WebSVN – HelenOS-historic – Blame – /kernel/trunk/generic/src/mm/as.c

Rev	Author	Line No.	Line
703	jermar	1	/*
		2	* Copyright (C) 2001-2006 Jakub Jermar
		3	* All rights reserved.
		4	*
		5	* Redistribution and use in source and binary forms, with or without
		6	* modification, are permitted provided that the following conditions
		7	* are met:
		8	*
		9	* - Redistributions of source code must retain the above copyright
		10	* notice, this list of conditions and the following disclaimer.
		11	* - Redistributions in binary form must reproduce the above copyright
		12	* notice, this list of conditions and the following disclaimer in the
		13	* documentation and/or other materials provided with the distribution.
		14	* - The name of the author may not be used to endorse or promote products
		15	* derived from this software without specific prior written permission.
		16	*
		17	* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
		18	* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
		19	* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
		20	* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
		21	* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
		22	* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
		23	* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
		24	* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
		25	* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
		26	* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
		27	*/
		28
1702	cejka	29	/** @defgroup mm Memory management
		30	* @ingroup kernel
		31	* @{
		32	* @}
		33	*/
		34
		35	/** @addtogroup genericmm generic
		36	* @ingroup mm
		37	* @{
		38	*/
		39
1248	jermar	40	/**
1702	cejka	41	* @file
1248	jermar	42	* @brief Address space related functions.
		43	*
703	jermar	44	* This file contains address space manipulation functions.
		45	* Roughly speaking, this is a higher-level client of
		46	* Virtual Address Translation (VAT) subsystem.
1248	jermar	47	*
		48	* Functionality provided by this file allows one to
		49	* create address space and create, resize and share
		50	* address space areas.
		51	*
		52	* @see page.c
		53	*
703	jermar	54	*/
		55
		56	#include <mm/as.h>
756	jermar	57	#include <arch/mm/as.h>
703	jermar	58	#include <mm/page.h>
		59	#include <mm/frame.h>
814	palkovsky	60	#include <mm/slab.h>
703	jermar	61	#include <mm/tlb.h>
		62	#include <arch/mm/page.h>
		63	#include <genarch/mm/page_pt.h>
1108	jermar	64	#include <genarch/mm/page_ht.h>
727	jermar	65	#include <mm/asid.h>
703	jermar	66	#include <arch/mm/asid.h>
		67	#include <synch/spinlock.h>
1380	jermar	68	#include <synch/mutex.h>
788	jermar	69	#include <adt/list.h>
1147	jermar	70	#include <adt/btree.h>
1235	jermar	71	#include <proc/task.h>
1288	jermar	72	#include <proc/thread.h>
1235	jermar	73	#include <arch/asm.h>
703	jermar	74	#include <panic.h>
		75	#include <debug.h>
1235	jermar	76	#include <print.h>
703	jermar	77	#include <memstr.h>
1070	jermar	78	#include <macros.h>
703	jermar	79	#include <arch.h>
1235	jermar	80	#include <errno.h>
		81	#include <config.h>
1387	jermar	82	#include <align.h>
1235	jermar	83	#include <arch/types.h>
		84	#include <typedefs.h>
1288	jermar	85	#include <syscall/copy.h>
		86	#include <arch/interrupt.h>
703	jermar	87
756	jermar	88	as_operations_t *as_operations = NULL;
703	jermar	89
1415	jermar	90	/** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
		91	SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
823	jermar	92
		93	/**
		94	* This list contains address spaces that are not active on any
		95	* processor and that have valid ASID.
		96	*/
		97	LIST_INITIALIZE(inactive_as_with_asid_head);
		98
757	jermar	99	/** Kernel address space. */
		100	as_t *AS_KERNEL = NULL;
		101
1235	jermar	102	static int area_flags_to_page_flags(int aflags);
977	jermar	103	static as_area_t find_area_and_lock(as_t as, __address va);
1048	jermar	104	static bool check_area_conflicts(as_t as, __address va, size_t size, as_area_t avoid_area);
1409	jermar	105	static void sh_info_remove_reference(share_info_t *sh_info);
703	jermar	106
756	jermar	107	/** Initialize address space subsystem. */
		108	void as_init(void)
		109	{
		110	as_arch_init();
789	palkovsky	111	AS_KERNEL = as_create(FLAG_AS_KERNEL);
1383	decky	112	if (!AS_KERNEL)
		113	panic("can't create kernel address space\n");
		114
756	jermar	115	}
		116
757	jermar	117	/** Create address space.
		118	*
		119	* @param flags Flags that influence way in wich the address space is created.
		120	*/
756	jermar	121	as_t *as_create(int flags)
703	jermar	122	{
		123	as_t *as;
		124
822	palkovsky	125	as = (as_t *) malloc(sizeof(as_t), 0);
823	jermar	126	link_initialize(&as->inactive_as_with_asid_link);
1380	jermar	127	mutex_initialize(&as->lock);
1147	jermar	128	btree_create(&as->as_area_btree);
822	palkovsky	129
		130	if (flags & FLAG_AS_KERNEL)
		131	as->asid = ASID_KERNEL;
		132	else
		133	as->asid = ASID_INVALID;
		134
1468	jermar	135	as->refcount = 0;
1415	jermar	136	as->cpu_refcount = 0;
822	palkovsky	137	as->page_table = page_table_create(flags);
703	jermar	138
		139	return as;
		140	}
		141
1468	jermar	142	/** Destroy adress space.
		143	*
		144	* When there are no tasks referencing this address space (i.e. its refcount is zero),
		145	* the address space can be destroyed.
		146	*/
		147	void as_destroy(as_t *as)
973	palkovsky	148	{
1468	jermar	149	ipl_t ipl;
1594	jermar	150	bool cond;
973	palkovsky	151
1468	jermar	152	ASSERT(as->refcount == 0);
		153
		154	/*
		155	* Since there is no reference to this area,
		156	* it is safe not to lock its mutex.
		157	*/
		158	ipl = interrupts_disable();
		159	spinlock_lock(&inactive_as_with_asid_lock);
1587	jermar	160	if (as->asid != ASID_INVALID && as != AS_KERNEL) {
1594	jermar	161	if (as != AS && as->cpu_refcount == 0)
1587	jermar	162	list_remove(&as->inactive_as_with_asid_link);
1468	jermar	163	asid_put(as->asid);
		164	}
		165	spinlock_unlock(&inactive_as_with_asid_lock);
		166
		167	/*
		168	* Destroy address space areas of the address space.
1594	jermar	169	* The B+tee must be walked carefully because it is
		170	* also being destroyed.
1468	jermar	171	*/
1594	jermar	172	for (cond = true; cond; ) {
1468	jermar	173	btree_node_t *node;
1594	jermar	174
		175	ASSERT(!list_empty(&as->as_area_btree.leaf_head));
		176	node = list_get_instance(as->as_area_btree.leaf_head.next, btree_node_t, leaf_link);
		177
		178	if ((cond = node->keys)) {
		179	as_area_destroy(as, node->key[0]);
		180	}
1468	jermar	181	}
1495	jermar	182
1483	jermar	183	btree_destroy(&as->as_area_btree);
1468	jermar	184	page_table_destroy(as->page_table);
		185
		186	interrupts_restore(ipl);
		187
973	palkovsky	188	free(as);
		189	}
		190
703	jermar	191	/** Create address space area of common attributes.
		192	*
		193	* The created address space area is added to the target address space.
		194	*
		195	* @param as Target address space.
