Subversion Repositories HelenOS

Rev

Rev 1889 | Rev 1891 | 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 <typedefs.h>
  78. #include <syscall/copy.h>
  79. #include <arch/interrupt.h>
  80.  
  81. /**
  82.  * Each architecture decides what functions will be used to carry out
  83.  * address space operations such as creating or locking page tables.
  84.  */
  85. as_operations_t *as_operations = NULL;
  86.  
  87. /**
  88.  * Slab for as_t objects.
  89.  */
  90. static slab_cache_t *as_slab;
  91.  
  92. /** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
  93. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  94.  
  95. /**
  96.  * This list contains address spaces that are not active on any
  97.  * processor and that have valid ASID.
  98.  */
  99. LIST_INITIALIZE(inactive_as_with_asid_head);
  100.  
  101. /** Kernel address space. */
  102. as_t *AS_KERNEL = NULL;
  103.  
  104. static int area_flags_to_page_flags(int aflags);
  105. static as_area_t *find_area_and_lock(as_t *as, uintptr_t va);
  106. static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size, as_area_t *avoid_area);
  107. static void sh_info_remove_reference(share_info_t *sh_info);
  108.  
  109. /** Initialize address space subsystem. */
  110. void as_init(void)
  111. {
  112.     as_arch_init();
  113.    
  114.     as_slab = slab_cache_create("as_slab", sizeof(as_t), 0, NULL, NULL, SLAB_CACHE_MAGDEFERRED);
  115.    
  116.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  117.     if (!AS_KERNEL)
  118.         panic("can't create kernel address space\n");
  119.    
  120. }
  121.  
  122. /** Create address space.
  123.  *
  124.  * @param flags Flags that influence way in wich the address space is created.
  125.  */
  126. as_t *as_create(int flags)
  127. {
  128.     as_t *as;
  129.  
  130.     as = (as_t *) slab_alloc(as_slab, 0);
  131.     link_initialize(&as->inactive_as_with_asid_link);
  132.     mutex_initialize(&as->lock);
  133.     btree_create(&as->as_area_btree);
  134.    
  135.     if (flags & FLAG_AS_KERNEL)
  136.         as->asid = ASID_KERNEL;
  137.     else
  138.         as->asid = ASID_INVALID;
  139.    
  140.     as->refcount = 0;
  141.     as->cpu_refcount = 0;
  142.     as->page_table = page_table_create(flags);
  143.  
  144.     return as;
  145. }
  146.  
  147. /** Destroy adress space.
  148.  *
  149.  * When there are no tasks referencing this address space (i.e. its refcount is zero),
  150.  * the address space can be destroyed.
  151.  */
  152. void as_destroy(as_t *as)
  153. {
  154.     ipl_t ipl;
  155.     bool cond;
  156.  
  157.     ASSERT(as->refcount == 0);
  158.    
  159.     /*
  160.      * Since there is no reference to this area,
  161.      * it is safe not to lock its mutex.
  162.      */
  163.     ipl = interrupts_disable();
  164.     spinlock_lock(&inactive_as_with_asid_lock);
  165.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  166.         if (as != AS && as->cpu_refcount == 0)
  167.             list_remove(&as->inactive_as_with_asid_link);
  168.         asid_put(as->asid);
  169.     }
  170.     spinlock_unlock(&inactive_as_with_asid_lock);
  171.  
  172.     /*
  173.      * Destroy address space areas of the address space.
  174.      * The B+tee must be walked carefully because it is
  175.      * also being destroyed.
  176.      */
  177.     for (cond = true; cond; ) {
  178.         btree_node_t *node;
  179.  
  180.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  181.         node = list_get_instance(as->as_area_btree.leaf_head.next, btree_node_t, leaf_link);
  182.  
  183.         if ((cond = node->keys)) {
  184.             as_area_destroy(as, node->key[0]);
  185.         }
  186.     }
  187.  
  188.     btree_destroy(&as->as_area_btree);
  189.     page_table_destroy(as->page_table);
  190.  
  191.     interrupts_restore(ipl);
  192.    
  193.     slab_free(as_slab, as);
  194. }
  195.  
  196. /** Create address space area of common attributes.
  197.  *
  198.  * The created address space area is added to the target address space.
  199.  *
  200.  * @param as Target address space.
  201.  * @param flags Flags of the area memory.
  202.  * @param size Size of area.
  203.  * @param base Base address of area.
  204.  * @param attrs Attributes of the area.
  205.  * @param backend Address space area backend. NULL if no backend is used.
  206.  * @param backend_data NULL or a pointer to an array holding two void *.
  207.  *
  208.  * @return Address space area on success or NULL on failure.
  209.  */
  210. as_area_t *as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
  211.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  212. {
  213.     ipl_t ipl;
  214.     as_area_t *a;
  215.    
  216.     if (base % PAGE_SIZE)
  217.         return NULL;
  218.  
  219.     if (!size)
  220.         return NULL;
  221.  
  222.     /* Writeable executable areas are not supported. */
  223.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  224.         return NULL;
  225.    
  226.     ipl = interrupts_disable();
  227.     mutex_lock(&as->lock);
  228.    
  229.     if (!check_area_conflicts(as, base, size, NULL)) {
  230.         mutex_unlock(&as->lock);
  231.         interrupts_restore(ipl);
  232.         return NULL;
  233.     }
  234.    
  235.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  236.  
  237.     mutex_initialize(&a->lock);
  238.    
  239.     a->as = as;
  240.     a->flags = flags;
  241.     a->attributes = attrs;
  242.     a->pages = SIZE2FRAMES(size);
  243.     a->base = base;
  244.     a->sh_info = NULL;
  245.     a->backend = backend;
  246.     if (backend_data)
  247.         a->backend_data = *backend_data;
  248.     else
  249.         memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data), 0);
  250.  
  251.     btree_create(&a->used_space);
  252.    
  253.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  254.  
  255.     mutex_unlock(&as->lock);
  256.     interrupts_restore(ipl);
  257.  
