xref: /illumos-gate/usr/src/uts/common/fs/zfs/zfs_vnops.c (revision 99d0d3f5)

/*
 * 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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
 * Copyright (c) 2014 Integros [integros.com]
 * Copyright 2020 Joyent, Inc.
 * Copyright 2017 Nexenta Systems, Inc.
 */

/* Portions Copyright 2007 Jeremy Teo */
/* Portions Copyright 2010 Robert Milkowski */

#include <sys/types.h>
#include <sys/param.h>
#include <sys/time.h>
#include <sys/systm.h>
#include <sys/sysmacros.h>
#include <sys/resource.h>
#include <sys/vfs.h>
#include <sys/vfs_opreg.h>
#include <sys/vnode.h>
#include <sys/file.h>
#include <sys/stat.h>
#include <sys/kmem.h>
#include <sys/taskq.h>
#include <sys/uio.h>
#include <sys/vmsystm.h>
#include <sys/atomic.h>
#include <sys/vm.h>
#include <vm/seg_vn.h>
#include <vm/pvn.h>
#include <vm/as.h>
#include <vm/kpm.h>
#include <vm/seg_kpm.h>
#include <sys/mman.h>
#include <sys/pathname.h>
#include <sys/cmn_err.h>
#include <sys/errno.h>
#include <sys/unistd.h>
#include <sys/zfs_dir.h>
#include <sys/zfs_acl.h>
#include <sys/zfs_ioctl.h>
#include <sys/fs/zfs.h>
#include <sys/dmu.h>
#include <sys/dmu_objset.h>
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/dbuf.h>
#include <sys/zap.h>
#include <sys/sa.h>
#include <sys/dirent.h>
#include <sys/policy.h>
#include <sys/sunddi.h>
#include <sys/filio.h>
#include <sys/sid.h>
#include "fs/fs_subr.h"
#include <sys/zfs_ctldir.h>
#include <sys/zfs_fuid.h>
#include <sys/zfs_sa.h>
#include <sys/dnlc.h>
#include <sys/zfs_rlock.h>
#include <sys/extdirent.h>
#include <sys/kidmap.h>
#include <sys/cred.h>
#include <sys/attr.h>
#include <sys/zil.h>
#include <sys/sa_impl.h>
#include <sys/zfs_project.h>

