsec_start && sector + nsec <= nsec_start + nsec_cache)
		return (cluster_cache + (sector - nsec_start) * SECSIZ);

	/* read disk sectors */
	head->fi_blocknum = sector;
	head->fi_count = nsec * SECSIZ;
	head->fi_memp = head->fi_buf;
	if (diskread(head)) {
		printf("failed to %d sectors at %d\n", nsec, sector);
		return (NULL);
	}

	return (head->fi_buf);
}

struct boot_fs_ops bpcfs_ops = {
	"boot_pcfs",
	bpcfs_mountroot,
	bpcfs_unmountroot,
	bpcfs_open,
	bpcfs_close,
	bpcfs_read,
	bpcfs_lseek,
	NULL
};
/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */

/*
 * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 * Copyright (c) 2016 by Delphix. All rights reserved.
 * Copyright 2022 Oxide Computer Company
 * Copyright 2025 MNX Cloud, Inc.
 */

#include <sys/types.h>
#include <sys/param.h>
#include <sys/vnode.h>
#include <sys/fs/ufs_fsdir.h>
#include <sys/fs/ufs_fs.h>
#include <sys/fs/ufs_inode.h>
#include <sys/sysmacros.h>
#include <sys/bootvfs.h>
#include <sys/filep.h>
#include <sys/kmem.h>
#include <sys/kobj.h>
#include <sys/sunddi.h>

extern void *bkmem_alloc(size_t);
extern void bkmem_free(void *, size_t);
extern int cf_check_compressed(fileid_t *);
extern void cf_close(fileid_t *);
extern void cf_seek(fileid_t *, off_t, int);
extern int cf_read(fileid_t *, caddr_t, size_t);

extern int bootrd_debug;

/*
 * This fd is used when talking to the device file itself.
 */
static fileid_t *head;

/* Only got one of these...ergo, only 1 fs open at once */
/* static */
devid_t		*ufs_devp;

struct dirinfo {
	int	loc;
	fileid_t *fi;
};

static	int	bufs_close(int);
static	void	bufs_closeall(int);
static	ino_t	find(fileid_t *filep, char *path);
static	ino_t	dlook(fileid_t *filep, char *path);
static	daddr32_t	sbmap(fileid_t *filep, daddr32_t bn);
static  struct direct *readdir(struct dirinfo *dstuff);
static	void set_cache(int, void *, uint_t);
static	void *get_cache(int);
static	void free_cache();


/*
 *	There is only 1 open (mounted) device at any given time.
 *	So we can keep a single, global devp file descriptor to
 *	use to index into the di[] array.  This is not true for the
 *	fi[] array.  We can have more than one file open at once,
 *	so there is no global fd for the fi[].
 *	The user program must save the fd passed back from open()
 *	and use it to do subsequent read()'s.
 */

static int
openi(fileid_t *filep, ino_t inode)
{
	struct dinode *dp;
	devid_t *devp = filep->fi_devp;

	filep->fi_inode = get_cache((int)inode);
	if (filep->fi_inode != 0)
		return (0);

	filep->fi_offset = 0;
	filep->fi_blocknum = fsbtodb(&devp->un_fs.di_fs,
	    itod(&devp->un_fs.di_fs, inode));

	/* never more than 1 disk block */
	filep->fi_count = devp->un_fs.di_fs.fs_bsize;
	filep->fi_memp = 0;		/* cached read */
	if (diskread(filep) != 0) {
		return (0);
	}

	dp = (struct dinode *)filep->fi_memp;
	filep->fi_inode = (struct inode *)
	    bkmem_alloc(sizeof (struct inode));
	bzero((char *)filep->fi_inode, sizeof (struct inode));
	filep->fi_inode->i_ic =
	    dp[itoo(&devp->un_fs.di_fs, inode)].di_un.di_icom;
	filep->fi_inode->i_number = inode;
	set_cache((int)inode, (void *)filep->fi_inode, sizeof (struct inode));
	return (0);
}

static fileid_t *
find_fp(int fd)
{
	fileid_t *filep = head;

	if (fd >= 0) {
		while ((filep = filep->fi_forw) != head)
			if (fd == filep->fi_filedes)
				return (filep->fi_taken ? filep : 0);
	}