  258.     return a;
  259. }
  260.  
  261. /** Find address space area and change it.
  262.  *
  263.  * @param as Address space.
  264.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  265.  * @param size New size of the virtual memory block starting at address.
  266.  * @param flags Flags influencing the remap operation. Currently unused.
  267.  *
  268.  * @return Zero on success or a value from @ref errno.h otherwise.
  269.  */
  270. int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
  271. {
  272.     as_area_t *area;
  273.     ipl_t ipl;
  274.     size_t pages;
  275.    
  276.     ipl = interrupts_disable();
  277.     mutex_lock(&as->lock);
  278.    
  279.     /*
  280.      * Locate the area.
  281.      */
  282.     area = find_area_and_lock(as, address);
  283.     if (!area) {
  284.         mutex_unlock(&as->lock);
  285.         interrupts_restore(ipl);
  286.         return ENOENT;
  287.     }
  288.  
  289.     if (area->backend == &phys_backend) {
  290.         /*
  291.          * Remapping of address space areas associated
  292.          * with memory mapped devices is not supported.
  293.          */
  294.         mutex_unlock(&area->lock);
  295.         mutex_unlock(&as->lock);
  296.         interrupts_restore(ipl);
  297.         return ENOTSUP;
  298.     }
  299.     if (area->sh_info) {
  300.         /*
  301.          * Remapping of shared address space areas
  302.          * is not supported.
  303.          */
  304.         mutex_unlock(&area->lock);
  305.         mutex_unlock(&as->lock);
  306.         interrupts_restore(ipl);
  307.         return ENOTSUP;
  308.     }
  309.  
  310.     pages = SIZE2FRAMES((address - area->base) + size);
  311.     if (!pages) {
  312.         /*
  313.          * Zero size address space areas are not allowed.
  314.          */
  315.         mutex_unlock(&area->lock);
  316.         mutex_unlock(&as->lock);
  317.         interrupts_restore(ipl);
  318.         return EPERM;
  319.     }
  320.    
  321.     if (pages < area->pages) {
  322.         bool cond;
  323.         uintptr_t start_free = area->base + pages*PAGE_SIZE;
  324.  
  325.         /*
  326.          * Shrinking the area.
  327.          * No need to check for overlaps.
  328.          */
  329.  
  330.         /*
  331.          * Start TLB shootdown sequence.
  332.          */
  333.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  334.  
  335.         /*
  336.          * Remove frames belonging to used space starting from
  337.          * the highest addresses downwards until an overlap with
  338.          * the resized address space area is found. Note that this
  339.          * is also the right way to remove part of the used_space
  340.          * B+tree leaf list.
  341.          */    
  342.         for (cond = true; cond;) {
  343.             btree_node_t *node;
  344.        
  345.             ASSERT(!list_empty(&area->used_space.leaf_head));
  346.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  347.             if ((cond = (bool) node->keys)) {
  348.                 uintptr_t b = node->key[node->keys - 1];
  349.                 count_t c = (count_t) node->value[node->keys - 1];
  350.                 int i = 0;
  351.            
  352.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  353.                    
  354.                     if (b + c*PAGE_SIZE <= start_free) {
  355.                         /*
  356.                          * The whole interval fits completely
  357.                          * in the resized address space area.
  358.                          */
  359.                         break;
  360.                     }
  361.        
  362.                     /*
  363.                      * Part of the interval corresponding to b and c
  364.                      * overlaps with the resized address space area.
  365.                      */
  366.        
  367.                     cond = false;   /* we are almost done */
  368.                     i = (start_free - b) >> PAGE_WIDTH;
  369.                     if (!used_space_remove(area, start_free, c - i))
  370.                         panic("Could not remove used space.\n");
  371.                 } else {
  372.                     /*
  373.                      * The interval of used space can be completely removed.
  374.                      */
  375.                     if (!used_space_remove(area, b, c))
  376.                         panic("Could not remove used space.\n");
  377.                 }
  378.            
  379.                 for (; i < c; i++) {
  380.                     pte_t *pte;
  381.            
  382.                     page_table_lock(as, false);
  383.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  384.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  385.                     if (area->backend && area->backend->frame_free) {
  386.                         area->backend->frame_free(area,
  387.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  388.                     }
  389.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  390.                     page_table_unlock(as, false);
  391.                 }
  392.             }
  393.         }
  394.  
  395.         /*
  396.          * Finish TLB shootdown sequence.
  397.          */
  398.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  399.         tlb_shootdown_finalize();
  400.        
  401.         /*
  402.          * Invalidate software translation caches (e.g. TSB on sparc64).
  403.          */
  404.         as_invalidate_translation_cache(as, area->base + pages*PAGE_SIZE, area->pages - pages);
  405.     } else {
  406.         /*
  407.          * Growing the area.
  408.          * Check for overlaps with other address space areas.
  409.          */
  410.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  411.             mutex_unlock(&area->lock);
  412.             mutex_unlock(&as->lock);       
  413.             interrupts_restore(ipl);
  414.             return EADDRNOTAVAIL;
  415.         }
  416.     }
  417.  
  418.     area->pages = pages;
  419.    
  420.     mutex_unlock(&area->lock);
  421.     mutex_unlock(&as->lock);
  422.     interrupts_restore(ipl);
  423.  
  424.     return 0;
  425. }
  426.  
  427. /** Destroy address space area.
  428.  *
  429.  * @param as Address space.
  430.  * @param address Address withing the area to be deleted.
  431.  *
  432.  * @return Zero on success or a value from @ref errno.h on failure.
  433.  */
  434. int as_area_destroy(as_t *as, uintptr_t address)
  435. {
  436.     as_area_t *area;
  437.     uintptr_t base;
  438.     link_t *cur;
  439.     ipl_t ipl;
  440.  
  441.     ipl = interrupts_disable();
  442.     mutex_lock(&as->lock);
  443.  
  444.     area = find_area_and_lock(as, address);
  445.     if (!area) {
  446.         mutex_unlock(&as->lock);
  447.         interrupts_restore(ipl);
  448.         return ENOENT;
  449.     }
  450.  