/*
 * Programming rules.
 *
 * Each vnode op performs some logical unit of work.  To do this, the ZPL must
 * properly lock its in-core state, create a DMU transaction, do the work,
 * record this work in the intent log (ZIL), commit the DMU transaction,
 * and wait for the intent log to commit if it is a synchronous operation.
 * Moreover, the vnode ops must work in both normal and log replay context.
 * The ordering of events is important to avoid deadlocks and references
 * to freed memory.  The example below illustrates the following Big Rules:
 *
 *  (1)	A check must be made in each zfs thread for a mounted file system.
 *	This is done avoiding races using ZFS_ENTER(zfsvfs).
 *	A ZFS_EXIT(zfsvfs) is needed before all returns.  Any znodes
 *	must be checked with ZFS_VERIFY_ZP(zp).  Both of these macros
 *	can return EIO from the calling function.
 *
 *  (2)	VN_RELE() should always be the last thing except for zil_commit()
 *	(if necessary) and ZFS_EXIT(). This is for 3 reasons:
 *	First, if it's the last reference, the vnode/znode
 *	can be freed, so the zp may point to freed memory.  Second, the last
 *	reference will call zfs_zinactive(), which may induce a lot of work --
 *	pushing cached pages (which acquires range locks) and syncing out
 *	cached atime changes.  Third, zfs_zinactive() may require a new tx,
 *	which could deadlock the system if you were already holding one.
 *	If you must call VN_RELE() within a tx then use VN_RELE_ASYNC().
 *
 *  (3)	All range locks must be grabbed before calling dmu_tx_assign(),
 *	as they can span dmu_tx_assign() calls.
 *
 *  (4) If ZPL locks are held, pass TXG_NOWAIT as the second argument to
 *      dmu_tx_assign().  This is critical because we don't want to block
 *      while holding locks.
 *
 *	If no ZPL locks are held (aside from ZFS_ENTER()), use TXG_WAIT.  This
 *	reduces lock contention and CPU usage when we must wait (note that if
 *	throughput is constrained by the storage, nearly every transaction
 *	must wait).
 *
 *      Note, in particular, that if a lock is sometimes acquired before
 *      the tx assigns, and sometimes after (e.g. z_lock), then failing
 *      to use a non-blocking assign can deadlock the system.  The scenario:
 *
 *	Thread A has grabbed a lock before calling dmu_tx_assign().
 *	Thread B is in an already-assigned tx, and blocks for this lock.
 *	Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open()
 *	forever, because the previous txg can't quiesce until B's tx commits.
 *
 *	If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT,
 *	then drop all locks, call dmu_tx_wait(), and try again.  On subsequent
 *	calls to dmu_tx_assign(), pass TXG_NOTHROTTLE in addition to TXG_NOWAIT,
 *	to indicate that this operation has already called dmu_tx_wait().
 *	This will ensure that we don't retry forever, waiting a short bit
 *	each time.
 *
 *  (5)	If the operation succeeded, generate the intent log entry for it
 *	before dropping locks.  This ensures that the ordering of events
 *	in the intent log matches the order in which they actually occurred.
 *	During ZIL replay the zfs_log_* functions will update the sequence
 *	number to indicate the zil transaction has replayed.
 *
 *  (6)	At the end of each vnode op, the DMU tx must always commit,
 *	regardless of whether there were any errors.
 *
 *  (7)	After dropping all locks, invoke zil_commit(zilog, foid)
 *	to ensure that synchronous semantics are provided when necessary.
 *
 * In general, this is how things should be ordered in each vnode op:
 *
 *	ZFS_ENTER(zfsvfs);		// exit if unmounted
 * top:
 *	zfs_dirent_lock(&dl, ...)	// lock directory entry (may VN_HOLD())
 *	rw_enter(...);			// grab any other locks you need
 *	tx = dmu_tx_create(...);	// get DMU tx
 *	dmu_tx_hold_*();		// hold each object you might modify
 *	error = dmu_tx_assign(tx, (waited ? TXG_NOTHROTTLE : 0) | TXG_NOWAIT);
 *	if (error) {
 *		rw_exit(...);		// drop locks
 *		zfs_dirent_unlock(dl);	// unlock directory entry
 *		VN_RELE(...);		// release held vnodes
 *		if (error == ERESTART) {
 *			waited = B_TRUE;
 *			dmu_tx_wait(tx);
 *			dmu_tx_abort(tx);
 *			goto top;
 *		}
 *		dmu_tx_abort(tx);	// abort DMU tx
 *		ZFS_EXIT(zfsvfs);	// finished in zfs
 *		return (error);		// really out of space
 *	}
 *	error = do_real_work();		// do whatever this VOP does
 *	if (error == 0)
 *		zfs_log_*(...);		// on success, make ZIL entry
 *	dmu_tx_commit(tx);		// commit DMU tx -- error or not
 *	rw_exit(...);			// drop locks
 *	zfs_dirent_unlock(dl);		// unlock directory entry
 *	VN_RELE(...);			// release held vnodes
 *	zil_commit(zilog, foid);	// synchronous when necessary
 *	ZFS_EXIT(zfsvfs);		// finished in zfs
 *	return (error);			// done, report error
 */

/* ARGSUSED */
static int
zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
{
	znode_t	*zp = VTOZ(*vpp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) &&
	    ((flag & FAPPEND) == 0)) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EPERM));
	}

	if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
	    ZTOV(zp)->v_type == VREG &&
	    !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) {
		if (fs_vscan(*vpp, cr, 0) != 0) {
			ZFS_EXIT(zfsvfs);
			return (SET_ERROR(EACCES));
		}
	}

	/* Keep a count of the synchronous opens in the znode */
	if (flag & (FSYNC | FDSYNC))
		atomic_inc_32(&zp->z_sync_cnt);

	ZFS_EXIT(zfsvfs);
	return (0);
}

/* ARGSUSED */
static int
zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
    caller_context_t *ct)
{
	znode_t	*zp = VTOZ(vp);
	zfsvfs_t *zfsvfs = zp->z_zfsvfs;

	/*
	 * Clean up any locks held by this process on the vp.
	 */
	cleanlocks(vp, ddi_get_pid(), 0);
	cleanshares(vp, ddi_get_pid());

