Subversion Repositories HelenOS

Rev

Rev 2125 | Rev 2131 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed

  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.  
  29. /** @addtogroup genericmm
  30.  * @{
  31.  */
  32.  
  33. /**
  34.  * @file
  35.  * @brief   Address space related functions.
  36.  *
  37.  * This file contains address space manipulation functions.
  38.  * Roughly speaking, this is a higher-level client of
  39.  * Virtual Address Translation (VAT) subsystem.
  40.  *
  41.  * Functionality provided by this file allows one to
  42.  * create address spaces and create, resize and share
  43.  * address space areas.
  44.  *
  45.  * @see page.c
  46.  *
  47.  */
  48.  
  49. #include <mm/as.h>
  50. #include <arch/mm/as.h>
  51. #include <mm/page.h>
  52. #include <mm/frame.h>
  53. #include <mm/slab.h>
  54. #include <mm/tlb.h>
  55. #include <arch/mm/page.h>
  56. #include <genarch/mm/page_pt.h>
  57. #include <genarch/mm/page_ht.h>
  58. #include <mm/asid.h>
  59. #include <arch/mm/asid.h>
  60. #include <synch/spinlock.h>
  61. #include <synch/mutex.h>
  62. #include <adt/list.h>
  63. #include <adt/btree.h>
  64. #include <proc/task.h>
  65. #include <proc/thread.h>
  66. #include <arch/asm.h>
  67. #include <panic.h>
  68. #include <debug.h>
  69. #include <print.h>
  70. #include <memstr.h>
  71. #include <macros.h>
  72. #include <arch.h>
  73. #include <errno.h>
  74. #include <config.h>
  75. #include <align.h>
  76. #include <arch/types.h>
  77. #include <syscall/copy.h>
  78. #include <arch/interrupt.h>
  79.  
  80. #ifdef CONFIG_VIRT_IDX_DCACHE
  81. #include <arch/mm/cache.h>
  82. #endif /* CONFIG_VIRT_IDX_DCACHE */
  83.  
  84. #ifndef __OBJC__
  85. /**
  86.  * Each architecture decides what functions will be used to carry out
  87.  * address space operations such as creating or locking page tables.
  88.  */
  89. as_operations_t *as_operations = NULL;
  90.  
  91. /**
  92.  * Slab for as_t objects.
  93.  */
  94. static slab_cache_t *as_slab;
  95. #endif
  96.  
  97. /**
  98.  * This lock protects inactive_as_with_asid_head list. It must be acquired
  99.  * before as_t mutex.
  100.  */
  101. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  102.  
  103. /**
  104.  * This list contains address spaces that are not active on any
  105.  * processor and that have valid ASID.
  106.  */
  107. LIST_INITIALIZE(inactive_as_with_asid_head);
  108.  
  109. /** Kernel address space. */
  110. as_t *AS_KERNEL = NULL;
  111.  
  112. static int area_flags_to_page_flags(int aflags);
  113. static as_area_t *find_area_and_lock(as_t *as, uintptr_t va);
  114. static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  115.     as_area_t *avoid_area);
  116. static void sh_info_remove_reference(share_info_t *sh_info);
  117.  
  118. #ifndef __OBJC__
  119. static int as_constructor(void *obj, int flags)
  120. {
  121.     as_t *as = (as_t *) obj;
  122.     int rc;
  123.  
  124.     link_initialize(&as->inactive_as_with_asid_link);
  125.     mutex_initialize(&as->lock);   
  126.    
  127.     rc = as_constructor_arch(as, flags);
  128.    
  129.     return rc;
  130. }
  131.  
  132. static int as_destructor(void *obj)
  133. {
  134.     as_t *as = (as_t *) obj;
  135.  
  136.     return as_destructor_arch(as);
  137. }
  138. #endif
  139.  
  140. /** Initialize address space subsystem. */
  141. void as_init(void)
  142. {
  143.     as_arch_init();
  144.  
  145. #ifndef __OBJC__
  146.     as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,
  147.         as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);
  148. #endif
  149.    
  150.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  151.     if (!AS_KERNEL)
  152.         panic("can't create kernel address space\n");
  153.    
  154. }
  155.  
  156. /** Create address space.
  157.  *
  158.  * @param flags Flags that influence way in wich the address space is created.
  159.  */
  160. as_t *as_create(int flags)
  161. {
  162.     as_t *as;
  163.  
  164. #ifdef __OBJC__
  165.     as = [as_t new];
  166.     link_initialize(&as->inactive_as_with_asid_link);
  167.     mutex_initialize(&as->lock);   
  168.     (void) as_constructor_arch(as, flags);
  169. #else
  170.     as = (as_t *) slab_alloc(as_slab, 0);
  171. #endif
  172.     (void) as_create_arch(as, 0);
  173.    
  174.     btree_create(&as->as_area_btree);
  175.    
  176.     if (flags & FLAG_AS_KERNEL)
  177.         as->asid = ASID_KERNEL;
  178.     else
  179.         as->asid = ASID_INVALID;
  180.    
  181.     as->refcount = 0;
  182.     as->cpu_refcount = 0;
  183. #ifdef AS_PAGE_TABLE
  184.     as->genarch.page_table = page_table_create(flags);
  185. #else
  186.     page_table_create(flags);
  187. #endif
  188.  
  189.     return as;
  190. }
  191.  
  192. /** Destroy adress space.
  193.  *
  194.  * When there are no tasks referencing this address space (i.e. its refcount is
  195.  * zero), the address space can be destroyed.
  196.  */
  197. void as_destroy(as_t *as)
  198. {
  199.     ipl_t ipl;
  200.     bool cond;
  201.  
  202.     ASSERT(as->refcount == 0);
  203.    
  204.     /*
  205.      * Since there is no reference to this area,
  206.      * it is safe not to lock its mutex.
  207.      */
  208.     ipl = interrupts_disable();
  209.     spinlock_lock(&inactive_as_with_asid_lock);
  210.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  211.         if (as != AS && as->cpu_refcount == 0)
  212.             list_remove(&as->inactive_as_with_asid_link);
  213.         asid_put(as->asid);
  214.     }
  215.     spinlock_unlock(&inactive_as_with_asid_lock);
  216.  
  217.     /*
  218.      * Destroy address space areas of the address space.
  219.      * The B+tree must be walked carefully because it is
  220.      * also being destroyed.
  221.      */
  222.     for (cond = true; cond; ) {
  223.         btree_node_t *node;
  224.  
  225.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  226.         node = list_get_instance(as->as_area_btree.leaf_head.next,
  227.             btree_node_t, leaf_link);
  228.  
  229.         if ((cond = node->keys)) {
  230.             as_area_destroy(as, node->key[0]);
  231.         }
  232.     }
  233.  
  234.     btree_destroy(&as->as_area_btree);
  235. #ifdef AS_PAGE_TABLE
  236.     page_table_destroy(as->genarch.page_table);
  237. #else
  238.     page_table_destroy(NULL);
  239. #endif
  240.  
  241.     interrupts_restore(ipl);
  242.  
  243. #ifdef __OBJC__
  244.     [as free];
  245. #else
  246.     slab_free(as_slab, as);
  247. #endif
  248. }
  249.  
  250. /** Create address space area of common attributes.
  251.  *
  252.  * The created address space area is added to the target address space.
  253.  *
  254.  * @param as Target address space.
  255.  * @param flags Flags of the area memory.
  256.  * @param size Size of area.
  257.  * @param base Base address of area.
  258.  * @param attrs Attributes of the area.
