xref: /onnv/onnv-gate/usr/src/uts/common/contract/device.c

/*
 * 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.
 */

#pragma ident	"%Z%%M%	%I%	%E% SMI"

#include <sys/mutex.h>
#include <sys/debug.h>
#include <sys/types.h>
#include <sys/param.h>
#include <sys/kmem.h>
#include <sys/thread.h>
#include <sys/id_space.h>
#include <sys/avl.h>
#include <sys/list.h>
#include <sys/sysmacros.h>
#include <sys/proc.h>
#include <sys/contract.h>
#include <sys/contract_impl.h>
#include <sys/contract/device.h>
#include <sys/contract/device_impl.h>
#include <sys/cmn_err.h>
#include <sys/nvpair.h>
#include <sys/policy.h>
#include <sys/ddi_impldefs.h>
#include <sys/ddi_implfuncs.h>
#include <sys/systm.h>
#include <sys/stat.h>
#include <sys/sunddi.h>
#include <sys/esunddi.h>
#include <sys/ddi.h>
#include <sys/fs/dv_node.h>
#include <sys/sunndi.h>
#undef ct_lock	/* needed because clnt.h defines ct_lock as a macro */

/*
 * Device Contracts
 * -----------------
 * This file contains the core code for the device contracts framework.
 * A device contract is an agreement or a contract between a process and
 * the kernel regarding the state of the device. A device contract may be
 * created when a relationship is formed between a device and a process
 * i.e. at open(2) time, or it may be created at some point after the device
 * has been opened. A device contract once formed may be broken by either party.
 * A device contract can be broken by the process by an explicit abandon of the
 * contract or by an implicit abandon when the process exits. A device contract
 * can be broken by the kernel either asynchronously (without negotiation) or
 * synchronously (with negotiation). Exactly which happens depends on the device
 * state transition. The following state diagram shows the transitions between
 * device states. Only device state transitions currently supported by device
 * contracts is shown.
 *
 *                              <-- A -->
 *                       /-----------------> DEGRADED
 *                       |                      |
 *                       |                      |
 *                       |                      | S
 *                       |                      | |
 *                       |                      | v
 *                       v       S -->          v
 *                      ONLINE ------------> OFFLINE
 *
 *
 * In the figure above, the arrows indicate the direction of transition. The
 * letter S refers to transitions which are inherently synchronous i.e.
 * require negotiation and the letter A indicates transitions which are
 * asynchronous i.e. are done without contract negotiations. A good example
 * of a synchronous transition is the ONLINE -> OFFLINE transition. This
 * transition cannot happen as long as there are consumers which have the
 * device open. Thus some form of negotiation needs to happen between the
 * consumers and the kernel to ensure that consumers either close devices
 * or disallow the move to OFFLINE. Certain other transitions such as
 * ONLINE --> DEGRADED for example, are inherently asynchronous i.e.
 * non-negotiable. A device that suffers a fault that degrades its
 * capabilities will become degraded irrespective of what consumers it has,
 * so a negotiation in this case is pointless.
 *
 * The following device states are currently defined for device contracts:
 *
 *      CT_DEV_EV_ONLINE
 *              The device is online and functioning normally
 *      CT_DEV_EV_DEGRADED
 *              The device is online but is functioning in a degraded capacity
 *      CT_DEV_EV_OFFLINE
 *              The device is offline and is no longer configured
 *
 * A typical consumer of device contracts starts out with a contract
 * template and adds terms to that template. These include the
 * "acceptable set" (A-set) term, which is a bitset of device states which
 * are guaranteed by the contract. If the device moves out of a state in
 * the A-set, the contract is broken. The breaking of the contract can
 * be asynchronous in which case a critical contract event is sent to the
 * contract holder but no negotiations take place. If the breaking of the
 * contract is synchronous, negotations are opened between the affected
 * consumer and the kernel. The kernel does this by sending a critical
 * event to the consumer with the CTE_NEG flag set indicating that this
 * is a negotiation event. The consumer can accept this change by sending
 * a ACK message to the kernel. Alternatively, if it has the necessary
 * privileges, it can send a NACK message to the kernel which will block
 * the device state change. To NACK a negotiable event, a process must
 * have the {PRIV_SYS_DEVICES} privilege asserted in its effective set.
 *
 * Other terms include the "minor path" term, specified explicitly if the
 * contract is not being created at open(2) time or specified implicitly
 * if the contract is being created at open time via an activated template.
 *
 * A contract event is sent on any state change to which the contract
 * owner has subscribed via the informative or critical event sets. Only
 * critical events are guaranteed to be delivered. Since all device state
 * changes are controlled by the kernel and cannot be arbitrarily generated
 * by a non-privileged user, the {PRIV_CONTRACT_EVENT} privilege does not
 * need to be asserted in a process's effective set to designate an event as
 * critical. To ensure privacy, a process must either have the same effective
 * userid as the contract holder or have the {PRIV_CONTRACT_OBSERVER} privilege
 * asserted in its effective set in order to observe device contract events
 * off the device contract type specific endpoint.
 *
 * Yet another term available with device contracts is the "non-negotiable"
 * term. This term is used to pre-specify a NACK to any contract negotiation.