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	/* Decrement the synchronous opens in the znode */
	if ((flag & (FSYNC | FDSYNC)) && (count == 1))
		atomic_dec_32(&zp->z_sync_cnt);

	if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan &&
	    ZTOV(zp)->v_type == VREG &&
	    !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0)
		VERIFY(fs_vscan(vp, cr, 1) == 0);

	ZFS_EXIT(zfsvfs);
	return (0);
}

/*
 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and
 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter.
 */
static int
zfs_holey(vnode_t *vp, int cmd, offset_t *off)
{
	znode_t	*zp = VTOZ(vp);
	uint64_t noff = (uint64_t)*off; /* new offset */
	uint64_t file_sz;
	int error;
	boolean_t hole;

	file_sz = zp->z_size;
	if (noff >= file_sz)  {
		return (SET_ERROR(ENXIO));
	}

	if (cmd == _FIO_SEEK_HOLE)
		hole = B_TRUE;
	else
		hole = B_FALSE;

	error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff);

	if (error == ESRCH)
		return (SET_ERROR(ENXIO));

	/*
	 * We could find a hole that begins after the logical end-of-file,
	 * because dmu_offset_next() only works on whole blocks.  If the
	 * EOF falls mid-block, then indicate that the "virtual hole"
	 * at the end of the file begins at the logical EOF, rather than
	 * at the end of the last block.
	 */
	if (noff > file_sz) {
		ASSERT(hole);
		noff = file_sz;
	}

	if (noff < *off)
		return (error);
	*off = noff;
	return (error);
}

static int
zfs_ioctl_getxattr(vnode_t *vp, intptr_t data, int flag, cred_t *cr,
    caller_context_t *ct)
{
	zfsxattr_t fsx = { 0 };
	znode_t *zp = VTOZ(vp);

	if (zp->z_pflags & ZFS_PROJINHERIT)
		fsx.fsx_xflags = ZFS_PROJINHERIT_FL;
	if (zp->z_pflags & ZFS_PROJID)
		fsx.fsx_projid = zp->z_projid;
	if (ddi_copyout(&fsx, (void *)data, sizeof (fsx), flag))
		return (SET_ERROR(EFAULT));

	return (0);
}

static int zfs_setattr(vnode_t *, vattr_t *, int, cred_t *, caller_context_t *);

static int
zfs_ioctl_setxattr(vnode_t *vp, intptr_t data, int flags, cred_t *cr,
    caller_context_t *ct)
{
	znode_t *zp = VTOZ(vp);
	zfsxattr_t fsx;
	xvattr_t xva;
	xoptattr_t *xoap;
	int err;

	if (ddi_copyin((void *)data, &fsx, sizeof (fsx), flags))
		return (SET_ERROR(EFAULT));

	if (!zpl_is_valid_projid(fsx.fsx_projid))
		return (SET_ERROR(EINVAL));

	if (fsx.fsx_xflags & ~ZFS_PROJINHERIT_FL)
		return (SET_ERROR(EOPNOTSUPP));

	xva_init(&xva);
	xoap = xva_getxoptattr(&xva);

	XVA_SET_REQ(&xva, XAT_PROJINHERIT);
	if (fsx.fsx_xflags & ZFS_PROJINHERIT_FL)
		xoap->xoa_projinherit = B_TRUE;

	XVA_SET_REQ(&xva, XAT_PROJID);
	xoap->xoa_projid = fsx.fsx_projid;

	return (zfs_setattr(vp, (vattr_t *)&xva, flags, cr, ct));
}

/* ARGSUSED */
static int
zfs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred,
    int *rvalp, caller_context_t *ct)
{
	offset_t off;
	offset_t ndata;
	dmu_object_info_t doi;
	int error;
	zfsvfs_t *zfsvfs;
	znode_t *zp;

	switch (com) {
	case _FIOFFS:
	{
		return (zfs_sync(vp->v_vfsp, 0, cred));