	return (0);
}

static ino_t
find(fileid_t *filep, char *path)
{
	char *q;
	char c;
	ino_t inode;
	char lpath[MAXPATHLEN];
	char *lpathp = lpath;
	int len, r;
	devid_t	*devp;

	inode = 0;
	if (path == NULL || *path == '\0') {
		kobj_printf("null path\n");
		return (inode);
	}

	if (bootrd_debug)
		kobj_printf("openi: %s\n", path);

	bzero(lpath, sizeof (lpath));
	bcopy(path, lpath, strlen(path));
	devp = filep->fi_devp;
	while (*lpathp) {
		/* if at the beginning of pathname get root inode */
		r = (lpathp == lpath);
		if (r && openi(filep, (ino_t)UFSROOTINO))
			return ((ino_t)0);
		while (*lpathp == '/')
			lpathp++;	/* skip leading slashes */
		q = lpathp;
		while (*q != '/' && *q != '\0')
			q++;		/* find end of component */
		c = *q;
		*q = '\0';		/* terminate component */

		/* Bail out early if opening root */
		if (r && (*lpathp == '\0'))
			return ((ino_t)UFSROOTINO);
		if ((inode = dlook(filep, lpathp)) != 0) {
			if (openi(filep, inode))
				return ((ino_t)0);
			if ((filep->fi_inode->i_smode & IFMT) == IFLNK) {
				filep->fi_blocknum =
				    fsbtodb(&devp->un_fs.di_fs,
				    filep->fi_inode->i_db[0]);
				filep->fi_count = DEV_BSIZE;
				filep->fi_memp = 0;
				if (diskread(filep) != 0)
					return ((ino_t)0);
				len = strlen(filep->fi_memp);
				if (filep->fi_memp[0] == '/')
					/* absolute link */
					lpathp = lpath;
				/* copy rest of unprocessed path up */
				bcopy(q, lpathp + len, strlen(q + 1) + 2);
				/* point to unprocessed path */
				*(lpathp + len) = c;
				/* prepend link in before unprocessed path */
				bcopy(filep->fi_memp, lpathp, len);
				lpathp = lpath;
				continue;
			} else
				*q = c;
			if (c == '\0')
				break;
			lpathp = q;
			continue;
		} else {
			return ((ino_t)0);
		}
	}
	return (inode);
}

static daddr32_t
sbmap(fileid_t *filep, daddr32_t bn)
{
	struct inode *inodep;
	int i, j, sh;
	daddr32_t nb, *bap;
	daddr32_t *db;
	devid_t	*devp;

	devp = filep->fi_devp;
	inodep = filep->fi_inode;
	db = inodep->i_db;

	/*
	 * blocks 0..NDADDR are direct blocks
	 */
	if (bn < NDADDR) {
		nb = db[bn];
		return (nb);
	}

	/*
	 * addresses NIADDR have single and double indirect blocks.
	 * the first step is to determine how many levels of indirection.
	 */
	sh = 1;
	bn -= NDADDR;
	for (j = NIADDR; j > 0; j--) {
		sh *= NINDIR(&devp->un_fs.di_fs);
		if (bn < sh)
			break;
		bn -= sh;
	}
	if (j == 0) {
		return ((daddr32_t)0);
	}

	/*
	 * fetch the first indirect block address from the inode
	 */
	nb = inodep->i_ib[NIADDR - j];
	if (nb == 0) {
		return ((daddr32_t)0);
	}

	/*
	 * fetch through the indirect blocks
	 */
	for (; j <= NIADDR; j++) {
		filep->fi_blocknum = fsbtodb(&devp->un_fs.di_fs, nb);
		filep->fi_count = devp->un_fs.di_fs.fs_bsize;
		filep->fi_memp = 0;
		if (diskread(filep) != 0)
			return (0);
		bap = (daddr32_t *)filep->fi_memp;
		sh /= NINDIR(&devp->un_fs.di_fs);
		i = (bn / sh) % NINDIR(&devp->un_fs.di_fs);
		nb = bap[i];
		if (nb == 0) {
			return ((daddr32_t)0);
		}
	}
	return (nb);
}

