xref: /illumos-gate/usr/src/uts/common/fs/pcfs/pc_vfsops.c (revision 80a3d255)

/*
 * 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 2009 Sun Microsystems, Inc.  All rights reserved.
 * Use is subject to license terms.
 */
/*
 * Copyright (c) 2017 by Delphix. All rights reserved.
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/kmem.h>
#include <sys/user.h>
#include <sys/proc.h>
#include <sys/cred.h>
#include <sys/disp.h>
#include <sys/buf.h>
#include <sys/vfs.h>
#include <sys/vfs_opreg.h>
#include <sys/vnode.h>
#include <sys/fdio.h>
#include <sys/file.h>
#include <sys/uio.h>
#include <sys/conf.h>
#include <sys/statvfs.h>
#include <sys/mount.h>
#include <sys/pathname.h>
#include <sys/cmn_err.h>
#include <sys/debug.h>
#include <sys/sysmacros.h>
#include <sys/conf.h>
#include <sys/mkdev.h>
#include <sys/swap.h>
#include <sys/sunddi.h>
#include <sys/sunldi.h>
#include <sys/dktp/fdisk.h>
#include <sys/fs/pc_label.h>
#include <sys/fs/pc_fs.h>
#include <sys/fs/pc_dir.h>
#include <sys/fs/pc_node.h>
#include <fs/fs_subr.h>
#include <sys/modctl.h>
#include <sys/dkio.h>
#include <sys/open.h>
#include <sys/mntent.h>
#include <sys/policy.h>
#include <sys/atomic.h>
#include <sys/sdt.h>

/*
 * The majority of PC media use a 512 sector size, but
 * occasionally you will run across a 1k sector size.
 * For media with a 1k sector size, fd_strategy() requires
 * the I/O size to be a 1k multiple; so when the sector size
 * is not yet known, always read 1k.
 */
#define	PC_SAFESECSIZE	(PC_SECSIZE * 2)

static int pcfs_pseudo_floppy(dev_t);

static int pcfsinit(int, char *);
static int pcfs_mount(struct vfs *, struct vnode *, struct mounta *,
	struct cred *);
static int pcfs_unmount(struct vfs *, int, struct cred *);
static int pcfs_root(struct vfs *, struct vnode **);
static int pcfs_statvfs(struct vfs *, struct statvfs64 *);
static int pc_syncfsnodes(struct pcfs *);
static int pcfs_sync(struct vfs *, short, struct cred *);
static int pcfs_vget(struct vfs *vfsp, struct vnode **vpp, struct fid *fidp);
static void pcfs_freevfs(vfs_t *vfsp);

static int pc_readfat(struct pcfs *fsp, uchar_t *fatp);
static int pc_writefat(struct pcfs *fsp, daddr_t start);

static int pc_getfattype(struct pcfs *fsp);
static void pcfs_parse_mntopts(struct pcfs *fsp);


/*
 * pcfs mount options table
 */

static char *nohidden_cancel[] = { MNTOPT_PCFS_HIDDEN, NULL };
static char *hidden_cancel[] = { MNTOPT_PCFS_NOHIDDEN, NULL };
static char *nofoldcase_cancel[] = { MNTOPT_PCFS_FOLDCASE, NULL };
static char *foldcase_cancel[] = { MNTOPT_PCFS_NOFOLDCASE, NULL };
static char *clamptime_cancel[] = { MNTOPT_PCFS_NOCLAMPTIME, NULL };
static char *noclamptime_cancel[] = { MNTOPT_PCFS_CLAMPTIME, NULL };
static char *atime_cancel[] = { MNTOPT_NOATIME, NULL };
static char *noatime_cancel[] = { MNTOPT_ATIME, NULL };

static mntopt_t mntopts[] = {
/*
 *	option name	cancel option	default arg	flags	opt data
 */
	{ MNTOPT_PCFS_NOHIDDEN, nohidden_cancel, NULL, 0, NULL },
	{ MNTOPT_PCFS_HIDDEN, hidden_cancel, NULL, MO_DEFAULT, NULL },
	{ MNTOPT_PCFS_NOFOLDCASE, nofoldcase_cancel, NULL, MO_DEFAULT, NULL },
	{ MNTOPT_PCFS_FOLDCASE, foldcase_cancel, NULL, 0, NULL },
	{ MNTOPT_PCFS_CLAMPTIME, clamptime_cancel, NULL, MO_DEFAULT, NULL },
	{ MNTOPT_PCFS_NOCLAMPTIME, noclamptime_cancel, NULL, 0, NULL },
	{ MNTOPT_NOATIME, noatime_cancel, NULL, 0, NULL },
	{ MNTOPT_ATIME, atime_cancel, NULL, 0, NULL },
	{ MNTOPT_PCFS_TIMEZONE, NULL, "+0", MO_DEFAULT | MO_HASVALUE, NULL },
	{ MNTOPT_PCFS_SECSIZE, NULL, NULL, MO_HASVALUE, NULL }
};