		/*
		 * The following two ioctls are used by bfu.  Faking out,
		 * necessary to avoid bfu errors.
		 */
	}
	case _FIOGDIO:
	case _FIOSDIO:
	{
		return (0);
	}

	case _FIODIRECTIO:
	{
		/*
		 * ZFS inherently provides the basic semantics for directio.
		 * This is the summary from the ZFS on Linux support for
		 * O_DIRECT, which is the common form of directio, and required
		 * no changes to ZFS.
		 *
		 * 1. Minimize cache effects of the I/O.
		 *
		 *    By design the ARC is already scan-resistant, which helps
		 *    mitigate the need for special O_DIRECT handling.
		 *
		 * 2. O_DIRECT _MAY_ impose restrictions on IO alignment and
		 *    length.
		 *
		 *    No additional alignment or length restrictions are
		 *    imposed by ZFS.
		 *
		 * 3. O_DIRECT _MAY_ perform unbuffered IO operations directly
		 *    between user memory and block device.
		 *
		 *    No unbuffered IO operations are currently supported. In
		 *    order to support features such as compression, encryption,
		 *    and checksumming a copy must be made to transform the
		 *    data.
		 *
		 * 4. O_DIRECT _MAY_ imply O_DSYNC (XFS).
		 *
		 *    O_DIRECT does not imply O_DSYNC for ZFS.
		 *
		 * 5. O_DIRECT _MAY_ disable file locking that serializes IO
		 *    operations.
		 *
		 *    All I/O in ZFS is locked for correctness and this locking
		 *    is not disabled by O_DIRECT.
		 */
		return (0);
	}

	case _FIO_SEEK_DATA:
	case _FIO_SEEK_HOLE:
	{
		if (ddi_copyin((void *)data, &off, sizeof (off), flag))
			return (SET_ERROR(EFAULT));

		zp = VTOZ(vp);
		zfsvfs = zp->z_zfsvfs;
		ZFS_ENTER(zfsvfs);
		ZFS_VERIFY_ZP(zp);

		/* offset parameter is in/out */
		error = zfs_holey(vp, com, &off);
		ZFS_EXIT(zfsvfs);
		if (error)
			return (error);
		if (ddi_copyout(&off, (void *)data, sizeof (off), flag))
			return (SET_ERROR(EFAULT));
		return (0);
	}
	case _FIO_COUNT_FILLED:
	{
		/*
		 * _FIO_COUNT_FILLED adds a new ioctl command which
		 * exposes the number of filled blocks in a
		 * ZFS object.
		 */
		zp = VTOZ(vp);
		zfsvfs = zp->z_zfsvfs;
		ZFS_ENTER(zfsvfs);
		ZFS_VERIFY_ZP(zp);

		/*
		 * Wait for all dirty blocks for this object
		 * to get synced out to disk, and the DMU info
		 * updated.
		 */
		error = dmu_object_wait_synced(zfsvfs->z_os, zp->z_id);
		if (error) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}

		/*
		 * Retrieve fill count from DMU object.
		 */
		error = dmu_object_info(zfsvfs->z_os, zp->z_id, &doi);
		if (error) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}

		ndata = doi.doi_fill_count;

		ZFS_EXIT(zfsvfs);
		if (ddi_copyout(&ndata, (void *)data, sizeof (ndata), flag))
			return (SET_ERROR(EFAULT));
		return (0);
	}
	case ZFS_IOC_FSGETXATTR:
		return (zfs_ioctl_getxattr(vp, data, flag, cred, ct));
	case ZFS_IOC_FSSETXATTR:
		return (zfs_ioctl_setxattr(vp, data, flag, cred, ct));
	}
	return (SET_ERROR(ENOTTY));
}

/*
 * Utility functions to map and unmap a single physical page.  These
 * are used to manage the mappable copies of ZFS file data, and therefore
 * do not update ref/mod bits.
 */
caddr_t
zfs_map_page(page_t *pp, enum seg_rw rw)
{
	if (kpm_enable)
		return (hat_kpm_mapin(pp, 0));
	ASSERT(rw == S_READ || rw == S_WRITE);
	return (ppmapin(pp, PROT_READ | ((rw == S_WRITE) ? PROT_WRITE : 0),
	    (caddr_t)-1));
}

void
zfs_unmap_page(page_t *pp, caddr_t addr)
{
	if (kpm_enable) {
		hat_kpm_mapout(pp, 0, addr);
	} else {
		ppmapout(addr);
	}
}