  259.  * @param backend Address space area backend. NULL if no backend is used.
  260.  * @param backend_data NULL or a pointer to an array holding two void *.
  261.  *
  262.  * @return Address space area on success or NULL on failure.
  263.  */
  264. as_area_t *
  265. as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
  266.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  267. {
  268.     ipl_t ipl;
  269.     as_area_t *a;
  270.    
  271.     if (base % PAGE_SIZE)
  272.         return NULL;
  273.  
  274.     if (!size)
  275.         return NULL;
  276.  
  277.     /* Writeable executable areas are not supported. */
  278.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  279.         return NULL;
  280.    
  281.     ipl = interrupts_disable();
  282.     mutex_lock(&as->lock);
  283.    
  284.     if (!check_area_conflicts(as, base, size, NULL)) {
  285.         mutex_unlock(&as->lock);
  286.         interrupts_restore(ipl);
  287.         return NULL;
  288.     }
  289.    
  290.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  291.  
  292.     mutex_initialize(&a->lock);
  293.    
  294.     a->as = as;
  295.     a->flags = flags;
  296.     a->attributes = attrs;
  297.     a->pages = SIZE2FRAMES(size);
  298.     a->base = base;
  299.     a->sh_info = NULL;
  300.     a->backend = backend;
  301.     if (backend_data)
  302.         a->backend_data = *backend_data;
  303.     else
  304.         memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data),
  305.             0);
  306.  
  307.     btree_create(&a->used_space);
  308.    
  309.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  310.  
  311.     mutex_unlock(&as->lock);
  312.     interrupts_restore(ipl);
  313.  
  314.     return a;
  315. }
  316.  
  317. /** Find address space area and change it.
  318.  *
  319.  * @param as Address space.
  320.  * @param address Virtual address belonging to the area to be changed. Must be
  321.  *     page-aligned.
  322.  * @param size New size of the virtual memory block starting at address.
  323.  * @param flags Flags influencing the remap operation. Currently unused.
  324.  *
  325.  * @return Zero on success or a value from @ref errno.h otherwise.
  326.  */
  327. int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
  328. {
  329.     as_area_t *area;
  330.     ipl_t ipl;
  331.     size_t pages;
  332.    
  333.     ipl = interrupts_disable();
  334.     mutex_lock(&as->lock);
  335.    
  336.     /*
  337.      * Locate the area.
  338.      */
  339.     area = find_area_and_lock(as, address);
  340.     if (!area) {
  341.         mutex_unlock(&as->lock);
  342.         interrupts_restore(ipl);
  343.         return ENOENT;
  344.     }
  345.  
  346.     if (area->backend == &phys_backend) {
  347.         /*
  348.          * Remapping of address space areas associated
  349.          * with memory mapped devices is not supported.
  350.          */
  351.         mutex_unlock(&area->lock);
  352.         mutex_unlock(&as->lock);
  353.         interrupts_restore(ipl);
  354.         return ENOTSUP;
  355.     }
  356.     if (area->sh_info) {
  357.         /*
  358.          * Remapping of shared address space areas
  359.          * is not supported.
  360.          */
  361.         mutex_unlock(&area->lock);
  362.         mutex_unlock(&as->lock);
  363.         interrupts_restore(ipl);
  364.         return ENOTSUP;
  365.     }
  366.  
  367.     pages = SIZE2FRAMES((address - area->base) + size);
  368.     if (!pages) {
  369.         /*
  370.          * Zero size address space areas are not allowed.
  371.          */
  372.         mutex_unlock(&area->lock);
  373.         mutex_unlock(&as->lock);
  374.         interrupts_restore(ipl);
  375.         return EPERM;
  376.     }
  377.    
  378.     if (pages < area->pages) {
  379.         bool cond;
  380.         uintptr_t start_free = area->base + pages*PAGE_SIZE;
  381.  
  382.         /*
  383.          * Shrinking the area.
  384.          * No need to check for overlaps.
  385.          */
  386.  
  387.         /*
  388.          * Start TLB shootdown sequence.
  389.          */
  390.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base +
  391.             pages * PAGE_SIZE, area->pages - pages);
  392.  
  393.         /*
  394.          * Remove frames belonging to used space starting from
  395.          * the highest addresses downwards until an overlap with
  396.          * the resized address space area is found. Note that this
  397.          * is also the right way to remove part of the used_space
  398.          * B+tree leaf list.
  399.          */    
  400.         for (cond = true; cond;) {
  401.             btree_node_t *node;
  402.        
  403.             ASSERT(!list_empty(&area->used_space.leaf_head));
  404.             node =
  405.                 list_get_instance(area->used_space.leaf_head.prev,
  406.                 btree_node_t, leaf_link);
  407.             if ((cond = (bool) node->keys)) {
  408.                 uintptr_t b = node->key[node->keys - 1];
  409.                 count_t c =
  410.                     (count_t) node->value[node->keys - 1];
  411.                 int i = 0;
  412.            
  413.                 if (overlaps(b, c * PAGE_SIZE, area->base,
  414.                     pages*PAGE_SIZE)) {
  415.                    
  416.                     if (b + c * PAGE_SIZE <= start_free) {
  417.                         /*
  418.                          * The whole interval fits
  419.                          * completely in the resized
  420.                          * address space area.
  421.                          */
  422.                         break;
  423.                     }
  424.        
  425.                     /*
  426.                      * Part of the interval corresponding
  427.                      * to b and c overlaps with the resized
  428.                      * address space area.
  429.                      */
  430.        
  431.                     cond = false;   /* we are almost done */
  432.                     i = (start_free - b) >> PAGE_WIDTH;
  433.                     if (!used_space_remove(area, start_free,
  434.                         c - i))
  435.                         panic("Could not remove used "
  436.                             "space.\n");
  437.                 } else {
  438.                     /*
  439.                      * The interval of used space can be
  440.                      * completely removed.
  441.                      */
  442.                     if (!used_space_remove(area, b, c))
  443.                         panic("Could not remove used "
  444.                             "space.\n");
  445.                 }
  446.            
  447.                 for (; i < c; i++) {
  448.                     pte_t *pte;
  449.            
  450.                     page_table_lock(as, false);
  451.                     pte = page_mapping_find(as, b +
  452.                         i * PAGE_SIZE);
  453.                     ASSERT(pte && PTE_VALID(pte) &&
  454.                         PTE_PRESENT(pte));
  455.                     if (area->backend &&
  456.                         area->backend->frame_free) {
  457.                         area->backend->frame_free(area,
  458.                             b + i * PAGE_SIZE,
  459.                             PTE_GET_FRAME(pte));
  460.                     }
  461.                     page_mapping_remove(as, b +
  462.                         i * PAGE_SIZE);
  463.                     page_table_unlock(as, false);
  464.                 }
  465.             }
  466.         }
  467.  
  468.         /*
  469.          * Finish TLB shootdown sequence.
  470.          */
  471.         tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,
  472.             area->pages - pages);
  473.         tlb_shootdown_finalize();
  474.        
  475.         /*
  476.          * Invalidate software translation caches (e.g. TSB on sparc64).
  477.          */
  478.         as_invalidate_translation_cache(as, area->base +
  479.             pages * PAGE_SIZE, area->pages - pages);
  480.     } else {
  481.         /*
  482.          * Growing the area.
  483.          * Check for overlaps with other address space areas.
