Subversion Repositories HelenOS

Rev

Rev 2069 | Rev 2087 | 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. #ifdef CONFIG_VIRT_IDX_DCACHE
  82. #include <arch/mm/cache.h>
  83. #endif /* CONFIG_VIRT_IDX_DCACHE */
  84.  
  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.  
  96. /** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
  97. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  98.  
  99. /**
  100.  * This list contains address spaces that are not active on any
  101.  * processor and that have valid ASID.
  102.  */
  103. LIST_INITIALIZE(inactive_as_with_asid_head);
  104.  
  105. /** Kernel address space. */
  106. as_t *AS_KERNEL = NULL;
  107.  
  108. static int area_flags_to_page_flags(int aflags);
  109. static as_area_t *find_area_and_lock(as_t *as, uintptr_t va);
  110. static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size, as_area_t *avoid_area);
  111. static void sh_info_remove_reference(share_info_t *sh_info);
  112.  
  113. static int as_constructor(void *obj, int flags)
  114. {
  115.     as_t *as = (as_t *) obj;
  116.     int rc;
  117.  
  118.     link_initialize(&as->inactive_as_with_asid_link);
  119.     mutex_initialize(&as->lock);   
  120.    
  121.     rc = as_constructor_arch(as, flags);
  122.    
  123.     return rc;
  124. }
  125.  
  126. static int as_destructor(void *obj)
  127. {
  128.     as_t *as = (as_t *) obj;
  129.  
  130.     return as_destructor_arch(as);
  131. }
  132.  
  133. /** Initialize address space subsystem. */
  134. void as_init(void)
  135. {
  136.     as_arch_init();
  137.    
  138.     as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,
  139.         as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);
  140.    
  141.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  142.     if (!AS_KERNEL)
  143.         panic("can't create kernel address space\n");
  144.    
  145. }
  146.  
  147. /** Create address space.
  148.  *
  149.  * @param flags Flags that influence way in wich the address space is created.
  150.  */
  151. as_t *as_create(int flags)
  152. {
  153.     as_t *as;
  154.  
  155.     as = (as_t *) slab_alloc(as_slab, 0);
  156.     (void) as_create_arch(as, 0);
  157.    
  158.     btree_create(&as->as_area_btree);
  159.    
  160.     if (flags & FLAG_AS_KERNEL)
  161.         as->asid = ASID_KERNEL;
  162.     else
  163.         as->asid = ASID_INVALID;
  164.    
  165.     as->refcount = 0;
  166.     as->cpu_refcount = 0;
  167.     as->page_table = page_table_create(flags);
  168.  
  169.     return as;
  170. }
  171.  
  172. /** Destroy adress space.
  173.  *
  174.  * When there are no tasks referencing this address space (i.e. its refcount is zero),
  175.  * the address space can be destroyed.
  176.  */
  177. void as_destroy(as_t *as)
  178. {
  179.     ipl_t ipl;
  180.     bool cond;
  181.  
  182.     ASSERT(as->refcount == 0);
  183.    
  184.     /*
  185.      * Since there is no reference to this area,
  186.      * it is safe not to lock its mutex.
  187.      */
  188.     ipl = interrupts_disable();
  189.     spinlock_lock(&inactive_as_with_asid_lock);
  190.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  191.         if (as != AS && as->cpu_refcount == 0)
  192.             list_remove(&as->inactive_as_with_asid_link);
  193.         asid_put(as->asid);
  194.     }
  195.     spinlock_unlock(&inactive_as_with_asid_lock);
  196.  
  197.     /*
  198.      * Destroy address space areas of the address space.
  199.      * The B+tree must be walked carefully because it is
  200.      * also being destroyed.
  201.      */
  202.     for (cond = true; cond; ) {
  203.         btree_node_t *node;
  204.  
  205.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  206.         node = list_get_instance(as->as_area_btree.leaf_head.next, btree_node_t, leaf_link);
  207.  
  208.         if ((cond = node->keys)) {
  209.             as_area_destroy(as, node->key[0]);
  210.         }
  211.     }
  212.  
  213.     btree_destroy(&as->as_area_btree);
  214.     page_table_destroy(as->page_table);
  215.  
  216.     interrupts_restore(ipl);
  217.    
  218.     slab_free(as_slab, as);
  219. }
  220.  
  221. /** Create address space area of common attributes.
  222.  *
  223.  * The created address space area is added to the target address space.
  224.  *
  225.  * @param as Target address space.
  226.  * @param flags Flags of the area memory.
  227.  * @param size Size of area.
  228.  * @param base Base address of area.
  229.  * @param attrs Attributes of the area.
  230.  * @param backend Address space area backend. NULL if no backend is used.
  231.  * @param backend_data NULL or a pointer to an array holding two void *.
  232.  *
  233.  * @return Address space area on success or NULL on failure.
  234.  */
  235. as_area_t *
  236. as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
  237.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  238. {
  239.     ipl_t ipl;
  240.     as_area_t *a;
  241.    
  242.     if (base % PAGE_SIZE)
  243.         return NULL;
  244.  
  245.     if (!size)
  246.         return NULL;
  247.  
  248.     /* Writeable executable areas are not supported. */
  249.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  250.         return NULL;
  251.    
  252.     ipl = interrupts_disable();
  253.     mutex_lock(&as->lock);
  254.    
  255.     if (!check_area_conflicts(as, base, size, NULL)) {
  256.         mutex_unlock(&as->lock);
  257.         interrupts_restore(ipl);
  258.         return NULL;
  259.     }
  260.    
  261.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  262.  
  263.     mutex_initialize(&a->lock);
  264.    
  265.     a->as = as;
  266.     a->flags = flags;
  267.     a->attributes = attrs;
  268.     a->pages = SIZE2FRAMES(size);
  269.     a->base = base;
  270.     a->sh_info = NULL;
  271.     a->backend = backend;
  272.     if (backend_data)
  273.         a->backend_data = *backend_data;
  274.     else
  275.         memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data), 0);
  276.  
  277.     btree_create(&a->used_space);
  278.    
  279.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  280.  
  281.     mutex_unlock(&as->lock);
  282.     interrupts_restore(ipl);
  283.  
  284.     return a;
  285. }
  286.  
  287. /** Find address space area and change it.
  288.  *
  289.  * @param as Address space.