static mntopts_t pcfs_mntopts = {
	sizeof (mntopts) / sizeof (mntopt_t),
	mntopts
};

int pcfsdebuglevel = 0;

/*
 * pcfslock:	protects the list of mounted pc filesystems "pc_mounttab.
 * pcfs_lock:	(inside per filesystem structure "pcfs")
 *		per filesystem lock. Most of the vfsops and vnodeops are
 *		protected by this lock.
 * pcnodes_lock: protects the pcnode hash table "pcdhead", "pcfhead".
 *
 * Lock hierarchy: pcfslock > pcfs_lock > pcnodes_lock
 *
 * pcfs_mountcount:	used to prevent module unloads while there is still
 *			pcfs state from a former mount hanging around. With
 *			forced umount support, the filesystem module must not
 *			be allowed to go away before the last VFS_FREEVFS()
 *			call has been made.
 *			Since this is just an atomic counter, there's no need
 *			for locking.
 */
kmutex_t	pcfslock;
krwlock_t	pcnodes_lock;
uint32_t	pcfs_mountcount;

static int pcfstype;

static vfsdef_t vfw = {
	VFSDEF_VERSION,
	"pcfs",
	pcfsinit,
	VSW_HASPROTO|VSW_CANREMOUNT|VSW_STATS|VSW_CANLOFI|VSW_MOUNTDEV,
	&pcfs_mntopts
};

extern struct mod_ops mod_fsops;

static struct modlfs modlfs = {
	&mod_fsops,
	"PC filesystem",
	&vfw
};

static struct modlinkage modlinkage = {
	MODREV_1,
	&modlfs,
	NULL
};

int
_init(void)
{
	int	error;

#if !defined(lint)
	/* make sure the on-disk structures are sane */
	ASSERT(sizeof (struct pcdir) == 32);
	ASSERT(sizeof (struct pcdir_lfn) == 32);
#endif
	mutex_init(&pcfslock, NULL, MUTEX_DEFAULT, NULL);
	rw_init(&pcnodes_lock, NULL, RW_DEFAULT, NULL);
	error = mod_install(&modlinkage);
	if (error) {
		mutex_destroy(&pcfslock);
		rw_destroy(&pcnodes_lock);
	}
	return (error);
}

int
_fini(void)
{
	int	error;

	/*
	 * If a forcedly unmounted instance is still hanging around,
	 * we cannot allow the module to be unloaded because that would
	 * cause panics once the VFS framework decides it's time to call
	 * into VFS_FREEVFS().
	 */
	if (pcfs_mountcount)
		return (EBUSY);

	error = mod_remove(&modlinkage);
	if (error)
		return (error);
	mutex_destroy(&pcfslock);
	rw_destroy(&pcnodes_lock);
	/*
	 * Tear down the operations vectors
	 */
	(void) vfs_freevfsops_by_type(pcfstype);
	vn_freevnodeops(pcfs_fvnodeops);
	vn_freevnodeops(pcfs_dvnodeops);
	return (0);
}

int
_info(struct modinfo *modinfop)
{
	return (mod_info(&modlinkage, modinfop));
}