static ino_t
dlook(fileid_t *filep, char *path)
{
	struct direct *dp;
	struct inode *ip;
	struct dirinfo dirp;
	int len;

	ip = filep->fi_inode;
	if (path == NULL || *path == '\0')
		return (0);

	if (bootrd_debug)
		kobj_printf("dlook: %s\n", path);

	if ((ip->i_smode & IFMT) != IFDIR) {
		return (0);
	}
	if (ip->i_size == 0) {
		return (0);
	}
	len = strlen(path);
	dirp.loc = 0;
	dirp.fi = filep;
	for (dp = readdir(&dirp); dp != NULL; dp = readdir(&dirp)) {
		if (dp->d_ino == 0)
			continue;
		if (dp->d_namlen == len && strcmp(path, dp->d_name) == 0) {
			return (dp->d_ino);
		}
		/* Allow "*" to print all names at that level, w/out match */
		if (strcmp(path, "*") == 0 && bootrd_debug)
			kobj_printf("%s\n", dp->d_name);
	}
	return (0);
}

/*
 * get next entry in a directory.
 */
struct direct *
readdir(struct dirinfo *dstuff)
{
	struct direct *dp;
	fileid_t *filep;
	daddr32_t lbn, d;
	int off;
	devid_t	*devp;

	filep = dstuff->fi;
	devp = filep->fi_devp;
	for (;;) {
		if (dstuff->loc >= filep->fi_inode->i_size) {
			return (NULL);
		}
		off = blkoff(&devp->un_fs.di_fs, dstuff->loc);
		if (bootrd_debug)
			kobj_printf("readdir: off = 0x%x\n", off);
		if (off == 0) {
			lbn = lblkno(&devp->un_fs.di_fs, dstuff->loc);
			d = sbmap(filep, lbn);

			if (d == 0)
				return (NULL);

			filep->fi_blocknum = fsbtodb(&devp->un_fs.di_fs, d);
			filep->fi_count =
			    blksize(&devp->un_fs.di_fs, filep->fi_inode, lbn);
			filep->fi_memp = 0;
			if (diskread(filep) != 0) {
				return (NULL);
			}
		}
		dp = (struct direct *)(filep->fi_memp + off);
		dstuff->loc += dp->d_reclen;
		if (dp->d_ino == 0)
			continue;
		if (bootrd_debug)
			kobj_printf("readdir: name = %s\n", dp->d_name);
		return (dp);
	}
}

/*
 * Get the next block of data from the file.  If possible, dma right into
 * user's buffer
 */
static int
getblock(fileid_t *filep, caddr_t buf, int count, int *rcount)
{
	struct fs *fs;
	caddr_t p;
	int off, size, diff;
	daddr32_t lbn;
	devid_t	*devp;

	if (bootrd_debug)
		kobj_printf("getblock: buf 0x%p, count 0x%x\n",
		    (void *)buf, count);

	devp = filep->fi_devp;
	p = filep->fi_memp;
	if ((signed)filep->fi_count <= 0) {

		/* find the amt left to be read in the file */
		diff = filep->fi_inode->i_size - filep->fi_offset;
		if (diff <= 0) {
			kobj_printf("Short read\n");
			return (-1);
		}

		fs = &devp->un_fs.di_fs;
		/* which block (or frag) in the file do we read? */
		lbn = lblkno(fs, filep->fi_offset);

		/* which physical block on the device do we read? */
		filep->fi_blocknum = fsbtodb(fs, sbmap(filep, lbn));

		off = blkoff(fs, filep->fi_offset);

		/* either blksize or fragsize */
		size = blksize(fs, filep->fi_inode, lbn);
		filep->fi_count = size;
		filep->fi_memp = filep->fi_buf;

		/*
		 * optimization if we are reading large blocks of data then
		 * we can go directly to user's buffer
		 */
		*rcount = 0;
		if (off == 0 && count >= size) {
			filep->fi_memp = buf;
			if (diskread(filep)) {
				return (-1);
			}
			*rcount = size;
			filep->fi_count = 0;
			return (0);
		} else if (diskread(filep))
			return (-1);

		if (filep->fi_offset - off + size >= filep->fi_inode->i_size)
			filep->fi_count = diff + off;
		filep->fi_count -= off;
		p = &filep->fi_memp[off];
	}
	filep->fi_memp = p;
	return (0);
}