1239	jermar	196	* @param flags Flags of the area memory.
1048	jermar	197	* @param size Size of area.
703	jermar	198	* @param base Base address of area.
1239	jermar	199	* @param attrs Attributes of the area.
1409	jermar	200	* @param backend Address space area backend. NULL if no backend is used.
		201	* @param backend_data NULL or a pointer to an array holding two void *.
703	jermar	202	*
		203	* @return Address space area on success or NULL on failure.
		204	*/
1409	jermar	205	as_area_t as_area_create(as_t as, int flags, size_t size, __address base, int attrs,
1424	jermar	206	mem_backend_t backend, mem_backend_data_t backend_data)
703	jermar	207	{
		208	ipl_t ipl;
		209	as_area_t *a;
		210
		211	if (base % PAGE_SIZE)
1048	jermar	212	return NULL;
		213
1233	jermar	214	if (!size)
		215	return NULL;
		216
1048	jermar	217	/* Writeable executable areas are not supported. */
		218	if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
		219	return NULL;
703	jermar	220
		221	ipl = interrupts_disable();
1380	jermar	222	mutex_lock(&as->lock);
703	jermar	223
1048	jermar	224	if (!check_area_conflicts(as, base, size, NULL)) {
1380	jermar	225	mutex_unlock(&as->lock);
1048	jermar	226	interrupts_restore(ipl);
		227	return NULL;
		228	}
703	jermar	229
822	palkovsky	230	a = (as_area_t *) malloc(sizeof(as_area_t), 0);
703	jermar	231
1380	jermar	232	mutex_initialize(&a->lock);
822	palkovsky	233
1424	jermar	234	a->as = as;
1026	jermar	235	a->flags = flags;
1239	jermar	236	a->attributes = attrs;
1048	jermar	237	a->pages = SIZE2FRAMES(size);
822	palkovsky	238	a->base = base;
1409	jermar	239	a->sh_info = NULL;
		240	a->backend = backend;
1424	jermar	241	if (backend_data)
		242	a->backend_data = *backend_data;
		243	else
		244	memsetb((__address) &a->backend_data, sizeof(a->backend_data), 0);
		245
1387	jermar	246	btree_create(&a->used_space);
822	palkovsky	247
1147	jermar	248	btree_insert(&as->as_area_btree, base, (void *) a, NULL);
822	palkovsky	249
1380	jermar	250	mutex_unlock(&as->lock);
703	jermar	251	interrupts_restore(ipl);
704	jermar	252
703	jermar	253	return a;
		254	}
		255
1235	jermar	256	/** Find address space area and change it.
		257	*
		258	* @param as Address space.
		259	* @param address Virtual address belonging to the area to be changed. Must be page-aligned.
		260	* @param size New size of the virtual memory block starting at address.
		261	* @param flags Flags influencing the remap operation. Currently unused.
		262	*
1306	jermar	263	* @return Zero on success or a value from @ref errno.h otherwise.
1235	jermar	264	*/
1306	jermar	265	int as_area_resize(as_t *as, __address address, size_t size, int flags)
1235	jermar	266	{
1306	jermar	267	as_area_t *area;
1235	jermar	268	ipl_t ipl;
		269	size_t pages;
		270
		271	ipl = interrupts_disable();
1380	jermar	272	mutex_lock(&as->lock);
1235	jermar	273
		274	/*
		275	* Locate the area.
		276	*/
		277	area = find_area_and_lock(as, address);
		278	if (!area) {
1380	jermar	279	mutex_unlock(&as->lock);
1235	jermar	280	interrupts_restore(ipl);
1306	jermar	281	return ENOENT;
1235	jermar	282	}
		283
1424	jermar	284	if (area->backend == &phys_backend) {
1235	jermar	285	/*
		286	* Remapping of address space areas associated
		287	* with memory mapped devices is not supported.
		288	*/
1380	jermar	289	mutex_unlock(&area->lock);
		290	mutex_unlock(&as->lock);
1235	jermar	291	interrupts_restore(ipl);
1306	jermar	292	return ENOTSUP;
1235	jermar	293	}
1409	jermar	294	if (area->sh_info) {
		295	/*
		296	* Remapping of shared address space areas
		297	* is not supported.
		298	*/
		299	mutex_unlock(&area->lock);
		300	mutex_unlock(&as->lock);
		301	interrupts_restore(ipl);
		302	return ENOTSUP;
		303	}
1235	jermar	304
		305	pages = SIZE2FRAMES((address - area->base) + size);
		306	if (!pages) {
		307	/*
		308	* Zero size address space areas are not allowed.
		309	*/
1380	jermar	310	mutex_unlock(&area->lock);
		311	mutex_unlock(&as->lock);
1235	jermar	312	interrupts_restore(ipl);
1306	jermar	313	return EPERM;
1235	jermar	314	}
		315
		316	if (pages < area->pages) {
1403	jermar	317	bool cond;
		318	__address start_free = area->base + pages*PAGE_SIZE;
1235	jermar	319
		320	/*
		321	* Shrinking the area.
		322	* No need to check for overlaps.
		323	*/
1403	jermar	324
		325	/*
1436	jermar	326	* Start TLB shootdown sequence.
		327	*/
		328	tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
		329
		330	/*
1403	jermar	331	* Remove frames belonging to used space starting from
		332	* the highest addresses downwards until an overlap with
		333	* the resized address space area is found. Note that this
		334	* is also the right way to remove part of the used_space
		335	* B+tree leaf list.
		336	*/
		337	for (cond = true; cond;) {
		338	btree_node_t *node;
		339
		340	ASSERT(!list_empty(&area->used_space.leaf_head));
		341	node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
		342	if ((cond = (bool) node->keys)) {
		343	__address b = node->key[node->keys - 1];
		344	count_t c = (count_t) node->value[node->keys - 1];
		345	int i = 0;
1235	jermar	346
1403	jermar	347	if (overlaps(b, cPAGE_SIZE, area->base, pagesPAGE_SIZE)) {
		348
		349	if (b + c*PAGE_SIZE <= start_free) {
		350	/*
		351	* The whole interval fits completely
		352	* in the resized address space area.
		353	*/
		354	break;
		355	}
		356
		357	/*
		358	* Part of the interval corresponding to b and c
		359	* overlaps with the resized address space area.
		360	*/
		361
		362	cond = false; /* we are almost done */
		363	i = (start_free - b) >> PAGE_WIDTH;
		364	if (!used_space_remove(area, start_free, c - i))
		365	panic("Could not remove used space.");
		366	} else {
		367	/*
		368	* The interval of used space can be completely removed.
		369	*/
		370	if (!used_space_remove(area, b, c))
		371	panic("Could not remove used space.\n");
		372	}
		373
		374	for (; i < c; i++) {
		375	pte_t *pte;
		376
		377	page_table_lock(as, false);
		378	pte = page_mapping_find(as, b + i*PAGE_SIZE);
		379	ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
1424	jermar	380	if (area->backend && area->backend->frame_free) {
		381	area->backend->frame_free(area,
1409	jermar	382	b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
		383	}
1403	jermar	384	page_mapping_remove(as, b + i*PAGE_SIZE);
		385	page_table_unlock(as, false);
		386	}
1235	jermar	387	}
		388	}
1436	jermar	389
1235	jermar	390	/*
1436	jermar	391	* Finish TLB shootdown sequence.
1235	jermar	392	*/
		393	tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
		394	tlb_shootdown_finalize();
		395	} else {
		396	/*
		397	* Growing the area.