/* ARGSUSED1 */
static int
pcfsinit(int fstype, char *name)
{
	static const fs_operation_def_t pcfs_vfsops_template[] = {
		VFSNAME_MOUNT,		{ .vfs_mount = pcfs_mount },
		VFSNAME_UNMOUNT,	{ .vfs_unmount = pcfs_unmount },
		VFSNAME_ROOT,		{ .vfs_root = pcfs_root },
		VFSNAME_STATVFS,	{ .vfs_statvfs = pcfs_statvfs },
		VFSNAME_SYNC,		{ .vfs_sync = pcfs_sync },
		VFSNAME_VGET,		{ .vfs_vget = pcfs_vget },
		VFSNAME_FREEVFS,	{ .vfs_freevfs = pcfs_freevfs },
		NULL,			NULL
	};
	int error;

	error = vfs_setfsops(fstype, pcfs_vfsops_template, NULL);
	if (error != 0) {
		cmn_err(CE_WARN, "pcfsinit: bad vfs ops template");
		return (error);
	}

	error = vn_make_ops("pcfs", pcfs_fvnodeops_template, &pcfs_fvnodeops);
	if (error != 0) {
		(void) vfs_freevfsops_by_type(fstype);
		cmn_err(CE_WARN, "pcfsinit: bad file vnode ops template");
		return (error);
	}

	error = vn_make_ops("pcfsd", pcfs_dvnodeops_template, &pcfs_dvnodeops);
	if (error != 0) {
		(void) vfs_freevfsops_by_type(fstype);
		vn_freevnodeops(pcfs_fvnodeops);
		cmn_err(CE_WARN, "pcfsinit: bad dir vnode ops template");
		return (error);
	}

	pcfstype = fstype;
	(void) pc_init();
	pcfs_mountcount = 0;
	return (0);
}

static struct pcfs *pc_mounttab = NULL;

extern struct pcfs_args pc_tz;

/*
 *  Define some special logical drives we use internal to this file.
 */
#define	BOOT_PARTITION_DRIVE	99
#define	PRIMARY_DOS_DRIVE	1
#define	UNPARTITIONED_DRIVE	0

static int
pcfs_device_identify(
	struct vfs *vfsp,
	struct mounta *uap,
	struct cred *cr,
	int *dos_ldrive,
	dev_t *xdev)
{
	struct pathname special;
	char *c;
	struct vnode *svp = NULL;
	struct vnode *lvp = NULL;
	int oflag, aflag;
	int error;

	/*
	 * Resolve path name of special file being mounted.
	 */
	if (error = pn_get(uap->spec, UIO_USERSPACE, &special)) {
		return (error);
	}

	*dos_ldrive = -1;

	if (error =
	    lookupname(special.pn_path, UIO_SYSSPACE, FOLLOW, NULLVPP, &svp)) {
		/*
		 * If there's no device node, the name specified most likely
		 * maps to a PCFS-style "partition specifier" to select a
		 * harddisk primary/logical partition. Disable floppy-specific
		 * checks in such cases unless an explicit :A or :B is
		 * requested.
		 */

		/*
		 * Split the pathname string at the last ':' separator.
		 * If there's no ':' in the device name, or the ':' is the
		 * last character in the string, the name is invalid and
		 * the error from the previous lookup will be returned.
		 */
		c = strrchr(special.pn_path, ':');
		if (c == NULL || strlen(c) == 0)
			goto devlookup_done;

		*c++ = '\0';

		/*
		 * PCFS partition name suffixes can be:
		 *	- "boot" to indicate the X86BOOT partition
		 *	- a drive letter [c-z] for the "DOS logical drive"
		 *	- a drive number 1..24 for the "DOS logical drive"
		 *	- a "floppy name letter", 'a' or 'b' (just strip this)
		 */
		if (strcasecmp(c, "boot") == 0) {
			/*
			 * The Solaris boot partition is requested.
			 */
			*dos_ldrive = BOOT_PARTITION_DRIVE;
		} else if (strspn(c, "0123456789") == strlen(c)) {
			/*
			 * All digits - parse the partition number.
			 */
			long drvnum = 0;

			if ((error = ddi_strtol(c, NULL, 10, &drvnum)) == 0) {
				/*
				 * A number alright - in the allowed range ?
				 */
				if (drvnum > 24 || drvnum == 0)
					error = ENXIO;
			}
			if (error)
				goto devlookup_done;
			*dos_ldrive = (int)drvnum;
		} else if (strlen(c) == 1) {
			/*
			 * A single trailing character was specified.
			 *	- [c-zC-Z] means a harddisk partition, and
			 *	  we retrieve the partition number.
			 *	- [abAB] means a floppy drive, so we swallow
			 *	  the "drive specifier" and test later
			 *	  whether the physical device is a floppy.
			 */
			*c = tolower(*c);
			if (*c == 'a' || *c == 'b') {
				*dos_ldrive = UNPARTITIONED_DRIVE;
			} else if (*c < 'c' || *c > 'z') {
				error = ENXIO;
				goto devlookup_done;
			} else {
				*dos_ldrive = 1 + *c - 'c';
			}
		} else {
			/*
			 * Can't parse this - pass through previous error.
			 */
			goto devlookup_done;
		}


		error = lookupname(special.pn_path, UIO_SYSSPACE, FOLLOW,
		    NULLVPP, &svp);
	} else {
		*dos_ldrive = UNPARTITIONED_DRIVE;
	}
devlookup_done:
	pn_free(&special);
	if (error)
		return (error);