  484.          */
  485.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE,
  486.             area)) {
  487.             mutex_unlock(&area->lock);
  488.             mutex_unlock(&as->lock);       
  489.             interrupts_restore(ipl);
  490.             return EADDRNOTAVAIL;
  491.         }
  492.     }
  493.  
  494.     area->pages = pages;
  495.    
  496.     mutex_unlock(&area->lock);
  497.     mutex_unlock(&as->lock);
  498.     interrupts_restore(ipl);
  499.  
  500.     return 0;
  501. }
  502.  
  503. /** Destroy address space area.
  504.  *
  505.  * @param as Address space.
  506.  * @param address Address withing the area to be deleted.
  507.  *
  508.  * @return Zero on success or a value from @ref errno.h on failure.
  509.  */
  510. int as_area_destroy(as_t *as, uintptr_t address)
  511. {
  512.     as_area_t *area;
  513.     uintptr_t base;
  514.     link_t *cur;
  515.     ipl_t ipl;
  516.  
  517.     ipl = interrupts_disable();
  518.     mutex_lock(&as->lock);
  519.  
  520.     area = find_area_and_lock(as, address);
  521.     if (!area) {
  522.         mutex_unlock(&as->lock);
  523.         interrupts_restore(ipl);
  524.         return ENOENT;
  525.     }
  526.  
  527.     base = area->base;
  528.  
  529.     /*
  530.      * Start TLB shootdown sequence.
  531.      */
  532.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  533.  
  534.     /*
  535.      * Visit only the pages mapped by used_space B+tree.
  536.      */
  537.     for (cur = area->used_space.leaf_head.next;
  538.         cur != &area->used_space.leaf_head; cur = cur->next) {
  539.         btree_node_t *node;
  540.         int i;
  541.        
  542.         node = list_get_instance(cur, btree_node_t, leaf_link);
  543.         for (i = 0; i < node->keys; i++) {
  544.             uintptr_t b = node->key[i];
  545.             count_t j;
  546.             pte_t *pte;
  547.            
  548.             for (j = 0; j < (count_t) node->value[i]; j++) {
  549.                 page_table_lock(as, false);
  550.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  551.                 ASSERT(pte && PTE_VALID(pte) &&
  552.                     PTE_PRESENT(pte));
  553.                 if (area->backend &&
  554.                     area->backend->frame_free) {
  555.                     area->backend->frame_free(area, b +
  556.                     j * PAGE_SIZE, PTE_GET_FRAME(pte));
  557.                 }
  558.                 page_mapping_remove(as, b + j * PAGE_SIZE);            
  559.                 page_table_unlock(as, false);
  560.             }
  561.         }
  562.     }
  563.  
  564.     /*
  565.      * Finish TLB shootdown sequence.
  566.      */
  567.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  568.     tlb_shootdown_finalize();
  569.    
  570.     /*
  571.      * Invalidate potential software translation caches (e.g. TSB on
  572.      * sparc64).
  573.      */
  574.     as_invalidate_translation_cache(as, area->base, area->pages);
  575.    
  576.     btree_destroy(&area->used_space);
  577.  
  578.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  579.    
  580.     if (area->sh_info)
  581.         sh_info_remove_reference(area->sh_info);
  582.        
  583.     mutex_unlock(&area->lock);
  584.  
  585.     /*
  586.      * Remove the empty area from address space.
  587.      */
  588.     btree_remove(&as->as_area_btree, base, NULL);
  589.    
  590.     free(area);
  591.    
  592.     mutex_unlock(&as->lock);
  593.     interrupts_restore(ipl);
  594.     return 0;
  595. }
  596.  
  597. /** Share address space area with another or the same address space.
  598.  *
  599.  * Address space area mapping is shared with a new address space area.
  600.  * If the source address space area has not been shared so far,
  601.  * a new sh_info is created. The new address space area simply gets the
  602.  * sh_info of the source area. The process of duplicating the
  603.  * mapping is done through the backend share function.
  604.  *
  605.  * @param src_as Pointer to source address space.
  606.  * @param src_base Base address of the source address space area.
  607.  * @param acc_size Expected size of the source area.
  608.  * @param dst_as Pointer to destination address space.
  609.  * @param dst_base Target base address.
  610.  * @param dst_flags_mask Destination address space area flags mask.
  611.  *
  612.  * @return Zero on success or ENOENT if there is no such task or if there is no
  613.  * such address space area, EPERM if there was a problem in accepting the area
  614.  * or ENOMEM if there was a problem in allocating destination address space
  615.  * area. ENOTSUP is returned if the address space area backend does not support
  616.  * sharing or if the kernel detects an attempt to create an illegal address
  617.  * alias.
  618.  */
  619. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  620.           as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  621. {
  622.     ipl_t ipl;
  623.     int src_flags;
  624.     size_t src_size;
  625.     as_area_t *src_area, *dst_area;
  626.     share_info_t *sh_info;
  627.     mem_backend_t *src_backend;
  628.     mem_backend_data_t src_backend_data;
  629.    
  630.     ipl = interrupts_disable();
  631.     mutex_lock(&src_as->lock);
  632.     src_area = find_area_and_lock(src_as, src_base);
  633.     if (!src_area) {
  634.         /*
  635.          * Could not find the source address space area.
  636.          */
  637.         mutex_unlock(&src_as->lock);
  638.         interrupts_restore(ipl);
  639.         return ENOENT;
  640.     }
  641.  
  642.     if (!src_area->backend || !src_area->backend->share) {
  643.         /*
  644.          * There is no backend or the backend does not
  645.          * know how to share the area.
  646.          */
  647.         mutex_unlock(&src_area->lock);
  648.         mutex_unlock(&src_as->lock);
  649.         interrupts_restore(ipl);
  650.         return ENOTSUP;
  651.     }
  652.    
  653.     src_size = src_area->pages * PAGE_SIZE;
  654.     src_flags = src_area->flags;
  655.     src_backend = src_area->backend;
  656.     src_backend_data = src_area->backend_data;
  657.  
  658.     /* Share the cacheable flag from the original mapping */
  659.     if (src_flags & AS_AREA_CACHEABLE)
  660.         dst_flags_mask |= AS_AREA_CACHEABLE;
  661.  
  662.     if (src_size != acc_size ||
  663.         (src_flags & dst_flags_mask) != dst_flags_mask) {
  664.         mutex_unlock(&src_area->lock);
  665.         mutex_unlock(&src_as->lock);
  666.         interrupts_restore(ipl);
  667.         return EPERM;
  668.     }
  669.  
  670. #ifdef CONFIG_VIRT_IDX_DCACHE
  671.     if (!(dst_flags_mask & AS_AREA_EXEC)) {
  672.         if (PAGE_COLOR(src_area->base) != PAGE_COLOR(dst_base)) {
  673.             /*
  674.              * Refuse to create an illegal address alias.
  675.              */
  676.             mutex_unlock(&src_area->lock);
  677.             mutex_unlock(&src_as->lock);
  678.             interrupts_restore(ipl);
  679.             return ENOTSUP;
  680.         }
  681.     }
  682. #endif /* CONFIG_VIRT_IDX_DCACHE */
  683.  
  684.     /*
  685.      * Now we are committed to sharing the area.
  686.      * First, prepare the area for sharing.
  687.      * Then it will be safe to unlock it.
  688.      */
  689.     sh_info = src_area->sh_info;
  690.     if (!sh_info) {
  691.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  692.         mutex_initialize(&sh_info->lock);
  693.         sh_info->refcount = 2;
  694.         btree_create(&sh_info->pagemap);
  695.         src_area->sh_info = sh_info;
  696.     } else {
  697.         mutex_lock(&sh_info->lock);
  698.         sh_info->refcount++;
  699.         mutex_unlock(&sh_info->lock);
  700.     }
  701.  