/*
 * Get the next block of data from the file.  Don't attempt to go directly
 * to user's buffer.
 */
static int
getblock_noopt(fileid_t *filep)
{
	struct fs *fs;
	caddr_t p;
	int off, size, diff;
	daddr32_t lbn;
	devid_t	*devp;

	if (bootrd_debug)
		kobj_printf("getblock_noopt: start\n");

	devp = filep->fi_devp;
	p = filep->fi_memp;
	if ((signed)filep->fi_count <= 0) {

		/* find the amt left to be read in the file */
		diff = filep->fi_inode->i_size - filep->fi_offset;
		if (diff <= 0) {
			kobj_printf("Short read\n");
			return (-1);
		}

		fs = &devp->un_fs.di_fs;
		/* which block (or frag) in the file do we read? */
		lbn = lblkno(fs, filep->fi_offset);

		/* which physical block on the device do we read? */
		filep->fi_blocknum = fsbtodb(fs, sbmap(filep, lbn));

		off = blkoff(fs, filep->fi_offset);

		/* either blksize or fragsize */
		size = blksize(fs, filep->fi_inode, lbn);
		filep->fi_count = size;
		/* reading on a ramdisk, just get a pointer to the data */
		filep->fi_memp = NULL;

		if (diskread(filep))
			return (-1);

		if (filep->fi_offset - off + size >= filep->fi_inode->i_size)
			filep->fi_count = diff + off;
		filep->fi_count -= off;
		p = &filep->fi_memp[off];
	}
	filep->fi_memp = p;
	return (0);
}


/*
 *  This is the high-level read function.  It works like this.
 *  We assume that our IO device buffers up some amount of
 *  data and that we can get a ptr to it.  Thus we need
 *  to actually call the device func about filesize/blocksize times
 *  and this greatly increases our IO speed.  When we already
 *  have data in the buffer, we just return that data (with bcopy() ).
 */

static ssize_t
bufs_read(int fd, caddr_t buf, size_t count)
{
	size_t i, j;
	caddr_t	n;
	int rcount;
	fileid_t *filep;

	if (!(filep = find_fp(fd))) {
		return (-1);
	}

	if ((filep->fi_flags & FI_COMPRESSED) == 0 &&
	    filep->fi_offset + count > filep->fi_inode->i_size)
		count = filep->fi_inode->i_size - filep->fi_offset;

	/* that was easy */
	if ((i = count) == 0)
		return (0);

	n = buf;
	while (i > 0) {
		if (filep->fi_flags & FI_COMPRESSED) {
			int rval;

			if ((rval = cf_read(filep, buf, count)) < 0)
				return (0); /* encountered an error */
			j = (size_t)rval;
			if (j < i)
				i = j; /* short read, must have hit EOF */
		} else {
			/* If we need to reload the buffer, do so */
			if ((j = filep->fi_count) == 0) {
				(void) getblock(filep, buf, i, &rcount);
				i -= rcount;
				buf += rcount;
				filep->fi_offset += rcount;
				continue;
			} else {
				/* else just bcopy from our buffer */
				j = MIN(i, j);
				bcopy(filep->fi_memp, buf, (unsigned)j);
			}
		}
		buf += j;
		filep->fi_memp += j;
		filep->fi_offset += j;
		filep->fi_count -= j;
		i -= j;
	}
	return (buf - n);
}

/*
 *	This routine will open a device as it is known by the V2 OBP.
 *	Interface Defn:
 *	err = mountroot(string);
 *		err = 0 on success
 *		err = -1 on failure
 *	string:	char string describing the properties of the device.
 *	We must not dork with any fi[]'s here.  Save that for later.
 */

static int
bufs_mountroot(char *str)
{
	if (ufs_devp)		/* already mounted */
		return (0);

	ufs_devp = (devid_t *)bkmem_alloc(sizeof (devid_t));
	ufs_devp->di_taken = 1;
	ufs_devp->di_dcookie = 0;
	ufs_devp->di_desc = (char *)bkmem_alloc(strlen(str) + 1);
	(void) strcpy(ufs_devp->di_desc, str);
	bzero(ufs_devp->un_fs.dummy, SBSIZE);
	head = (fileid_t *)bkmem_alloc(sizeof (fileid_t));
	head->fi_back = head->fi_forw = head;
	head->fi_filedes = 0;
	head->fi_taken = 0;

