[apple/xnu.git] / osfmk / default_pager / dp_memory_object.c

/*
 * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_LICENSE_HEADER_START@
 * 
 * The contents of this file constitute Original Code as defined in and
 * are subject to the Apple Public Source License Version 1.1 (the
 * "License").  You may not use this file except in compliance with the
 * License.  Please obtain a copy of the License at
 * http://www.apple.com/publicsource and read it before using this file.
 * 
 * This Original Code and all software distributed under the License are
 * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
 * License for the specific language governing rights and limitations
 * under the License.
 * 
 * @APPLE_LICENSE_HEADER_END@
 */
/*
 * @OSF_COPYRIGHT@
 */
/* 
 * Mach Operating System
 * Copyright (c) 1991,1990,1989 Carnegie Mellon University
 * All Rights Reserved.
 * 
 * Permission to use, copy, modify and distribute this software and its
 * documentation is hereby granted, provided that both the copyright
 * notice and this permission notice appear in all copies of the
 * software, derivative works or modified versions, and any portions
 * thereof, and that both notices appear in supporting documentation.
 * 
 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
 * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
 * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
 * 
 * Carnegie Mellon requests users of this software to return to
 * 
 *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
 *  School of Computer Science
 *  Carnegie Mellon University
 *  Pittsburgh PA 15213-3890
 * 
 * any improvements or extensions that they make and grant Carnegie Mellon
 * the rights to redistribute these changes.
 */

/*
 *      Default Pager.
 *              Memory Object Management.
 */

#include "default_pager_internal.h"
#include <default_pager/default_pager_object_server.h>
#include <mach/memory_object_default_server.h>
#include <mach/memory_object_control.h>
#include <mach/memory_object_types.h>
#include <mach/memory_object_server.h>
#include <mach/upl.h>
#include <mach/vm_map.h>
#include <vm/memory_object.h>
#include <vm/vm_pageout.h> 
#include <vm/vm_map.h>
#include <vm/vm_protos.h>

/* forward declaration */
vstruct_t vs_object_create(vm_size_t size);

/*
 * List of all vstructs.  A specific vstruct is
 * found directly via its port, this list is
 * only used for monitoring purposes by the
 * default_pager_object* calls and by ps_delete
 * when abstract memory objects must be scanned
 * to remove any live storage on a segment which
 * is to be removed.
 */
struct vstruct_list_head        vstruct_list;

__private_extern__ void
vstruct_list_insert(
        vstruct_t vs)
{
        VSL_LOCK();
        queue_enter(&vstruct_list.vsl_queue, vs, vstruct_t, vs_links);
        vstruct_list.vsl_count++;
        VSL_UNLOCK();
}


__private_extern__ void
vstruct_list_delete(
        vstruct_t vs)
{
        queue_remove(&vstruct_list.vsl_queue, vs, vstruct_t, vs_links);
        vstruct_list.vsl_count--;
}

/*
 * We use the sequence numbers on requests to regulate
 * our parallelism.  In general, we allow multiple reads and writes
 * to proceed in parallel, with the exception that reads must
 * wait for previous writes to finish.  (Because the kernel might
 * generate a data-request for a page on the heels of a data-write
 * for the same page, and we must avoid returning stale data.)
 * terminate requests wait for proceeding reads and writes to finish.
 */

static unsigned int     default_pager_total = 0;                /* debugging */
static unsigned int     default_pager_wait_seqno = 0;           /* debugging */
static unsigned int     default_pager_wait_read = 0;            /* debugging */
static unsigned int     default_pager_wait_write = 0;           /* debugging */

__private_extern__ void
vs_async_wait(
        vstruct_t       vs)
{

        ASSERT(vs->vs_async_pending >= 0);
        while (vs->vs_async_pending > 0) {
                vs->vs_waiting_async = TRUE;
                assert_wait(&vs->vs_async_pending, THREAD_UNINT);
                VS_UNLOCK(vs);
                thread_block(THREAD_CONTINUE_NULL);
                VS_LOCK(vs);
        }
        ASSERT(vs->vs_async_pending == 0);
}