 * This term is ignored for asynchronous state changes. For example, a
 * provcess may have the A-set {ONLINE|DEGRADED} and make the contract
 * non-negotiable. In this case, the device contract framework assumes a
 * NACK for any transition to OFFLINE and blocks the offline. If the A-set
 * is {ONLINE} and the non-negotiable term is set, transitions to OFFLINE
 * are NACKed but transitions to DEGRADE succeed.
 *
 * The OFFLINE negotiation (if OFFLINE state is not in the A-set for a contract)
 * happens just before the I/O framework attempts to offline a device
 * (i.e. detach a device and set the offline flag so that it cannot be
 * reattached). A device contract holder is expected to either NACK the offline
 * (if privileged) or release the device and allow the offline to proceed.
 *
 * The DEGRADE contract event (if DEGRADE is not in the A-set for a contract)
 * is generated just before the I/O framework transitions the device state
 * to "degraded" (i.e. DEVI_DEVICE_DEGRADED in I/O framework terminology).
 *
 * The contract holder is expected to ACK or NACK a negotiation event
 * within a certain period of time. If the ACK/NACK is not received
 * within the timeout period, the device contract framework will behave
 * as if the contract does not exist and will proceed with the event.
 *
 * Unlike a process contract a device contract does not need to exist
 * once it is abandoned, since it does not define a fault boundary. It
 * merely represents an agreement between a process and the kernel
 * regarding the state of the device. Once the process has abandoned
 * the contract (either implicitly via a process exit or explicitly)
 * the kernel has no reason to retain the contract. As a result
 * device contracts are neither inheritable nor need to exist in an
 * orphan state.
 *
 * A device unlike a process may exist in multiple contracts and has
 * a "life" outside a device contract. A device unlike a process
 * may exist without an associated contract. Unlike a process contract
 * a device contract may be formed after a binding relationship is
 * formed between a process and a device.
 *
 *	IMPLEMENTATION NOTES
 *	====================
 * DATA STRUCTURES
 * ----------------
 * 	The heart of the device contracts implementation is the device contract
 * 	private cont_device_t (or ctd for short) data structure. It encapsulates
 * 	the generic contract_t data structure and has a number of private
 *	fields.
 * 	These include:
 *		cond_minor: The minor device that is the subject of the contract
 *		cond_aset:  The bitset of states which are guaranteed by the
 *			   contract
 *		cond_noneg: If set, indicates that the result of negotiation has
 *			    been predefined to be a NACK
 * 	In addition, there are other device identifiers such the devinfo node,
 * 	dev_t and spec_type of the minor node. There are also a few fields that
 * 	are used during negotiation to maintain state. See
 *		uts/common/sys/contract/device_impl.h
 * 	for details.
 * 	The ctd structure represents the device private part of a contract of
 * 	type "device"
 *
 * 	Another data structure used by device contracts is ctmpl_device. It is
 * 	the device contracts private part of the contract template structure. It
 *	encapsulates the generic template structure "ct_template_t" and includes
 *	the following device contract specific fields
 *		ctd_aset:   The bitset of states that should be guaranteed by a
 *			    contract
 *		ctd_noneg:  If set, indicates that contract should NACK a
 *			    negotiation
 *		ctd_minor:  The devfs_path (without the /devices prefix) of the
 *			    minor node that is the subject of the contract.
 *
 * ALGORITHMS
 * ---------
 * There are three sets of routines in this file
 * 	Template related routines
 * 	-------------------------
 *	These routines provide support for template related operations initated
 *	via the generic template operations. These include routines that dup
 *	a template, free it, and set various terms in the template
 *	(such as the minor node path, the acceptable state set (or A-set)
 *	and the non-negotiable term) as well as a routine to query the
 *	device specific portion of the template for the abovementioned terms.
 *	There is also a routine to create (ctmpl_device_create) that is used to
 *	create a contract from a template. This routine calls (after initial
 *	setup) the common function used to create a device contract
 *	(contract_device_create).
 *
 *	core device contract implementation
 *	----------------------------------
 *	These routines support the generic contract framework to provide
 *	functionality that allows contracts to be created, managed and
 *	destroyed. The contract_device_create() routine is a routine used
 *	to create a contract from a template (either via an explicit create
 *	operation on a template or implicitly via an open with an
 *	activated template.). The contract_device_free() routine assists
 *	in freeing the device contract specific parts. There are routines
 *	used to abandon (contract_device_abandon) a device contract as well
 *	as a routine to destroy (which despite its name does not destroy,
 *	it only moves a contract to a dead state) a contract.
 *	There is also a routine to return status information about a
 *	contract - the level of detail depends on what is requested by the
 *	user. A value of CTD_FIXED only returns fixed length fields such
 *	as the A-set, state of device and value of the "noneg" term. If
 *	CTD_ALL is specified, the minor node path is returned as well.
 *
 *	In addition there are interfaces (contract_device_ack/nack) which
 *	are used to support negotiation between userland processes and
 *	device contracts. These interfaces record the acknowledgement
 *	or lack thereof for negotiation events and help determine if the
 *	negotiated event should occur.