  290.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  291.  * @param size New size of the virtual memory block starting at address.
  292.  * @param flags Flags influencing the remap operation. Currently unused.
  293.  *
  294.  * @return Zero on success or a value from @ref errno.h otherwise.
  295.  */
  296. int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
  297. {
  298.     as_area_t *area;
  299.     ipl_t ipl;
  300.     size_t pages;
  301.    
  302.     ipl = interrupts_disable();
  303.     mutex_lock(&as->lock);
  304.    
  305.     /*
  306.      * Locate the area.
  307.      */
  308.     area = find_area_and_lock(as, address);
  309.     if (!area) {
  310.         mutex_unlock(&as->lock);
  311.         interrupts_restore(ipl);
  312.         return ENOENT;
  313.     }
  314.  
  315.     if (area->backend == &phys_backend) {
  316.         /*
  317.          * Remapping of address space areas associated
  318.          * with memory mapped devices is not supported.
  319.          */
  320.         mutex_unlock(&area->lock);
  321.         mutex_unlock(&as->lock);
  322.         interrupts_restore(ipl);
  323.         return ENOTSUP;
  324.     }
  325.     if (area->sh_info) {
  326.         /*
  327.          * Remapping of shared address space areas
  328.          * is not supported.
  329.          */
  330.         mutex_unlock(&area->lock);
  331.         mutex_unlock(&as->lock);
  332.         interrupts_restore(ipl);
  333.         return ENOTSUP;
  334.     }
  335.  
  336.     pages = SIZE2FRAMES((address - area->base) + size);
  337.     if (!pages) {
  338.         /*
  339.          * Zero size address space areas are not allowed.
  340.          */
  341.         mutex_unlock(&area->lock);
  342.         mutex_unlock(&as->lock);
  343.         interrupts_restore(ipl);
  344.         return EPERM;
  345.     }
  346.    
  347.     if (pages < area->pages) {
  348.         bool cond;
  349.         uintptr_t start_free = area->base + pages*PAGE_SIZE;
  350.  
  351.         /*
  352.          * Shrinking the area.
  353.          * No need to check for overlaps.
  354.          */
  355.  
  356.         /*
  357.          * Start TLB shootdown sequence.
  358.          */
  359.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  360.  
  361.         /*
  362.          * Remove frames belonging to used space starting from
  363.          * the highest addresses downwards until an overlap with
  364.          * the resized address space area is found. Note that this
  365.          * is also the right way to remove part of the used_space
  366.          * B+tree leaf list.
  367.          */    
  368.         for (cond = true; cond;) {
  369.             btree_node_t *node;
  370.        
  371.             ASSERT(!list_empty(&area->used_space.leaf_head));
  372.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  373.             if ((cond = (bool) node->keys)) {
  374.                 uintptr_t b = node->key[node->keys - 1];
  375.                 count_t c = (count_t) node->value[node->keys - 1];
  376.                 int i = 0;
  377.            
  378.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  379.                    
  380.                     if (b + c*PAGE_SIZE <= start_free) {
  381.                         /*
  382.                          * The whole interval fits completely
  383.                          * in the resized address space area.
  384.                          */
  385.                         break;
  386.                     }
  387.        
  388.                     /*
  389.                      * Part of the interval corresponding to b and c
  390.                      * overlaps with the resized address space area.
  391.                      */
  392.        
  393.                     cond = false;   /* we are almost done */
  394.                     i = (start_free - b) >> PAGE_WIDTH;
  395.                     if (!used_space_remove(area, start_free, c - i))
  396.                         panic("Could not remove used space.\n");
  397.                 } else {
  398.                     /*
  399.                      * The interval of used space can be completely removed.
  400.                      */
  401.                     if (!used_space_remove(area, b, c))
  402.                         panic("Could not remove used space.\n");
  403.                 }
  404.            
  405.                 for (; i < c; i++) {
  406.                     pte_t *pte;
  407.            
  408.                     page_table_lock(as, false);
  409.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  410.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  411.                     if (area->backend && area->backend->frame_free) {
  412.                         area->backend->frame_free(area,
  413.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  414.                     }
  415.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  416.                     page_table_unlock(as, false);
  417.                 }
  418.             }
  419.         }
  420.  
  421.         /*
  422.          * Finish TLB shootdown sequence.
  423.          */
  424.         tlb_invalidate_pages(as->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  425.         tlb_shootdown_finalize();
  426.        
  427.         /*
  428.          * Invalidate software translation caches (e.g. TSB on sparc64).
  429.          */
  430.         as_invalidate_translation_cache(as, area->base + pages*PAGE_SIZE, area->pages - pages);
  431.     } else {
  432.         /*
  433.          * Growing the area.
  434.          * Check for overlaps with other address space areas.
  435.          */
  436.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  437.             mutex_unlock(&area->lock);
  438.             mutex_unlock(&as->lock);       
  439.             interrupts_restore(ipl);
  440.             return EADDRNOTAVAIL;
  441.         }
  442.     }
  443.  
  444.     area->pages = pages;
  445.    
  446.     mutex_unlock(&area->lock);
  447.     mutex_unlock(&as->lock);
  448.     interrupts_restore(ipl);
  449.  
  450.     return 0;
  451. }
  452.  
  453. /** Destroy address space area.
  454.  *
  455.  * @param as Address space.
  456.  * @param address Address withing the area to be deleted.
  457.  *
  458.  * @return Zero on success or a value from @ref errno.h on failure.
  459.  */
  460. int as_area_destroy(as_t *as, uintptr_t address)
  461. {
  462.     as_area_t *area;
  463.     uintptr_t base;
  464.     link_t *cur;
  465.     ipl_t ipl;
  466.  
  467.     ipl = interrupts_disable();
  468.     mutex_lock(&as->lock);
  469.  
  470.     area = find_area_and_lock(as, address);
  471.     if (!area) {
  472.         mutex_unlock(&as->lock);
  473.         interrupts_restore(ipl);
  474.         return ENOENT;
  475.     }
  476.  
  477.     base = area->base;
  478.  
  479.     /*
  480.      * Start TLB shootdown sequence.
  481.      */
  482.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  483.  
  484.     /*
  485.      * Visit only the pages mapped by used_space B+tree.