	/* Setup read of the superblock */
	head->fi_devp = ufs_devp;
	head->fi_blocknum = SBLOCK;
	head->fi_count = (uint_t)SBSIZE;
	head->fi_memp = (caddr_t)&(ufs_devp->un_fs.di_fs);
	head->fi_offset = 0;

	if (diskread(head)) {
		kobj_printf("failed to read superblock\n");
		(void) bufs_closeall(1);
		return (-1);
	}

	if (ufs_devp->un_fs.di_fs.fs_magic != FS_MAGIC) {
		if (bootrd_debug)
			kobj_printf("fs magic = 0x%x\n",
			    ufs_devp->un_fs.di_fs.fs_magic);
		(void) bufs_closeall(1);
		return (-1);
	}
	if (bootrd_debug)
		kobj_printf("mountroot succeeded\n");
	return (0);
}

/*
 * Unmount the currently mounted root fs.  In practice, this means
 * closing all open files and releasing resources.  All of this
 * is done by closeall().
 */

static int
bufs_unmountroot(void)
{
	if (ufs_devp == NULL)
		return (-1);

	(void) bufs_closeall(1);

	return (0);
}

/*
 *	We allocate an fd here for use when talking
 *	to the file itself.
 */

static int
bufs_open(char *filename, int flags __unused)
{
	fileid_t	*filep;
	ino_t	inode;
	static int	filedes = 1;

	if (bootrd_debug)
		kobj_printf("open: %s\n", filename);

	/* build and link a new file descriptor */
	filep = (fileid_t *)bkmem_alloc(sizeof (fileid_t));
	filep->fi_back = head->fi_back;
	filep->fi_forw = head;
	head->fi_back->fi_forw = filep;
	head->fi_back = filep;
	filep->fi_filedes = filedes++;
	filep->fi_taken = 1;
	filep->fi_path = (char *)bkmem_alloc(strlen(filename) + 1);
	(void) strcpy(filep->fi_path, filename);
	filep->fi_devp = ufs_devp; /* dev is already "mounted" */
	filep->fi_inode = NULL;
	bzero(filep->fi_buf, MAXBSIZE);
	filep->fi_getblock = getblock_noopt;
	filep->fi_flags = 0;

	inode = find(filep, (char *)filename);
	if (inode == 0) {
		if (bootrd_debug)
			kobj_printf("open: cannot find %s\n", filename);
		(void) bufs_close(filep->fi_filedes);
		return (-1);
	}
	if (openi(filep, inode)) {
		kobj_printf("open: cannot open %s\n", filename);
		(void) bufs_close(filep->fi_filedes);
		return (-1);
	}

	filep->fi_offset = filep->fi_count = 0;

	if (cf_check_compressed(filep) != 0)
		return (-1);
	return (filep->fi_filedes);
}

/*
 *  We don't do any IO here.
 *  We just play games with the device pointers.
 */

static off_t
bufs_lseek(int fd, off_t addr, int whence)
{
	fileid_t *filep;

	/* Make sure user knows what file they are talking to */
	if (!(filep = find_fp(fd)))
		return (-1);

	if (filep->fi_flags & FI_COMPRESSED) {
		cf_seek(filep, addr, whence);
	} else {
		switch (whence) {
		case SEEK_CUR:
			filep->fi_offset += addr;
			break;
		case SEEK_SET:
			filep->fi_offset = addr;
			break;
		default:
		case SEEK_END:
			kobj_printf("lseek(): invalid whence value %d\n",
			    whence);
			break;
		}
		filep->fi_blocknum = addr / DEV_BSIZE;
	}

	filep->fi_count = 0;

	return (0);
}


int
bufs_fstat(int fd, struct bootstat *stp)
{
	fileid_t	*filep;
	struct inode	*ip;