  451.     base = area->base;
  452.  
  453.     /*
  454.      * Start TLB shootdown sequence.
  455.      */
  456.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  457.  
  458.     /*
  459.      * Visit only the pages mapped by used_space B+tree.
  460.      */
  461.     for (cur = area->used_space.leaf_head.next; cur != &area->used_space.leaf_head; cur = cur->next) {
  462.         btree_node_t *node;
  463.         int i;
  464.        
  465.         node = list_get_instance(cur, btree_node_t, leaf_link);
  466.         for (i = 0; i < node->keys; i++) {
  467.             uintptr_t b = node->key[i];
  468.             count_t j;
  469.             pte_t *pte;
  470.            
  471.             for (j = 0; j < (count_t) node->value[i]; j++) {
  472.                 page_table_lock(as, false);
  473.                 pte = page_mapping_find(as, b + j*PAGE_SIZE);
  474.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  475.                 if (area->backend && area->backend->frame_free) {
  476.                     area->backend->frame_free(area,
  477.                         b + j*PAGE_SIZE, PTE_GET_FRAME(pte));
  478.                 }
  479.                 page_mapping_remove(as, b + j*PAGE_SIZE);              
  480.                 page_table_unlock(as, false);
  481.             }
  482.         }
  483.     }
  484.  
  485.     /*
  486.      * Finish TLB shootdown sequence.
  487.      */
  488.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  489.     tlb_shootdown_finalize();
  490.    
  491.     /*
  492.      * Invalidate potential software translation caches (e.g. TSB on sparc64).
  493.      */
  494.     as_invalidate_translation_cache(as, area->base, area->pages);
  495.    
  496.     btree_destroy(&area->used_space);
  497.  
  498.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  499.    
  500.     if (area->sh_info)
  501.         sh_info_remove_reference(area->sh_info);
  502.        
  503.     mutex_unlock(&area->lock);
  504.  
  505.     /*
  506.      * Remove the empty area from address space.
  507.      */
  508.     btree_remove(&as->as_area_btree, base, NULL);
  509.    
  510.     free(area);
  511.    
  512.     mutex_unlock(&as->lock);
  513.     interrupts_restore(ipl);
  514.     return 0;
  515. }
  516.  
  517. /** Share address space area with another or the same address space.
  518.  *
  519.  * Address space area mapping is shared with a new address space area.
  520.  * If the source address space area has not been shared so far,
  521.  * a new sh_info is created. The new address space area simply gets the
  522.  * sh_info of the source area. The process of duplicating the
  523.  * mapping is done through the backend share function.
  524.  *
  525.  * @param src_as Pointer to source address space.
  526.  * @param src_base Base address of the source address space area.
  527.  * @param acc_size Expected size of the source area.
  528.  * @param dst_as Pointer to destination address space.
  529.  * @param dst_base Target base address.
  530.  * @param dst_flags_mask Destination address space area flags mask.
  531.  *
  532.  * @return Zero on success or ENOENT if there is no such task or
  533.  *     if there is no such address space area,
  534.  *     EPERM if there was a problem in accepting the area or
  535.  *     ENOMEM if there was a problem in allocating destination
  536.  *     address space area. ENOTSUP is returned if an attempt
  537.  *     to share non-anonymous address space area is detected.
  538.  */
  539. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  540.           as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  541. {
  542.     ipl_t ipl;
  543.     int src_flags;
  544.     size_t src_size;
  545.     as_area_t *src_area, *dst_area;
  546.     share_info_t *sh_info;
  547.     mem_backend_t *src_backend;
  548.     mem_backend_data_t src_backend_data;
  549.    
  550.     ipl = interrupts_disable();
  551.     mutex_lock(&src_as->lock);
  552.     src_area = find_area_and_lock(src_as, src_base);
  553.     if (!src_area) {
  554.         /*
  555.          * Could not find the source address space area.
  556.          */
  557.         mutex_unlock(&src_as->lock);
  558.         interrupts_restore(ipl);
  559.         return ENOENT;
  560.     }
  561.    
  562.     if (!src_area->backend || !src_area->backend->share) {
  563.         /*
  564.          * There is no backend or the backend does not
  565.          * know how to share the area.
  566.          */
  567.         mutex_unlock(&src_area->lock);
  568.         mutex_unlock(&src_as->lock);
  569.         interrupts_restore(ipl);
  570.         return ENOTSUP;
  571.     }
  572.    
  573.     src_size = src_area->pages * PAGE_SIZE;
  574.     src_flags = src_area->flags;
  575.     src_backend = src_area->backend;
  576.     src_backend_data = src_area->backend_data;
  577.  
  578.     /* Share the cacheable flag from the original mapping */
  579.     if (src_flags & AS_AREA_CACHEABLE)
  580.         dst_flags_mask |= AS_AREA_CACHEABLE;
  581.  
  582.     if (src_size != acc_size || (src_flags & dst_flags_mask) != dst_flags_mask) {
  583.         mutex_unlock(&src_area->lock);
  584.         mutex_unlock(&src_as->lock);
  585.         interrupts_restore(ipl);
  586.         return EPERM;
  587.     }
  588.  
  589.     /*
  590.      * Now we are committed to sharing the area.
  591.      * First prepare the area for sharing.
  592.      * Then it will be safe to unlock it.
  593.      */
  594.     sh_info = src_area->sh_info;
  595.     if (!sh_info) {
  596.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  597.         mutex_initialize(&sh_info->lock);
  598.         sh_info->refcount = 2;
  599.         btree_create(&sh_info->pagemap);
  600.         src_area->sh_info = sh_info;
  601.     } else {
  602.         mutex_lock(&sh_info->lock);
  603.         sh_info->refcount++;
  604.         mutex_unlock(&sh_info->lock);
  605.     }
  606.  
  607.     src_area->backend->share(src_area);
  608.  
  609.     mutex_unlock(&src_area->lock);
  610.     mutex_unlock(&src_as->lock);
  611.  