 *
 *	"backend routines"
 *	-----------------
 *	The backend routines form the interface between the I/O framework
 *	and the device contract subsystem. These routines, allow the I/O
 *	framework to call into the device contract subsystem to notify it of
 *	impending changes to a device state as well as to inform of the
 *	final disposition of such attempted state changes. Routines in this
 *	class include contract_device_offline() that indicates an attempt to
 *	offline a device, contract_device_degrade() that indicates that
 *	a device is moving to the degraded state and contract_device_negend()
 *	that is used by the I/O framework to inform the contracts subsystem of
 *	the final disposition of an attempted operation.
 *
 *	SUMMARY
 *	-------
 *      A contract starts its life as a template. A process allocates a device
 *	contract template and sets various terms:
 *		The A-set
 *		The device minor node
 *		Critical and informative events
 *		The noneg i.e. no negotition term
 *	Setting of these terms in the template is done via the
 *	ctmpl_device_set() entry point in this file. A process can query a
 *	template to determine the terms already set in the template - this is
 *	facilitated by the ctmpl_device_get() routine.
 *
 *	Once all the appropriate terms are set, the contract is instantiated via
 *	one of two methods
 *	- via an explicit create operation - this is facilitated by the
 *	  ctmpl_device_create() entry point
 *	- synchronously with the open(2) system call - this is achieved via the
 *	  contract_device_open() routine.
 *	The core work for both these above functions is done by
 *	contract_device_create()
 *
 *	A contract once created can be queried for its status. Support for
 *	status info is provided by both the common contracts framework and by
 *	the "device" contract type. If the level of detail requested is
 *	CTD_COMMON, only the common contract framework data is used. Higher
 *	levels of detail result in calls to contract_device_status() to supply
 *	device contract type specific status information.
 *
 *	A contract once created may be abandoned either explicitly or implictly.
 *	In either case, the contract_device_abandon() function is invoked. This
 * 	function merely calls contract_destroy() which moves the contract to
 *	the DEAD state. The device contract portion of destroy processing is
 *	provided by contract_device_destroy() which merely disassociates the
 *	contract from its device devinfo node. A contract in the DEAD state is
 *	not freed. It hanbgs around until all references to the contract are
 *	gone. When that happens, the contract is finally deallocated. The
 *	device contract specific portion of the free is done by
 *	contract_device_free() which finally frees the device contract specific
 *	data structure (cont_device_t).
 *
 *	When a device undergoes a state change, the I/O framework calls the
 *	corresponding device contract entry point. For example, when a device
 *	is about to go OFFLINE, the routine contract_device_offline() is
 *	invoked. Similarly if a device moves to DEGRADED state, the routine
 *	contract_device_degrade() function is called. These functions call the
 *	core routine contract_device_publish(). This function determines via
 *	the function is_sync_neg() whether an event is a synchronous (i.e.
 *	negotiable) event or not. In the former case contract_device_publish()
 *	publishes a CTE_NEG event and then waits in wait_for_acks() for ACKs
 *	and/or NACKs from contract holders. In the latter case, it simply
 *	publishes the event and does not wait. In the negotiation case, ACKs or
 *	NACKs from userland consumers results in contract_device_ack_nack()
 *	being called where the result of the negotiation is recorded in the
 *	contract data structure. Once all outstanding contract owners have
 *	responded, the device contract code in wait_for_acks() determines the
 *	final result of the negotiation. A single NACK overrides all other ACKs
 *	If there is no NACK, then a single ACK will result in an overall ACK
 *	result. If there are no ACKs or NACKs, then the result CT_NONE is
 *	returned back to the I/O framework. Once the event is permitted or
 *	blocked, the I/O framework proceeds or aborts the state change. The
 *	I/O framework then calls contract_device_negend() with a result code
 *	indicating final disposition of the event. This call releases the
 *	barrier and other state associated with the previous negotiation,
 *	which permits the next event (if any) to come into the device contract
 *	framework.
 *
 *	Finally, a device that has outstanding contracts may be removed from
 *	the system which results in its devinfo node being freed. The devinfo
 *	free routine in the I/O framework, calls into the device contract
 *	function - contract_device_remove_dip(). This routine, disassociates
 *	the dip from all contracts associated with the contract being freed,
 *	allowing the devinfo node to be freed.
 *
 * LOCKING
 * ---------
 * 	There are four sets of data that need to be protected by locks
 *
 *	i) device contract specific portion of the contract template - This data
 *	is protected by the template lock ctmpl_lock.
 *
 *	ii) device contract specific portion of the contract - This data is
 *	protected by the contract lock ct_lock
 *
 *	iii) The linked list of contracts hanging off a devinfo node - This
 *	list is protected by the per-devinfo node lock devi_ct_lock
 *
 *	iv) Finally there is a barrier, controlled by devi_ct_lock, devi_ct_cv
 *	and devi_ct_count that controls state changes to a dip
 *
 *	The template lock is independent in that none of the other locks in this
 *	file may be taken while holding the template lock (and vice versa).
 *
 *	The remaining three locks have the following lock order
 *
 *	devi_ct_lock  -> ct_count barrier ->  ct_lock
 *
 */