#if     PARALLEL
/* 
 * Waits for correct sequence number.  Leaves pager locked.
 *
 * JMM - Sequence numbers guarantee ordering of requests generated
 *       by a single thread if the receiver is multithreaded and
 *       the interfaces are asynchronous (i.e. sender can generate
 *       more than one request before the first is received in the
 *       pager).  Normally, IPC would generate these number in that
 *       case.  But we are trying to avoid using IPC for the in-kernel
 *       scenario. Since these are actually invoked synchronously
 *       anyway (in-kernel), we can just fake the sequence number
 *       generation here (thus avoiding the dependence on IPC).
 */
__private_extern__ void
vs_lock(
        vstruct_t               vs)
{
        mach_port_seqno_t       seqno;

        default_pager_total++;
        VS_LOCK(vs);

        seqno = vs->vs_next_seqno++;

        while (vs->vs_seqno != seqno) {
                default_pager_wait_seqno++;
                vs->vs_waiting_seqno = TRUE;
                assert_wait(&vs->vs_seqno, THREAD_UNINT);
                VS_UNLOCK(vs);
                thread_block(THREAD_CONTINUE_NULL);
                VS_LOCK(vs);
        }
}

/*
 * Increments sequence number and unlocks pager.
 */
__private_extern__ void
vs_unlock(vstruct_t vs)
{
        vs->vs_seqno++;
        if (vs->vs_waiting_seqno) {
                vs->vs_waiting_seqno = FALSE;
                VS_UNLOCK(vs);
                thread_wakeup(&vs->vs_seqno);
                return;
        }
        VS_UNLOCK(vs);
}

/* 
 * Start a read - one more reader.  Pager must be locked.
 */
__private_extern__ void
vs_start_read(
        vstruct_t vs)
{
        vs->vs_readers++;
}

/*
 * Wait for readers.  Unlocks and relocks pager if wait needed.
 */
__private_extern__ void
vs_wait_for_readers(
        vstruct_t vs)
{
        while (vs->vs_readers != 0) {
                default_pager_wait_read++;
                vs->vs_waiting_read = TRUE;
                assert_wait(&vs->vs_readers, THREAD_UNINT);
                VS_UNLOCK(vs);
                thread_block(THREAD_CONTINUE_NULL);
                VS_LOCK(vs);
        }
}

/*
 * Finish a read.  Pager is unlocked and returns unlocked.
 */
__private_extern__ void
vs_finish_read(
        vstruct_t vs)
{
        VS_LOCK(vs);
        if (--vs->vs_readers == 0 && vs->vs_waiting_read) {
                vs->vs_waiting_read = FALSE;
                VS_UNLOCK(vs);
                thread_wakeup(&vs->vs_readers);
                return;
        }
        VS_UNLOCK(vs);
}

/*
 * Start a write - one more writer.  Pager must be locked.
 */
__private_extern__ void
vs_start_write(
        vstruct_t vs)
{
        vs->vs_writers++;
}

/* 
 * Wait for writers.  Unlocks and relocks pager if wait needed.
 */
__private_extern__ void
vs_wait_for_writers(
        vstruct_t vs)
{
        while (vs->vs_writers != 0) {
                default_pager_wait_write++;
                vs->vs_waiting_write = TRUE;
                assert_wait(&vs->vs_writers, THREAD_UNINT);
                VS_UNLOCK(vs);
                thread_block(THREAD_CONTINUE_NULL);
                VS_LOCK(vs);
        }
        vs_async_wait(vs);
}

/* This is to be used for the transfer from segment code ONLY */
/* The transfer code holds off vs destruction by keeping the  */
/* vs_async_wait count non-zero.  It will not ocnflict with   */
/* other writers on an async basis because it only writes on  */
/* a cluster basis into fresh (as of sync time) cluster locations */

__private_extern__ void 
vs_wait_for_sync_writers(
        vstruct_t vs)
{
        while (vs->vs_writers != 0) {
                default_pager_wait_write++;
                vs->vs_waiting_write = TRUE;
                assert_wait(&vs->vs_writers, THREAD_UNINT);
                VS_UNLOCK(vs);
                thread_block(THREAD_CONTINUE_NULL);
                VS_LOCK(vs);
        }
}       