  612.     /*
  613.      * Create copy of the source address space area.
  614.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  615.      * attribute set which prevents race condition with
  616.      * preliminary as_page_fault() calls.
  617.      * The flags of the source area are masked against dst_flags_mask
  618.      * to support sharing in less privileged mode.
  619.      */
  620.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  621.                   AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  622.     if (!dst_area) {
  623.         /*
  624.          * Destination address space area could not be created.
  625.          */
  626.         sh_info_remove_reference(sh_info);
  627.        
  628.         interrupts_restore(ipl);
  629.         return ENOMEM;
  630.     }
  631.    
  632.     /*
  633.      * Now the destination address space area has been
  634.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  635.      * attribute and set the sh_info.
  636.      */
  637.     mutex_lock(&dst_area->lock);
  638.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  639.     dst_area->sh_info = sh_info;
  640.     mutex_unlock(&dst_area->lock);
  641.    
  642.     interrupts_restore(ipl);
  643.    
  644.     return 0;
  645. }
  646.  
  647. /** Check access mode for address space area.
  648.  *
  649.  * The address space area must be locked prior to this call.
  650.  *
  651.  * @param area Address space area.
  652.  * @param access Access mode.
  653.  *
  654.  * @return False if access violates area's permissions, true otherwise.
  655.  */
  656. bool as_area_check_access(as_area_t *area, pf_access_t access)
  657. {
  658.     int flagmap[] = {
  659.         [PF_ACCESS_READ] = AS_AREA_READ,
  660.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  661.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  662.     };
  663.  
  664.     if (!(area->flags & flagmap[access]))
  665.         return false;
  666.    
  667.     return true;
  668. }
  669.  
  670. /** Handle page fault within the current address space.
  671.  *
  672.  * This is the high-level page fault handler. It decides
  673.  * whether the page fault can be resolved by any backend
  674.  * and if so, it invokes the backend to resolve the page
  675.  * fault.
  676.  *
  677.  * Interrupts are assumed disabled.
  678.  *
  679.  * @param page Faulting page.
  680.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  681.  * @param istate Pointer to interrupted state.
  682.  *
  683.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  684.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  685.  */
  686. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  687. {
  688.     pte_t *pte;
  689.     as_area_t *area;
  690.    
  691.     if (!THREAD)
  692.         return AS_PF_FAULT;
  693.        
  694.     ASSERT(AS);
  695.  
  696.     mutex_lock(&AS->lock);
  697.     area = find_area_and_lock(AS, page);   
  698.     if (!area) {
  699.         /*
  700.          * No area contained mapping for 'page'.
  701.          * Signal page fault to low-level handler.
  702.          */
  703.         mutex_unlock(&AS->lock);
  704.         goto page_fault;
  705.     }
  706.  
  707.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  708.         /*
  709.          * The address space area is not fully initialized.
  710.          * Avoid possible race by returning error.
  711.          */
  712.         mutex_unlock(&area->lock);
  713.         mutex_unlock(&AS->lock);
  714.         goto page_fault;       
  715.     }
  716.  
  717.     if (!area->backend || !area->backend->page_fault) {
  718.         /*
  719.          * The address space area is not backed by any backend
  720.          * or the backend cannot handle page faults.
  721.          */
  722.         mutex_unlock(&area->lock);
  723.         mutex_unlock(&AS->lock);
  724.         goto page_fault;       
  725.     }
  726.  
  727.     page_table_lock(AS, false);
  728.    
  729.     /*
  730.      * To avoid race condition between two page faults
  731.      * on the same address, we need to make sure
  732.      * the mapping has not been already inserted.
  733.      */
  734.     if ((pte = page_mapping_find(AS, page))) {
  735.         if (PTE_PRESENT(pte)) {
  736.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  737.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  738.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  739.                 page_table_unlock(AS, false);
  740.                 mutex_unlock(&area->lock);
  741.                 mutex_unlock(&AS->lock);
  742.                 return AS_PF_OK;
  743.             }
  744.         }
  745.     }
  746.    
  747.     /*
  748.      * Resort to the backend page fault handler.
  749.      */
  750.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  751.         page_table_unlock(AS, false);
  752.         mutex_unlock(&area->lock);
  753.         mutex_unlock(&AS->lock);
  754.         goto page_fault;
  755.     }
  756.    
  757.     page_table_unlock(AS, false);
  758.     mutex_unlock(&area->lock);
  759.     mutex_unlock(&AS->lock);
  760.     return AS_PF_OK;
  761.  
  762. page_fault:
  763.     if (THREAD->in_copy_from_uspace) {
  764.         THREAD->in_copy_from_uspace = false;
  765.         istate_set_retaddr(istate, (uintptr_t) &memcpy_from_uspace_failover_address);
  766.     } else if (THREAD->in_copy_to_uspace) {
  767.         THREAD->in_copy_to_uspace = false;
  768.         istate_set_retaddr(istate, (uintptr_t) &memcpy_to_uspace_failover_address);
  769.     } else {
  770.         return AS_PF_FAULT;
  771.     }
  772.  
  773.     return AS_PF_DEFER;
  774. }
  775.  
  776. /** Switch address spaces.
  777.  *
  778.  * Note that this function cannot sleep as it is essentially a part of
  779.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  780.  *
  781.  * @param old Old address space or NULL.
  782.  * @param new New address space.
  783.  */
  784. void as_switch(as_t *old, as_t *new)
  785. {
  786.     ipl_t ipl;
  787.     bool needs_asid = false;
  788.    
  789.     ipl = interrupts_disable();
  790.     spinlock_lock(&inactive_as_with_asid_lock);
  791.  
  792.     /*
  793.      * First, take care of the old address space.
  794.      */
  795.     if (old) {
  796.         mutex_lock_active(&old->lock);
  797.         ASSERT(old->cpu_refcount);
  798.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  799.             /*
  800.              * The old address space is no longer active on
  801.              * any processor. It can be appended to the
  802.              * list of inactive address spaces with assigned
  803.              * ASID.