static cont_device_t *contract_device_create(ctmpl_device_t *dtmpl, dev_t dev,
    int spec_type, proc_t *owner, int *errorp);

/* barrier routines */
static void ct_barrier_acquire(dev_info_t *dip);
static void ct_barrier_release(dev_info_t *dip);
static int ct_barrier_held(dev_info_t *dip);
static int ct_barrier_empty(dev_info_t *dip);
static void ct_barrier_wait_for_release(dev_info_t *dip);
static int ct_barrier_wait_for_empty(dev_info_t *dip, int secs);
static void ct_barrier_decr(dev_info_t *dip);
static void ct_barrier_incr(dev_info_t *dip);

ct_type_t *device_type;

/*
 * Macro predicates for determining when events should be sent and how.
 */
#define	EVSENDP(ctd, flag) \
	((ctd->cond_contract.ct_ev_info | ctd->cond_contract.ct_ev_crit) & flag)

#define	EVINFOP(ctd, flag) \
	((ctd->cond_contract.ct_ev_crit & flag) == 0)

/*
 * State transition table showing which transitions are synchronous and which
 * are not.
 */
struct ct_dev_negtable {
	uint_t	st_old;
	uint_t	st_new;
	uint_t	st_neg;
} ct_dev_negtable[] = {
	{CT_DEV_EV_ONLINE, CT_DEV_EV_OFFLINE,	1},
	{CT_DEV_EV_ONLINE, CT_DEV_EV_DEGRADED,	0},
	{CT_DEV_EV_DEGRADED, CT_DEV_EV_ONLINE,	0},
	{CT_DEV_EV_DEGRADED, CT_DEV_EV_OFFLINE,	1},
	{0}
};

/*
 * Device contract template implementation
 */

/*
 * ctmpl_device_dup
 *
 * The device contract template dup entry point.
 * This simply copies all the fields (generic as well as device contract
 * specific) fields of the original.
 */
static struct ct_template *
ctmpl_device_dup(struct ct_template *template)
{
	ctmpl_device_t *new;
	ctmpl_device_t *old = template->ctmpl_data;
	char *buf;
	char *minor;

	new = kmem_zalloc(sizeof (ctmpl_device_t), KM_SLEEP);
	buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);

	/*
	 * copy generic fields.
	 * ctmpl_copy returns with old template lock held
	 */
	ctmpl_copy(&new->ctd_ctmpl, template);

	new->ctd_ctmpl.ctmpl_data = new;
	new->ctd_aset = old->ctd_aset;
	new->ctd_minor = NULL;
	new->ctd_noneg = old->ctd_noneg;

	if (old->ctd_minor) {
		ASSERT(strlen(old->ctd_minor) + 1 <= MAXPATHLEN);
		bcopy(old->ctd_minor, buf, strlen(old->ctd_minor) + 1);
	} else {
		kmem_free(buf, MAXPATHLEN);
		buf = NULL;
	}

	mutex_exit(&template->ctmpl_lock);
	if (buf) {
		minor = i_ddi_strdup(buf, KM_SLEEP);
		kmem_free(buf, MAXPATHLEN);
		buf = NULL;
	} else {
		minor = NULL;
	}
	mutex_enter(&template->ctmpl_lock);

	if (minor) {
		new->ctd_minor = minor;
	}

	ASSERT(buf == NULL);
	return (&new->ctd_ctmpl);
}

/*
 * ctmpl_device_free
 *
 * The device contract template free entry point.  Just
 * frees the template.
 */
static void
ctmpl_device_free(struct ct_template *template)
{
	ctmpl_device_t *dtmpl = template->ctmpl_data;

	if (dtmpl->ctd_minor)
		kmem_free(dtmpl->ctd_minor, strlen(dtmpl->ctd_minor) + 1);

	kmem_free(dtmpl, sizeof (ctmpl_device_t));
}

/*
 * SAFE_EV is the set of events which a non-privileged process is
 * allowed to make critical. An unprivileged device contract owner has
 * no control over when a device changes state, so all device events
 * can be in the critical set.
 *
 * EXCESS tells us if "value", a critical event set, requires
 * additional privilege. For device contracts EXCESS currently
 * evaluates to 0.
 */
#define	SAFE_EV		(CT_DEV_ALLEVENT)
#define	EXCESS(value)	((value) & ~SAFE_EV)


/*
 * ctmpl_device_set
 *
 * The device contract template set entry point. Sets various terms in the
 * template. The non-negotiable  term can only be set if the process has
 * the {PRIV_SYS_DEVICES} privilege asserted in its effective set.
 */
static int
ctmpl_device_set(struct ct_template *tmpl, ct_param_t *param, const cred_t *cr)
{
	ctmpl_device_t *dtmpl = tmpl->ctmpl_data;
	int error;
	dev_info_t *dip;
	int spec_type;
	uint64_t param_value;
	char *str_value;

	ASSERT(MUTEX_HELD(&tmpl->ctmpl_lock));

	if (param->ctpm_id == CTDP_MINOR) {
		str_value = (char *)param->ctpm_value;
		str_value[param->ctpm_size - 1] = '\0';
	} else {
		if (param->ctpm_size < sizeof (uint64_t))
			return (EINVAL);
		param_value = *(uint64_t *)param->ctpm_value;
	}

	switch (param->ctpm_id) {
	case CTDP_ACCEPT:
		if (param_value & ~CT_DEV_ALLEVENT)
			return (EINVAL);
		if (param_value == 0)
			return (EINVAL);
		if (param_value == CT_DEV_ALLEVENT)
			return (EINVAL);

		dtmpl->ctd_aset = param_value;
		break;
	case CTDP_NONEG:
		if (param_value != CTDP_NONEG_SET &&
		    param_value != CTDP_NONEG_CLEAR)
			return (EINVAL);