/*
 * Finish a write.  Pager is unlocked and returns unlocked.
 */
__private_extern__ void
vs_finish_write(
        vstruct_t vs)
{
        VS_LOCK(vs);
        if (--vs->vs_writers == 0 && vs->vs_waiting_write) {
                vs->vs_waiting_write = FALSE;
                VS_UNLOCK(vs);
                thread_wakeup(&vs->vs_writers);
                return;
        }
        VS_UNLOCK(vs);
}
#endif  /* PARALLEL */

vstruct_t
vs_object_create(
        vm_size_t size)
{
        vstruct_t       vs;

        /*
         * Allocate a vstruct. If there are any problems, then report them
         * to the console.
         */
        vs = ps_vstruct_create(size);
        if (vs == VSTRUCT_NULL) {
                dprintf(("vs_object_create: unable to allocate %s\n",
                         "-- either run swapon command or reboot"));
                return VSTRUCT_NULL;
        }

        return vs;
}

#if 0
void default_pager_add(vstruct_t, boolean_t);   /* forward */

void
default_pager_add(
        vstruct_t vs,
        boolean_t internal)
{
        memory_object_t         mem_obj = vs->vs_mem_obj;
        mach_port_t             pset;
        mach_port_mscount_t     sync;
        mach_port_t             previous;
        kern_return_t           kr;
        static char             here[] = "default_pager_add";

        /*
         * The port currently has a make-send count of zero,
         * because either we just created the port or we just
         * received the port in a memory_object_create request.
         */

        if (internal) {
                /* possibly generate an immediate no-senders notification */
                sync = 0;
                pset = default_pager_internal_set;
        } else {
                /* delay notification till send right is created */
                sync = 1;
                pset = default_pager_external_set;
        }

        ipc_port_make_sonce(mem_obj);
        ip_lock(mem_obj);  /* unlocked in nsrequest below */
        ipc_port_nsrequest(mem_obj, sync, mem_obj, &previous);
}

#endif

const struct memory_object_pager_ops default_pager_ops = {
        dp_memory_object_reference,
        dp_memory_object_deallocate,
        dp_memory_object_init,
        dp_memory_object_terminate,
        dp_memory_object_data_request,
        dp_memory_object_data_return,
        dp_memory_object_data_initialize,
        dp_memory_object_data_unlock,
        dp_memory_object_synchronize,
        dp_memory_object_unmap,
        "default pager"
};

kern_return_t
dp_memory_object_init(
        memory_object_t         mem_obj,
        memory_object_control_t control,
        __unused vm_size_t pager_page_size)
{
        vstruct_t               vs;

        assert(pager_page_size == vm_page_size);

        memory_object_control_reference(control);

        vs_lookup(mem_obj, vs);
        vs_lock(vs);

        if (vs->vs_control != MEMORY_OBJECT_CONTROL_NULL)
                Panic("bad request");

        vs->vs_control = control;
        vs_unlock(vs);

        return KERN_SUCCESS;
}

kern_return_t
dp_memory_object_synchronize(
        memory_object_t         mem_obj,
        memory_object_offset_t  offset,
        vm_size_t               length,
        __unused vm_sync_t              flags)
{
        vstruct_t       vs;

        vs_lookup(mem_obj, vs);
        vs_lock(vs);
        vs_unlock(vs);

        memory_object_synchronize_completed(vs->vs_control, offset, length);

        return KERN_SUCCESS;
}

kern_return_t
dp_memory_object_unmap(
        __unused memory_object_t                mem_obj)
{
        panic("dp_memory_object_unmap");

        return KERN_FAILURE;
}

kern_return_t
dp_memory_object_terminate(
        memory_object_t         mem_obj)
{
        memory_object_control_t control;
        vstruct_t               vs;

        /* 
         * control port is a receive right, not a send right.
         */

        vs_lookup(mem_obj, vs);
        vs_lock(vs);

        /*
         * Wait for read and write requests to terminate.
         */

        vs_wait_for_readers(vs);
        vs_wait_for_writers(vs);