  804.              */
  805.              ASSERT(old->asid != ASID_INVALID);
  806.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  807.         }
  808.         mutex_unlock(&old->lock);
  809.  
  810.         /*
  811.          * Perform architecture-specific tasks when the address space
  812.          * is being removed from the CPU.
  813.          */
  814.         as_deinstall_arch(old);
  815.     }
  816.  
  817.     /*
  818.      * Second, prepare the new address space.
  819.      */
  820.     mutex_lock_active(&new->lock);
  821.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  822.         if (new->asid != ASID_INVALID)
  823.             list_remove(&new->inactive_as_with_asid_link);
  824.         else
  825.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  826.     }
  827.     SET_PTL0_ADDRESS(new->page_table);
  828.     mutex_unlock(&new->lock);
  829.  
  830.     if (needs_asid) {
  831.         /*
  832.          * Allocation of new ASID was deferred
  833.          * until now in order to avoid deadlock.
  834.          */
  835.         asid_t asid;
  836.        
  837.         asid = asid_get();
  838.         mutex_lock_active(&new->lock);
  839.         new->asid = asid;
  840.         mutex_unlock(&new->lock);
  841.     }
  842.     spinlock_unlock(&inactive_as_with_asid_lock);
  843.     interrupts_restore(ipl);
  844.    
  845.     /*
  846.      * Perform architecture-specific steps.
  847.      * (e.g. write ASID to hardware register etc.)
  848.      */
  849.     as_install_arch(new);
  850.    
  851.     AS = new;
  852. }
  853.  
  854. /** Convert address space area flags to page flags.
  855.  *
  856.  * @param aflags Flags of some address space area.
  857.  *
  858.  * @return Flags to be passed to page_mapping_insert().
  859.  */
  860. int area_flags_to_page_flags(int aflags)
  861. {
  862.     int flags;
  863.  
  864.     flags = PAGE_USER | PAGE_PRESENT;
  865.    
  866.     if (aflags & AS_AREA_READ)
  867.         flags |= PAGE_READ;
  868.        
  869.     if (aflags & AS_AREA_WRITE)
  870.         flags |= PAGE_WRITE;
  871.    
  872.     if (aflags & AS_AREA_EXEC)
  873.         flags |= PAGE_EXEC;
  874.    
  875.     if (aflags & AS_AREA_CACHEABLE)
  876.         flags |= PAGE_CACHEABLE;
  877.        
  878.     return flags;
  879. }
  880.  
  881. /** Compute flags for virtual address translation subsytem.
  882.  *
  883.  * The address space area must be locked.
  884.  * Interrupts must be disabled.
  885.  *
  886.  * @param a Address space area.
  887.  *
  888.  * @return Flags to be used in page_mapping_insert().
  889.  */
  890. int as_area_get_flags(as_area_t *a)
  891. {
  892.     return area_flags_to_page_flags(a->flags);
  893. }
  894.  
  895. /** Create page table.
  896.  *
  897.  * Depending on architecture, create either address space
  898.  * private or global page table.
  899.  *
  900.  * @param flags Flags saying whether the page table is for kernel address space.
  901.  *
  902.  * @return First entry of the page table.
  903.  */
  904. pte_t *page_table_create(int flags)
  905. {
  906.         ASSERT(as_operations);
  907.         ASSERT(as_operations->page_table_create);
  908.  
  909.         return as_operations->page_table_create(flags);
  910. }
  911.  
  912. /** Destroy page table.
  913.  *
  914.  * Destroy page table in architecture specific way.
  915.  *
  916.  * @param page_table Physical address of PTL0.
  917.  */
  918. void page_table_destroy(pte_t *page_table)
  919. {
  920.         ASSERT(as_operations);
  921.         ASSERT(as_operations->page_table_destroy);
  922.  
  923.         as_operations->page_table_destroy(page_table);
  924. }
  925.  
  926. /** Lock page table.
  927.  *
  928.  * This function should be called before any page_mapping_insert(),
  929.  * page_mapping_remove() and page_mapping_find().
  930.  *
  931.  * Locking order is such that address space areas must be locked
  932.  * prior to this call. Address space can be locked prior to this
  933.  * call in which case the lock argument is false.
  934.  *
  935.  * @param as Address space.
  936.  * @param lock If false, do not attempt to lock as->lock.
  937.  */
  938. void page_table_lock(as_t *as, bool lock)
  939. {
  940.     ASSERT(as_operations);
  941.     ASSERT(as_operations->page_table_lock);
  942.  
  943.     as_operations->page_table_lock(as, lock);
  944. }
  945.  
  946. /** Unlock page table.
  947.  *
  948.  * @param as Address space.
  949.  * @param unlock If false, do not attempt to unlock as->lock.
  950.  */
  951. void page_table_unlock(as_t *as, bool unlock)
  952. {
  953.     ASSERT(as_operations);
  954.     ASSERT(as_operations->page_table_unlock);
  955.  
  956.     as_operations->page_table_unlock(as, unlock);
  957. }
  958.  
  959.  
  960. /** Find address space area and lock it.
  961.  *
  962.  * The address space must be locked and interrupts must be disabled.
  963.  *
  964.  * @param as Address space.
  965.  * @param va Virtual address.
  966.  *
  967.  * @return Locked address space area containing va on success or NULL on failure.
  968.  */
  969. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  970. {
  971.     as_area_t *a;
  972.     btree_node_t *leaf, *lnode;
  973.     int i;
  974.    
  975.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  976.     if (a) {
  977.         /* va is the base address of an address space area */
  978.         mutex_lock(&a->lock);
  979.         return a;
  980.     }
  981.    
  982.     /*
  983.      * Search the leaf node and the righmost record of its left neighbour
  984.      * to find out whether this is a miss or va belongs to an address
  985.      * space area found there.
  986.      */
  987.    
  988.     /* First, search the leaf node itself. */
  989.     for (i = 0; i < leaf->keys; i++) {
  990.         a = (as_area_t *) leaf->value[i];
  991.         mutex_lock(&a->lock);
  992.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  993.             return a;
  994.         }
  995.         mutex_unlock(&a->lock);
  996.     }
  997.  