		398	* Check for overlaps with other address space areas.
		399	*/
		400	if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
1380	jermar	401	mutex_unlock(&area->lock);
		402	mutex_unlock(&as->lock);
1235	jermar	403	interrupts_restore(ipl);
1306	jermar	404	return EADDRNOTAVAIL;
1235	jermar	405	}
		406	}
		407
		408	area->pages = pages;
		409
1380	jermar	410	mutex_unlock(&area->lock);
		411	mutex_unlock(&as->lock);
1235	jermar	412	interrupts_restore(ipl);
		413
1306	jermar	414	return 0;
1235	jermar	415	}
		416
1306	jermar	417	/** Destroy address space area.
		418	*
		419	* @param as Address space.
		420	* @param address Address withing the area to be deleted.
		421	*
		422	* @return Zero on success or a value from @ref errno.h on failure.
		423	*/
		424	int as_area_destroy(as_t *as, __address address)
		425	{
		426	as_area_t *area;
		427	__address base;
1495	jermar	428	link_t *cur;
1306	jermar	429	ipl_t ipl;
		430
		431	ipl = interrupts_disable();
1380	jermar	432	mutex_lock(&as->lock);
1306	jermar	433
		434	area = find_area_and_lock(as, address);
		435	if (!area) {
1380	jermar	436	mutex_unlock(&as->lock);
1306	jermar	437	interrupts_restore(ipl);
		438	return ENOENT;
		439	}
		440
1403	jermar	441	base = area->base;
		442
1411	jermar	443	/*
1436	jermar	444	* Start TLB shootdown sequence.
		445	*/
		446	tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
		447
		448	/*
1411	jermar	449	* Visit only the pages mapped by used_space B+tree.
		450	*/
1495	jermar	451	for (cur = area->used_space.leaf_head.next; cur != &area->used_space.leaf_head; cur = cur->next) {
1411	jermar	452	btree_node_t *node;
1495	jermar	453	int i;
1403	jermar	454
1495	jermar	455	node = list_get_instance(cur, btree_node_t, leaf_link);
		456	for (i = 0; i < node->keys; i++) {
		457	__address b = node->key[i];
		458	count_t j;
1411	jermar	459	pte_t *pte;
1403	jermar	460
1495	jermar	461	for (j = 0; j < (count_t) node->value[i]; j++) {
1411	jermar	462	page_table_lock(as, false);
1495	jermar	463	pte = page_mapping_find(as, b + j*PAGE_SIZE);
1411	jermar	464	ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
1424	jermar	465	if (area->backend && area->backend->frame_free) {
		466	area->backend->frame_free(area,
1495	jermar	467	b + j*PAGE_SIZE, PTE_GET_FRAME(pte));
1403	jermar	468	}
1495	jermar	469	page_mapping_remove(as, b + j*PAGE_SIZE);
1411	jermar	470	page_table_unlock(as, false);
1306	jermar	471	}
		472	}
		473	}
1403	jermar	474
1306	jermar	475	/*
1436	jermar	476	* Finish TLB shootdown sequence.
1306	jermar	477	*/
		478	tlb_invalidate_pages(AS->asid, area->base, area->pages);
		479	tlb_shootdown_finalize();
1436	jermar	480
		481	btree_destroy(&area->used_space);
1306	jermar	482
1309	jermar	483	area->attributes \|= AS_AREA_ATTR_PARTIAL;
1409	jermar	484
		485	if (area->sh_info)
		486	sh_info_remove_reference(area->sh_info);
		487
1380	jermar	488	mutex_unlock(&area->lock);
1306	jermar	489
		490	/*
		491	* Remove the empty area from address space.
		492	*/
		493	btree_remove(&AS->as_area_btree, base, NULL);
		494
1309	jermar	495	free(area);
		496
1380	jermar	497	mutex_unlock(&AS->lock);
1306	jermar	498	interrupts_restore(ipl);
		499	return 0;
		500	}
		501
1413	jermar	502	/** Share address space area with another or the same address space.
1235	jermar	503	*
1424	jermar	504	* Address space area mapping is shared with a new address space area.
		505	* If the source address space area has not been shared so far,
		506	* a new sh_info is created. The new address space area simply gets the
		507	* sh_info of the source area. The process of duplicating the
		508	* mapping is done through the backend share function.
1413	jermar	509	*
1417	jermar	510	* @param src_as Pointer to source address space.
1239	jermar	511	* @param src_base Base address of the source address space area.
1417	jermar	512	* @param acc_size Expected size of the source area.
1428	palkovsky	513	* @param dst_as Pointer to destination address space.
1417	jermar	514	* @param dst_base Target base address.
		515	* @param dst_flags_mask Destination address space area flags mask.
1235	jermar	516	*
1306	jermar	517	* @return Zero on success or ENOENT if there is no such task or
1235	jermar	518	* if there is no such address space area,
		519	* EPERM if there was a problem in accepting the area or
		520	* ENOMEM if there was a problem in allocating destination
1413	jermar	521	* address space area. ENOTSUP is returned if an attempt
		522	* to share non-anonymous address space area is detected.
1235	jermar	523	*/
1413	jermar	524	int as_area_share(as_t *src_as, __address src_base, size_t acc_size,
1428	palkovsky	525	as_t *dst_as, __address dst_base, int dst_flags_mask)
1235	jermar	526	{
		527	ipl_t ipl;
1239	jermar	528	int src_flags;
		529	size_t src_size;
		530	as_area_t src_area, dst_area;
1413	jermar	531	share_info_t *sh_info;
1424	jermar	532	mem_backend_t *src_backend;
		533	mem_backend_data_t src_backend_data;
1434	palkovsky	534
1235	jermar	535	ipl = interrupts_disable();
1380	jermar	536	mutex_lock(&src_as->lock);
1329	palkovsky	537	src_area = find_area_and_lock(src_as, src_base);
1239	jermar	538	if (!src_area) {
1238	jermar	539	/*
		540	* Could not find the source address space area.
		541	*/
1380	jermar	542	mutex_unlock(&src_as->lock);
1238	jermar	543	interrupts_restore(ipl);
		544	return ENOENT;
		545	}
1413	jermar	546
1424	jermar	547	if (!src_area->backend \|\| !src_area->backend->share) {
1413	jermar	548	/*
1424	jermar	549	* There is now backend or the backend does not
		550	* know how to share the area.
1413	jermar	551	*/
		552	mutex_unlock(&src_area->lock);
		553	mutex_unlock(&src_as->lock);
		554	interrupts_restore(ipl);
		555	return ENOTSUP;
		556	}
		557
1239	jermar	558	src_size = src_area->pages * PAGE_SIZE;
		559	src_flags = src_area->flags;
1424	jermar	560	src_backend = src_area->backend;
		561	src_backend_data = src_area->backend_data;
1544	palkovsky	562
		563	/* Share the cacheable flag from the original mapping */
		564	if (src_flags & AS_AREA_CACHEABLE)
		565	dst_flags_mask \|= AS_AREA_CACHEABLE;
		566
1461	palkovsky	567	if (src_size != acc_size \|\| (src_flags & dst_flags_mask) != dst_flags_mask) {
1413	jermar	568	mutex_unlock(&src_area->lock);
		569	mutex_unlock(&src_as->lock);
1235	jermar	570	interrupts_restore(ipl);
		571	return EPERM;
		572	}
1413	jermar	573
1235	jermar	574	/*
1413	jermar	575	* Now we are committed to sharing the area.
		576	* First prepare the area for sharing.