  702.     src_area->backend->share(src_area);
  703.  
  704.     mutex_unlock(&src_area->lock);
  705.     mutex_unlock(&src_as->lock);
  706.  
  707.     /*
  708.      * Create copy of the source address space area.
  709.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  710.      * attribute set which prevents race condition with
  711.      * preliminary as_page_fault() calls.
  712.      * The flags of the source area are masked against dst_flags_mask
  713.      * to support sharing in less privileged mode.
  714.      */
  715.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  716.         AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  717.     if (!dst_area) {
  718.         /*
  719.          * Destination address space area could not be created.
  720.          */
  721.         sh_info_remove_reference(sh_info);
  722.        
  723.         interrupts_restore(ipl);
  724.         return ENOMEM;
  725.     }
  726.  
  727.     /*
  728.      * Now the destination address space area has been
  729.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  730.      * attribute and set the sh_info.
  731.      */
  732.     mutex_lock(&dst_as->lock); 
  733.     mutex_lock(&dst_area->lock);
  734.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  735.     dst_area->sh_info = sh_info;
  736.     mutex_unlock(&dst_area->lock);
  737.     mutex_unlock(&dst_as->lock);   
  738.  
  739.     interrupts_restore(ipl);
  740.    
  741.     return 0;
  742. }
  743.  
  744. /** Check access mode for address space area.
  745.  *
  746.  * The address space area must be locked prior to this call.
  747.  *
  748.  * @param area Address space area.
  749.  * @param access Access mode.
  750.  *
  751.  * @return False if access violates area's permissions, true otherwise.
  752.  */
  753. bool as_area_check_access(as_area_t *area, pf_access_t access)
  754. {
  755.     int flagmap[] = {
  756.         [PF_ACCESS_READ] = AS_AREA_READ,
  757.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  758.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  759.     };
  760.  
  761.     if (!(area->flags & flagmap[access]))
  762.         return false;
  763.    
  764.     return true;
  765. }
  766.  
  767. /** Handle page fault within the current address space.
  768.  *
  769.  * This is the high-level page fault handler. It decides
  770.  * whether the page fault can be resolved by any backend
  771.  * and if so, it invokes the backend to resolve the page
  772.  * fault.
  773.  *
  774.  * Interrupts are assumed disabled.
  775.  *
  776.  * @param page Faulting page.
  777.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  778.  * @param istate Pointer to interrupted state.
  779.  *
  780.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  781.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  782.  */
  783. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  784. {
  785.     pte_t *pte;
  786.     as_area_t *area;
  787.    
  788.     if (!THREAD)
  789.         return AS_PF_FAULT;
  790.        
  791.     ASSERT(AS);
  792.  
  793.     mutex_lock(&AS->lock);
  794.     area = find_area_and_lock(AS, page);   
  795.     if (!area) {
  796.         /*
  797.          * No area contained mapping for 'page'.
  798.          * Signal page fault to low-level handler.
  799.          */
  800.         mutex_unlock(&AS->lock);
  801.         goto page_fault;
  802.     }
  803.  
  804.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  805.         /*
  806.          * The address space area is not fully initialized.
  807.          * Avoid possible race by returning error.
  808.          */
  809.         mutex_unlock(&area->lock);
  810.         mutex_unlock(&AS->lock);
  811.         goto page_fault;       
  812.     }
  813.  
  814.     if (!area->backend || !area->backend->page_fault) {
  815.         /*
  816.          * The address space area is not backed by any backend
  817.          * or the backend cannot handle page faults.
  818.          */
  819.         mutex_unlock(&area->lock);
  820.         mutex_unlock(&AS->lock);
  821.         goto page_fault;       
  822.     }
  823.  
  824.     page_table_lock(AS, false);
  825.    
  826.     /*
  827.      * To avoid race condition between two page faults
  828.      * on the same address, we need to make sure
  829.      * the mapping has not been already inserted.
  830.      */
  831.     if ((pte = page_mapping_find(AS, page))) {
  832.         if (PTE_PRESENT(pte)) {
  833.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  834.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  835.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  836.                 page_table_unlock(AS, false);
  837.                 mutex_unlock(&area->lock);
  838.                 mutex_unlock(&AS->lock);
  839.                 return AS_PF_OK;
  840.             }
  841.         }
  842.     }
  843.    
  844.     /*
  845.      * Resort to the backend page fault handler.
  846.      */
  847.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  848.         page_table_unlock(AS, false);
  849.         mutex_unlock(&area->lock);
  850.         mutex_unlock(&AS->lock);
  851.         goto page_fault;
  852.     }
  853.    
  854.     page_table_unlock(AS, false);
  855.     mutex_unlock(&area->lock);
  856.     mutex_unlock(&AS->lock);
  857.     return AS_PF_OK;
  858.  
  859. page_fault:
  860.     if (THREAD->in_copy_from_uspace) {
  861.         THREAD->in_copy_from_uspace = false;
  862.         istate_set_retaddr(istate,
  863.             (uintptr_t) &memcpy_from_uspace_failover_address);
  864.     } else if (THREAD->in_copy_to_uspace) {
  865.         THREAD->in_copy_to_uspace = false;
  866.         istate_set_retaddr(istate,
  867.             (uintptr_t) &memcpy_to_uspace_failover_address);
  868.     } else {
  869.         return AS_PF_FAULT;
  870.     }
  871.  
  872.     return AS_PF_DEFER;
  873. }
  874.  
  875. /** Switch address spaces.
  876.  *
  877.  * Note that this function cannot sleep as it is essentially a part of
  878.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  879.  *
  880.  * @param old Old address space or NULL.
  881.  * @param new New address space.
  882.  */
  883. void as_switch(as_t *old_as, as_t *new_as)
  884. {
  885.     ipl_t ipl;
  886.     bool needs_asid = false;
  887.    
  888.     ipl = interrupts_disable();
  889.     spinlock_lock(&inactive_as_with_asid_lock);
  890.  
  891.     /*
  892.      * First, take care of the old address space.
  893.      */
  894.     if (old_as) {
  895.         mutex_lock_active(&old_as->lock);
  896.         ASSERT(old_as->cpu_refcount);
  897.         if((--old_as->cpu_refcount == 0) && (old_as != AS_KERNEL)) {
  898.             /*
  899.              * The old address space is no longer active on
  900.              * any processor. It can be appended to the
  901.              * list of inactive address spaces with assigned
  902.              * ASID.
  903.              */
  904.              ASSERT(old_as->asid != ASID_INVALID);
  905.              list_append(&old_as->inactive_as_with_asid_link,
  906.                  &inactive_as_with_asid_head);
  907.         }
  908.         mutex_unlock(&old_as->lock);
  909.  
  910.         /*
  911.          * Perform architecture-specific tasks when the address space
  912.          * is being removed from the CPU.
  913.          */
  914.         as_deinstall_arch(old_as);
  915.     }
  916.  
  917.     /*
  918.      * Second, prepare the new address space.
  919.      */
  920.     mutex_lock_active(&new_as->lock);
  921.     if ((new_as->cpu_refcount++ == 0) && (new_as != AS_KERNEL)) {
  922.         if (new_as->asid != ASID_INVALID) {
  923.             list_remove(&new_as->inactive_as_with_asid_link);
  924.         } else {
  925.             /*
  926.              * Defer call to asid_get() until new_as->lock is released.