  486.      */
  487.     for (cur = area->used_space.leaf_head.next; cur != &area->used_space.leaf_head; cur = cur->next) {
  488.         btree_node_t *node;
  489.         int i;
  490.        
  491.         node = list_get_instance(cur, btree_node_t, leaf_link);
  492.         for (i = 0; i < node->keys; i++) {
  493.             uintptr_t b = node->key[i];
  494.             count_t j;
  495.             pte_t *pte;
  496.            
  497.             for (j = 0; j < (count_t) node->value[i]; j++) {
  498.                 page_table_lock(as, false);
  499.                 pte = page_mapping_find(as, b + j*PAGE_SIZE);
  500.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  501.                 if (area->backend && area->backend->frame_free) {
  502.                     area->backend->frame_free(area,
  503.                         b + j*PAGE_SIZE, PTE_GET_FRAME(pte));
  504.                 }
  505.                 page_mapping_remove(as, b + j*PAGE_SIZE);              
  506.                 page_table_unlock(as, false);
  507.             }
  508.         }
  509.     }
  510.  
  511.     /*
  512.      * Finish TLB shootdown sequence.
  513.      */
  514.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  515.     tlb_shootdown_finalize();
  516.    
  517.     /*
  518.      * Invalidate potential software translation caches (e.g. TSB on sparc64).
  519.      */
  520.     as_invalidate_translation_cache(as, area->base, area->pages);
  521.    
  522.     btree_destroy(&area->used_space);
  523.  
  524.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  525.    
  526.     if (area->sh_info)
  527.         sh_info_remove_reference(area->sh_info);
  528.        
  529.     mutex_unlock(&area->lock);
  530.  
  531.     /*
  532.      * Remove the empty area from address space.
  533.      */
  534.     btree_remove(&as->as_area_btree, base, NULL);
  535.    
  536.     free(area);
  537.    
  538.     mutex_unlock(&as->lock);
  539.     interrupts_restore(ipl);
  540.     return 0;
  541. }
  542.  
  543. /** Share address space area with another or the same address space.
  544.  *
  545.  * Address space area mapping is shared with a new address space area.
  546.  * If the source address space area has not been shared so far,
  547.  * a new sh_info is created. The new address space area simply gets the
  548.  * sh_info of the source area. The process of duplicating the
  549.  * mapping is done through the backend share function.
  550.  *
  551.  * @param src_as Pointer to source address space.
  552.  * @param src_base Base address of the source address space area.
  553.  * @param acc_size Expected size of the source area.
  554.  * @param dst_as Pointer to destination address space.
  555.  * @param dst_base Target base address.
  556.  * @param dst_flags_mask Destination address space area flags mask.
  557.  *
  558.  * @return Zero on success or ENOENT if there is no such task or if there is no
  559.  * such address space area, EPERM if there was a problem in accepting the area
  560.  * or ENOMEM if there was a problem in allocating destination address space
  561.  * area. ENOTSUP is returned if the address space area backend does not support
  562.  * sharing or if the kernel detects an attempt to create an illegal address
  563.  * alias.
  564.  */
  565. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  566.           as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  567. {
  568.     ipl_t ipl;
  569.     int src_flags;
  570.     size_t src_size;
  571.     as_area_t *src_area, *dst_area;
  572.     share_info_t *sh_info;
  573.     mem_backend_t *src_backend;
  574.     mem_backend_data_t src_backend_data;
  575.    
  576.     ipl = interrupts_disable();
  577.     mutex_lock(&src_as->lock);
  578.     src_area = find_area_and_lock(src_as, src_base);
  579.     if (!src_area) {
  580.         /*
  581.          * Could not find the source address space area.
  582.          */
  583.         mutex_unlock(&src_as->lock);
  584.         interrupts_restore(ipl);
  585.         return ENOENT;
  586.     }
  587.  
  588.     if (!src_area->backend || !src_area->backend->share) {
  589.         /*
  590.          * There is no backend or the backend does not
  591.          * know how to share the area.
  592.          */
  593.         mutex_unlock(&src_area->lock);
  594.         mutex_unlock(&src_as->lock);
  595.         interrupts_restore(ipl);
  596.         return ENOTSUP;
  597.     }
  598.    
  599.     src_size = src_area->pages * PAGE_SIZE;
  600.     src_flags = src_area->flags;
  601.     src_backend = src_area->backend;
  602.     src_backend_data = src_area->backend_data;
  603.  
  604.     /* Share the cacheable flag from the original mapping */
  605.     if (src_flags & AS_AREA_CACHEABLE)
  606.         dst_flags_mask |= AS_AREA_CACHEABLE;
  607.  
  608.     if (src_size != acc_size || (src_flags & dst_flags_mask) != dst_flags_mask) {
  609.         mutex_unlock(&src_area->lock);
  610.         mutex_unlock(&src_as->lock);
  611.         interrupts_restore(ipl);
  612.         return EPERM;
  613.     }
  614.  
  615. #ifdef CONFIG_VIRT_IDX_DCACHE
  616.     if (!(dst_flags_mask & AS_AREA_EXEC)) {
  617.         if (PAGE_COLOR(src_area->base) != PAGE_COLOR(dst_base)) {
  618.             /*
  619.              * Refuse to create an illegal address alias.
  620.              */
  621.             mutex_unlock(&src_area->lock);
  622.             mutex_unlock(&src_as->lock);
  623.             interrupts_restore(ipl);
  624.             return ENOTSUP;
  625.         }
  626.     }
  627. #endif /* CONFIG_VIRT_IDX_DCACHE */
  628.  
  629.     /*
  630.      * Now we are committed to sharing the area.
  631.      * First, prepare the area for sharing.
  632.      * Then it will be safe to unlock it.
  633.      */
  634.     sh_info = src_area->sh_info;
  635.     if (!sh_info) {
  636.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  637.         mutex_initialize(&sh_info->lock);
  638.         sh_info->refcount = 2;
  639.         btree_create(&sh_info->pagemap);
  640.         src_area->sh_info = sh_info;
  641.     } else {
  642.         mutex_lock(&sh_info->lock);
  643.         sh_info->refcount++;
  644.         mutex_unlock(&sh_info->lock);
  645.     }
  646.  
  647.     src_area->backend->share(src_area);
  648.  
  649.     mutex_unlock(&src_area->lock);
  650.     mutex_unlock(&src_as->lock);
  651.  
