Subversion Repositories HelenOS-historic

Rev

Rev 1403 | Rev 1411 | 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. /**
  30.  * @file    as.c
  31.  * @brief   Address space related functions.
  32.  *
  33.  * This file contains address space manipulation functions.
  34.  * Roughly speaking, this is a higher-level client of
  35.  * Virtual Address Translation (VAT) subsystem.
  36.  *
  37.  * Functionality provided by this file allows one to
  38.  * create address space and create, resize and share
  39.  * address space areas.
  40.  *
  41.  * @see page.c
  42.  *
  43.  */
  44.  
  45. #include <mm/as.h>
  46. #include <arch/mm/as.h>
  47. #include <mm/page.h>
  48. #include <mm/frame.h>
  49. #include <mm/slab.h>
  50. #include <mm/tlb.h>
  51. #include <arch/mm/page.h>
  52. #include <genarch/mm/page_pt.h>
  53. #include <genarch/mm/page_ht.h>
  54. #include <mm/asid.h>
  55. #include <arch/mm/asid.h>
  56. #include <synch/spinlock.h>
  57. #include <synch/mutex.h>
  58. #include <adt/list.h>
  59. #include <adt/btree.h>
  60. #include <proc/task.h>
  61. #include <proc/thread.h>
  62. #include <arch/asm.h>
  63. #include <panic.h>
  64. #include <debug.h>
  65. #include <print.h>
  66. #include <memstr.h>
  67. #include <macros.h>
  68. #include <arch.h>
  69. #include <errno.h>
  70. #include <config.h>
  71. #include <align.h>
  72. #include <arch/types.h>
  73. #include <typedefs.h>
  74. #include <syscall/copy.h>
  75. #include <arch/interrupt.h>
  76.  
  77. /** This structure contains information associated with the shared address space area. */
  78. struct share_info {
  79.     mutex_t lock;       /**< This lock must be acquired only when the as_area lock is held. */
  80.     count_t refcount;   /**< This structure can be deallocated if refcount drops to 0. */
  81.     btree_t pagemap;    /**< B+tree containing complete map of anonymous pages of the shared area. */
  82. };
  83.  
  84. as_operations_t *as_operations = NULL;
  85.  
  86. /** Address space lock. It protects inactive_as_with_asid_head. Must be acquired before as_t mutex. */
  87. SPINLOCK_INITIALIZE(as_lock);
  88.  
  89. /**
  90.  * This list contains address spaces that are not active on any
  91.  * processor and that have valid ASID.
  92.  */
  93. LIST_INITIALIZE(inactive_as_with_asid_head);
  94.  
  95. /** Kernel address space. */
  96. as_t *AS_KERNEL = NULL;
  97.  
  98. static int area_flags_to_page_flags(int aflags);
  99. static as_area_t *find_area_and_lock(as_t *as, __address va);
  100. static bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area);
  101. static void sh_info_remove_reference(share_info_t *sh_info);
  102.  
  103. /** Initialize address space subsystem. */
  104. void as_init(void)
  105. {
  106.     as_arch_init();
  107.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  108.     if (!AS_KERNEL)
  109.         panic("can't create kernel address space\n");
  110.    
  111. }
  112.  
  113. /** Create address space.
  114.  *
  115.  * @param flags Flags that influence way in wich the address space is created.
  116.  */
  117. as_t *as_create(int flags)
  118. {
  119.     as_t *as;
  120.  
  121.     as = (as_t *) malloc(sizeof(as_t), 0);
  122.     link_initialize(&as->inactive_as_with_asid_link);
  123.     mutex_initialize(&as->lock);
  124.     btree_create(&as->as_area_btree);
  125.    
  126.     if (flags & FLAG_AS_KERNEL)
  127.         as->asid = ASID_KERNEL;
  128.     else
  129.         as->asid = ASID_INVALID;
  130.    
  131.     as->refcount = 0;
  132.     as->page_table = page_table_create(flags);
  133.  
  134.     return as;
  135. }
  136.  
  137. /** Free Adress space */
  138. void as_free(as_t *as)
  139. {
  140.     ASSERT(as->refcount == 0);
  141.  