  998.     /*
  999.      * Second, locate the left neighbour and test its last record.
  1000.      * Because of its position in the B+tree, it must have base < va.
  1001.      */
  1002.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1003.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  1004.         mutex_lock(&a->lock);
  1005.         if (va < a->base + a->pages * PAGE_SIZE) {
  1006.             return a;
  1007.         }
  1008.         mutex_unlock(&a->lock);
  1009.     }
  1010.  
  1011.     return NULL;
  1012. }
  1013.  
  1014. /** Check area conflicts with other areas.
  1015.  *
  1016.  * The address space must be locked and interrupts must be disabled.
  1017.  *
  1018.  * @param as Address space.
  1019.  * @param va Starting virtual address of the area being tested.
  1020.  * @param size Size of the area being tested.
  1021.  * @param avoid_area Do not touch this area.
  1022.  *
  1023.  * @return True if there is no conflict, false otherwise.
  1024.  */
  1025. bool check_area_conflicts(as_t *as, uintptr_t va, size_t size, as_area_t *avoid_area)
  1026. {
  1027.     as_area_t *a;
  1028.     btree_node_t *leaf, *node;
  1029.     int i;
  1030.    
  1031.     /*
  1032.      * We don't want any area to have conflicts with NULL page.
  1033.      */
  1034.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1035.         return false;
  1036.    
  1037.     /*
  1038.      * The leaf node is found in O(log n), where n is proportional to
  1039.      * the number of address space areas belonging to as.
  1040.      * The check for conflicts is then attempted on the rightmost
  1041.      * record in the left neighbour, the leftmost record in the right
  1042.      * neighbour and all records in the leaf node itself.
  1043.      */
  1044.    
  1045.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1046.         if (a != avoid_area)
  1047.             return false;
  1048.     }
  1049.    
  1050.     /* First, check the two border cases. */
  1051.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1052.         a = (as_area_t *) node->value[node->keys - 1];
  1053.         mutex_lock(&a->lock);
  1054.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1055.             mutex_unlock(&a->lock);
  1056.             return false;
  1057.         }
  1058.         mutex_unlock(&a->lock);
  1059.     }
  1060.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1061.         a = (as_area_t *) node->value[0];
  1062.         mutex_lock(&a->lock);
  1063.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1064.             mutex_unlock(&a->lock);
  1065.             return false;
  1066.         }
  1067.         mutex_unlock(&a->lock);
  1068.     }
  1069.    
  1070.     /* Second, check the leaf node. */
  1071.     for (i = 0; i < leaf->keys; i++) {
  1072.         a = (as_area_t *) leaf->value[i];
  1073.    
  1074.         if (a == avoid_area)
  1075.             continue;
  1076.    
  1077.         mutex_lock(&a->lock);
  1078.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1079.             mutex_unlock(&a->lock);
  1080.             return false;
  1081.         }
  1082.         mutex_unlock(&a->lock);
  1083.     }
  1084.  
  1085.     /*
  1086.      * So far, the area does not conflict with other areas.
  1087.      * Check if it doesn't conflict with kernel address space.
  1088.      */  
  1089.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1090.         return !overlaps(va, size,
  1091.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1092.     }
  1093.  
  1094.     return true;
  1095. }
  1096.  
  1097. /** Return size of the address space area with given base.  */
  1098. size_t as_get_size(uintptr_t base)
  1099. {
  1100.     ipl_t ipl;
  1101.     as_area_t *src_area;
  1102.     size_t size;
  1103.  
  1104.     ipl = interrupts_disable();
  1105.     src_area = find_area_and_lock(AS, base);
  1106.     if (src_area){
  1107.         size = src_area->pages * PAGE_SIZE;
  1108.         mutex_unlock(&src_area->lock);
  1109.     } else {
  1110.         size = 0;
  1111.     }
  1112.     interrupts_restore(ipl);
  1113.     return size;
  1114. }
  1115.  
  1116. /** Mark portion of address space area as used.
  1117.  *
  1118.  * The address space area must be already locked.
  1119.  *
  1120.  * @param a Address space area.
  1121.  * @param page First page to be marked.
  1122.  * @param count Number of page to be marked.
  1123.  *
  1124.  * @return 0 on failure and 1 on success.
  1125.  */
  1126. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1127. {
  1128.     btree_node_t *leaf, *node;
  1129.     count_t pages;
  1130.     int i;
  1131.  
  1132.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1133.     ASSERT(count);
  1134.  
  1135.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1136.     if (pages) {
  1137.         /*
  1138.          * We hit the beginning of some used space.
  1139.          */
  1140.         return 0;
  1141.     }
  1142.  
  1143.     if (!leaf->keys) {
  1144.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1145.         return 1;
  1146.     }
  1147.  
  1148.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1149.     if (node) {
  1150.         uintptr_t left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1151.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1152.        
  1153.         /*
  1154.          * Examine the possibility that the interval fits
  1155.          * somewhere between the rightmost interval of
  1156.          * the left neigbour and the first interval of the leaf.
  1157.          */
  1158.          
  1159.         if (page >= right_pg) {
  1160.             /* Do nothing. */
  1161.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1162.             /* The interval intersects with the left interval. */
  1163.             return 0;
  1164.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1165.             /* The interval intersects with the right interval. */
  1166.             return 0;          
  1167.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1168.             /* The interval can be added by merging the two already present intervals. */
  1169.             node->value[node->keys - 1] += count + right_cnt;
  1170.             btree_remove(&a->used_space, right_pg, leaf);
  1171.             return 1;
  1172.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1173.             /* The interval can be added by simply growing the left interval. */
  1174.             node->value[node->keys - 1] += count;
  1175.             return 1;
  1176.         } else if (page + count*PAGE_SIZE == right_pg) {
  1177.             /*
  1178.              * The interval can be addded by simply moving base of the right
  1179.              * interval down and increasing its size accordingly.