        /*
         * After memory_object_terminate both memory_object_init
         * and a no-senders notification are possible, so we need
         * to clean up our reference to the memory_object_control
         * to prepare for a new init.
         */

        control = vs->vs_control;
        vs->vs_control = MEMORY_OBJECT_CONTROL_NULL;

        /* a bit of special case ugliness here.  Wakeup any waiting reads */
        /* these data requests had to be removed from the seqno traffic   */
        /* based on a performance bottleneck with large memory objects    */
        /* the problem will right itself with the new component based     */
        /* synchronous interface.  The new async will be able to return   */
        /* failure during its sync phase.   In the mean time ... */

        thread_wakeup(&vs->vs_writers);
        thread_wakeup(&vs->vs_async_pending);

        vs_unlock(vs);

        /*
         * Now we deallocate our reference on the control.
         */
        memory_object_control_deallocate(control);
        return KERN_SUCCESS;
}

void
dp_memory_object_reference(
        memory_object_t         mem_obj)
{
        vstruct_t               vs;

        vs_lookup_safe(mem_obj, vs);
        if (vs == VSTRUCT_NULL)
                return;

        VS_LOCK(vs);
        assert(vs->vs_references > 0);
        vs->vs_references++;
        VS_UNLOCK(vs);
}

void
dp_memory_object_deallocate(
        memory_object_t         mem_obj)
{
        vstruct_t               vs;
        mach_port_seqno_t       seqno;

        /*
         * Because we don't give out multiple first references
         * for a memory object, there can't be a race
         * between getting a deallocate call and creating
         * a new reference for the object.
         */

        vs_lookup_safe(mem_obj, vs);
        if (vs == VSTRUCT_NULL)
                return;

        VS_LOCK(vs);
        if (--vs->vs_references > 0) {
                VS_UNLOCK(vs);
                return;
        }

        seqno = vs->vs_next_seqno++;
        while (vs->vs_seqno != seqno) {
                default_pager_wait_seqno++;
                vs->vs_waiting_seqno = TRUE;
                assert_wait(&vs->vs_seqno, THREAD_UNINT);
                VS_UNLOCK(vs);
                thread_block(THREAD_CONTINUE_NULL);
                VS_LOCK(vs);
        }

        vs_async_wait(vs);      /* wait for pending async IO */

        /* do not delete the vs structure until the referencing pointers */
        /* in the vstruct list have been expunged */

        /* get VSL_LOCK out of order by using TRY mechanism */
        while(!VSL_LOCK_TRY()) {
                VS_UNLOCK(vs);
                VSL_LOCK();
                VSL_UNLOCK();
                VS_LOCK(vs);
                vs_async_wait(vs);      /* wait for pending async IO */
        }


        /*
         * We shouldn't get a deallocation call
         * when the kernel has the object cached.
         */
        if (vs->vs_control != MEMORY_OBJECT_CONTROL_NULL)
                Panic("bad request");

        /*
         * Unlock the pager (though there should be no one
         * waiting for it).
         */
        VS_UNLOCK(vs);

        /* Lock out paging segment removal for the duration of this */
        /* call.  We are vulnerable to losing a paging segment we rely */
        /* on as soon as we remove ourselves from the VSL and unlock */

        /* Keep our thread from blocking on attempt to trigger backing */
        /* store release */
        backing_store_release_trigger_disable += 1;

        /*
         * Remove the memory object port association, and then
         * the destroy the port itself.  We must remove the object
         * from the port list before deallocating the pager,
         * because of default_pager_objects.
         */
        vstruct_list_delete(vs);
        VSL_UNLOCK();

        ps_vstruct_dealloc(vs);

        VSL_LOCK();
        backing_store_release_trigger_disable -= 1;
        if(backing_store_release_trigger_disable == 0) {
                thread_wakeup((event_t)&backing_store_release_trigger_disable);
        }
        VSL_UNLOCK();
}

kern_return_t
dp_memory_object_data_request(
        memory_object_t         mem_obj,
        memory_object_offset_t  offset,
        vm_size_t               length,
        __unused vm_prot_t              protection_required)
{
        vstruct_t               vs;

        GSTAT(global_stats.gs_pagein_calls++);