  652.     /*
  653.      * Create copy of the source address space area.
  654.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  655.      * attribute set which prevents race condition with
  656.      * preliminary as_page_fault() calls.
  657.      * The flags of the source area are masked against dst_flags_mask
  658.      * to support sharing in less privileged mode.
  659.      */
  660.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  661.                   AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  662.     if (!dst_area) {
  663.         /*
  664.          * Destination address space area could not be created.
  665.          */
  666.         sh_info_remove_reference(sh_info);
  667.        
  668.         interrupts_restore(ipl);
  669.         return ENOMEM;
  670.     }
  671.  
  672.     /*
  673.      * Now the destination address space area has been
  674.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  675.      * attribute and set the sh_info.
  676.      */
  677.     mutex_lock(&dst_as->lock); 
  678.     mutex_lock(&dst_area->lock);
  679.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  680.     dst_area->sh_info = sh_info;
  681.     mutex_unlock(&dst_area->lock);
  682.     mutex_unlock(&dst_as->lock);   
  683.  
  684.     interrupts_restore(ipl);
  685.    
  686.     return 0;
  687. }
  688.  
  689. /** Check access mode for address space area.
  690.  *
  691.  * The address space area must be locked prior to this call.
  692.  *
  693.  * @param area Address space area.
  694.  * @param access Access mode.
  695.  *
  696.  * @return False if access violates area's permissions, true otherwise.
  697.  */
  698. bool as_area_check_access(as_area_t *area, pf_access_t access)
  699. {
  700.     int flagmap[] = {
  701.         [PF_ACCESS_READ] = AS_AREA_READ,
  702.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  703.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  704.     };
  705.  
  706.     if (!(area->flags & flagmap[access]))
  707.         return false;
  708.    
  709.     return true;
  710. }
  711.  
  712. /** Handle page fault within the current address space.
  713.  *
  714.  * This is the high-level page fault handler. It decides
  715.  * whether the page fault can be resolved by any backend
  716.  * and if so, it invokes the backend to resolve the page
  717.  * fault.
  718.  *
  719.  * Interrupts are assumed disabled.
  720.  *
  721.  * @param page Faulting page.
  722.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  723.  * @param istate Pointer to interrupted state.
  724.  *
  725.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  726.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  727.  */
  728. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  729. {
  730.     pte_t *pte;
  731.     as_area_t *area;
  732.    
  733.     if (!THREAD)
  734.         return AS_PF_FAULT;
  735.        
  736.     ASSERT(AS);
  737.  
  738.     mutex_lock(&AS->lock);
  739.     area = find_area_and_lock(AS, page);   
  740.     if (!area) {
  741.         /*
  742.          * No area contained mapping for 'page'.
  743.          * Signal page fault to low-level handler.
  744.          */
  745.         mutex_unlock(&AS->lock);
  746.         goto page_fault;
  747.     }
  748.  
  749.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  750.         /*
  751.          * The address space area is not fully initialized.
  752.          * Avoid possible race by returning error.
  753.          */
  754.         mutex_unlock(&area->lock);
  755.         mutex_unlock(&AS->lock);
  756.         goto page_fault;       
  757.     }
  758.  
  759.     if (!area->backend || !area->backend->page_fault) {
  760.         /*
  761.          * The address space area is not backed by any backend
  762.          * or the backend cannot handle page faults.
  763.          */
  764.         mutex_unlock(&area->lock);
  765.         mutex_unlock(&AS->lock);
  766.         goto page_fault;       
  767.     }
  768.  
  769.     page_table_lock(AS, false);
  770.    
  771.     /*
  772.      * To avoid race condition between two page faults
  773.      * on the same address, we need to make sure
  774.      * the mapping has not been already inserted.
  775.      */
  776.     if ((pte = page_mapping_find(AS, page))) {
  777.         if (PTE_PRESENT(pte)) {
  778.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  779.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  780.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  781.                 page_table_unlock(AS, false);
  782.                 mutex_unlock(&area->lock);
  783.                 mutex_unlock(&AS->lock);
  784.                 return AS_PF_OK;
  785.             }
  786.         }
  787.     }
  788.    
  789.     /*
  790.      * Resort to the backend page fault handler.
  791.      */
  792.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  793.         page_table_unlock(AS, false);
  794.         mutex_unlock(&area->lock);
  795.         mutex_unlock(&AS->lock);
  796.         goto page_fault;
  797.     }
  798.    
  799.     page_table_unlock(AS, false);
  800.     mutex_unlock(&area->lock);
  801.     mutex_unlock(&AS->lock);
  802.     return AS_PF_OK;
  803.  
  804. page_fault:
  805.     if (THREAD->in_copy_from_uspace) {
  806.         THREAD->in_copy_from_uspace = false;
  807.         istate_set_retaddr(istate, (uintptr_t) &memcpy_from_uspace_failover_address);
  808.     } else if (THREAD->in_copy_to_uspace) {
  809.         THREAD->in_copy_to_uspace = false;
  810.         istate_set_retaddr(istate, (uintptr_t) &memcpy_to_uspace_failover_address);
  811.     } else {
  812.         return AS_PF_FAULT;
  813.     }
  814.  
  815.     return AS_PF_DEFER;
  816. }
  817.  
  818. /** Switch address spaces.
  819.  *
  820.  * Note that this function cannot sleep as it is essentially a part of
  821.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  822.  *
  823.  * @param old Old address space or NULL.
  824.  * @param new New address space.
  825.  */
  826. void as_switch(as_t *old, as_t *new)
  827. {
  828.     ipl_t ipl;
  829.     bool needs_asid = false;
  830.    
  831.     ipl = interrupts_disable();
  832.     spinlock_lock(&inactive_as_with_asid_lock);
  833.  
  834.     /*
  835.      * First, take care of the old address space.
  836.      */
  837.     if (old) {
  838.         mutex_lock_active(&old->lock);
  839.         ASSERT(old->cpu_refcount);
  840.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  841.             /*
  842.              * The old address space is no longer active on
  843.              * any processor. It can be appended to the
  844.              * list of inactive address spaces with assigned
  845.              * ASID.
  846.              */
  847.              ASSERT(old->asid != ASID_INVALID);
  848.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  849.         }
  850.         mutex_unlock(&old->lock);
  851.  