		577	* Then it will be safe to unlock it.
		578	*/
		579	sh_info = src_area->sh_info;
		580	if (!sh_info) {
		581	sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
		582	mutex_initialize(&sh_info->lock);
		583	sh_info->refcount = 2;
		584	btree_create(&sh_info->pagemap);
		585	src_area->sh_info = sh_info;
		586	} else {
		587	mutex_lock(&sh_info->lock);
		588	sh_info->refcount++;
		589	mutex_unlock(&sh_info->lock);
		590	}
		591
1424	jermar	592	src_area->backend->share(src_area);
1413	jermar	593
		594	mutex_unlock(&src_area->lock);
		595	mutex_unlock(&src_as->lock);
		596
		597	/*
1239	jermar	598	* Create copy of the source address space area.
		599	* The destination area is created with AS_AREA_ATTR_PARTIAL
		600	* attribute set which prevents race condition with
		601	* preliminary as_page_fault() calls.
1417	jermar	602	* The flags of the source area are masked against dst_flags_mask
		603	* to support sharing in less privileged mode.
1235	jermar	604	*/
1461	palkovsky	605	dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
1424	jermar	606	AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
1239	jermar	607	if (!dst_area) {
1235	jermar	608	/*
		609	* Destination address space area could not be created.
		610	*/
1413	jermar	611	sh_info_remove_reference(sh_info);
		612
1235	jermar	613	interrupts_restore(ipl);
		614	return ENOMEM;
		615	}
		616
		617	/*
1239	jermar	618	* Now the destination address space area has been
		619	* fully initialized. Clear the AS_AREA_ATTR_PARTIAL
1413	jermar	620	* attribute and set the sh_info.
1239	jermar	621	*/
1380	jermar	622	mutex_lock(&dst_area->lock);
1239	jermar	623	dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
1413	jermar	624	dst_area->sh_info = sh_info;
1380	jermar	625	mutex_unlock(&dst_area->lock);
1235	jermar	626
		627	interrupts_restore(ipl);
		628
		629	return 0;
		630	}
		631
1423	jermar	632	/** Check access mode for address space area.
		633	*
		634	* The address space area must be locked prior to this call.
		635	*
		636	* @param area Address space area.
		637	* @param access Access mode.
		638	*
		639	* @return False if access violates area's permissions, true otherwise.
		640	*/
		641	bool as_area_check_access(as_area_t *area, pf_access_t access)
		642	{
		643	int flagmap[] = {
		644	[PF_ACCESS_READ] = AS_AREA_READ,
		645	[PF_ACCESS_WRITE] = AS_AREA_WRITE,
		646	[PF_ACCESS_EXEC] = AS_AREA_EXEC
		647	};
		648
		649	if (!(area->flags & flagmap[access]))
		650	return false;
		651
		652	return true;
		653	}
		654
703	jermar	655	/** Handle page fault within the current address space.
		656	*
1409	jermar	657	* This is the high-level page fault handler. It decides
		658	* whether the page fault can be resolved by any backend
		659	* and if so, it invokes the backend to resolve the page
		660	* fault.
		661	*
703	jermar	662	* Interrupts are assumed disabled.
		663	*
		664	* @param page Faulting page.
1411	jermar	665	* @param access Access mode that caused the fault (i.e. read/write/exec).
1288	jermar	666	* @param istate Pointer to interrupted state.
703	jermar	667	*
1409	jermar	668	* @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
		669	* fault was caused by copy_to_uspace() or copy_from_uspace().
703	jermar	670	*/
1411	jermar	671	int as_page_fault(__address page, pf_access_t access, istate_t *istate)
703	jermar	672	{
1044	jermar	673	pte_t *pte;
977	jermar	674	as_area_t *area;
703	jermar	675
1380	jermar	676	if (!THREAD)
1409	jermar	677	return AS_PF_FAULT;
1380	jermar	678
703	jermar	679	ASSERT(AS);
1044	jermar	680
1380	jermar	681	mutex_lock(&AS->lock);
977	jermar	682	area = find_area_and_lock(AS, page);
703	jermar	683	if (!area) {
		684	/*
		685	* No area contained mapping for 'page'.
		686	* Signal page fault to low-level handler.
		687	*/
1380	jermar	688	mutex_unlock(&AS->lock);
1288	jermar	689	goto page_fault;
703	jermar	690	}
		691
1239	jermar	692	if (area->attributes & AS_AREA_ATTR_PARTIAL) {
		693	/*
		694	* The address space area is not fully initialized.
		695	* Avoid possible race by returning error.
		696	*/
1380	jermar	697	mutex_unlock(&area->lock);
		698	mutex_unlock(&AS->lock);
1288	jermar	699	goto page_fault;
1239	jermar	700	}
		701
1424	jermar	702	if (!area->backend \|\| !area->backend->page_fault) {
1409	jermar	703	/*
		704	* The address space area is not backed by any backend
		705	* or the backend cannot handle page faults.
		706	*/
		707	mutex_unlock(&area->lock);
		708	mutex_unlock(&AS->lock);
		709	goto page_fault;
		710	}
1179	jermar	711
1044	jermar	712	page_table_lock(AS, false);
		713
703	jermar	714	/*
1044	jermar	715	* To avoid race condition between two page faults
		716	* on the same address, we need to make sure
		717	* the mapping has not been already inserted.
		718	*/
		719	if ((pte = page_mapping_find(AS, page))) {
		720	if (PTE_PRESENT(pte)) {
1423	jermar	721	if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) \|\|
		722	(access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) \|\|
		723	(access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
		724	page_table_unlock(AS, false);
		725	mutex_unlock(&area->lock);
		726	mutex_unlock(&AS->lock);
		727	return AS_PF_OK;
		728	}
1044	jermar	729	}
		730	}
1409	jermar	731
1044	jermar	732	/*
1409	jermar	733	* Resort to the backend page fault handler.
703	jermar	734	*/
1424	jermar	735	if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
1409	jermar	736	page_table_unlock(AS, false);
		737	mutex_unlock(&area->lock);
		738	mutex_unlock(&AS->lock);
		739	goto page_fault;
		740	}
703	jermar	741
1044	jermar	742	page_table_unlock(AS, false);
1380	jermar	743	mutex_unlock(&area->lock);
		744	mutex_unlock(&AS->lock);
1288	jermar	745	return AS_PF_OK;
		746
		747	page_fault:
		748	if (THREAD->in_copy_from_uspace) {
		749	THREAD->in_copy_from_uspace = false;
		750	istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
		751	} else if (THREAD->in_copy_to_uspace) {
		752	THREAD->in_copy_to_uspace = false;
		753	istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
		754	} else {
		755	return AS_PF_FAULT;
		756	}
		757
		758	return AS_PF_DEFER;
703	jermar	759	}
		760
823	jermar	761	/** Switch address spaces.
703	jermar	762	*
1380	jermar	763	* Note that this function cannot sleep as it is essentially a part of
1415	jermar	764	* scheduling. Sleeping here would lead to deadlock on wakeup.
1380	jermar	765	*
823	jermar	766	* @param old Old address space or NULL.
		767	* @param new New address space.
703	jermar	768	*/
823	jermar	769	void as_switch(as_t old, as_t new)
703	jermar	770	{
		771	ipl_t ipl;
823	jermar	772	bool needs_asid = false;
703	jermar	773
		774	ipl = interrupts_disable();
1415	jermar	775	spinlock_lock(&inactive_as_with_asid_lock);
703	jermar	776
		777	/*
823	jermar	778	* First, take care of the old address space.