        /* CDY at this moment vs_lookup panics when presented with the wrong */
        /* port.  As we are expanding this pager to support user interfaces */
        /* this should be changed to return kern_failure */
        vs_lookup(mem_obj, vs);
        vs_lock(vs);

        /* We are going to relax the strict sequencing here for performance */
        /* reasons.  We can do this because we know that the read and */
        /* write threads are different and we rely on synchronization */
        /* of read and write requests at the cache memory_object level */
        /* break out wait_for_writers, all of this goes away when */
        /* we get real control of seqno with the new component interface */

        if (vs->vs_writers != 0) {
                /* you can't hold on to the seqno and go */
                /* to sleep like that */
                vs_unlock(vs);  /* bump internal count of seqno */
                VS_LOCK(vs);
                while (vs->vs_writers != 0) {
                        default_pager_wait_write++;
                        vs->vs_waiting_write = TRUE;
                        assert_wait(&vs->vs_writers, THREAD_UNINT);
                        VS_UNLOCK(vs);
                        thread_block(THREAD_CONTINUE_NULL);
                        VS_LOCK(vs);
                        vs_async_wait(vs);
                }
                if(vs->vs_control == MEMORY_OBJECT_CONTROL_NULL) {
                        VS_UNLOCK(vs);
                        return KERN_FAILURE;
                }
                vs_start_read(vs);
                VS_UNLOCK(vs);
        } else {
                vs_start_read(vs);
                vs_unlock(vs);
        }

        /*
         * Request must be on a page boundary and a multiple of pages.
         */
        if ((offset & vm_page_mask) != 0 || (length & vm_page_mask) != 0)
                Panic("bad alignment");

        pvs_cluster_read(vs, (vm_offset_t)offset, length);

        vs_finish_read(vs);

        return KERN_SUCCESS;
}

/*
 * memory_object_data_initialize: check whether we already have each page, and
 * write it if we do not.  The implementation is far from optimized, and
 * also assumes that the default_pager is single-threaded.
 */
/*  It is questionable whether or not a pager should decide what is relevant */
/* and what is not in data sent from the kernel.  Data initialize has been */
/* changed to copy back all data sent to it in preparation for its eventual */
/* merge with data return.  It is the kernel that should decide what pages */
/* to write back.  As of the writing of this note, this is indeed the case */
/* the kernel writes back one page at a time through this interface */

kern_return_t
dp_memory_object_data_initialize(
        memory_object_t         mem_obj,
        memory_object_offset_t  offset,
        vm_size_t               size)
{
        vstruct_t       vs;

        DP_DEBUG(DEBUG_MO_EXTERNAL,
                 ("mem_obj=0x%x,offset=0x%x,cnt=0x%x\n",
                  (int)mem_obj, (int)offset, (int)size));
        GSTAT(global_stats.gs_pages_init += atop_32(size));

        vs_lookup(mem_obj, vs);
        vs_lock(vs);
        vs_start_write(vs);
        vs_unlock(vs);

        /*
         * Write the data via clustered writes. vs_cluster_write will
         * loop if the address range specified crosses cluster
         * boundaries.
         */
        vs_cluster_write(vs, 0, (vm_offset_t)offset, size, FALSE, 0);

        vs_finish_write(vs);

        return KERN_SUCCESS;
}

kern_return_t
dp_memory_object_data_unlock(
        __unused memory_object_t                mem_obj,
        __unused memory_object_offset_t offset,
        __unused vm_size_t              size,
        __unused vm_prot_t              desired_access)
{
        Panic("dp_memory_object_data_unlock: illegal");
        return KERN_FAILURE;
}


/*ARGSUSED8*/
kern_return_t
dp_memory_object_data_return(
        memory_object_t         mem_obj,
        memory_object_offset_t  offset,
        vm_size_t                       size,
        __unused memory_object_offset_t *resid_offset,
        __unused int            *io_error,
        __unused boolean_t      dirty,
        __unused boolean_t      kernel_copy,
        __unused int    upl_flags)
{
        vstruct_t       vs;

        DP_DEBUG(DEBUG_MO_EXTERNAL,
                 ("mem_obj=0x%x,offset=0x%x,size=0x%x\n",
                  (int)mem_obj, (int)offset, (int)size));
        GSTAT(global_stats.gs_pageout_calls++);