		/*
		 * only privileged processes can designate a contract
		 * non-negotiatble.
		 */
		if (param_value == CTDP_NONEG_SET &&
		    (error = secpolicy_sys_devices(cr)) != 0) {
			return (error);
		}

		dtmpl->ctd_noneg = param_value;
		break;

	case CTDP_MINOR:
		if (*str_value != '/' ||
		    strncmp(str_value, "/devices/",
		    strlen("/devices/")) == 0 ||
		    strstr(str_value, "../devices/") != NULL ||
		    strchr(str_value, ':') == NULL) {
			return (EINVAL);
		}

		spec_type = 0;
		dip = NULL;
		if (resolve_pathname(str_value, &dip, NULL, &spec_type) != 0) {
			return (ERANGE);
		}
		ddi_release_devi(dip);

		if (spec_type != S_IFCHR && spec_type != S_IFBLK) {
			return (EINVAL);
		}

		if (dtmpl->ctd_minor != NULL) {
			kmem_free(dtmpl->ctd_minor,
			    strlen(dtmpl->ctd_minor) + 1);
		}
		dtmpl->ctd_minor = i_ddi_strdup(str_value, KM_SLEEP);
		break;
	case CTP_EV_CRITICAL:
		/*
		 * Currently for device contracts, any event
		 * may be added to the critical set. We retain the
		 * following code however for future enhancements.
		 */
		if (EXCESS(param_value) &&
		    (error = secpolicy_contract_event(cr)) != 0)
			return (error);
		tmpl->ctmpl_ev_crit = param_value;
		break;
	default:
		return (EINVAL);
	}

	return (0);
}

/*
 * ctmpl_device_get
 *
 * The device contract template get entry point.  Simply fetches and
 * returns the value of the requested term.
 */
static int
ctmpl_device_get(struct ct_template *template, ct_param_t *param)
{
	ctmpl_device_t *dtmpl = template->ctmpl_data;
	uint64_t *param_value = param->ctpm_value;

	ASSERT(MUTEX_HELD(&template->ctmpl_lock));

	if (param->ctpm_id == CTDP_ACCEPT ||
	    param->ctpm_id == CTDP_NONEG) {
		if (param->ctpm_size < sizeof (uint64_t))
			return (EINVAL);
		param->ctpm_size = sizeof (uint64_t);
	}

	switch (param->ctpm_id) {
	case CTDP_ACCEPT:
		*param_value = dtmpl->ctd_aset;
		break;
	case CTDP_NONEG:
		*param_value = dtmpl->ctd_noneg;
		break;
	case CTDP_MINOR:
		if (dtmpl->ctd_minor) {
			param->ctpm_size = strlcpy((char *)param->ctpm_value,
			    dtmpl->ctd_minor, param->ctpm_size);
			param->ctpm_size++;
		} else {
			return (ENOENT);
		}
		break;
	default:
		return (EINVAL);
	}

	return (0);
}

/*
 * Device contract type specific portion of creating a contract using
 * a specified template
 */
/*ARGSUSED*/
int
ctmpl_device_create(ct_template_t *template, ctid_t *ctidp)
{
	ctmpl_device_t *dtmpl;
	char *buf;
	dev_t dev;
	int spec_type;
	int error;
	cont_device_t *ctd;

	if (ctidp == NULL)
		return (EINVAL);

	buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);

	dtmpl = template->ctmpl_data;

	mutex_enter(&template->ctmpl_lock);
	if (dtmpl->ctd_minor == NULL) {
		/* incomplete template */
		mutex_exit(&template->ctmpl_lock);
		kmem_free(buf, MAXPATHLEN);
		return (EINVAL);
	} else {
		ASSERT(strlen(dtmpl->ctd_minor) < MAXPATHLEN);
		bcopy(dtmpl->ctd_minor, buf, strlen(dtmpl->ctd_minor) + 1);
	}
	mutex_exit(&template->ctmpl_lock);

	spec_type = 0;
	dev = NODEV;
	if (resolve_pathname(buf, NULL, &dev, &spec_type) != 0 ||
	    dev == NODEV || dev == DDI_DEV_T_ANY || dev == DDI_DEV_T_NONE ||
	    (spec_type != S_IFCHR && spec_type != S_IFBLK)) {
		CT_DEBUG((CE_WARN,
		    "tmpl_create: failed to find device: %s", buf));
		kmem_free(buf, MAXPATHLEN);
		return (ERANGE);
	}
	kmem_free(buf, MAXPATHLEN);

	ctd = contract_device_create(template->ctmpl_data,
	    dev, spec_type, curproc, &error);

	if (ctd == NULL) {
		CT_DEBUG((CE_WARN, "Failed to create device contract for "
		    "process (%d) with device (devt = %lu, spec_type = %s)",
		    curproc->p_pid, dev,
		    spec_type == S_IFCHR ? "S_IFCHR" : "S_IFBLK"));
		return (error);
	}

	mutex_enter(&ctd->cond_contract.ct_lock);
	*ctidp = ctd->cond_contract.ct_id;
	mutex_exit(&ctd->cond_contract.ct_lock);

	return (0);
}

/*
 * Device contract specific template entry points
 */
static ctmplops_t ctmpl_device_ops = {
	ctmpl_device_dup,		/* ctop_dup */
	ctmpl_device_free,		/* ctop_free */
	ctmpl_device_set,		/* ctop_set */
	ctmpl_device_get,		/* ctop_get */
	ctmpl_device_create,		/* ctop_create */
	CT_DEV_ALLEVENT			/* all device events bitmask */
};