		779	*/
		780	if (old) {
1380	jermar	781	mutex_lock_active(&old->lock);
1415	jermar	782	ASSERT(old->cpu_refcount);
		783	if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
823	jermar	784	/*
		785	* The old address space is no longer active on
		786	* any processor. It can be appended to the
		787	* list of inactive address spaces with assigned
		788	* ASID.
		789	*/
		790	ASSERT(old->asid != ASID_INVALID);
		791	list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
		792	}
1380	jermar	793	mutex_unlock(&old->lock);
823	jermar	794	}
		795
		796	/*
		797	* Second, prepare the new address space.
		798	*/
1380	jermar	799	mutex_lock_active(&new->lock);
1415	jermar	800	if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
823	jermar	801	if (new->asid != ASID_INVALID)
		802	list_remove(&new->inactive_as_with_asid_link);
		803	else
		804	needs_asid = true; /* defer call to asid_get() until new->lock is released */
		805	}
		806	SET_PTL0_ADDRESS(new->page_table);
1380	jermar	807	mutex_unlock(&new->lock);
823	jermar	808
		809	if (needs_asid) {
		810	/*
		811	* Allocation of new ASID was deferred
		812	* until now in order to avoid deadlock.
		813	*/
		814	asid_t asid;
		815
		816	asid = asid_get();
1380	jermar	817	mutex_lock_active(&new->lock);
823	jermar	818	new->asid = asid;
1380	jermar	819	mutex_unlock(&new->lock);
823	jermar	820	}
1415	jermar	821	spinlock_unlock(&inactive_as_with_asid_lock);
823	jermar	822	interrupts_restore(ipl);
		823
		824	/*
703	jermar	825	* Perform architecture-specific steps.
727	jermar	826	* (e.g. write ASID to hardware register etc.)
703	jermar	827	*/
823	jermar	828	as_install_arch(new);
703	jermar	829
823	jermar	830	AS = new;
703	jermar	831	}
754	jermar	832
1235	jermar	833	/** Convert address space area flags to page flags.
754	jermar	834	*
1235	jermar	835	* @param aflags Flags of some address space area.
754	jermar	836	*
1235	jermar	837	* @return Flags to be passed to page_mapping_insert().
754	jermar	838	*/
1235	jermar	839	int area_flags_to_page_flags(int aflags)
754	jermar	840	{
		841	int flags;
		842
1178	jermar	843	flags = PAGE_USER \| PAGE_PRESENT;
754	jermar	844
1235	jermar	845	if (aflags & AS_AREA_READ)
1026	jermar	846	flags \|= PAGE_READ;
		847
1235	jermar	848	if (aflags & AS_AREA_WRITE)
1026	jermar	849	flags \|= PAGE_WRITE;
		850
1235	jermar	851	if (aflags & AS_AREA_EXEC)
1026	jermar	852	flags \|= PAGE_EXEC;
		853
1424	jermar	854	if (aflags & AS_AREA_CACHEABLE)
1178	jermar	855	flags \|= PAGE_CACHEABLE;
		856
754	jermar	857	return flags;
		858	}
756	jermar	859
1235	jermar	860	/** Compute flags for virtual address translation subsytem.
		861	*
		862	* The address space area must be locked.
		863	* Interrupts must be disabled.
		864	*
		865	* @param a Address space area.
		866	*
		867	* @return Flags to be used in page_mapping_insert().
		868	*/
1409	jermar	869	int as_area_get_flags(as_area_t *a)
1235	jermar	870	{
		871	return area_flags_to_page_flags(a->flags);
		872	}
		873
756	jermar	874	/** Create page table.
		875	*
		876	* Depending on architecture, create either address space
		877	* private or global page table.
		878	*
		879	* @param flags Flags saying whether the page table is for kernel address space.
		880	*
		881	* @return First entry of the page table.
		882	*/
		883	pte_t *page_table_create(int flags)
		884	{
		885	ASSERT(as_operations);
		886	ASSERT(as_operations->page_table_create);
		887
		888	return as_operations->page_table_create(flags);
		889	}
977	jermar	890
1468	jermar	891	/** Destroy page table.
		892	*
		893	* Destroy page table in architecture specific way.
		894	*
		895	* @param page_table Physical address of PTL0.
		896	*/
		897	void page_table_destroy(pte_t *page_table)
		898	{
		899	ASSERT(as_operations);
		900	ASSERT(as_operations->page_table_destroy);
		901
		902	as_operations->page_table_destroy(page_table);
		903	}
		904
1044	jermar	905	/** Lock page table.
		906	*
		907	* This function should be called before any page_mapping_insert(),
		908	* page_mapping_remove() and page_mapping_find().
		909	*
		910	* Locking order is such that address space areas must be locked
		911	* prior to this call. Address space can be locked prior to this
		912	* call in which case the lock argument is false.
		913	*
		914	* @param as Address space.
1248	jermar	915	* @param lock If false, do not attempt to lock as->lock.
1044	jermar	916	*/
		917	void page_table_lock(as_t *as, bool lock)
		918	{
		919	ASSERT(as_operations);
		920	ASSERT(as_operations->page_table_lock);
		921
		922	as_operations->page_table_lock(as, lock);
		923	}
		924
		925	/** Unlock page table.
		926	*
		927	* @param as Address space.
1248	jermar	928	* @param unlock If false, do not attempt to unlock as->lock.
1044	jermar	929	*/
		930	void page_table_unlock(as_t *as, bool unlock)
		931	{
		932	ASSERT(as_operations);
		933	ASSERT(as_operations->page_table_unlock);
		934
		935	as_operations->page_table_unlock(as, unlock);
		936	}
		937
977	jermar	938
		939	/** Find address space area and lock it.
		940	*
		941	* The address space must be locked and interrupts must be disabled.
		942	*
		943	* @param as Address space.
		944	* @param va Virtual address.
		945	*
		946	* @return Locked address space area containing va on success or NULL on failure.
		947	*/
		948	as_area_t find_area_and_lock(as_t as, __address va)
		949	{
		950	as_area_t *a;
1147	jermar	951	btree_node_t leaf, lnode;
		952	int i;
977	jermar	953
1147	jermar	954	a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
		955	if (a) {
		956	/* va is the base address of an address space area */
1380	jermar	957	mutex_lock(&a->lock);
1147	jermar	958	return a;
		959	}
		960
		961	/*
1150	jermar	962	* Search the leaf node and the righmost record of its left neighbour
1147	jermar	963	* to find out whether this is a miss or va belongs to an address
		964	* space area found there.
		965	*/
		966
		967	/* First, search the leaf node itself. */
		968	for (i = 0; i < leaf->keys; i++) {
		969	a = (as_area_t *) leaf->value[i];
1380	jermar	970	mutex_lock(&a->lock);
1147	jermar	971	if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
		972	return a;
		973	}
1380	jermar	974	mutex_unlock(&a->lock);
1147	jermar	975	}
977	jermar	976
1147	jermar	977	/*
1150	jermar	978	* Second, locate the left neighbour and test its last record.
1148	jermar	979	* Because of its position in the B+tree, it must have base < va.
1147	jermar	980	*/
1150	jermar	981	if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
1147	jermar	982	a = (as_area_t *) lnode->value[lnode->keys - 1];
1380	jermar	983	mutex_lock(&a->lock);
1147	jermar	984	if (va < a->base + a->pages * PAGE_SIZE) {
1048	jermar	985	return a;
1147	jermar	986	}
1380	jermar	987	mutex_unlock(&a->lock);
977	jermar	988	}
		989
		990	return NULL;
		991	}
1048	jermar	992
		993	/** Check area conflicts with other areas.