  852.         /*
  853.          * Perform architecture-specific tasks when the address space
  854.          * is being removed from the CPU.
  855.          */
  856.         as_deinstall_arch(old);
  857.     }
  858.  
  859.     /*
  860.      * Second, prepare the new address space.
  861.      */
  862.     mutex_lock_active(&new->lock);
  863.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  864.         if (new->asid != ASID_INVALID)
  865.             list_remove(&new->inactive_as_with_asid_link);
  866.         else
  867.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  868.     }
  869.     SET_PTL0_ADDRESS(new->page_table);
  870.     mutex_unlock(&new->lock);
  871.  
  872.     if (needs_asid) {
  873.         /*
  874.          * Allocation of new ASID was deferred
  875.          * until now in order to avoid deadlock.
  876.          */
  877.         asid_t asid;
  878.        
  879.         asid = asid_get();
  880.         mutex_lock_active(&new->lock);
  881.         new->asid = asid;
  882.         mutex_unlock(&new->lock);
  883.     }
  884.     spinlock_unlock(&inactive_as_with_asid_lock);
  885.     interrupts_restore(ipl);
  886.    
  887.     /*
  888.      * Perform architecture-specific steps.
  889.      * (e.g. write ASID to hardware register etc.)
  890.      */
  891.     as_install_arch(new);
  892.    
  893.     AS = new;
  894. }
  895.  
  896. /** Convert address space area flags to page flags.
  897.  *
  898.  * @param aflags Flags of some address space area.
  899.  *
  900.  * @return Flags to be passed to page_mapping_insert().
  901.  */
  902. int area_flags_to_page_flags(int aflags)
  903. {
  904.     int flags;
  905.  
  906.     flags = PAGE_USER | PAGE_PRESENT;
  907.    
  908.     if (aflags & AS_AREA_READ)
  909.         flags |= PAGE_READ;
  910.        
  911.     if (aflags & AS_AREA_WRITE)
  912.         flags |= PAGE_WRITE;
  913.    
  914.     if (aflags & AS_AREA_EXEC)
  915.         flags |= PAGE_EXEC;
  916.    
  917.     if (aflags & AS_AREA_CACHEABLE)
  918.         flags |= PAGE_CACHEABLE;
  919.        
  920.     return flags;
  921. }
  922.  
  923. /** Compute flags for virtual address translation subsytem.
  924.  *
  925.  * The address space area must be locked.
  926.  * Interrupts must be disabled.
  927.  *
  928.  * @param a Address space area.
  929.  *
  930.  * @return Flags to be used in page_mapping_insert().
  931.  */
  932. int as_area_get_flags(as_area_t *a)
  933. {
  934.     return area_flags_to_page_flags(a->flags);
  935. }
  936.  
  937. /** Create page table.
  938.  *
  939.  * Depending on architecture, create either address space
  940.  * private or global page table.
  941.  *
  942.  * @param flags Flags saying whether the page table is for kernel address space.
  943.  *
  944.  * @return First entry of the page table.
  945.  */
  946. pte_t *page_table_create(int flags)
  947. {
  948.         ASSERT(as_operations);
  949.         ASSERT(as_operations->page_table_create);
  950.  
  951.         return as_operations->page_table_create(flags);
  952. }
  953.  
  954. /** Destroy page table.
  955.  *
  956.  * Destroy page table in architecture specific way.
  957.  *
  958.  * @param page_table Physical address of PTL0.
  959.  */
  960. void page_table_destroy(pte_t *page_table)
  961. {
  962.         ASSERT(as_operations);
  963.         ASSERT(as_operations->page_table_destroy);
  964.  
  965.         as_operations->page_table_destroy(page_table);
  966. }
  967.  
  968. /** Lock page table.
  969.  *
  970.  * This function should be called before any page_mapping_insert(),
  971.  * page_mapping_remove() and page_mapping_find().
  972.  *
  973.  * Locking order is such that address space areas must be locked
  974.  * prior to this call. Address space can be locked prior to this
  975.  * call in which case the lock argument is false.
  976.  *
  977.  * @param as Address space.
  978.  * @param lock If false, do not attempt to lock as->lock.
  979.  */
  980. void page_table_lock(as_t *as, bool lock)
  981. {
  982.     ASSERT(as_operations);
  983.     ASSERT(as_operations->page_table_lock);
  984.  
  985.     as_operations->page_table_lock(as, lock);
  986. }
  987.  
  988. /** Unlock page table.
  989.  *
  990.  * @param as Address space.
  991.  * @param unlock If false, do not attempt to unlock as->lock.
  992.  */
  993. void page_table_unlock(as_t *as, bool unlock)
  994. {
  995.     ASSERT(as_operations);
  996.     ASSERT(as_operations->page_table_unlock);
  997.  
  998.     as_operations->page_table_unlock(as, unlock);
  999. }
  1000.  
  1001.  
  1002. /** Find address space area and lock it.
  1003.  *
  1004.  * The address space must be locked and interrupts must be disabled.
  1005.  *
  1006.  * @param as Address space.
  1007.  * @param va Virtual address.
  1008.  *
  1009.  * @return Locked address space area containing va on success or NULL on failure.
  1010.  */
  1011. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  1012. {
  1013.     as_area_t *a;
  1014.     btree_node_t *leaf, *lnode;
  1015.     int i;
  1016.    
  1017.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  1018.     if (a) {
  1019.         /* va is the base address of an address space area */
  1020.         mutex_lock(&a->lock);
  1021.         return a;
  1022.     }
  1023.    
  1024.     /*
  1025.      * Search the leaf node and the righmost record of its left neighbour
  1026.      * to find out whether this is a miss or va belongs to an address
  1027.      * space area found there.
  1028.      */
  1029.    
  1030.     /* First, search the leaf node itself. */
  1031.     for (i = 0; i < leaf->keys; i++) {
  1032.         a = (as_area_t *) leaf->value[i];
  1033.         mutex_lock(&a->lock);
  1034.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  1035.             return a;
  1036.         }
  1037.         mutex_unlock(&a->lock);
  1038.     }
  1039.  
  1040.     /*
  1041.      * Second, locate the left neighbour and test its last record.
  1042.      * Because of its position in the B+tree, it must have base < va.