  927.              */
  928.             needs_asid = true;
  929.         }
  930.     }
  931. #ifdef AS_PAGE_TABLE
  932.     SET_PTL0_ADDRESS(new_as->genarch.page_table);
  933. #endif
  934.     mutex_unlock(&new_as->lock);
  935.  
  936.     if (needs_asid) {
  937.         /*
  938.          * Allocation of new ASID was deferred
  939.          * until now in order to avoid deadlock.
  940.          */
  941.         asid_t asid;
  942.        
  943.         asid = asid_get();
  944.         mutex_lock_active(&new_as->lock);
  945.         new_as->asid = asid;
  946.         mutex_unlock(&new_as->lock);
  947.     }
  948.     spinlock_unlock(&inactive_as_with_asid_lock);
  949.     interrupts_restore(ipl);
  950.    
  951.     /*
  952.      * Perform architecture-specific steps.
  953.      * (e.g. write ASID to hardware register etc.)
  954.      */
  955.     as_install_arch(new_as);
  956.    
  957.     AS = new_as;
  958. }
  959.  
  960. /** Convert address space area flags to page flags.
  961.  *
  962.  * @param aflags Flags of some address space area.
  963.  *
  964.  * @return Flags to be passed to page_mapping_insert().
  965.  */
  966. int area_flags_to_page_flags(int aflags)
  967. {
  968.     int flags;
  969.  
  970.     flags = PAGE_USER | PAGE_PRESENT;
  971.    
  972.     if (aflags & AS_AREA_READ)
  973.         flags |= PAGE_READ;
  974.        
  975.     if (aflags & AS_AREA_WRITE)
  976.         flags |= PAGE_WRITE;
  977.    
  978.     if (aflags & AS_AREA_EXEC)
  979.         flags |= PAGE_EXEC;
  980.    
  981.     if (aflags & AS_AREA_CACHEABLE)
  982.         flags |= PAGE_CACHEABLE;
  983.        
  984.     return flags;
  985. }
  986.  
  987. /** Compute flags for virtual address translation subsytem.
  988.  *
  989.  * The address space area must be locked.
  990.  * Interrupts must be disabled.
  991.  *
  992.  * @param a Address space area.
  993.  *
  994.  * @return Flags to be used in page_mapping_insert().
  995.  */
  996. int as_area_get_flags(as_area_t *a)
  997. {
  998.     return area_flags_to_page_flags(a->flags);
  999. }
  1000.  
  1001. /** Create page table.
  1002.  *
  1003.  * Depending on architecture, create either address space
  1004.  * private or global page table.
  1005.  *
  1006.  * @param flags Flags saying whether the page table is for kernel address space.
  1007.  *
  1008.  * @return First entry of the page table.
  1009.  */
  1010. pte_t *page_table_create(int flags)
  1011. {
  1012. #ifdef __OBJC__
  1013.     return [as_t page_table_create: flags];
  1014. #else
  1015.     ASSERT(as_operations);
  1016.     ASSERT(as_operations->page_table_create);
  1017.    
  1018.     return as_operations->page_table_create(flags);
  1019. #endif
  1020. }
  1021.  
  1022. /** Destroy page table.
  1023.  *
  1024.  * Destroy page table in architecture specific way.
  1025.  *
  1026.  * @param page_table Physical address of PTL0.
  1027.  */
  1028. void page_table_destroy(pte_t *page_table)
  1029. {
  1030. #ifdef __OBJC__
  1031.     return [as_t page_table_destroy: page_table];
  1032. #else
  1033.     ASSERT(as_operations);
  1034.     ASSERT(as_operations->page_table_destroy);
  1035.    
  1036.     as_operations->page_table_destroy(page_table);
  1037. #endif
  1038. }
  1039.  
  1040. /** Lock page table.
  1041.  *
  1042.  * This function should be called before any page_mapping_insert(),
  1043.  * page_mapping_remove() and page_mapping_find().
  1044.  *
  1045.  * Locking order is such that address space areas must be locked
  1046.  * prior to this call. Address space can be locked prior to this
  1047.  * call in which case the lock argument is false.
  1048.  *
  1049.  * @param as Address space.
  1050.  * @param lock If false, do not attempt to lock as->lock.
  1051.  */
  1052. void page_table_lock(as_t *as, bool lock)
  1053. {
  1054. #ifdef __OBJC__
  1055.     [as page_table_lock: lock];
  1056. #else
  1057.     ASSERT(as_operations);
  1058.     ASSERT(as_operations->page_table_lock);
  1059.    
  1060.     as_operations->page_table_lock(as, lock);
  1061. #endif
  1062. }
  1063.  
  1064. /** Unlock page table.
  1065.  *
  1066.  * @param as Address space.
  1067.  * @param unlock If false, do not attempt to unlock as->lock.
  1068.  */
  1069. void page_table_unlock(as_t *as, bool unlock)
  1070. {
  1071. #ifdef __OBJC__
  1072.     [as page_table_unlock: unlock];
  1073. #else
  1074.     ASSERT(as_operations);
  1075.     ASSERT(as_operations->page_table_unlock);
  1076.    
  1077.     as_operations->page_table_unlock(as, unlock);
  1078. #endif
  1079. }
  1080.  
  1081.  
  1082. /** Find address space area and lock it.
  1083.  *
  1084.  * The address space must be locked and interrupts must be disabled.
  1085.  *
  1086.  * @param as Address space.
  1087.  * @param va Virtual address.
  1088.  *
  1089.  * @return Locked address space area containing va on success or NULL on
  1090.  *     failure.
  1091.  */
  1092. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  1093. {
  1094.     as_area_t *a;
  1095.     btree_node_t *leaf, *lnode;
  1096.     int i;
  1097.    
  1098.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  1099.     if (a) {
  1100.         /* va is the base address of an address space area */
  1101.         mutex_lock(&a->lock);
  1102.         return a;
  1103.     }
  1104.    
  1105.     /*
  1106.      * Search the leaf node and the righmost record of its left neighbour
  1107.      * to find out whether this is a miss or va belongs to an address
  1108.      * space area found there.
  1109.      */
  1110.    
  1111.     /* First, search the leaf node itself. */
  1112.     for (i = 0; i < leaf->keys; i++) {
  1113.         a = (as_area_t *) leaf->value[i];
  1114.         mutex_lock(&a->lock);
  1115.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  1116.             return a;
  1117.         }
  1118.         mutex_unlock(&a->lock);
  1119.     }
  1120.  
  1121.     /*
  1122.      * Second, locate the left neighbour and test its last record.
  1123.      * Because of its position in the B+tree, it must have base < va.
  1124.      */
  1125.     lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf);
  1126.     if (lnode) {
  1127.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  1128.         mutex_lock(&a->lock);
  1129.         if (va < a->base + a->pages * PAGE_SIZE) {
  1130.             return a;
  1131.         }
  1132.         mutex_unlock(&a->lock);
  1133.     }
  1134.  
  1135.     return NULL;
  1136. }
  1137.  
  1138. /** Check area conflicts with other areas.
  1139.  *
  1140.  * The address space must be locked and interrupts must be disabled.
  1141.  *
  1142.  * @param as Address space.
  1143.  * @param va Starting virtual address of the area being tested.
  1144.  * @param size Size of the area being tested.
  1145.  * @param avoid_area Do not touch this area.
  1146.  *
  1147.  * @return True if there is no conflict, false otherwise.