/*
 * Device contract implementation
 */

/*
 * contract_device_default
 *
 * The device contract default template entry point.  Creates a
 * device contract template with a default A-set and no "noneg" ,
 * with informative degrade events and critical offline events.
 * There is no default minor path.
 */
static ct_template_t *
contract_device_default(void)
{
	ctmpl_device_t *new;

	new = kmem_zalloc(sizeof (ctmpl_device_t), KM_SLEEP);
	ctmpl_init(&new->ctd_ctmpl, &ctmpl_device_ops, device_type, new);

	new->ctd_aset = CT_DEV_EV_ONLINE | CT_DEV_EV_DEGRADED;
	new->ctd_noneg = 0;
	new->ctd_ctmpl.ctmpl_ev_info = CT_DEV_EV_DEGRADED;
	new->ctd_ctmpl.ctmpl_ev_crit = CT_DEV_EV_OFFLINE;

	return (&new->ctd_ctmpl);
}

/*
 * contract_device_free
 *
 * Destroys the device contract specific portion of a contract and
 * frees the contract.
 */
static void
contract_device_free(contract_t *ct)
{
	cont_device_t *ctd = ct->ct_data;

	ASSERT(ctd->cond_minor);
	ASSERT(strlen(ctd->cond_minor) < MAXPATHLEN);
	kmem_free(ctd->cond_minor, strlen(ctd->cond_minor) + 1);

	ASSERT(ctd->cond_devt != DDI_DEV_T_ANY &&
	    ctd->cond_devt != DDI_DEV_T_NONE && ctd->cond_devt != NODEV);

	ASSERT(ctd->cond_spec == S_IFBLK || ctd->cond_spec == S_IFCHR);

	ASSERT(!(ctd->cond_aset & ~CT_DEV_ALLEVENT));
	ASSERT(ctd->cond_noneg == 0 || ctd->cond_noneg == 1);

	ASSERT(!(ctd->cond_currev_type & ~CT_DEV_ALLEVENT));
	ASSERT(!(ctd->cond_currev_ack & ~(CT_ACK | CT_NACK)));

	ASSERT((ctd->cond_currev_id > 0) ^ (ctd->cond_currev_type == 0));
	ASSERT((ctd->cond_currev_id > 0) || (ctd->cond_currev_ack == 0));

	ASSERT(!list_link_active(&ctd->cond_next));

	kmem_free(ctd, sizeof (cont_device_t));
}

/*
 * contract_device_abandon
 *
 * The device contract abandon entry point.
 */
static void
contract_device_abandon(contract_t *ct)
{
	ASSERT(MUTEX_HELD(&ct->ct_lock));

	/*
	 * device contracts cannot be inherited or orphaned.
	 * Move the contract to the DEAD_STATE. It will be freed
	 * once all references to it are gone.
	 */
	contract_destroy(ct);
}

/*
 * contract_device_destroy
 *
 * The device contract destroy entry point.
 * Called from contract_destroy() to do any type specific destroy. Note
 * that destroy is a misnomer - this does not free the contract, it only
 * moves it to the dead state. A contract is actually freed via
 * 	contract_rele() -> contract_dtor(), contop_free()
 */
static void
contract_device_destroy(contract_t *ct)
{
	cont_device_t	*ctd = ct->ct_data;
	dev_info_t	*dip = ctd->cond_dip;

	ASSERT(MUTEX_HELD(&ct->ct_lock));

	if (dip == NULL) {
		/*
		 * The dip has been removed, this is a dangling contract
		 * Check that dip linkages are NULL
		 */
		ASSERT(!list_link_active(&ctd->cond_next));
		CT_DEBUG((CE_NOTE, "contract_device_destroy: contract has no "
		    "devinfo node. contract ctid : %d", ct->ct_id));
		return;
	}

	/*
	 * Need to have lock order: devi_ct_lock -> ct_count barrier -> ct_lock
	 */
	mutex_exit(&ct->ct_lock);

	/*
	 * Waiting for the barrier to be released is strictly speaking not
	 * necessary. But it simplifies the implementation of
	 * contract_device_publish() by establishing the invariant that
	 * device contracts cannot go away during negotiation.
	 */
	mutex_enter(&(DEVI(dip)->devi_ct_lock));
	ct_barrier_wait_for_release(dip);
	mutex_enter(&ct->ct_lock);

	list_remove(&(DEVI(dip)->devi_ct), ctd);
	ctd->cond_dip = NULL; /* no longer linked to dip */
	contract_rele(ct);	/* remove hold for dip linkage */

	mutex_exit(&ct->ct_lock);
	mutex_exit(&(DEVI(dip)->devi_ct_lock));
	mutex_enter(&ct->ct_lock);
}