  142.     /* TODO: free as_areas and other resources held by as */
  143.     /* TODO: free page table */
  144.     free(as);
  145. }
  146.  
  147. /** Create address space area of common attributes.
  148.  *
  149.  * The created address space area is added to the target address space.
  150.  *
  151.  * @param as Target address space.
  152.  * @param flags Flags of the area memory.
  153.  * @param size Size of area.
  154.  * @param base Base address of area.
  155.  * @param attrs Attributes of the area.
  156.  * @param backend Address space area backend. NULL if no backend is used.
  157.  * @param backend_data NULL or a pointer to an array holding two void *.
  158.  *
  159.  * @return Address space area on success or NULL on failure.
  160.  */
  161. as_area_t *as_area_create(as_t *as, int flags, size_t size, __address base, int attrs,
  162.            mem_backend_t *backend, void **backend_data)
  163. {
  164.     ipl_t ipl;
  165.     as_area_t *a;
  166.    
  167.     if (base % PAGE_SIZE)
  168.         return NULL;
  169.  
  170.     if (!size)
  171.         return NULL;
  172.  
  173.     /* Writeable executable areas are not supported. */
  174.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  175.         return NULL;
  176.    
  177.     ipl = interrupts_disable();
  178.     mutex_lock(&as->lock);
  179.    
  180.     if (!check_area_conflicts(as, base, size, NULL)) {
  181.         mutex_unlock(&as->lock);
  182.         interrupts_restore(ipl);
  183.         return NULL;
  184.     }
  185.    
  186.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  187.  
  188.     mutex_initialize(&a->lock);
  189.    
  190.     a->flags = flags;
  191.     a->attributes = attrs;
  192.     a->pages = SIZE2FRAMES(size);
  193.     a->base = base;
  194.     a->sh_info = NULL;
  195.     a->backend = backend;
  196.     if (backend_data) {
  197.         a->backend_data[0] = backend_data[0];
  198.         a->backend_data[1] = backend_data[1];
  199.     }
  200.     btree_create(&a->used_space);
  201.    
  202.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  203.  
  204.     mutex_unlock(&as->lock);
  205.     interrupts_restore(ipl);
  206.  
  207.     return a;
  208. }
  209.  
  210. /** Find address space area and change it.
  211.  *
  212.  * @param as Address space.
  213.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  214.  * @param size New size of the virtual memory block starting at address.
  215.  * @param flags Flags influencing the remap operation. Currently unused.
  216.  *
  217.  * @return Zero on success or a value from @ref errno.h otherwise.
  218.  */
  219. int as_area_resize(as_t *as, __address address, size_t size, int flags)
  220. {
  221.     as_area_t *area;
  222.     ipl_t ipl;
  223.     size_t pages;
  224.    
  225.     ipl = interrupts_disable();
  226.     mutex_lock(&as->lock);
  227.    
  228.     /*
  229.      * Locate the area.
  230.      */
  231.     area = find_area_and_lock(as, address);
  232.     if (!area) {
  233.         mutex_unlock(&as->lock);
  234.         interrupts_restore(ipl);
  235.         return ENOENT;
  236.     }
  237.  
  238.     if (area->flags & AS_AREA_DEVICE) {
  239.         /*
  240.          * Remapping of address space areas associated
  241.          * with memory mapped devices is not supported.
  242.          */
  243.         mutex_unlock(&area->lock);
  244.         mutex_unlock(&as->lock);
  245.         interrupts_restore(ipl);
  246.         return ENOTSUP;
  247.     }
  248.     if (area->sh_info) {
  249.         /*
  250.          * Remapping of shared address space areas
  251.          * is not supported.
  252.          */
  253.         mutex_unlock(&area->lock);
  254.         mutex_unlock(&as->lock);
  255.         interrupts_restore(ipl);
  256.         return ENOTSUP;
  257.     }
  258.  
  259.     pages = SIZE2FRAMES((address - area->base) + size);
  260.     if (!pages) {
  261.         /*
  262.          * Zero size address space areas are not allowed.