  1148.  */
  1149. bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  1150.               as_area_t *avoid_area)
  1151. {
  1152.     as_area_t *a;
  1153.     btree_node_t *leaf, *node;
  1154.     int i;
  1155.    
  1156.     /*
  1157.      * We don't want any area to have conflicts with NULL page.
  1158.      */
  1159.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1160.         return false;
  1161.    
  1162.     /*
  1163.      * The leaf node is found in O(log n), where n is proportional to
  1164.      * the number of address space areas belonging to as.
  1165.      * The check for conflicts is then attempted on the rightmost
  1166.      * record in the left neighbour, the leftmost record in the right
  1167.      * neighbour and all records in the leaf node itself.
  1168.      */
  1169.    
  1170.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1171.         if (a != avoid_area)
  1172.             return false;
  1173.     }
  1174.    
  1175.     /* First, check the two border cases. */
  1176.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1177.         a = (as_area_t *) node->value[node->keys - 1];
  1178.         mutex_lock(&a->lock);
  1179.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1180.             mutex_unlock(&a->lock);
  1181.             return false;
  1182.         }
  1183.         mutex_unlock(&a->lock);
  1184.     }
  1185.     node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf);
  1186.     if (node) {
  1187.         a = (as_area_t *) node->value[0];
  1188.         mutex_lock(&a->lock);
  1189.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1190.             mutex_unlock(&a->lock);
  1191.             return false;
  1192.         }
  1193.         mutex_unlock(&a->lock);
  1194.     }
  1195.    
  1196.     /* Second, check the leaf node. */
  1197.     for (i = 0; i < leaf->keys; i++) {
  1198.         a = (as_area_t *) leaf->value[i];
  1199.    
  1200.         if (a == avoid_area)
  1201.             continue;
  1202.    
  1203.         mutex_lock(&a->lock);
  1204.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1205.             mutex_unlock(&a->lock);
  1206.             return false;
  1207.         }
  1208.         mutex_unlock(&a->lock);
  1209.     }
  1210.  
  1211.     /*
  1212.      * So far, the area does not conflict with other areas.
  1213.      * Check if it doesn't conflict with kernel address space.
  1214.      */  
  1215.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1216.         return !overlaps(va, size,
  1217.             KERNEL_ADDRESS_SPACE_START,
  1218.             KERNEL_ADDRESS_SPACE_END - KERNEL_ADDRESS_SPACE_START);
  1219.     }
  1220.  
  1221.     return true;
  1222. }
  1223.  
  1224. /** Return size of the address space area with given base.  */
  1225. size_t as_get_size(uintptr_t base)
  1226. {
  1227.     ipl_t ipl;
  1228.     as_area_t *src_area;
  1229.     size_t size;
  1230.  
  1231.     ipl = interrupts_disable();
  1232.     src_area = find_area_and_lock(AS, base);
  1233.     if (src_area){
  1234.         size = src_area->pages * PAGE_SIZE;
  1235.         mutex_unlock(&src_area->lock);
  1236.     } else {
  1237.         size = 0;
  1238.     }
  1239.     interrupts_restore(ipl);
  1240.     return size;
  1241. }
  1242.  
  1243. /** Mark portion of address space area as used.
  1244.  *
  1245.  * The address space area must be already locked.
  1246.  *
  1247.  * @param a Address space area.
  1248.  * @param page First page to be marked.
  1249.  * @param count Number of page to be marked.
  1250.  *
  1251.  * @return 0 on failure and 1 on success.
  1252.  */
  1253. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1254. {
  1255.     btree_node_t *leaf, *node;
  1256.     count_t pages;
  1257.     int i;
  1258.  
  1259.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1260.     ASSERT(count);
  1261.  
  1262.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1263.     if (pages) {
  1264.         /*
  1265.          * We hit the beginning of some used space.
  1266.          */
  1267.         return 0;
  1268.     }
  1269.  
  1270.     if (!leaf->keys) {
  1271.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1272.         return 1;
  1273.     }
  1274.  
  1275.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1276.     if (node) {
  1277.         uintptr_t left_pg = node->key[node->keys - 1];
  1278.         uintptr_t right_pg = leaf->key[0];
  1279.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1280.         count_t right_cnt = (count_t) leaf->value[0];
  1281.        
  1282.         /*
  1283.          * Examine the possibility that the interval fits
  1284.          * somewhere between the rightmost interval of
  1285.          * the left neigbour and the first interval of the leaf.
  1286.          */
  1287.          
  1288.         if (page >= right_pg) {
  1289.             /* Do nothing. */
  1290.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1291.             left_cnt * PAGE_SIZE)) {
  1292.             /* The interval intersects with the left interval. */
  1293.             return 0;
  1294.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1295.             right_cnt * PAGE_SIZE)) {
  1296.             /* The interval intersects with the right interval. */
  1297.             return 0;          
  1298.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1299.             (page + count * PAGE_SIZE == right_pg)) {
  1300.             /*
  1301.              * The interval can be added by merging the two already
  1302.              * present intervals.
  1303.              */
  1304.             node->value[node->keys - 1] += count + right_cnt;
  1305.             btree_remove(&a->used_space, right_pg, leaf);
  1306.             return 1;
  1307.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1308.             /*
  1309.              * The interval can be added by simply growing the left
  1310.              * interval.
  1311.              */
  1312.             node->value[node->keys - 1] += count;
  1313.             return 1;
  1314.         } else if (page + count * PAGE_SIZE == right_pg) {
  1315.             /*
  1316.              * The interval can be addded by simply moving base of
  1317.              * the right interval down and increasing its size
  1318.              * accordingly.
  1319.              */
  1320.             leaf->value[0] += count;
  1321.             leaf->key[0] = page;
  1322.             return 1;
  1323.         } else {
  1324.             /*
  1325.              * The interval is between both neigbouring intervals,
  1326.              * but cannot be merged with any of them.
  1327.              */
  1328.             btree_insert(&a->used_space, page, (void *) count,
  1329.                 leaf);
  1330.             return 1;
  1331.         }
  1332.     } else if (page < leaf->key[0]) {
  1333.         uintptr_t right_pg = leaf->key[0];
  1334.         count_t right_cnt = (count_t) leaf->value[0];
  1335.    
  1336.         /*
  1337.          * Investigate the border case in which the left neighbour does
  1338.          * not exist but the interval fits from the left.
  1339.          */
  1340.          
  1341.         if (overlaps(page, count * PAGE_SIZE, right_pg,
  1342.             right_cnt * PAGE_SIZE)) {
  1343.             /* The interval intersects with the right interval. */
  1344.             return 0;
  1345.         } else if (page + count * PAGE_SIZE == right_pg) {
  1346.             /*
  1347.              * The interval can be added by moving the base of the
  1348.              * right interval down and increasing its size
  1349.              * accordingly.
  1350.              */
  1351.             leaf->key[0] = page;
  1352.             leaf->value[0] += count;
  1353.             return 1;
  1354.         } else {
  1355.             /*
  1356.              * The interval doesn't adjoin with the right interval.
  1357.              * It must be added individually.
  1358.              */
  1359.             btree_insert(&a->used_space, page, (void *) count,
  1360.                 leaf);
  1361.             return 1;
  1362.         }
  1363.     }
  1364.  
  1365.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1366.     if (node) {
  1367.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1368.         uintptr_t right_pg = node->key[0];
  1369.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1370.         count_t right_cnt = (count_t) node->value[0];
  1371.        
  1372.         /*
  1373.          * Examine the possibility that the interval fits
  1374.          * somewhere between the leftmost interval of
  1375.          * the right neigbour and the last interval of the leaf.