  1043.      */
  1044.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1045.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  1046.         mutex_lock(&a->lock);
  1047.         if (va < a->base + a->pages * PAGE_SIZE) {
  1048.             return a;
  1049.         }
  1050.         mutex_unlock(&a->lock);
  1051.     }
  1052.  
  1053.     return NULL;
  1054. }
  1055.  
  1056. /** Check area conflicts with other areas.
  1057.  *
  1058.  * The address space must be locked and interrupts must be disabled.
  1059.  *
  1060.  * @param as Address space.
  1061.  * @param va Starting virtual address of the area being tested.
  1062.  * @param size Size of the area being tested.
  1063.  * @param avoid_area Do not touch this area.
  1064.  *
  1065.  * @return True if there is no conflict, false otherwise.
  1066.  */
  1067. bool check_area_conflicts(as_t *as, uintptr_t va, size_t size, as_area_t *avoid_area)
  1068. {
  1069.     as_area_t *a;
  1070.     btree_node_t *leaf, *node;
  1071.     int i;
  1072.    
  1073.     /*
  1074.      * We don't want any area to have conflicts with NULL page.
  1075.      */
  1076.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1077.         return false;
  1078.    
  1079.     /*
  1080.      * The leaf node is found in O(log n), where n is proportional to
  1081.      * the number of address space areas belonging to as.
  1082.      * The check for conflicts is then attempted on the rightmost
  1083.      * record in the left neighbour, the leftmost record in the right
  1084.      * neighbour and all records in the leaf node itself.
  1085.      */
  1086.    
  1087.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1088.         if (a != avoid_area)
  1089.             return false;
  1090.     }
  1091.    
  1092.     /* First, check the two border cases. */
  1093.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1094.         a = (as_area_t *) node->value[node->keys - 1];
  1095.         mutex_lock(&a->lock);
  1096.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1097.             mutex_unlock(&a->lock);
  1098.             return false;
  1099.         }
  1100.         mutex_unlock(&a->lock);
  1101.     }
  1102.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1103.         a = (as_area_t *) node->value[0];
  1104.         mutex_lock(&a->lock);
  1105.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1106.             mutex_unlock(&a->lock);
  1107.             return false;
  1108.         }
  1109.         mutex_unlock(&a->lock);
  1110.     }
  1111.    
  1112.     /* Second, check the leaf node. */
  1113.     for (i = 0; i < leaf->keys; i++) {
  1114.         a = (as_area_t *) leaf->value[i];
  1115.    
  1116.         if (a == avoid_area)
  1117.             continue;
  1118.    
  1119.         mutex_lock(&a->lock);
  1120.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1121.             mutex_unlock(&a->lock);
  1122.             return false;
  1123.         }
  1124.         mutex_unlock(&a->lock);
  1125.     }
  1126.  
  1127.     /*
  1128.      * So far, the area does not conflict with other areas.
  1129.      * Check if it doesn't conflict with kernel address space.
  1130.      */  
  1131.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1132.         return !overlaps(va, size,
  1133.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1134.     }
  1135.  
  1136.     return true;
  1137. }
  1138.  
  1139. /** Return size of the address space area with given base.  */
  1140. size_t as_get_size(uintptr_t base)
  1141. {
  1142.     ipl_t ipl;
  1143.     as_area_t *src_area;
  1144.     size_t size;
  1145.  
  1146.     ipl = interrupts_disable();
  1147.     src_area = find_area_and_lock(AS, base);
  1148.     if (src_area){
  1149.         size = src_area->pages * PAGE_SIZE;
  1150.         mutex_unlock(&src_area->lock);
  1151.     } else {
  1152.         size = 0;
  1153.     }
  1154.     interrupts_restore(ipl);
  1155.     return size;
  1156. }
  1157.  
  1158. /** Mark portion of address space area as used.
  1159.  *
  1160.  * The address space area must be already locked.
  1161.  *
  1162.  * @param a Address space area.
  1163.  * @param page First page to be marked.
  1164.  * @param count Number of page to be marked.
  1165.  *
  1166.  * @return 0 on failure and 1 on success.
  1167.  */
  1168. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1169. {
  1170.     btree_node_t *leaf, *node;
  1171.     count_t pages;
  1172.     int i;
  1173.  
  1174.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1175.     ASSERT(count);
  1176.  
  1177.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1178.     if (pages) {
  1179.         /*
  1180.          * We hit the beginning of some used space.
  1181.          */
  1182.         return 0;
  1183.     }
  1184.  
  1185.     if (!leaf->keys) {
  1186.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1187.         return 1;
  1188.     }
  1189.  
  1190.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1191.     if (node) {
  1192.         uintptr_t left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1193.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1194.        
  1195.         /*
  1196.          * Examine the possibility that the interval fits
  1197.          * somewhere between the rightmost interval of
  1198.          * the left neigbour and the first interval of the leaf.
  1199.          */
  1200.          
  1201.         if (page >= right_pg) {
  1202.             /* Do nothing. */
  1203.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1204.             /* The interval intersects with the left interval. */
  1205.             return 0;
  1206.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1207.             /* The interval intersects with the right interval. */
  1208.             return 0;          
  1209.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1210.             /* The interval can be added by merging the two already present intervals. */
  1211.             node->value[node->keys - 1] += count + right_cnt;
  1212.             btree_remove(&a->used_space, right_pg, leaf);
  1213.             return 1;
  1214.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1215.             /* The interval can be added by simply growing the left interval. */
  1216.             node->value[node->keys - 1] += count;
  1217.             return 1;
  1218.         } else if (page + count*PAGE_SIZE == right_pg) {
  1219.             /*
  1220.              * The interval can be addded by simply moving base of the right
  1221.              * interval down and increasing its size accordingly.
  1222.              */
  1223.             leaf->value[0] += count;
  1224.             leaf->key[0] = page;
  1225.             return 1;
  1226.         } else {
  1227.             /*
  1228.              * The interval is between both neigbouring intervals,
  1229.              * but cannot be merged with any of them.
  1230.              */
  1231.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1232.             return 1;
  1233.         }
  1234.     } else if (page < leaf->key[0]) {
  1235.         uintptr_t right_pg = leaf->key[0];
  1236.         count_t right_cnt = (count_t) leaf->value[0];
  1237.    
  1238.         /*
  1239.          * Investigate the border case in which the left neighbour does not
  1240.          * exist but the interval fits from the left.
  1241.          */
  1242.          