		994	*
		995	* The address space must be locked and interrupts must be disabled.
		996	*
		997	* @param as Address space.
		998	* @param va Starting virtual address of the area being tested.
		999	* @param size Size of the area being tested.
		1000	* @param avoid_area Do not touch this area.
		1001	*
		1002	* @return True if there is no conflict, false otherwise.
		1003	*/
		1004	bool check_area_conflicts(as_t as, __address va, size_t size, as_area_t avoid_area)
		1005	{
		1006	as_area_t *a;
1147	jermar	1007	btree_node_t leaf, node;
		1008	int i;
1048	jermar	1009
1070	jermar	1010	/*
		1011	* We don't want any area to have conflicts with NULL page.
		1012	*/
		1013	if (overlaps(va, size, NULL, PAGE_SIZE))
		1014	return false;
		1015
1147	jermar	1016	/*
		1017	* The leaf node is found in O(log n), where n is proportional to
		1018	* the number of address space areas belonging to as.
		1019	* The check for conflicts is then attempted on the rightmost
1150	jermar	1020	* record in the left neighbour, the leftmost record in the right
		1021	* neighbour and all records in the leaf node itself.
1147	jermar	1022	*/
1048	jermar	1023
1147	jermar	1024	if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
		1025	if (a != avoid_area)
		1026	return false;
		1027	}
		1028
		1029	/* First, check the two border cases. */
1150	jermar	1030	if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
1147	jermar	1031	a = (as_area_t *) node->value[node->keys - 1];
1380	jermar	1032	mutex_lock(&a->lock);
1147	jermar	1033	if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
1380	jermar	1034	mutex_unlock(&a->lock);
1147	jermar	1035	return false;
		1036	}
1380	jermar	1037	mutex_unlock(&a->lock);
1147	jermar	1038	}
1150	jermar	1039	if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
1147	jermar	1040	a = (as_area_t *) node->value[0];
1380	jermar	1041	mutex_lock(&a->lock);
1147	jermar	1042	if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
1380	jermar	1043	mutex_unlock(&a->lock);
1147	jermar	1044	return false;
		1045	}
1380	jermar	1046	mutex_unlock(&a->lock);
1147	jermar	1047	}
		1048
		1049	/* Second, check the leaf node. */
		1050	for (i = 0; i < leaf->keys; i++) {
		1051	a = (as_area_t *) leaf->value[i];
		1052
1048	jermar	1053	if (a == avoid_area)
		1054	continue;
1147	jermar	1055
1380	jermar	1056	mutex_lock(&a->lock);
1147	jermar	1057	if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
1380	jermar	1058	mutex_unlock(&a->lock);
1147	jermar	1059	return false;
		1060	}
1380	jermar	1061	mutex_unlock(&a->lock);
1048	jermar	1062	}
		1063
1070	jermar	1064	/*
		1065	* So far, the area does not conflict with other areas.
		1066	* Check if it doesn't conflict with kernel address space.
		1067	*/
		1068	if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
		1069	return !overlaps(va, size,
		1070	KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
		1071	}
		1072
1048	jermar	1073	return true;
		1074	}
1235	jermar	1075
1380	jermar	1076	/** Return size of the address space area with given base. */
1329	palkovsky	1077	size_t as_get_size(__address base)
		1078	{
		1079	ipl_t ipl;
		1080	as_area_t *src_area;
		1081	size_t size;
		1082
		1083	ipl = interrupts_disable();
		1084	src_area = find_area_and_lock(AS, base);
		1085	if (src_area){
		1086	size = src_area->pages * PAGE_SIZE;
1380	jermar	1087	mutex_unlock(&src_area->lock);
1329	palkovsky	1088	} else {
		1089	size = 0;
		1090	}
		1091	interrupts_restore(ipl);
		1092	return size;
		1093	}
		1094
1387	jermar	1095	/** Mark portion of address space area as used.
		1096	*
		1097	* The address space area must be already locked.
		1098	*
		1099	* @param a Address space area.
		1100	* @param page First page to be marked.
		1101	* @param count Number of page to be marked.
		1102	*
		1103	* @return 0 on failure and 1 on success.
		1104	*/
		1105	int used_space_insert(as_area_t *a, __address page, count_t count)
		1106	{
		1107	btree_node_t leaf, node;
		1108	count_t pages;
		1109	int i;
		1110
		1111	ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
		1112	ASSERT(count);
		1113
		1114	pages = (count_t) btree_search(&a->used_space, page, &leaf);
		1115	if (pages) {
		1116	/*
		1117	* We hit the beginning of some used space.
		1118	*/
		1119	return 0;
		1120	}
		1121
1437	jermar	1122	if (!leaf->keys) {
		1123	btree_insert(&a->used_space, page, (void *) count, leaf);
		1124	return 1;
		1125	}
		1126
1387	jermar	1127	node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
		1128	if (node) {
		1129	__address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
		1130	count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
		1131
		1132	/*
		1133	* Examine the possibility that the interval fits
		1134	* somewhere between the rightmost interval of
		1135	* the left neigbour and the first interval of the leaf.
		1136	*/
		1137
		1138	if (page >= right_pg) {
		1139	/* Do nothing. */
		1140	} else if (overlaps(page, countPAGE_SIZE, left_pg, left_cntPAGE_SIZE)) {
		1141	/* The interval intersects with the left interval. */
		1142	return 0;
		1143	} else if (overlaps(page, countPAGE_SIZE, right_pg, right_cntPAGE_SIZE)) {
		1144	/* The interval intersects with the right interval. */
		1145	return 0;
		1146	} else if ((page == left_pg + left_cntPAGE_SIZE) && (page + countPAGE_SIZE == right_pg)) {
		1147	/* The interval can be added by merging the two already present intervals. */
1403	jermar	1148	node->value[node->keys - 1] += count + right_cnt;
1387	jermar	1149	btree_remove(&a->used_space, right_pg, leaf);
		1150	return 1;
		1151	} else if (page == left_pg + left_cnt*PAGE_SIZE) {
		1152	/* The interval can be added by simply growing the left interval. */
1403	jermar	1153	node->value[node->keys - 1] += count;
1387	jermar	1154	return 1;
		1155	} else if (page + count*PAGE_SIZE == right_pg) {
		1156	/*
		1157	* The interval can be addded by simply moving base of the right
		1158	* interval down and increasing its size accordingly.
		1159	*/
1403	jermar	1160	leaf->value[0] += count;
1387	jermar	1161	leaf->key[0] = page;
		1162	return 1;
		1163	} else {
		1164	/*
		1165	* The interval is between both neigbouring intervals,
		1166	* but cannot be merged with any of them.
		1167	*/
		1168	btree_insert(&a->used_space, page, (void *) count, leaf);
		1169	return 1;
		1170	}
		1171	} else if (page < leaf->key[0]) {
		1172	__address right_pg = leaf->key[0];
		1173	count_t right_cnt = (count_t) leaf->value[0];
		1174
		1175	/*
		1176	* Investigate the border case in which the left neighbour does not
		1177	* exist but the interval fits from the left.