/*
 * When a file is memory mapped, we must keep the IO data synchronized
 * between the DMU cache and the memory mapped pages.  What this means:
 *
 * On Write:	If we find a memory mapped page, we write to *both*
 *		the page and the dmu buffer.
 */
static void
update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid)
{
	int64_t	off;

	off = start & PAGEOFFSET;
	for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
		page_t *pp;
		uint64_t nbytes = MIN(PAGESIZE - off, len);

		if (pp = page_lookup(vp, start, SE_SHARED)) {
			caddr_t va;

			va = zfs_map_page(pp, S_WRITE);
			(void) dmu_read(os, oid, start+off, nbytes, va+off,
			    DMU_READ_PREFETCH);
			zfs_unmap_page(pp, va);
			page_unlock(pp);
		}
		len -= nbytes;
		off = 0;
	}
}

/*
 * When a file is memory mapped, we must keep the IO data synchronized
 * between the DMU cache and the memory mapped pages.  What this means:
 *
 * On Read:	We "read" preferentially from memory mapped pages,
 *		else we default from the dmu buffer.
 *
 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when
 *	 the file is memory mapped.
 */
static int
mappedread(vnode_t *vp, int nbytes, uio_t *uio)
{
	znode_t *zp = VTOZ(vp);
	int64_t	start, off;
	int len = nbytes;
	int error = 0;

	start = uio->uio_loffset;
	off = start & PAGEOFFSET;
	for (start &= PAGEMASK; len > 0; start += PAGESIZE) {
		page_t *pp;
		uint64_t bytes = MIN(PAGESIZE - off, len);

		if (pp = page_lookup(vp, start, SE_SHARED)) {
			caddr_t va;

			va = zfs_map_page(pp, S_READ);
			error = uiomove(va + off, bytes, UIO_READ, uio);
			zfs_unmap_page(pp, va);
			page_unlock(pp);
		} else {
			error = dmu_read_uio_dbuf(sa_get_db(zp->z_sa_hdl),
			    uio, bytes);
		}
		len -= bytes;
		off = 0;
		if (error)
			break;
	}
	return (error);
}

offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */

/*
 * Read bytes from specified file into supplied buffer.
 *
 *	IN:	vp	- vnode of file to be read from.
 *		uio	- structure supplying read location, range info,
 *			  and return buffer.
 *		ioflag	- SYNC flags; used to provide FRSYNC semantics.
 *		cr	- credentials of caller.
 *		ct	- caller context
 *
 *	OUT:	uio	- updated offset and range, buffer filled.
 *
 *	RETURN:	0 on success, error code on failure.
 *
 * Side Effects:
 *	vp - atime updated if byte count > 0
 */
/* ARGSUSED */
static int
zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct)
{
	znode_t		*zp = VTOZ(vp);
	zfsvfs_t	*zfsvfs = zp->z_zfsvfs;
	ssize_t		n, nbytes;
	int		error = 0;
	boolean_t	frsync = B_FALSE;
	xuio_t		*xuio = NULL;

	ZFS_ENTER(zfsvfs);
	ZFS_VERIFY_ZP(zp);

	if (zp->z_pflags & ZFS_AV_QUARANTINED) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EACCES));
	}

	/*
	 * Validate file offset
	 */
	if (uio->uio_loffset < (offset_t)0) {
		ZFS_EXIT(zfsvfs);
		return (SET_ERROR(EINVAL));
	}

	/*
	 * Fasttrack empty reads
	 */
	if (uio->uio_resid == 0) {
		ZFS_EXIT(zfsvfs);
		return (0);
	}

	/*
	 * Check for mandatory locks
	 */
	if (MANDMODE(zp->z_mode)) {
		if (error = chklock(vp, FREAD,
		    uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) {
			ZFS_EXIT(zfsvfs);
			return (error);
		}
	}

#ifdef FRSYNC
	/*
	 * If we're in FRSYNC mode, sync out this znode before reading it.
	 * Only do this for non-snapshots.
	 *
	 * Some platforms do not support FRSYNC and instead map it
	 * to FSYNC, which results in unnecessary calls to zil_commit. We
	 * only honor FRSYNC requests on platforms which support it.
	 */
	frsync = !!(ioflag & FRSYNC);
#endif

	if (zfsvfs->z_log &&
	    (frsync || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS))
		zil_commit(zfsvfs->z_log, zp->z_id);

	/*
	 * Lock the range against changes.
	 */
	locked_range_t *lr = rangelock_enter(&zp->z_rangelock,
	    uio->uio_loffset, uio->uio_resid, RL_READER);