        /* This routine is called by the pageout thread.  The pageout thread */
        /* cannot be blocked by read activities unless the read activities   */
        /* Therefore the grant of vs lock must be done on a try versus a      */
        /* blocking basis.  The code below relies on the fact that the       */
        /* interface is synchronous.  Should this interface be again async   */
        /* for some type  of pager in the future the pages will have to be   */
        /* returned through a separate, asynchronous path.                   */

        vs_lookup(mem_obj, vs);

        default_pager_total++;
        if(!VS_TRY_LOCK(vs)) {
                /* the call below will not be done by caller when we have */
                /* a synchronous interface */
                /* return KERN_LOCK_OWNED; */
                upl_t           upl;
                unsigned int    page_list_count = 0;
                memory_object_super_upl_request(vs->vs_control,
                                        (memory_object_offset_t)offset,
                                        size, size,
                                        &upl, NULL, &page_list_count,
                                        UPL_NOBLOCK | UPL_CLEAN_IN_PLACE 
                                        | UPL_NO_SYNC | UPL_COPYOUT_FROM);
                upl_abort(upl,0);
                upl_deallocate(upl);
                return KERN_SUCCESS;
        }

        if ((vs->vs_seqno != vs->vs_next_seqno++)
                        || (vs->vs_readers)
                        || (vs->vs_xfer_pending)) {
                upl_t           upl;
                unsigned int    page_list_count = 0;

                vs->vs_next_seqno--;
                VS_UNLOCK(vs);

                /* the call below will not be done by caller when we have */
                /* a synchronous interface */
                /* return KERN_LOCK_OWNED; */
                memory_object_super_upl_request(vs->vs_control,
                                (memory_object_offset_t)offset,
                                size, size,
                                &upl, NULL, &page_list_count,
                                UPL_NOBLOCK | UPL_CLEAN_IN_PLACE 
                                        | UPL_NO_SYNC | UPL_COPYOUT_FROM);
                upl_abort(upl,0);
                upl_deallocate(upl);
                return KERN_SUCCESS;
        }

        if ((size % vm_page_size) != 0)
                Panic("bad alignment");

        vs_start_write(vs);


        vs->vs_async_pending += 1;  /* protect from backing store contraction */
        vs_unlock(vs);

        /*
         * Write the data via clustered writes. vs_cluster_write will
         * loop if the address range specified crosses cluster
         * boundaries.
         */
        vs_cluster_write(vs, 0, (vm_offset_t)offset, size, FALSE, 0);

        vs_finish_write(vs);

        /* temporary, need a finer lock based on cluster */

        VS_LOCK(vs);
        vs->vs_async_pending -= 1;  /* release vs_async_wait */
        if (vs->vs_async_pending == 0 && vs->vs_waiting_async) {
                vs->vs_waiting_async = FALSE;
                VS_UNLOCK(vs);
                thread_wakeup(&vs->vs_async_pending);
        } else {
                VS_UNLOCK(vs);
        }


        return KERN_SUCCESS;
}

/*
 * Routine:     default_pager_memory_object_create
 * Purpose:
 *      Handle requests for memory objects from the
 *      kernel.
 * Notes:
 *      Because we only give out the default memory
 *      manager port to the kernel, we don't have to
 *      be so paranoid about the contents.
 */
kern_return_t
default_pager_memory_object_create(
        __unused memory_object_default_t        dmm,
        vm_size_t               new_size,
        memory_object_t         *new_mem_obj)
{
        vstruct_t               vs;

        assert(dmm == default_pager_object);

        vs = vs_object_create(new_size);
        if (vs == VSTRUCT_NULL)
                return KERN_RESOURCE_SHORTAGE;

        vs->vs_next_seqno = 0;

        /*
         * Set up associations between this memory object
         * and this default_pager structure
         */

        vs->vs_pager_ops = &default_pager_ops;
        vs->vs_mem_obj_ikot = IKOT_MEMORY_OBJECT;