		1178	*/
		1179
		1180	if (overlaps(page, countPAGE_SIZE, right_pg, right_cntPAGE_SIZE)) {
		1181	/* The interval intersects with the right interval. */
		1182	return 0;
		1183	} else if (page + count*PAGE_SIZE == right_pg) {
		1184	/*
		1185	* The interval can be added by moving the base of the right interval down
		1186	* and increasing its size accordingly.
		1187	*/
		1188	leaf->key[0] = page;
1403	jermar	1189	leaf->value[0] += count;
1387	jermar	1190	return 1;
		1191	} else {
		1192	/*
		1193	* The interval doesn't adjoin with the right interval.
		1194	* It must be added individually.
		1195	*/
		1196	btree_insert(&a->used_space, page, (void *) count, leaf);
		1197	return 1;
		1198	}
		1199	}
		1200
		1201	node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
		1202	if (node) {
		1203	__address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
		1204	count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
		1205
		1206	/*
		1207	* Examine the possibility that the interval fits
		1208	* somewhere between the leftmost interval of
		1209	* the right neigbour and the last interval of the leaf.
		1210	*/
		1211
		1212	if (page < left_pg) {
		1213	/* Do nothing. */
		1214	} else if (overlaps(page, countPAGE_SIZE, left_pg, left_cntPAGE_SIZE)) {
		1215	/* The interval intersects with the left interval. */
		1216	return 0;
		1217	} else if (overlaps(page, countPAGE_SIZE, right_pg, right_cntPAGE_SIZE)) {
		1218	/* The interval intersects with the right interval. */
		1219	return 0;
		1220	} else if ((page == left_pg + left_cntPAGE_SIZE) && (page + countPAGE_SIZE == right_pg)) {
		1221	/* The interval can be added by merging the two already present intervals. */
1403	jermar	1222	leaf->value[leaf->keys - 1] += count + right_cnt;
1387	jermar	1223	btree_remove(&a->used_space, right_pg, node);
		1224	return 1;
		1225	} else if (page == left_pg + left_cnt*PAGE_SIZE) {
		1226	/* The interval can be added by simply growing the left interval. */
1403	jermar	1227	leaf->value[leaf->keys - 1] += count;
1387	jermar	1228	return 1;
		1229	} else if (page + count*PAGE_SIZE == right_pg) {
		1230	/*
		1231	* The interval can be addded by simply moving base of the right
		1232	* interval down and increasing its size accordingly.
		1233	*/
1403	jermar	1234	node->value[0] += count;
1387	jermar	1235	node->key[0] = page;
		1236	return 1;
		1237	} else {
		1238	/*
		1239	* The interval is between both neigbouring intervals,
		1240	* but cannot be merged with any of them.
		1241	*/
		1242	btree_insert(&a->used_space, page, (void *) count, leaf);
		1243	return 1;
		1244	}
		1245	} else if (page >= leaf->key[leaf->keys - 1]) {
		1246	__address left_pg = leaf->key[leaf->keys - 1];
		1247	count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
		1248
		1249	/*
		1250	* Investigate the border case in which the right neighbour does not
		1251	* exist but the interval fits from the right.
		1252	*/
		1253
		1254	if (overlaps(page, countPAGE_SIZE, left_pg, left_cntPAGE_SIZE)) {
1403	jermar	1255	/* The interval intersects with the left interval. */
1387	jermar	1256	return 0;
		1257	} else if (left_pg + left_cnt*PAGE_SIZE == page) {
		1258	/* The interval can be added by growing the left interval. */
1403	jermar	1259	leaf->value[leaf->keys - 1] += count;
1387	jermar	1260	return 1;
		1261	} else {
		1262	/*
		1263	* The interval doesn't adjoin with the left interval.
		1264	* It must be added individually.
		1265	*/
		1266	btree_insert(&a->used_space, page, (void *) count, leaf);
		1267	return 1;
		1268	}
		1269	}
		1270
		1271	/*
		1272	* Note that if the algorithm made it thus far, the interval can fit only
		1273	* between two other intervals of the leaf. The two border cases were already
		1274	* resolved.
		1275	*/
		1276	for (i = 1; i < leaf->keys; i++) {
		1277	if (page < leaf->key[i]) {
		1278	__address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
		1279	count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
		1280
		1281	/*
		1282	* The interval fits between left_pg and right_pg.
		1283	*/
		1284
		1285	if (overlaps(page, countPAGE_SIZE, left_pg, left_cntPAGE_SIZE)) {
		1286	/* The interval intersects with the left interval. */
		1287	return 0;
		1288	} else if (overlaps(page, countPAGE_SIZE, right_pg, right_cntPAGE_SIZE)) {
		1289	/* The interval intersects with the right interval. */
		1290	return 0;
		1291	} else if ((page == left_pg + left_cntPAGE_SIZE) && (page + countPAGE_SIZE == right_pg)) {
		1292	/* The interval can be added by merging the two already present intervals. */
1403	jermar	1293	leaf->value[i - 1] += count + right_cnt;
1387	jermar	1294	btree_remove(&a->used_space, right_pg, leaf);
		1295	return 1;
		1296	} else if (page == left_pg + left_cnt*PAGE_SIZE) {
		1297	/* The interval can be added by simply growing the left interval. */
1403	jermar	1298	leaf->value[i - 1] += count;
1387	jermar	1299	return 1;
		1300	} else if (page + count*PAGE_SIZE == right_pg) {
		1301	/*
		1302	* The interval can be addded by simply moving base of the right
		1303	* interval down and increasing its size accordingly.
		1304	*/
1403	jermar	1305	leaf->value[i] += count;
1387	jermar	1306	leaf->key[i] = page;
		1307	return 1;
		1308	} else {
		1309	/*
		1310	* The interval is between both neigbouring intervals,
		1311	* but cannot be merged with any of them.
		1312	*/
		1313	btree_insert(&a->used_space, page, (void *) count, leaf);
		1314	return 1;
		1315	}
		1316	}
		1317	}
		1318
		1319	panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
		1320	}
		1321
		1322	/** Mark portion of address space area as unused.
		1323	*
		1324	* The address space area must be already locked.
		1325	*
		1326	* @param a Address space area.
		1327	* @param page First page to be marked.
		1328	* @param count Number of page to be marked.
		1329	*
		1330	* @return 0 on failure and 1 on success.
		1331	*/
		1332	int used_space_remove(as_area_t *a, __address page, count_t count)
		1333	{
		1334	btree_node_t leaf, node;
		1335	count_t pages;
		1336	int i;
		1337
		1338	ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
		1339	ASSERT(count);
		1340
		1341	pages = (count_t) btree_search(&a->used_space, page, &leaf);
		1342	if (pages) {
		1343	/*
		1344	* We are lucky, page is the beginning of some interval.
		1345	*/
		1346	if (count > pages) {
		1347	return 0;
		1348	} else if (count == pages) {
		1349	btree_remove(&a->used_space, page, leaf);
1403	jermar	1350	return 1;
1387	jermar	1351	} else {
		1352	/*
		1353	* Find the respective interval.
		1354	* Decrease its size and relocate its start address.
		1355	*/
		1356	for (i = 0; i < leaf->keys; i++) {
		1357	if (leaf->key[i] == page) {
		1358	leaf->key[i] += count*PAGE_SIZE;
1403	jermar	1359	leaf->value[i] -= count;
1387	jermar	1360	return 1;
		1361	}
		1362	}
		1363	goto error;
		1364	}
		1365	}
		1366
		1367	node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
		1368	if (node && page < leaf->key[0]) {
		1369	__address left_pg = node->key[node->keys - 1];
		1370	count_t left_cnt = (count_t) node->value[node->keys - 1];
		1371
		1372	if (overlaps(left_pg, left_cntPAGE_SIZE, page, countPAGE_SIZE)) {
		1373	if (page + countPAGE_SIZE == left_pg + left_cntPAGE_SIZE) {
		1374	/*
		1375	* The interval is contained in the rightmost interval
		1376	* of the left neighbour and can be removed by
		1377	* updating the size of the bigger interval.