  263.          */
  264.         mutex_unlock(&area->lock);
  265.         mutex_unlock(&as->lock);
  266.         interrupts_restore(ipl);
  267.         return EPERM;
  268.     }
  269.    
  270.     if (pages < area->pages) {
  271.         bool cond;
  272.         __address start_free = area->base + pages*PAGE_SIZE;
  273.  
  274.         /*
  275.          * Shrinking the area.
  276.          * No need to check for overlaps.
  277.          */
  278.  
  279.         /*
  280.          * Remove frames belonging to used space starting from
  281.          * the highest addresses downwards until an overlap with
  282.          * the resized address space area is found. Note that this
  283.          * is also the right way to remove part of the used_space
  284.          * B+tree leaf list.
  285.          */    
  286.         for (cond = true; cond;) {
  287.             btree_node_t *node;
  288.        
  289.             ASSERT(!list_empty(&area->used_space.leaf_head));
  290.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  291.             if ((cond = (bool) node->keys)) {
  292.                 __address b = node->key[node->keys - 1];
  293.                 count_t c = (count_t) node->value[node->keys - 1];
  294.                 int i = 0;
  295.            
  296.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  297.                    
  298.                     if (b + c*PAGE_SIZE <= start_free) {
  299.                         /*
  300.                          * The whole interval fits completely
  301.                          * in the resized address space area.
  302.                          */
  303.                         break;
  304.                     }
  305.        
  306.                     /*
  307.                      * Part of the interval corresponding to b and c
  308.                      * overlaps with the resized address space area.
  309.                      */
  310.        
  311.                     cond = false;   /* we are almost done */
  312.                     i = (start_free - b) >> PAGE_WIDTH;
  313.                     if (!used_space_remove(area, start_free, c - i))
  314.                         panic("Could not remove used space.");
  315.                 } else {
  316.                     /*
  317.                      * The interval of used space can be completely removed.
  318.                      */
  319.                     if (!used_space_remove(area, b, c))
  320.                         panic("Could not remove used space.\n");
  321.                 }
  322.            
  323.                 for (; i < c; i++) {
  324.                     pte_t *pte;
  325.            
  326.                     page_table_lock(as, false);
  327.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  328.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  329.                     if (area->backend && area->backend->backend_frame_free) {
  330.                         area->backend->backend_frame_free(area,
  331.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  332.                     }
  333.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  334.                     page_table_unlock(as, false);
  335.                 }
  336.             }
  337.         }
  338.         /*
  339.          * Invalidate TLB's.
  340.          */
  341.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  342.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  343.         tlb_shootdown_finalize();
  344.     } else {
  345.         /*
  346.          * Growing the area.
  347.          * Check for overlaps with other address space areas.
  348.          */
  349.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  350.             mutex_unlock(&area->lock);
  351.             mutex_unlock(&as->lock);       
  352.             interrupts_restore(ipl);
  353.             return EADDRNOTAVAIL;
  354.         }
  355.     }
  356.  
  357.     area->pages = pages;
  358.    
  359.     mutex_unlock(&area->lock);
  360.     mutex_unlock(&as->lock);
  361.     interrupts_restore(ipl);
  362.  
  363.     return 0;
  364. }
  365.  
  366. /** Destroy address space area.
  367.  *
  368.  * @param as Address space.
  369.  * @param address Address withing the area to be deleted.
  370.  *
  371.  * @return Zero on success or a value from @ref errno.h on failure.
  372.  */
  373. int as_area_destroy(as_t *as, __address address)
  374. {
  375.     as_area_t *area;
  376.     __address base;
  377.     ipl_t ipl;
  378.  
  379.     ipl = interrupts_disable();
  380.     mutex_lock(&as->lock);
  381.  
  382.     area = find_area_and_lock(as, address);
  383.     if (!area) {
  384.         mutex_unlock(&as->lock);
  385.         interrupts_restore(ipl);
  386.         return ENOENT;
  387.     }
  388.  
  389.     base = area->base;
  390.     if (!(area->flags & AS_AREA_DEVICE)) {
  391.         bool cond; 
  392.    