        /*
         * After this, other threads might receive requests
         * for this memory object or find it in the port list.
         */

        vstruct_list_insert(vs);
        *new_mem_obj = vs_to_mem_obj(vs);
        return KERN_SUCCESS;
}

/*
 * Create an external object.
 */
kern_return_t
default_pager_object_create(
        default_pager_t default_pager,
        vm_size_t       size,
        memory_object_t *mem_objp)
{
        vstruct_t       vs;

        if (default_pager != default_pager_object)
                return KERN_INVALID_ARGUMENT;

        vs = vs_object_create(size);
        if (vs == VSTRUCT_NULL)
                return KERN_RESOURCE_SHORTAGE;

        /*
         * Set up associations between the default pager
         * and this vstruct structure
         */
        vs->vs_pager_ops = &default_pager_ops;
        vstruct_list_insert(vs);
        *mem_objp = vs_to_mem_obj(vs);
        return KERN_SUCCESS;
}

kern_return_t
default_pager_objects(
        default_pager_t                 default_pager,
        default_pager_object_array_t    *objectsp,
        mach_msg_type_number_t          *ocountp,
        mach_port_array_t               *portsp,
        mach_msg_type_number_t          *pcountp)
{
        vm_offset_t             oaddr = 0;      /* memory for objects */
        vm_size_t               osize = 0;      /* current size */
        default_pager_object_t  * objects;
        unsigned int            opotential = 0;

        vm_map_copy_t           pcopy = 0;      /* copy handle for pagers */
        vm_size_t               psize = 0;      /* current size */
        memory_object_t         * pagers;
        unsigned int            ppotential = 0;

        unsigned int            actual;
        unsigned int            num_objects;
        kern_return_t           kr;
        vstruct_t               entry;

        if (default_pager != default_pager_object)
                return KERN_INVALID_ARGUMENT;

        /*
         * We will send no more than this many
         */
        actual = vstruct_list.vsl_count;

        /*
         * Out out-of-line port arrays are simply kalloc'ed.
         */
        psize = round_page(actual * sizeof * pagers);
        ppotential = psize / sizeof * pagers;
        pagers = (memory_object_t *)kalloc(psize);
        if (0 == pagers)
                return KERN_RESOURCE_SHORTAGE;
                
        /*
         * returned out of line data must be allocated out
         * the ipc_kernel_map, wired down, filled in, and
         * then "copied in" as if it had been sent by a
         * user process.
         */
        osize = round_page(actual * sizeof * objects);
        opotential = osize / sizeof * objects;
        kr = kmem_alloc(ipc_kernel_map, &oaddr, osize);
        if (KERN_SUCCESS != kr) {
                kfree(pagers, psize);
                return KERN_RESOURCE_SHORTAGE;
        }
        objects = (default_pager_object_t *)oaddr;


        /*
         * Now scan the list.
         */

        VSL_LOCK();

        num_objects = 0;
        queue_iterate(&vstruct_list.vsl_queue, entry, vstruct_t, vs_links) {

                memory_object_t                 pager;
                vm_size_t                       size;

                if ((num_objects >= opotential) ||
                    (num_objects >= ppotential)) {

                        /*
                         * This should be rare.  In any case,
                         * we will only miss recent objects,
                         * because they are added at the end.
                         */
                        break;
                }

                /*
                 * Avoid interfering with normal operations
                 */
                if (!VS_MAP_TRY_LOCK(entry))
                        goto not_this_one;
                size = ps_vstruct_allocated_size(entry);
                VS_MAP_UNLOCK(entry);

                VS_LOCK(entry);

                /*
                 * We need a reference for our caller.  Adding this
                 * reference through the linked list could race with
                 * destruction of the object.  If we find the object
                 * has no references, just give up on it.
                 */
                VS_LOCK(entry);
                if (entry->vs_references == 0) {
                        VS_UNLOCK(entry);
                        goto not_this_one;
                }
                pager = vs_to_mem_obj(entry);
                dp_memory_object_reference(pager);
                VS_UNLOCK(entry);