  1243.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1244.             /* The interval intersects with the right interval. */
  1245.             return 0;
  1246.         } else if (page + count*PAGE_SIZE == right_pg) {
  1247.             /*
  1248.              * The interval can be added by moving the base of the right interval down
  1249.              * and increasing its size accordingly.
  1250.              */
  1251.             leaf->key[0] = page;
  1252.             leaf->value[0] += count;
  1253.             return 1;
  1254.         } else {
  1255.             /*
  1256.              * The interval doesn't adjoin with the right interval.
  1257.              * It must be added individually.
  1258.              */
  1259.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1260.             return 1;
  1261.         }
  1262.     }
  1263.  
  1264.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1265.     if (node) {
  1266.         uintptr_t left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1267.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1268.        
  1269.         /*
  1270.          * Examine the possibility that the interval fits
  1271.          * somewhere between the leftmost interval of
  1272.          * the right neigbour and the last interval of the leaf.
  1273.          */
  1274.  
  1275.         if (page < left_pg) {
  1276.             /* Do nothing. */
  1277.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1278.             /* The interval intersects with the left interval. */
  1279.             return 0;
  1280.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1281.             /* The interval intersects with the right interval. */
  1282.             return 0;          
  1283.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1284.             /* The interval can be added by merging the two already present intervals. */
  1285.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1286.             btree_remove(&a->used_space, right_pg, node);
  1287.             return 1;
  1288.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1289.             /* The interval can be added by simply growing the left interval. */
  1290.             leaf->value[leaf->keys - 1] +=  count;
  1291.             return 1;
  1292.         } else if (page + count*PAGE_SIZE == right_pg) {
  1293.             /*
  1294.              * The interval can be addded by simply moving base of the right
  1295.              * interval down and increasing its size accordingly.
  1296.              */
  1297.             node->value[0] += count;
  1298.             node->key[0] = page;
  1299.             return 1;
  1300.         } else {
  1301.             /*
  1302.              * The interval is between both neigbouring intervals,
  1303.              * but cannot be merged with any of them.
  1304.              */
  1305.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1306.             return 1;
  1307.         }
  1308.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1309.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1310.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1311.    
  1312.         /*
  1313.          * Investigate the border case in which the right neighbour does not
  1314.          * exist but the interval fits from the right.
  1315.          */
  1316.          
  1317.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1318.             /* The interval intersects with the left interval. */
  1319.             return 0;
  1320.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1321.             /* The interval can be added by growing the left interval. */
  1322.             leaf->value[leaf->keys - 1] += count;
  1323.             return 1;
  1324.         } else {
  1325.             /*
  1326.              * The interval doesn't adjoin with the left interval.
  1327.              * It must be added individually.
  1328.              */
  1329.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1330.             return 1;
  1331.         }
  1332.     }
  1333.    
  1334.     /*
  1335.      * Note that if the algorithm made it thus far, the interval can fit only
  1336.      * between two other intervals of the leaf. The two border cases were already
  1337.      * resolved.
  1338.      */
  1339.     for (i = 1; i < leaf->keys; i++) {
  1340.         if (page < leaf->key[i]) {
  1341.             uintptr_t left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1342.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1343.  
  1344.             /*
  1345.              * The interval fits between left_pg and right_pg.
  1346.              */
  1347.  
  1348.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1349.                 /* The interval intersects with the left interval. */
  1350.                 return 0;
  1351.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1352.                 /* The interval intersects with the right interval. */
  1353.                 return 0;          
  1354.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1355.                 /* The interval can be added by merging the two already present intervals. */
  1356.                 leaf->value[i - 1] += count + right_cnt;
  1357.                 btree_remove(&a->used_space, right_pg, leaf);
  1358.                 return 1;
  1359.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1360.                 /* The interval can be added by simply growing the left interval. */
  1361.                 leaf->value[i - 1] += count;
  1362.                 return 1;
  1363.             } else if (page + count*PAGE_SIZE == right_pg) {
  1364.                 /*
  1365.                      * The interval can be addded by simply moving base of the right
  1366.                  * interval down and increasing its size accordingly.
  1367.                  */
  1368.                 leaf->value[i] += count;
  1369.                 leaf->key[i] = page;
  1370.                 return 1;
  1371.             } else {
  1372.                 /*
  1373.                  * The interval is between both neigbouring intervals,
  1374.                  * but cannot be merged with any of them.
  1375.                  */
  1376.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1377.                 return 1;
  1378.             }
  1379.         }
  1380.     }
  1381.  
  1382.     panic("Inconsistency detected while adding %d pages of used space at %p.\n", count, page);
  1383. }
  1384.  
  1385. /** Mark portion of address space area as unused.
  1386.  *
  1387.  * The address space area must be already locked.
  1388.  *
  1389.  * @param a Address space area.
  1390.  * @param page First page to be marked.
  1391.  * @param count Number of page to be marked.
  1392.  *
  1393.  * @return 0 on failure and 1 on success.
  1394.  */
  1395. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1396. {
  1397.     btree_node_t *leaf, *node;
  1398.     count_t pages;
  1399.     int i;
  1400.  
  1401.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1402.     ASSERT(count);
  1403.  
  1404.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1405.     if (pages) {
  1406.         /*
  1407.          * We are lucky, page is the beginning of some interval.
  1408.          */
  1409.         if (count > pages) {
  1410.             return 0;
  1411.         } else if (count == pages) {
  1412.             btree_remove(&a->used_space, page, leaf);
  1413.             return 1;
  1414.         } else {
  1415.             /*
  1416.              * Find the respective interval.
  1417.              * Decrease its size and relocate its start address.
  1418.              */
  1419.             for (i = 0; i < leaf->keys; i++) {
  1420.                 if (leaf->key[i] == page) {
  1421.                     leaf->key[i] += count*PAGE_SIZE;
  1422.                     leaf->value[i] -= count;
  1423.                     return 1;
  1424.                 }
  1425.             }
  1426.             goto error;
  1427.         }
  1428.     }
  1429.  
  1430.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1431.     if (node && page < leaf->key[0]) {
  1432.         uintptr_t left_pg = node->key[node->keys - 1];
  1433.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1434.  
  1435.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1436.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1437.                 /*
  1438.                  * The interval is contained in the rightmost interval
  1439.                  * of the left neighbour and can be removed by
  1440.                  * updating the size of the bigger interval.