  1376.          */
  1377.  
  1378.         if (page < left_pg) {
  1379.             /* Do nothing. */
  1380.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1381.             left_cnt * PAGE_SIZE)) {
  1382.             /* The interval intersects with the left interval. */
  1383.             return 0;
  1384.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1385.             right_cnt * PAGE_SIZE)) {
  1386.             /* The interval intersects with the right interval. */
  1387.             return 0;          
  1388.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1389.             (page + count * PAGE_SIZE == right_pg)) {
  1390.             /*
  1391.              * The interval can be added by merging the two already
  1392.              * present intervals.
  1393.              * */
  1394.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1395.             btree_remove(&a->used_space, right_pg, node);
  1396.             return 1;
  1397.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1398.             /*
  1399.              * The interval can be added by simply growing the left
  1400.              * interval.
  1401.              * */
  1402.             leaf->value[leaf->keys - 1] +=  count;
  1403.             return 1;
  1404.         } else if (page + count * PAGE_SIZE == right_pg) {
  1405.             /*
  1406.              * The interval can be addded by simply moving base of
  1407.              * the right interval down and increasing its size
  1408.              * accordingly.
  1409.              */
  1410.             node->value[0] += count;
  1411.             node->key[0] = page;
  1412.             return 1;
  1413.         } else {
  1414.             /*
  1415.              * The interval is between both neigbouring intervals,
  1416.              * but cannot be merged with any of them.
  1417.              */
  1418.             btree_insert(&a->used_space, page, (void *) count,
  1419.                 leaf);
  1420.             return 1;
  1421.         }
  1422.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1423.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1424.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1425.    
  1426.         /*
  1427.          * Investigate the border case in which the right neighbour
  1428.          * does not exist but the interval fits from the right.
  1429.          */
  1430.          
  1431.         if (overlaps(page, count * PAGE_SIZE, left_pg,
  1432.             left_cnt * PAGE_SIZE)) {
  1433.             /* The interval intersects with the left interval. */
  1434.             return 0;
  1435.         } else if (left_pg + left_cnt * PAGE_SIZE == page) {
  1436.             /*
  1437.              * The interval can be added by growing the left
  1438.              * interval.
  1439.              */
  1440.             leaf->value[leaf->keys - 1] += count;
  1441.             return 1;
  1442.         } else {
  1443.             /*
  1444.              * The interval doesn't adjoin with the left interval.
  1445.              * It must be added individually.
  1446.              */
  1447.             btree_insert(&a->used_space, page, (void *) count,
  1448.                 leaf);
  1449.             return 1;
  1450.         }
  1451.     }
  1452.    
  1453.     /*
  1454.      * Note that if the algorithm made it thus far, the interval can fit
  1455.      * only between two other intervals of the leaf. The two border cases
  1456.      * were already resolved.
  1457.      */
  1458.     for (i = 1; i < leaf->keys; i++) {
  1459.         if (page < leaf->key[i]) {
  1460.             uintptr_t left_pg = leaf->key[i - 1];
  1461.             uintptr_t right_pg = leaf->key[i];
  1462.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1463.             count_t right_cnt = (count_t) leaf->value[i];
  1464.  
  1465.             /*
  1466.              * The interval fits between left_pg and right_pg.
  1467.              */
  1468.  
  1469.             if (overlaps(page, count * PAGE_SIZE, left_pg,
  1470.                 left_cnt * PAGE_SIZE)) {
  1471.                 /*
  1472.                  * The interval intersects with the left
  1473.                  * interval.
  1474.                  */
  1475.                 return 0;
  1476.             } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1477.                 right_cnt * PAGE_SIZE)) {
  1478.                 /*
  1479.                  * The interval intersects with the right
  1480.                  * interval.
  1481.                  */
  1482.                 return 0;          
  1483.             } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1484.                 (page + count * PAGE_SIZE == right_pg)) {
  1485.                 /*
  1486.                  * The interval can be added by merging the two
  1487.                  * already present intervals.
  1488.                  */
  1489.                 leaf->value[i - 1] += count + right_cnt;
  1490.                 btree_remove(&a->used_space, right_pg, leaf);
  1491.                 return 1;
  1492.             } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1493.                 /*
  1494.                  * The interval can be added by simply growing
  1495.                  * the left interval.
  1496.                  */
  1497.                 leaf->value[i - 1] += count;
  1498.                 return 1;
  1499.             } else if (page + count * PAGE_SIZE == right_pg) {
  1500.                 /*
  1501.                      * The interval can be addded by simply moving
  1502.                  * base of the right interval down and
  1503.                  * increasing its size accordingly.
  1504.                  */
  1505.                 leaf->value[i] += count;
  1506.                 leaf->key[i] = page;
  1507.                 return 1;
  1508.             } else {
  1509.                 /*
  1510.                  * The interval is between both neigbouring
  1511.                  * intervals, but cannot be merged with any of
  1512.                  * them.
  1513.                  */
  1514.                 btree_insert(&a->used_space, page,
  1515.                     (void *) count, leaf);
  1516.                 return 1;
  1517.             }
  1518.         }
  1519.     }
  1520.  
  1521.     panic("Inconsistency detected while adding %d pages of used space at "
  1522.         "%p.\n", count, page);
  1523. }
  1524.  
  1525. /** Mark portion of address space area as unused.
  1526.  *
  1527.  * The address space area must be already locked.
  1528.  *
  1529.  * @param a Address space area.
  1530.  * @param page First page to be marked.
  1531.  * @param count Number of page to be marked.
  1532.  *
  1533.  * @return 0 on failure and 1 on success.
  1534.  */
  1535. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1536. {
  1537.     btree_node_t *leaf, *node;
  1538.     count_t pages;
  1539.     int i;
  1540.  
  1541.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1542.     ASSERT(count);
  1543.  
  1544.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1545.     if (pages) {
  1546.         /*
  1547.          * We are lucky, page is the beginning of some interval.
  1548.          */
  1549.         if (count > pages) {
  1550.             return 0;
  1551.         } else if (count == pages) {
  1552.             btree_remove(&a->used_space, page, leaf);
  1553.             return 1;
  1554.         } else {
  1555.             /*
  1556.              * Find the respective interval.
  1557.              * Decrease its size and relocate its start address.
  1558.              */
  1559.             for (i = 0; i < leaf->keys; i++) {
  1560.                 if (leaf->key[i] == page) {
  1561.                     leaf->key[i] += count * PAGE_SIZE;
  1562.                     leaf->value[i] -= count;
  1563.                     return 1;
  1564.                 }
  1565.             }
  1566.             goto error;
  1567.         }
  1568.     }
  1569.  
  1570.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1571.     if (node && page < leaf->key[0]) {
  1572.         uintptr_t left_pg = node->key[node->keys - 1];
  1573.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1574.  
  1575.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1576.             count * PAGE_SIZE)) {
  1577.             if (page + count * PAGE_SIZE ==
  1578.                 left_pg + left_cnt * PAGE_SIZE) {
  1579.                 /*
  1580.                  * The interval is contained in the rightmost
  1581.                  * interval of the left neighbour and can be
  1582.                  * removed by updating the size of the bigger
  1583.                  * interval.
  1584.                  */
  1585.                 node->value[node->keys - 1] -= count;
  1586.                 return 1;
  1587.             } else if (page + count * PAGE_SIZE <
  1588.                 left_pg + left_cnt*PAGE_SIZE) {
  1589.                 count_t new_cnt;
  1590.                