  393.         /*
  394.          * Releasing physical memory.
  395.          * Areas mapping memory-mapped devices are treated differently than
  396.          * areas backing frame_alloc()'ed memory.
  397.          */
  398.  
  399.         /*
  400.          * Visit only the pages mapped by used_space B+tree.
  401.          * Note that we must be very careful when walking the tree
  402.          * leaf list and removing used space as the leaf list changes
  403.          * unpredictibly after each remove. The solution is to actually
  404.          * not walk the tree at all, but to remove items from the head
  405.          * of the leaf list until there are some keys left.
  406.          */
  407.         for (cond = true; cond;) {
  408.             btree_node_t *node;
  409.        
  410.             ASSERT(!list_empty(&area->used_space.leaf_head));
  411.             node = list_get_instance(area->used_space.leaf_head.next, btree_node_t, leaf_link);
  412.             if ((cond = (bool) node->keys)) {
  413.                 __address b = node->key[0];
  414.                 count_t i;
  415.                 pte_t *pte;
  416.            
  417.                 for (i = 0; i < (count_t) node->value[0]; i++) {
  418.                     page_table_lock(as, false);
  419.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  420.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  421.                     if (area->backend && area->backend->backend_frame_free) {
  422.                         area->backend->backend_frame_free(area,
  423.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  424.                     }
  425.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  426.                     page_table_unlock(as, false);
  427.                 }
  428.                 if (!used_space_remove(area, b, i))
  429.                     panic("Could not remove used space.\n");
  430.             }
  431.         }
  432.     }
  433.     btree_destroy(&area->used_space);
  434.  
  435.     /*
  436.      * Invalidate TLB's.
  437.      */
  438.     tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
  439.     tlb_invalidate_pages(AS->asid, area->base, area->pages);
  440.     tlb_shootdown_finalize();
  441.  
  442.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  443.    
  444.     if (area->sh_info)
  445.         sh_info_remove_reference(area->sh_info);
  446.        
  447.     mutex_unlock(&area->lock);
  448.  
  449.     /*
  450.      * Remove the empty area from address space.
  451.      */
  452.     btree_remove(&AS->as_area_btree, base, NULL);
  453.    
  454.     free(area);
  455.    
  456.     mutex_unlock(&AS->lock);
  457.     interrupts_restore(ipl);
  458.     return 0;
  459. }
  460.  
  461. /** Steal address space area from another task.
  462.  *
  463.  * Address space area is stolen from another task
  464.  * Moreover, any existing mapping
  465.  * is copied as well, providing thus a mechanism
  466.  * for sharing group of pages. The source address
  467.  * space area and any associated mapping is preserved.
  468.  *
  469.  * @param src_task Pointer of source task
  470.  * @param src_base Base address of the source address space area.
  471.  * @param acc_size Expected size of the source area
  472.  * @param dst_base Target base address
  473.  *
  474.  * @return Zero on success or ENOENT if there is no such task or
  475.  *     if there is no such address space area,
  476.  *     EPERM if there was a problem in accepting the area or
  477.  *     ENOMEM if there was a problem in allocating destination
  478.  *     address space area.
  479.  */
  480. int as_area_steal(task_t *src_task, __address src_base, size_t acc_size,
  481.           __address dst_base)
  482. {
  483.     ipl_t ipl;
  484.     count_t i;
  485.     as_t *src_as;      
  486.     int src_flags;
  487.     size_t src_size;
  488.     as_area_t *src_area, *dst_area;
  489.  
  490.     ipl = interrupts_disable();
  491.     spinlock_lock(&src_task->lock);
  492.     src_as = src_task->as;
  493.    
  494.     mutex_lock(&src_as->lock);
  495.     src_area = find_area_and_lock(src_as, src_base);
  496.     if (!src_area) {
  497.         /*
  498.          * Could not find the source address space area.
  499.          */
  500.         spinlock_unlock(&src_task->lock);
  501.         mutex_unlock(&src_as->lock);
  502.         interrupts_restore(ipl);
  503.         return ENOENT;
  504.     }
  505.     src_size = src_area->pages * PAGE_SIZE;
  506.     src_flags = src_area->flags;
  507.     mutex_unlock(&src_area->lock);
  508.     mutex_unlock(&src_as->lock);
  509.  