/*
 * contract_device_status
 *
 * The device contract status entry point. Called when level of "detail"
 * is either CTD_FIXED or CTD_ALL
 *
 */
static void
contract_device_status(contract_t *ct, zone_t *zone, int detail, nvlist_t *nvl,
    void *status, model_t model)
{
	cont_device_t *ctd = ct->ct_data;

	ASSERT(detail == CTD_FIXED || detail == CTD_ALL);

	mutex_enter(&ct->ct_lock);
	contract_status_common(ct, zone, status, model);

	/*
	 * There's no need to hold the contract lock while accessing static
	 * data like aset or noneg. But since we need the lock to access other
	 * data like state, we hold it anyway.
	 */
	VERIFY(nvlist_add_uint32(nvl, CTDS_STATE, ctd->cond_state) == 0);
	VERIFY(nvlist_add_uint32(nvl, CTDS_ASET, ctd->cond_aset) == 0);
	VERIFY(nvlist_add_uint32(nvl, CTDS_NONEG, ctd->cond_noneg) == 0);

	if (detail == CTD_FIXED) {
		mutex_exit(&ct->ct_lock);
		return;
	}

	ASSERT(ctd->cond_minor);
	VERIFY(nvlist_add_string(nvl, CTDS_MINOR, ctd->cond_minor) == 0);

	mutex_exit(&ct->ct_lock);
}

/*
 * Converts a result integer into the corresponding string. Used for printing
 * messages
 */
static char *
result_str(uint_t result)
{
	switch (result) {
	case CT_ACK:
		return ("CT_ACK");
	case CT_NACK:
		return ("CT_NACK");
	case CT_NONE:
		return ("CT_NONE");
	default:
		return ("UNKNOWN");
	}
}

/*
 * Converts a device state integer constant into the corresponding string.
 * Used to print messages.
 */
static char *
state_str(uint_t state)
{
	switch (state) {
	case CT_DEV_EV_ONLINE:
		return ("ONLINE");
	case CT_DEV_EV_DEGRADED:
		return ("DEGRADED");
	case CT_DEV_EV_OFFLINE:
		return ("OFFLINE");
	default:
		return ("UNKNOWN");
	}
}

/*
 * Routine that determines if a particular CT_DEV_EV_? event corresponds to a
 * synchronous state change or not.
 */
static int
is_sync_neg(uint_t old, uint_t new)
{
	int	i;

	ASSERT(old & CT_DEV_ALLEVENT);
	ASSERT(new & CT_DEV_ALLEVENT);

	if (old == new) {
		CT_DEBUG((CE_WARN, "is_sync_neg: transition to same state: %s",
		    state_str(new)));
		return (-2);
	}

	for (i = 0; ct_dev_negtable[i].st_new != 0; i++) {
		if (old == ct_dev_negtable[i].st_old &&
		    new == ct_dev_negtable[i].st_new) {
			return (ct_dev_negtable[i].st_neg);
		}
	}

	CT_DEBUG((CE_WARN, "is_sync_neg: Unsupported state transition: "
	    "old = %s -> new = %s", state_str(old), state_str(new)));

	return (-1);
}

/*
 * Used to cleanup cached dv_nodes so that when a device is released by
 * a contract holder, its devinfo node can be successfully detached.
 */
static int
contract_device_dvclean(dev_info_t *dip)
{
	char		*devnm;
	dev_info_t	*pdip;
	int		error;

	ASSERT(dip);

	/* pdip can be NULL if we have contracts against the root dip */
	pdip = ddi_get_parent(dip);

	if (pdip && DEVI_BUSY_OWNED(pdip) || !pdip && DEVI_BUSY_OWNED(dip)) {
		char		*path;

		path = kmem_alloc(MAXPATHLEN, KM_SLEEP);
		(void) ddi_pathname(dip, path);
		CT_DEBUG((CE_WARN, "ct_dv_clean: Parent node is busy owned, "
		    "device=%s", path));
		kmem_free(path, MAXPATHLEN);
		return (EDEADLOCK);
	}

	if (pdip) {
		devnm = kmem_alloc(MAXNAMELEN + 1, KM_SLEEP);
		(void) ddi_deviname(dip, devnm);
		error = devfs_clean(pdip, devnm + 1, DV_CLEAN_FORCE);
		kmem_free(devnm, MAXNAMELEN + 1);
	} else {
		error = devfs_clean(dip, NULL, DV_CLEAN_FORCE);
	}

	return (error);
}

/*
 * Endpoint of a ct_ctl_ack() or ct_ctl_nack() call from userland.
 * Results in the ACK or NACK being recorded on the dip for one particular
 * contract. The device contracts framework evaluates the ACK/NACKs for all
 * contracts against a device to determine if a particular device state change
 * should be allowed.
 */
static int
contract_device_ack_nack(contract_t *ct, uint_t evtype, uint64_t evid,
    uint_t cmd)
{
	cont_device_t *ctd = ct->ct_data;
	dev_info_t *dip;
	ctid_t	ctid;
	int error;

	ctid = ct->ct_id;

	CT_DEBUG((CE_NOTE, "ack_nack: entered: ctid %d", ctid));

	mutex_enter(&ct->ct_lock);
	CT_DEBUG((CE_NOTE, "ack_nack: contract lock acquired: %d", ctid));

	dip = ctd->cond_dip;

	ASSERT(ctd->cond_minor);
	ASSERT(strlen(ctd->cond_minor) < MAXPATHLEN);