  1591.                 /*
  1592.                  * The interval is contained in the rightmost
  1593.                  * interval of the left neighbour but its
  1594.                  * removal requires both updating the size of
  1595.                  * the original interval and also inserting a
  1596.                  * new interval.
  1597.                  */
  1598.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1599.                     (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1600.                 node->value[node->keys - 1] -= count + new_cnt;
  1601.                 btree_insert(&a->used_space, page +
  1602.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1603.                 return 1;
  1604.             }
  1605.         }
  1606.         return 0;
  1607.     } else if (page < leaf->key[0]) {
  1608.         return 0;
  1609.     }
  1610.    
  1611.     if (page > leaf->key[leaf->keys - 1]) {
  1612.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1613.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1614.  
  1615.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1616.             count * PAGE_SIZE)) {
  1617.             if (page + count * PAGE_SIZE ==
  1618.                 left_pg + left_cnt * PAGE_SIZE) {
  1619.                 /*
  1620.                  * The interval is contained in the rightmost
  1621.                  * interval of the leaf and can be removed by
  1622.                  * updating the size of the bigger interval.
  1623.                  */
  1624.                 leaf->value[leaf->keys - 1] -= count;
  1625.                 return 1;
  1626.             } else if (page + count * PAGE_SIZE < left_pg +
  1627.                 left_cnt * PAGE_SIZE) {
  1628.                 count_t new_cnt;
  1629.                
  1630.                 /*
  1631.                  * The interval is contained in the rightmost
  1632.                  * interval of the leaf but its removal
  1633.                  * requires both updating the size of the
  1634.                  * original interval and also inserting a new
  1635.                  * interval.
  1636.                  */
  1637.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1638.                     (page + count * PAGE_SIZE)) >> PAGE_WIDTH;
  1639.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1640.                 btree_insert(&a->used_space, page +
  1641.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1642.                 return 1;
  1643.             }
  1644.         }
  1645.         return 0;
  1646.     }  
  1647.    
  1648.     /*
  1649.      * The border cases have been already resolved.
  1650.      * Now the interval can be only between intervals of the leaf.
  1651.      */
  1652.     for (i = 1; i < leaf->keys - 1; i++) {
  1653.         if (page < leaf->key[i]) {
  1654.             uintptr_t left_pg = leaf->key[i - 1];
  1655.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1656.  
  1657.             /*
  1658.              * Now the interval is between intervals corresponding
  1659.              * to (i - 1) and i.
  1660.              */
  1661.             if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1662.                 count * PAGE_SIZE)) {
  1663.                 if (page + count * PAGE_SIZE ==
  1664.                     left_pg + left_cnt*PAGE_SIZE) {
  1665.                     /*
  1666.                      * The interval is contained in the
  1667.                      * interval (i - 1) of the leaf and can
  1668.                      * be removed by updating the size of
  1669.                      * the bigger interval.
  1670.                      */
  1671.                     leaf->value[i - 1] -= count;
  1672.                     return 1;
  1673.                 } else if (page + count * PAGE_SIZE <
  1674.                     left_pg + left_cnt * PAGE_SIZE) {
  1675.                     count_t new_cnt;
  1676.                
  1677.                     /*
  1678.                      * The interval is contained in the
  1679.                      * interval (i - 1) of the leaf but its
  1680.                      * removal requires both updating the
  1681.                      * size of the original interval and
  1682.                      * also inserting a new interval.
  1683.                      */
  1684.                     new_cnt = ((left_pg +
  1685.                         left_cnt * PAGE_SIZE) -
  1686.                         (page + count * PAGE_SIZE)) >>
  1687.                         PAGE_WIDTH;
  1688.                     leaf->value[i - 1] -= count + new_cnt;
  1689.                     btree_insert(&a->used_space, page +
  1690.                         count * PAGE_SIZE, (void *) new_cnt,
  1691.                         leaf);
  1692.                     return 1;
  1693.                 }
  1694.             }
  1695.             return 0;
  1696.         }
  1697.     }
  1698.  
  1699. error:
  1700.     panic("Inconsistency detected while removing %d pages of used space "
  1701.         "from %p.\n", count, page);
  1702. }
  1703.  
  1704. /** Remove reference to address space area share info.
  1705.  *
  1706.  * If the reference count drops to 0, the sh_info is deallocated.
  1707.  *
  1708.  * @param sh_info Pointer to address space area share info.
  1709.  */
  1710. void sh_info_remove_reference(share_info_t *sh_info)
  1711. {
  1712.     bool dealloc = false;
  1713.  
  1714.     mutex_lock(&sh_info->lock);
  1715.     ASSERT(sh_info->refcount);
  1716.     if (--sh_info->refcount == 0) {
  1717.         dealloc = true;
  1718.         link_t *cur;
  1719.        
  1720.         /*
  1721.          * Now walk carefully the pagemap B+tree and free/remove
  1722.          * reference from all frames found there.
  1723.          */
  1724.         for (cur = sh_info->pagemap.leaf_head.next;
  1725.             cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1726.             btree_node_t *node;
  1727.             int i;
  1728.            
  1729.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1730.             for (i = 0; i < node->keys; i++)
  1731.                 frame_free((uintptr_t) node->value[i]);
  1732.         }
  1733.        
  1734.     }
  1735.     mutex_unlock(&sh_info->lock);
  1736.    
  1737.     if (dealloc) {
  1738.         btree_destroy(&sh_info->pagemap);
  1739.         free(sh_info);
  1740.     }
  1741. }
  1742.  
  1743. /*
  1744.  * Address space related syscalls.
  1745.  */
  1746.  
  1747. /** Wrapper for as_area_create(). */
  1748. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1749. {
  1750.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address,
  1751.         AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1752.         return (unative_t) address;
  1753.     else
  1754.         return (unative_t) -1;
  1755. }
  1756.  
  1757. /** Wrapper for as_area_resize(). */
  1758. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1759. {
  1760.     return (unative_t) as_area_resize(AS, address, size, 0);
  1761. }
  1762.  
  1763. /** Wrapper for as_area_destroy(). */
  1764. unative_t sys_as_area_destroy(uintptr_t address)
  1765. {
  1766.     return (unative_t) as_area_destroy(AS, address);
  1767. }
  1768.  
  1769. /** Print out information about address space.
  1770.  *
  1771.  * @param as Address space.
  1772.  */
  1773. void as_print(as_t *as)
  1774. {
  1775.     ipl_t ipl;
  1776.    
  1777.     ipl = interrupts_disable();
  1778.     mutex_lock(&as->lock);
  1779.    
  1780.     /* print out info about address space areas */
  1781.     link_t *cur;
  1782.     for (cur = as->as_area_btree.leaf_head.next;
  1783.         cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  1784.         btree_node_t *node;
  1785.        
  1786.         node = list_get_instance(cur, btree_node_t, leaf_link);
  1787.        
  1788.         int i;
  1789.         for (i = 0; i < node->keys; i++) {
  1790.             as_area_t *area = node->value[i];
  1791.        
  1792.             mutex_lock(&area->lock);
  1793.             printf("as_area: %p, base=%p, pages=%d (%p - %p)\n",
  1794.                 area, area->base, area->pages, area->base,
  1795.                 area->base + area->pages*PAGE_SIZE);
  1796.             mutex_unlock(&area->lock);
  1797.         }
  1798.     }
  1799.    
  1800.     mutex_unlock(&as->lock);
  1801.     interrupts_restore(ipl);
  1802. }
  1803.  
  1804. /** @}
  1805.  */
  1806.