  1180.              */
  1181.             leaf->value[0] += count;
  1182.             leaf->key[0] = page;
  1183.             return 1;
  1184.         } else {
  1185.             /*
  1186.              * The interval is between both neigbouring intervals,
  1187.              * but cannot be merged with any of them.
  1188.              */
  1189.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1190.             return 1;
  1191.         }
  1192.     } else if (page < leaf->key[0]) {
  1193.         uintptr_t right_pg = leaf->key[0];
  1194.         count_t right_cnt = (count_t) leaf->value[0];
  1195.    
  1196.         /*
  1197.          * Investigate the border case in which the left neighbour does not
  1198.          * exist but the interval fits from the left.
  1199.          */
  1200.          
  1201.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1202.             /* The interval intersects with the right interval. */
  1203.             return 0;
  1204.         } else if (page + count*PAGE_SIZE == right_pg) {
  1205.             /*
  1206.              * The interval can be added by moving the base of the right interval down
  1207.              * and increasing its size accordingly.
  1208.              */
  1209.             leaf->key[0] = page;
  1210.             leaf->value[0] += count;
  1211.             return 1;
  1212.         } else {
  1213.             /*
  1214.              * The interval doesn't adjoin with the right interval.
  1215.              * It must be added individually.
  1216.              */
  1217.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1218.             return 1;
  1219.         }
  1220.     }
  1221.  
  1222.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1223.     if (node) {
  1224.         uintptr_t left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1225.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1226.        
  1227.         /*
  1228.          * Examine the possibility that the interval fits
  1229.          * somewhere between the leftmost interval of
  1230.          * the right neigbour and the last interval of the leaf.
  1231.          */
  1232.  
  1233.         if (page < left_pg) {
  1234.             /* Do nothing. */
  1235.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1236.             /* The interval intersects with the left interval. */
  1237.             return 0;
  1238.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1239.             /* The interval intersects with the right interval. */
  1240.             return 0;          
  1241.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1242.             /* The interval can be added by merging the two already present intervals. */
  1243.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1244.             btree_remove(&a->used_space, right_pg, node);
  1245.             return 1;
  1246.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1247.             /* The interval can be added by simply growing the left interval. */
  1248.             leaf->value[leaf->keys - 1] +=  count;
  1249.             return 1;
  1250.         } else if (page + count*PAGE_SIZE == right_pg) {
  1251.             /*
  1252.              * The interval can be addded by simply moving base of the right
  1253.              * interval down and increasing its size accordingly.
  1254.              */
  1255.             node->value[0] += count;
  1256.             node->key[0] = page;
  1257.             return 1;
  1258.         } else {
  1259.             /*
  1260.              * The interval is between both neigbouring intervals,
  1261.              * but cannot be merged with any of them.
  1262.              */
  1263.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1264.             return 1;
  1265.         }
  1266.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1267.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1268.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1269.    
  1270.         /*
  1271.          * Investigate the border case in which the right neighbour does not
  1272.          * exist but the interval fits from the right.
  1273.          */
  1274.          
  1275.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1276.             /* The interval intersects with the left interval. */
  1277.             return 0;
  1278.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1279.             /* The interval can be added by growing the left interval. */
  1280.             leaf->value[leaf->keys - 1] += count;
  1281.             return 1;
  1282.         } else {
  1283.             /*
  1284.              * The interval doesn't adjoin with the left interval.
  1285.              * It must be added individually.
  1286.              */
  1287.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1288.             return 1;
  1289.         }
  1290.     }
  1291.    
  1292.     /*
  1293.      * Note that if the algorithm made it thus far, the interval can fit only
  1294.      * between two other intervals of the leaf. The two border cases were already
  1295.      * resolved.
  1296.      */
  1297.     for (i = 1; i < leaf->keys; i++) {
  1298.         if (page < leaf->key[i]) {
  1299.             uintptr_t left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1300.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1301.  
  1302.             /*
  1303.              * The interval fits between left_pg and right_pg.
  1304.              */
  1305.  
  1306.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1307.                 /* The interval intersects with the left interval. */
  1308.                 return 0;
  1309.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1310.                 /* The interval intersects with the right interval. */
  1311.                 return 0;          
  1312.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1313.                 /* The interval can be added by merging the two already present intervals. */
  1314.                 leaf->value[i - 1] += count + right_cnt;
  1315.                 btree_remove(&a->used_space, right_pg, leaf);
  1316.                 return 1;
  1317.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1318.                 /* The interval can be added by simply growing the left interval. */
  1319.                 leaf->value[i - 1] += count;
  1320.                 return 1;
  1321.             } else if (page + count*PAGE_SIZE == right_pg) {
  1322.                 /*
  1323.                      * The interval can be addded by simply moving base of the right
  1324.                  * interval down and increasing its size accordingly.
  1325.                  */
  1326.                 leaf->value[i] += count;
  1327.                 leaf->key[i] = page;
  1328.                 return 1;
  1329.             } else {
  1330.                 /*
  1331.                  * The interval is between both neigbouring intervals,
  1332.                  * but cannot be merged with any of them.
  1333.                  */
  1334.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1335.                 return 1;
  1336.             }
  1337.         }
  1338.     }
  1339.  
  1340.     panic("Inconsistency detected while adding %d pages of used space at %p.\n", count, page);
  1341. }
  1342.  
  1343. /** Mark portion of address space area as unused.
  1344.  *
  1345.  * The address space area must be already locked.
  1346.  *
  1347.  * @param a Address space area.
  1348.  * @param page First page to be marked.
  1349.  * @param count Number of page to be marked.
  1350.  *
  1351.  * @return 0 on failure and 1 on success.
  1352.  */
  1353. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1354. {
  1355.     btree_node_t *leaf, *node;
  1356.     count_t pages;
  1357.     int i;
  1358.  
  1359.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1360.     ASSERT(count);
  1361.  
  1362.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1363.     if (pages) {
  1364.         /*
  1365.          * We are lucky, page is the beginning of some interval.