	/*
	 * Negotiation only if new state is not in A-set
	 */
	ASSERT(!(ctd->cond_aset & evtype));

	/*
	 * Negotiation only if transition is synchronous
	 */
	ASSERT(is_sync_neg(ctd->cond_state, evtype));

	/*
	 * We shouldn't be negotiating if the "noneg" flag is set
	 */
	ASSERT(!ctd->cond_noneg);

	if (dip)
		ndi_hold_devi(dip);

	mutex_exit(&ct->ct_lock);

	/*
	 * dv_clean only if !NACK and offline state change
	 */
	if (cmd != CT_NACK && evtype == CT_DEV_EV_OFFLINE && dip) {
		CT_DEBUG((CE_NOTE, "ack_nack: dv_clean: %d", ctid));
		error = contract_device_dvclean(dip);
		if (error != 0) {
			CT_DEBUG((CE_NOTE, "ack_nack: dv_clean: failed: %d",
			    ctid));
			ddi_release_devi(dip);
		}
	}

	mutex_enter(&ct->ct_lock);

	if (dip)
		ddi_release_devi(dip);

	if (dip == NULL) {
		if (ctd->cond_currev_id != evid) {
			CT_DEBUG((CE_WARN, "%sACK for non-current event "
			    "(type=%s, id=%llu) on removed device",
			    cmd == CT_NACK ? "N" : "",
			    state_str(evtype), (unsigned long long)evid));
			CT_DEBUG((CE_NOTE, "ack_nack: error: ESRCH, ctid: %d",
			    ctid));
		} else {
			ASSERT(ctd->cond_currev_type == evtype);
			CT_DEBUG((CE_WARN, "contract_ack: no such device: "
			    "ctid: %d", ctid));
		}
		error = (ct->ct_state == CTS_DEAD) ? ESRCH :
		    ((cmd == CT_NACK) ? ETIMEDOUT : 0);
		mutex_exit(&ct->ct_lock);
		return (error);
	}

	/*
	 * Must follow lock order: devi_ct_lock -> ct_count barrier - >ct_lock
	 */
	mutex_exit(&ct->ct_lock);

	mutex_enter(&DEVI(dip)->devi_ct_lock);
	mutex_enter(&ct->ct_lock);
	if (ctd->cond_currev_id != evid) {
		char *buf;
		mutex_exit(&ct->ct_lock);
		mutex_exit(&DEVI(dip)->devi_ct_lock);
		ndi_hold_devi(dip);
		buf = kmem_alloc(MAXPATHLEN, KM_SLEEP);
		(void) ddi_pathname(dip, buf);
		ddi_release_devi(dip);
		CT_DEBUG((CE_WARN, "%sACK for non-current event"
		    "(type=%s, id=%llu) on device %s",
		    cmd == CT_NACK ? "N" : "",
		    state_str(evtype), (unsigned long long)evid, buf));
		kmem_free(buf, MAXPATHLEN);
		CT_DEBUG((CE_NOTE, "ack_nack: error: %d, ctid: %d",
		    cmd == CT_NACK ? ETIMEDOUT : 0, ctid));
		return (cmd == CT_ACK ? 0 : ETIMEDOUT);
	}

	ASSERT(ctd->cond_currev_type == evtype);
	ASSERT(cmd == CT_ACK || cmd == CT_NACK);

	CT_DEBUG((CE_NOTE, "ack_nack: setting %sACK for ctid: %d",
	    cmd == CT_NACK ? "N" : "", ctid));

	ctd->cond_currev_ack = cmd;
	mutex_exit(&ct->ct_lock);

	ct_barrier_decr(dip);
	mutex_exit(&DEVI(dip)->devi_ct_lock);

	CT_DEBUG((CE_NOTE, "ack_nack: normal exit: ctid: %d", ctid));

	return (0);
}

/*
 * Invoked when a userland contract holder approves (i.e. ACKs) a state change
 */
static int
contract_device_ack(contract_t *ct, uint_t evtype, uint64_t evid)
{
	return (contract_device_ack_nack(ct, evtype, evid, CT_ACK));
}

/*
 * Invoked when a userland contract holder blocks (i.e. NACKs) a state change
 */
static int
contract_device_nack(contract_t *ct, uint_t evtype, uint64_t evid)
{
	return (contract_device_ack_nack(ct, evtype, evid, CT_NACK));
}

/*
 * Creates a new contract synchronously with the breaking of an existing
 * contract. Currently not supported.
 */
/*ARGSUSED*/
static int
contract_device_newct(contract_t *ct)
{
	return (ENOTSUP);
}

/*
 * Core device contract implementation entry points
 */
static contops_t contract_device_ops = {
	contract_device_free,		/* contop_free */
	contract_device_abandon,	/* contop_abandon */
	contract_device_destroy,	/* contop_destroy */
	contract_device_status,		/* contop_status */
	contract_device_ack,		/* contop_ack */
	contract_device_nack,		/* contop_nack */
	contract_qack_notsup,		/* contop_qack */
	contract_device_newct		/* contop_newct */
};

/*
 * contract_device_init
 *
 * Initializes the device contract type.
 */
void
contract_device_init(void)
{
	device_type = contract_type_init(CTT_DEVICE, "device",
	    &contract_device_ops,