	if (!(filep = find_fp(fd)))
		return (-1);

	ip = filep->fi_inode;

	stp->st_mode = 0;
	stp->st_size = 0;

	if (ip == NULL)
		return (0);

	switch (ip->i_smode & IFMT) {
	case IFLNK:
		stp->st_mode = S_IFLNK;
		break;
	case IFREG:
		stp->st_mode = S_IFREG;
		break;
	default:
		break;
	}
	/*
	 * NOTE: this size will be the compressed size for a compressed file
	 * This could confuse the caller since we decompress the file behind
	 * the scenes when the file is read.
	 */
	stp->st_size = ip->i_size;
	stp->st_atim.tv_sec = ip->i_atime.tv_sec;
	stp->st_atim.tv_nsec = ip->i_atime.tv_usec * 1000;
	stp->st_mtim.tv_sec = ip->i_mtime.tv_sec;
	stp->st_mtim.tv_nsec = ip->i_mtime.tv_usec * 1000;
	stp->st_ctim.tv_sec = ip->i_ctime.tv_sec;
	stp->st_ctim.tv_nsec = ip->i_ctime.tv_usec * 1000;

	return (0);
}


static int
bufs_close(int fd)
{
	fileid_t *filep;

	/* Make sure user knows what file they are talking to */
	if (!(filep = find_fp(fd)))
		return (-1);

	if (filep->fi_taken && (filep != head)) {
		/* Clear the ranks */
		bkmem_free(filep->fi_path, strlen(filep->fi_path)+1);
		filep->fi_blocknum = filep->fi_count = filep->fi_offset = 0;
		filep->fi_memp = (caddr_t)0;
		filep->fi_devp = 0;
		filep->fi_taken = 0;

		/* unlink and deallocate node */
		filep->fi_forw->fi_back = filep->fi_back;
		filep->fi_back->fi_forw = filep->fi_forw;
		cf_close(filep);
		bkmem_free((char *)filep, sizeof (fileid_t));

		/*
		 * Some files are opened and closed in early boot, for example
		 * when doing a microcode update on the boot CPU. In that case
		 * the inode cache will contain memory allocated from boot
		 * pages, which will be invalid once kmem is initialised.
		 * Until kmem is ready, clear the inode cache when closing a
		 * file.
		 */
		if (kmem_ready == 0)
			free_cache();

		return (0);
	} else {
		/* Big problem */
		kobj_printf("\nFile descrip %d not allocated!", fd);
		return (-1);
	}
}

static void
bufs_closeall(int flag __unused)
{
	fileid_t *filep = head;

	while ((filep = filep->fi_forw) != head)
		if (filep->fi_taken)
			if (bufs_close(filep->fi_filedes))
				kobj_printf(
				    "Filesystem may be inconsistent.\n");

	ufs_devp->di_taken = 0;
	bkmem_free((char *)ufs_devp, sizeof (devid_t));
	bkmem_free((char *)head, sizeof (fileid_t));
	ufs_devp = NULL;
	head = NULL;
	free_cache();
}

static struct cache {
	struct cache *next;
	void *data;
	int key;
	uint_t size;
} *icache;

void
set_cache(int key, void *data, uint_t size)
{
	struct cache *entry = bkmem_alloc(sizeof (*entry));
	entry->key = key;
	entry->data = data;
	entry->size = size;
	if (icache) {
		entry->next = icache;
		icache = entry;
	} else {
		icache = entry;
		entry->next = 0;
	}
}

void *
get_cache(int key)
{
	struct cache *entry = icache;
	while (entry) {
		if (entry->key == key)
			return (entry->data);
		entry = entry->next;
	}
	return (NULL);
}

void
free_cache()
{
	struct cache *next, *entry = icache;
	while (entry) {
		next = entry->next;
		bkmem_free(entry->data, entry->size);
		bkmem_free(entry, sizeof (*entry));
		entry = next;
	}
	icache = 0;
}

struct boot_fs_ops bufs_ops = {
	"boot_ufs",
	bufs_mountroot,
	bufs_unmountroot,
	bufs_open,
	bufs_close,
	bufs_read,
	bufs_lseek,
	bufs_fstat,
	NULL
};