  1441.                  */
  1442.                 node->value[node->keys - 1] -= count;
  1443.                 return 1;
  1444.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1445.                 count_t new_cnt;
  1446.                
  1447.                 /*
  1448.                  * The interval is contained in the rightmost interval
  1449.                  * of the left neighbour but its removal requires
  1450.                  * both updating the size of the original interval and
  1451.                  * also inserting a new interval.
  1452.                  */
  1453.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1454.                 node->value[node->keys - 1] -= count + new_cnt;
  1455.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1456.                 return 1;
  1457.             }
  1458.         }
  1459.         return 0;
  1460.     } else if (page < leaf->key[0]) {
  1461.         return 0;
  1462.     }
  1463.    
  1464.     if (page > leaf->key[leaf->keys - 1]) {
  1465.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1466.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1467.  
  1468.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1469.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1470.                 /*
  1471.                  * The interval is contained in the rightmost interval
  1472.                  * of the leaf and can be removed by updating the size
  1473.                  * of the bigger interval.
  1474.                  */
  1475.                 leaf->value[leaf->keys - 1] -= count;
  1476.                 return 1;
  1477.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1478.                 count_t new_cnt;
  1479.                
  1480.                 /*
  1481.                  * The interval is contained in the rightmost interval
  1482.                  * of the leaf but its removal requires both updating
  1483.                  * the size of the original interval and
  1484.                  * also inserting a new interval.
  1485.                  */
  1486.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1487.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1488.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1489.                 return 1;
  1490.             }
  1491.         }
  1492.         return 0;
  1493.     }  
  1494.    
  1495.     /*
  1496.      * The border cases have been already resolved.
  1497.      * Now the interval can be only between intervals of the leaf.
  1498.      */
  1499.     for (i = 1; i < leaf->keys - 1; i++) {
  1500.         if (page < leaf->key[i]) {
  1501.             uintptr_t left_pg = leaf->key[i - 1];
  1502.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1503.  
  1504.             /*
  1505.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1506.              */
  1507.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1508.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1509.                     /*
  1510.                     * The interval is contained in the interval (i - 1)
  1511.                      * of the leaf and can be removed by updating the size
  1512.                      * of the bigger interval.
  1513.                      */
  1514.                     leaf->value[i - 1] -= count;
  1515.                     return 1;
  1516.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1517.                     count_t new_cnt;
  1518.                
  1519.                     /*
  1520.                      * The interval is contained in the interval (i - 1)
  1521.                      * of the leaf but its removal requires both updating
  1522.                      * the size of the original interval and
  1523.                      * also inserting a new interval.
  1524.                      */
  1525.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1526.                     leaf->value[i - 1] -= count + new_cnt;
  1527.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1528.                     return 1;
  1529.                 }
  1530.             }
  1531.             return 0;
  1532.         }
  1533.     }
  1534.  
  1535. error:
  1536.     panic("Inconsistency detected while removing %d pages of used space from %p.\n", count, page);
  1537. }
  1538.  
  1539. /** Remove reference to address space area share info.
  1540.  *
  1541.  * If the reference count drops to 0, the sh_info is deallocated.
  1542.  *
  1543.  * @param sh_info Pointer to address space area share info.
  1544.  */
  1545. void sh_info_remove_reference(share_info_t *sh_info)
  1546. {
  1547.     bool dealloc = false;
  1548.  
  1549.     mutex_lock(&sh_info->lock);
  1550.     ASSERT(sh_info->refcount);
  1551.     if (--sh_info->refcount == 0) {
  1552.         dealloc = true;
  1553.         link_t *cur;
  1554.        
  1555.         /*
  1556.          * Now walk carefully the pagemap B+tree and free/remove
  1557.          * reference from all frames found there.
  1558.          */
  1559.         for (cur = sh_info->pagemap.leaf_head.next; cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1560.             btree_node_t *node;
  1561.             int i;
  1562.            
  1563.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1564.             for (i = 0; i < node->keys; i++)
  1565.                 frame_free((uintptr_t) node->value[i]);
  1566.         }
  1567.        
  1568.     }
  1569.     mutex_unlock(&sh_info->lock);
  1570.    
  1571.     if (dealloc) {
  1572.         btree_destroy(&sh_info->pagemap);
  1573.         free(sh_info);
  1574.     }
  1575. }
  1576.  
  1577. /*
  1578.  * Address space related syscalls.
  1579.  */
  1580.  
  1581. /** Wrapper for as_area_create(). */
  1582. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1583. {
  1584.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1585.         return (unative_t) address;
  1586.     else
  1587.         return (unative_t) -1;
  1588. }
  1589.  
  1590. /** Wrapper for as_area_resize(). */
  1591. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1592. {
  1593.     return (unative_t) as_area_resize(AS, address, size, 0);
  1594. }
  1595.  
  1596. /** Wrapper for as_area_destroy(). */
  1597. unative_t sys_as_area_destroy(uintptr_t address)
  1598. {
  1599.     return (unative_t) as_area_destroy(AS, address);
  1600. }
  1601.  
  1602. /** Print out information about address space.
  1603.  *
  1604.  * @param as Address space.
  1605.  */
  1606. void as_print(as_t *as)
  1607. {
  1608.     ipl_t ipl;
  1609.    
  1610.     ipl = interrupts_disable();
  1611.     mutex_lock(&as->lock);
  1612.    
  1613.     /* print out info about address space areas */
  1614.     link_t *cur;
  1615.     for (cur = as->as_area_btree.leaf_head.next; cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  1616.         btree_node_t *node = list_get_instance(cur, btree_node_t, leaf_link);
  1617.        
  1618.         int i;
  1619.         for (i = 0; i < node->keys; i++) {
  1620.             as_area_t *area = node->value[i];
  1621.        
  1622.             mutex_lock(&area->lock);
  1623.             printf("as_area: %p, base=%p, pages=%d (%p - %p)\n",
  1624.                 area, area->base, area->pages, area->base, area->base + area->pages*PAGE_SIZE);
  1625.             mutex_unlock(&area->lock);
  1626.         }
  1627.     }
  1628.    
  1629.     mutex_unlock(&as->lock);
  1630.     interrupts_restore(ipl);
  1631. }
  1632.  
  1633. /** @}
  1634.  */
  1635.