  1366.          */
  1367.         if (count > pages) {
  1368.             return 0;
  1369.         } else if (count == pages) {
  1370.             btree_remove(&a->used_space, page, leaf);
  1371.             return 1;
  1372.         } else {
  1373.             /*
  1374.              * Find the respective interval.
  1375.              * Decrease its size and relocate its start address.
  1376.              */
  1377.             for (i = 0; i < leaf->keys; i++) {
  1378.                 if (leaf->key[i] == page) {
  1379.                     leaf->key[i] += count*PAGE_SIZE;
  1380.                     leaf->value[i] -= count;
  1381.                     return 1;
  1382.                 }
  1383.             }
  1384.             goto error;
  1385.         }
  1386.     }
  1387.  
  1388.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1389.     if (node && page < leaf->key[0]) {
  1390.         uintptr_t left_pg = node->key[node->keys - 1];
  1391.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1392.  
  1393.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1394.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1395.                 /*
  1396.                  * The interval is contained in the rightmost interval
  1397.                  * of the left neighbour and can be removed by
  1398.                  * updating the size of the bigger interval.
  1399.                  */
  1400.                 node->value[node->keys - 1] -= count;
  1401.                 return 1;
  1402.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1403.                 count_t new_cnt;
  1404.                
  1405.                 /*
  1406.                  * The interval is contained in the rightmost interval
  1407.                  * of the left neighbour but its removal requires
  1408.                  * both updating the size of the original interval and
  1409.                  * also inserting a new interval.
  1410.                  */
  1411.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1412.                 node->value[node->keys - 1] -= count + new_cnt;
  1413.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1414.                 return 1;
  1415.             }
  1416.         }
  1417.         return 0;
  1418.     } else if (page < leaf->key[0]) {
  1419.         return 0;
  1420.     }
  1421.    
  1422.     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.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1427.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1428.                 /*
  1429.                  * The interval is contained in the rightmost interval
  1430.                  * of the leaf and can be removed by updating the size
  1431.                  * of the bigger interval.
  1432.                  */
  1433.                 leaf->value[leaf->keys - 1] -= count;
  1434.                 return 1;
  1435.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1436.                 count_t new_cnt;
  1437.                
  1438.                 /*
  1439.                  * The interval is contained in the rightmost interval
  1440.                  * of the leaf but its removal requires both updating
  1441.                  * the size of the original interval and
  1442.                  * also inserting a new interval.
  1443.                  */
  1444.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1445.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1446.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1447.                 return 1;
  1448.             }
  1449.         }
  1450.         return 0;
  1451.     }  
  1452.    
  1453.     /*
  1454.      * The border cases have been already resolved.
  1455.      * Now the interval can be only between intervals of the leaf.
  1456.      */
  1457.     for (i = 1; i < leaf->keys - 1; i++) {
  1458.         if (page < leaf->key[i]) {
  1459.             uintptr_t left_pg = leaf->key[i - 1];
  1460.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1461.  
  1462.             /*
  1463.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1464.              */
  1465.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1466.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1467.                     /*
  1468.                     * The interval is contained in the interval (i - 1)
  1469.                      * of the leaf and can be removed by updating the size
  1470.                      * of the bigger interval.
  1471.                      */
  1472.                     leaf->value[i - 1] -= count;
  1473.                     return 1;
  1474.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1475.                     count_t new_cnt;
  1476.                
  1477.                     /*
  1478.                      * The interval is contained in the interval (i - 1)
  1479.                      * of the leaf but its removal requires both updating
  1480.                      * the size of the original interval and
  1481.                      * also inserting a new interval.
  1482.                      */
  1483.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1484.                     leaf->value[i - 1] -= count + new_cnt;
  1485.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1486.                     return 1;
  1487.                 }
  1488.             }
  1489.             return 0;
  1490.         }
  1491.     }
  1492.  
  1493. error:
  1494.     panic("Inconsistency detected while removing %d pages of used space from %p.\n", count, page);
  1495. }
  1496.  
  1497. /** Remove reference to address space area share info.
  1498.  *
  1499.  * If the reference count drops to 0, the sh_info is deallocated.
  1500.  *
  1501.  * @param sh_info Pointer to address space area share info.
  1502.  */
  1503. void sh_info_remove_reference(share_info_t *sh_info)
  1504. {
  1505.     bool dealloc = false;
  1506.  
  1507.     mutex_lock(&sh_info->lock);
  1508.     ASSERT(sh_info->refcount);
  1509.     if (--sh_info->refcount == 0) {
  1510.         dealloc = true;
  1511.         link_t *cur;
  1512.        
  1513.         /*
  1514.          * Now walk carefully the pagemap B+tree and free/remove
  1515.          * reference from all frames found there.
  1516.          */
  1517.         for (cur = sh_info->pagemap.leaf_head.next; cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1518.             btree_node_t *node;
  1519.             int i;
  1520.            
  1521.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1522.             for (i = 0; i < node->keys; i++)
  1523.                 frame_free((uintptr_t) node->value[i]);
  1524.         }
  1525.        
  1526.     }
  1527.     mutex_unlock(&sh_info->lock);
  1528.    
  1529.     if (dealloc) {
  1530.         btree_destroy(&sh_info->pagemap);
  1531.         free(sh_info);
  1532.     }
  1533. }
  1534.  
  1535. /*
  1536.  * Address space related syscalls.
  1537.  */
  1538.  
  1539. /** Wrapper for as_area_create(). */
  1540. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1541. {
  1542.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1543.         return (unative_t) address;
  1544.     else
  1545.         return (unative_t) -1;
  1546. }
  1547.  
  1548. /** Wrapper for as_area_resize(). */
  1549. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1550. {
  1551.     return (unative_t) as_area_resize(AS, address, size, 0);
  1552. }
  1553.  
  1554. /** Wrapper for as_area_destroy(). */
  1555. unative_t sys_as_area_destroy(uintptr_t address)
  1556. {
  1557.     return (unative_t) as_area_destroy(AS, address);
  1558. }
  1559.  
  1560. /** @}
  1561.  */
  1562.