		1378	*/
1403	jermar	1379	node->value[node->keys - 1] -= count;
1387	jermar	1380	return 1;
		1381	} else if (page + countPAGE_SIZE < left_pg + left_cntPAGE_SIZE) {
1403	jermar	1382	count_t new_cnt;
1387	jermar	1383
		1384	/*
		1385	* The interval is contained in the rightmost interval
		1386	* of the left neighbour but its removal requires
		1387	* both updating the size of the original interval and
		1388	* also inserting a new interval.
		1389	*/
1403	jermar	1390	new_cnt = ((left_pg + left_cntPAGE_SIZE) - (page + countPAGE_SIZE)) >> PAGE_WIDTH;
		1391	node->value[node->keys - 1] -= count + new_cnt;
1387	jermar	1392	btree_insert(&a->used_space, page + countPAGE_SIZE, (void ) new_cnt, leaf);
		1393	return 1;
		1394	}
		1395	}
		1396	return 0;
		1397	} else if (page < leaf->key[0]) {
		1398	return 0;
		1399	}
		1400
		1401	if (page > leaf->key[leaf->keys - 1]) {
		1402	__address left_pg = leaf->key[leaf->keys - 1];
		1403	count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
		1404
		1405	if (overlaps(left_pg, left_cntPAGE_SIZE, page, countPAGE_SIZE)) {
		1406	if (page + countPAGE_SIZE == left_pg + left_cntPAGE_SIZE) {
		1407	/*
		1408	* The interval is contained in the rightmost interval
		1409	* of the leaf and can be removed by updating the size
		1410	* of the bigger interval.
		1411	*/
1403	jermar	1412	leaf->value[leaf->keys - 1] -= count;
1387	jermar	1413	return 1;
		1414	} else if (page + countPAGE_SIZE < left_pg + left_cntPAGE_SIZE) {
1403	jermar	1415	count_t new_cnt;
1387	jermar	1416
		1417	/*
		1418	* The interval is contained in the rightmost interval
		1419	* of the leaf but its removal requires both updating
		1420	* the size of the original interval and
		1421	* also inserting a new interval.
		1422	*/
1403	jermar	1423	new_cnt = ((left_pg + left_cntPAGE_SIZE) - (page + countPAGE_SIZE)) >> PAGE_WIDTH;
		1424	leaf->value[leaf->keys - 1] -= count + new_cnt;
1387	jermar	1425	btree_insert(&a->used_space, page + countPAGE_SIZE, (void ) new_cnt, leaf);
		1426	return 1;
		1427	}
		1428	}
		1429	return 0;
		1430	}
		1431
		1432	/*
		1433	* The border cases have been already resolved.
		1434	* Now the interval can be only between intervals of the leaf.
		1435	*/
		1436	for (i = 1; i < leaf->keys - 1; i++) {
		1437	if (page < leaf->key[i]) {
		1438	__address left_pg = leaf->key[i - 1];
		1439	count_t left_cnt = (count_t) leaf->value[i - 1];
		1440
		1441	/*
		1442	* Now the interval is between intervals corresponding to (i - 1) and i.
		1443	*/
		1444	if (overlaps(left_pg, left_cntPAGE_SIZE, page, countPAGE_SIZE)) {
		1445	if (page + countPAGE_SIZE == left_pg + left_cntPAGE_SIZE) {
		1446	/*
		1447	* The interval is contained in the interval (i - 1)
		1448	* of the leaf and can be removed by updating the size
		1449	* of the bigger interval.
		1450	*/
1403	jermar	1451	leaf->value[i - 1] -= count;
1387	jermar	1452	return 1;
		1453	} else if (page + countPAGE_SIZE < left_pg + left_cntPAGE_SIZE) {
1403	jermar	1454	count_t new_cnt;
1387	jermar	1455
		1456	/*
		1457	* The interval is contained in the interval (i - 1)
		1458	* of the leaf but its removal requires both updating
		1459	* the size of the original interval and
		1460	* also inserting a new interval.
		1461	*/
1403	jermar	1462	new_cnt = ((left_pg + left_cntPAGE_SIZE) - (page + countPAGE_SIZE)) >> PAGE_WIDTH;
		1463	leaf->value[i - 1] -= count + new_cnt;
1387	jermar	1464	btree_insert(&a->used_space, page + countPAGE_SIZE, (void ) new_cnt, leaf);
		1465	return 1;
		1466	}
		1467	}
		1468	return 0;
		1469	}
		1470	}
		1471
		1472	error:
		1473	panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
		1474	}
		1475
1409	jermar	1476	/** Remove reference to address space area share info.
		1477	*
		1478	* If the reference count drops to 0, the sh_info is deallocated.
		1479	*
		1480	* @param sh_info Pointer to address space area share info.
		1481	*/
		1482	void sh_info_remove_reference(share_info_t *sh_info)
		1483	{
		1484	bool dealloc = false;
		1485
		1486	mutex_lock(&sh_info->lock);
		1487	ASSERT(sh_info->refcount);
		1488	if (--sh_info->refcount == 0) {
		1489	dealloc = true;
1495	jermar	1490	link_t *cur;
1409	jermar	1491
		1492	/*
		1493	* Now walk carefully the pagemap B+tree and free/remove
		1494	* reference from all frames found there.
		1495	*/
1495	jermar	1496	for (cur = sh_info->pagemap.leaf_head.next; cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
1409	jermar	1497	btree_node_t *node;
1495	jermar	1498	int i;
1409	jermar	1499
1495	jermar	1500	node = list_get_instance(cur, btree_node_t, leaf_link);
		1501	for (i = 0; i < node->keys; i++)
		1502	frame_free(ADDR2PFN((__address) node->value[i]));
1409	jermar	1503	}
		1504
		1505	}
		1506	mutex_unlock(&sh_info->lock);
		1507
		1508	if (dealloc) {
		1509	btree_destroy(&sh_info->pagemap);
		1510	free(sh_info);
		1511	}
		1512	}
		1513
1235	jermar	1514	/*
		1515	* Address space related syscalls.
		1516	*/
		1517
		1518	/** Wrapper for as_area_create(). */
		1519	__native sys_as_area_create(__address address, size_t size, int flags)
		1520	{
1424	jermar	1521	if (as_area_create(AS, flags \| AS_AREA_CACHEABLE, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
1235	jermar	1522	return (__native) address;
		1523	else
		1524	return (__native) -1;
		1525	}
		1526
		1527	/** Wrapper for as_area_resize. */
		1528	__native sys_as_area_resize(__address address, size_t size, int flags)
		1529	{
1306	jermar	1530	return (__native) as_area_resize(AS, address, size, 0);
1235	jermar	1531	}
		1532
1306	jermar	1533	/** Wrapper for as_area_destroy. */
		1534	__native sys_as_area_destroy(__address address)
		1535	{
		1536	return (__native) as_area_destroy(AS, address);
		1537	}
1702	cejka	1538
		1539	/** @}
		1540	*/
		1541

Subversion Repositories HelenOS-historic

(root)/kernel/trunk/generic/src/mm/as.c – Rev 1702