                /* the arrays are wired, so no deadlock worries */

                objects[num_objects].dpo_object = (vm_offset_t) entry;
                objects[num_objects].dpo_size = size;
                pagers [num_objects++] = pager;
                continue;

            not_this_one:
                /*
                 * Do not return garbage
                 */
                objects[num_objects].dpo_object = (vm_offset_t) 0;
                objects[num_objects].dpo_size = 0;
                pagers[num_objects++] = MEMORY_OBJECT_NULL;

        }

        VSL_UNLOCK();

        /* clear out any excess allocation */
        while (num_objects < opotential) {
                objects[--opotential].dpo_object = (vm_offset_t) 0;
                objects[opotential].dpo_size = 0;
        }
        while (num_objects < ppotential) {
                pagers[--ppotential] = MEMORY_OBJECT_NULL;
        }

        kr = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(oaddr),
                           vm_map_round_page(oaddr + osize), FALSE);
        assert(KERN_SUCCESS == kr);
        kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)oaddr,
                           (vm_map_size_t)osize, TRUE, &pcopy);
        assert(KERN_SUCCESS == kr);

        *objectsp = (default_pager_object_array_t)objects;
        *ocountp = num_objects;
        *portsp = (mach_port_array_t)pcopy;
        *pcountp = num_objects;

        return KERN_SUCCESS;
}

kern_return_t
default_pager_object_pages(
        default_pager_t         default_pager,
        mach_port_t                     memory_object,
        default_pager_page_array_t      *pagesp,
        mach_msg_type_number_t          *countp)
{
        vm_offset_t                     addr = 0; /* memory for page offsets */
        vm_size_t                       size = 0; /* current memory size */
        vm_map_copy_t                   copy;
        default_pager_page_t            * pages = 0;
        unsigned int                    potential;
        unsigned int                    actual;
        kern_return_t                   kr;
        memory_object_t                 object;

        if (default_pager != default_pager_object)
                return KERN_INVALID_ARGUMENT;

        object = (memory_object_t) memory_object;

        potential = 0;
        for (;;) {
                vstruct_t       entry;

                VSL_LOCK();
                queue_iterate(&vstruct_list.vsl_queue, entry, vstruct_t,
                              vs_links) {
                        VS_LOCK(entry);
                        if (vs_to_mem_obj(entry) == object) {
                                VSL_UNLOCK();
                                goto found_object;
                        }
                        VS_UNLOCK(entry);
                }
                VSL_UNLOCK();

                /* did not find the object */
                if (0 != addr)
                        kmem_free(ipc_kernel_map, addr, size);

                return KERN_INVALID_ARGUMENT;

            found_object:

                if (!VS_MAP_TRY_LOCK(entry)) {
                        /* oh well bad luck */
                        int wresult;

                        VS_UNLOCK(entry);

                        assert_wait_timeout((event_t)assert_wait_timeout, THREAD_UNINT, 1, 1000*NSEC_PER_USEC);
                        wresult = thread_block(THREAD_CONTINUE_NULL);
                        assert(wresult == THREAD_TIMED_OUT);
                        continue;
                }

                actual = ps_vstruct_allocated_pages(entry, pages, potential);
                VS_MAP_UNLOCK(entry);
                VS_UNLOCK(entry);

                if (actual <= potential)
                        break;

                /* allocate more memory */
                if (0 != addr)
                        kmem_free(ipc_kernel_map, addr, size);

                size = round_page(actual * sizeof * pages);
                kr = kmem_alloc(ipc_kernel_map, &addr, size);
                if (KERN_SUCCESS != kr)
                        return KERN_RESOURCE_SHORTAGE;

                pages = (default_pager_page_t *)addr;
                potential = size / sizeof * pages;
        }

        /*
         * Clear unused memory.
         */
        while (actual < potential)
                pages[--potential].dpp_offset = 0;

        kr = vm_map_unwire(ipc_kernel_map, vm_map_trunc_page(addr),
                           vm_map_round_page(addr + size), FALSE);
        assert(KERN_SUCCESS == kr);
        kr = vm_map_copyin(ipc_kernel_map, (vm_map_address_t)addr,
                           (vm_map_size_t)size, TRUE, &copy);
        assert(KERN_SUCCESS == kr);

        
        *pagesp = (default_pager_page_array_t)copy;
        *countp = actual;
        return KERN_SUCCESS;
}