	ASSERT(*dos_ldrive >= UNPARTITIONED_DRIVE);

	/*
	 * Verify caller's permission to open the device special file.
	 */
	if ((vfsp->vfs_flag & VFS_RDONLY) != 0 ||
	    ((uap->flags & MS_RDONLY) != 0)) {
		oflag = FREAD;
		aflag = VREAD;
	} else {
		oflag = FREAD | FWRITE;
		aflag = VREAD | VWRITE;
	}

	error = vfs_get_lofi(vfsp, &lvp);

	if (error > 0) {
		if (error == ENOENT)
			error = ENODEV;
		goto out;
	} else if (error == 0) {
		*xdev = lvp->v_rdev;
	} else {
		*xdev = svp->v_rdev;

		if (svp->v_type != VBLK) {
			error = ENOTBLK;
			goto out;
		}

		if ((error = secpolicy_spec_open(cr, svp, oflag)) != 0)
			goto out;
	}

	if (getmajor(*xdev) >= devcnt) {
		error = ENXIO;
		goto out;
	}

	if ((error = VOP_ACCESS(svp, aflag, 0, cr, NULL)) != 0)
		goto out;

out:
	if (svp != NULL)
		VN_RELE(svp);
	if (lvp != NULL)
		VN_RELE(lvp);
	return (error);
}

static int
pcfs_device_ismounted(
	struct vfs *vfsp,
	int dos_ldrive,
	dev_t xdev,
	int *remounting,
	dev_t *pseudodev)
{
	struct pcfs *fsp;
	int remount = *remounting;

	/*
	 * Ensure that this logical drive isn't already mounted, unless
	 * this is a REMOUNT request.
	 * Note: The framework will perform this check if the "...:c"
	 * PCFS-style "logical drive" syntax has not been used and an
	 * actually existing physical device is backing this filesystem.
	 * Once all block device drivers support PC-style partitioning,
	 * this codeblock can be dropped.
	 */
	*pseudodev = xdev;

	if (dos_ldrive) {
		mutex_enter(&pcfslock);
		for (fsp = pc_mounttab; fsp; fsp = fsp->pcfs_nxt)
			if (fsp->pcfs_xdev == xdev &&
			    fsp->pcfs_ldrive == dos_ldrive) {
				mutex_exit(&pcfslock);
				if (remount) {
					return (0);
				} else {
					return (EBUSY);
				}
			}
		/*
		 * Assign a unique device number for the vfs
		 * The old way (getudev() + a constantly incrementing
		 * major number) was wrong because it changes vfs_dev
		 * across mounts and reboots, which breaks nfs file handles.
		 * UFS just uses the real dev_t. We can't do that because
		 * of the way pcfs opens fdisk partitons (the :c and :d
		 * partitions are on the same dev_t). Though that _might_
		 * actually be ok, since the file handle contains an
		 * absolute block number, it's probably better to make them
		 * different. So I think we should retain the original
		 * dev_t, but come up with a different minor number based
		 * on the logical drive that will _always_ come up the same.
		 * For now, we steal the upper 6 bits.
		 */
#ifdef notdef
		/* what should we do here? */
		if (((getminor(xdev) >> 12) & 0x3F) != 0)
			printf("whoops - upper bits used!\n");
#endif
		*pseudodev = makedevice(getmajor(xdev),
		    ((dos_ldrive << 12) | getminor(xdev)) & MAXMIN32);
		if (vfs_devmounting(*pseudodev, vfsp)) {
			mutex_exit(&pcfslock);
			return (EBUSY);
		}
		if (vfs_devismounted(*pseudodev)) {
			mutex_exit(&pcfslock);
			if (remount) {
				return (0);
			} else {
				return (EBUSY);
			}
		}
		mutex_exit(&pcfslock);
	} else {
		*pseudodev = xdev;
		if (vfs_devmounting(*pseudodev, vfsp)) {
			return (EBUSY);
		}
		if (vfs_devismounted(*pseudodev))
			if (remount) {
				return (0);
			} else {
				return (EBUSY);
			}
	}

	/*
	 * This is not a remount. Even if MS_REMOUNT was requested,
	 * the caller needs to proceed as it would on an ordinary
	 * mount.
	 */
	*remounting = 0;

	ASSERT(*pseudodev);
	return (0);
}

/*
 * Get the PCFS-specific mount options from the VFS framework.
 * For "timezone" and "secsize", we need to parse the number
 * ourselves and ensure its validity.
 * Note: "secsize" is deliberately undocumented at this time,
 * it's a workaround for devices (particularly: lofi image files)
 * that don't support the DKIOCGMEDIAINFO ioctl for autodetection.
 */
static void
pcfs_parse_mntopts(struct pcfs *fsp)
{
	char *c;
	char *endptr;
	long l;
	struct vfs *vfsp = fsp->pcfs_vfs;

	ASSERT(fsp->pcfs_secondswest == 0);
	ASSERT(fsp->pcfs_secsize == 0);

	if (vfs_optionisset(vfsp, MNTOPT_PCFS_HIDDEN, NULL))
		fsp->pcfs_flags |= PCFS_HIDDEN;
	if (vfs_optionisset(vfsp, MNTOPT_PCFS_FOLDCASE, NULL))
		fsp->pcfs_flags |= PCFS_FOLDCASE;
	if (vfs_optionisset(vfsp, MNTOPT_PCFS_NOCLAMPTIME, NULL))
		fsp->pcfs_flags |= PCFS_NOCLAMPTIME;
	if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL))
		fsp->pcfs_flags |= PCFS_NOATIME;

	if (vfs_optionisset(vfsp, MNTOPT_PCFS_TIMEZONE, &c)) {
		if (ddi_strtol(c, &endptr, 10, &l) == 0 &&
		    endptr == c + strlen(c)) {
			/*
			 * A number alright - in the allowed range ?
			 */
			if (l <= -12*3600 || l >= 12*3600) {
				cmn_err(CE_WARN, "!pcfs: invalid use of "
				    "'timezone' mount option - %ld "
				    "is out of range. Assuming 0.", l);
				l = 0;
			}
		} else {
			cmn_err(CE_WARN, "!pcfs: invalid use of "
			    "'timezone' mount option - argument %s "
			    "is not a valid number. Assuming 0.", c);
			l = 0;
		}
		fsp->pcfs_secondswest = l;
	}

	/*
	 * The "secsize=..." mount option is a workaround for the lack of
	 * lofi(7d) support for DKIOCGMEDIAINFO. If PCFS wants to parse the
	 * partition table of a disk image and it has been partitioned with
	 * sector sizes other than 512 bytes, we'd fail on loopback'ed disk
	 * images.
	 * That should really be fixed in lofi ... this is a workaround.
	 */
	if (vfs_optionisset(vfsp, MNTOPT_PCFS_SECSIZE, &c)) {
		if (ddi_strtol(c, &endptr, 10, &l) == 0 &&
		    endptr == c + strlen(c)) {
			/*
			 * A number alright - a valid sector size as well ?
			 */
			if (!VALID_SECSIZE(l)) {
				cmn_err(CE_WARN, "!pcfs: invalid use of "
				    "'secsize' mount option - %ld is "
				    "unsupported. Autodetecting.", l);
				l = 0;
			}
		} else {
			cmn_err(CE_WARN, "!pcfs: invalid use of "
			    "'secsize' mount option - argument %s "
			    "is not a valid number. Autodetecting.", c);
			l = 0;
		}
		fsp->pcfs_secsize = l;
		fsp->pcfs_sdshift = ddi_ffs(l / DEV_BSIZE) - 1;
	}
}

/*
 * vfs operations
 */

/*
 * pcfs_mount - backend for VFS_MOUNT() on PCFS.
 */
static int
pcfs_mount(
	struct vfs *vfsp,
	struct vnode *mvp,
	struct mounta *uap,
	struct cred *cr)
{
	struct pcfs *fsp;
	struct vnode *devvp;
	dev_t pseudodev;
	dev_t xdev;
	int dos_ldrive = 0;
	int error;
	int remounting;