  510.     if (src_size != acc_size) {
  511.         spinlock_unlock(&src_task->lock);
  512.         interrupts_restore(ipl);
  513.         return EPERM;
  514.     }
  515.     /*
  516.      * Create copy of the source address space area.
  517.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  518.      * attribute set which prevents race condition with
  519.      * preliminary as_page_fault() calls.
  520.      */
  521.     dst_area = as_area_create(AS, src_flags, src_size, dst_base, AS_AREA_ATTR_PARTIAL, &anon_backend, NULL);
  522.     if (!dst_area) {
  523.         /*
  524.          * Destination address space area could not be created.
  525.          */
  526.         spinlock_unlock(&src_task->lock);
  527.         interrupts_restore(ipl);
  528.         return ENOMEM;
  529.     }
  530.    
  531.     spinlock_unlock(&src_task->lock);
  532.    
  533.     /*
  534.      * Avoid deadlock by first locking the address space with lower address.
  535.      */
  536.     if (AS < src_as) {
  537.         mutex_lock(&AS->lock);
  538.         mutex_lock(&src_as->lock);
  539.     } else {
  540.         mutex_lock(&AS->lock);
  541.         mutex_lock(&src_as->lock);
  542.     }
  543.    
  544.     for (i = 0; i < SIZE2FRAMES(src_size); i++) {
  545.         pte_t *pte;
  546.         __address frame;
  547.            
  548.         page_table_lock(src_as, false);
  549.         pte = page_mapping_find(src_as, src_base + i*PAGE_SIZE);
  550.         if (pte && PTE_VALID(pte)) {
  551.             ASSERT(PTE_PRESENT(pte));
  552.             frame = PTE_GET_FRAME(pte);
  553.             if (!(src_flags & AS_AREA_DEVICE))
  554.                 frame_reference_add(ADDR2PFN(frame));
  555.             page_table_unlock(src_as, false);
  556.         } else {
  557.             page_table_unlock(src_as, false);
  558.             continue;
  559.         }
  560.        
  561.         page_table_lock(AS, false);
  562.         page_mapping_insert(AS, dst_base + i*PAGE_SIZE, frame, area_flags_to_page_flags(src_flags));
  563.         page_table_unlock(AS, false);
  564.     }
  565.  
  566.     /*
  567.      * Now the destination address space area has been
  568.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  569.      * attribute.
  570.      */
  571.     mutex_lock(&dst_area->lock);
  572.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  573.     mutex_unlock(&dst_area->lock);
  574.    
  575.     mutex_unlock(&AS->lock);
  576.     mutex_unlock(&src_as->lock);
  577.     interrupts_restore(ipl);
  578.    
  579.     return 0;
  580. }
  581.  
  582. /** Initialize mapping for one page of address space.
  583.  *
  584.  * This functions maps 'page' to 'frame' according
  585.  * to attributes of the address space area to
  586.  * wich 'page' belongs.
  587.  *
  588.  * @param as Target address space.
  589.  * @param page Virtual page within the area.
  590.  * @param frame Physical frame to which page will be mapped.
  591.  */
  592. void as_set_mapping(as_t *as, __address page, __address frame)
  593. {
  594.     as_area_t *area;
  595.     ipl_t ipl;
  596.    
  597.     ipl = interrupts_disable();
  598.     page_table_lock(as, true);
  599.    
  600.     area = find_area_and_lock(as, page);
  601.     if (!area) {
  602.         panic("Page not part of any as_area.\n");
  603.     }
  604.  
  605.     ASSERT(!area->backend);
  606.    
  607.     page_mapping_insert(as, page, frame, as_area_get_flags(area));
  608.     if (!used_space_insert(area, page, 1))
  609.         panic("Could not insert used space.\n");
  610.    
  611.     mutex_unlock(&area->lock);
  612.     page_table_unlock(as, true);
  613.     interrupts_restore(ipl);
  614. }
  615.  
  616. /** Handle page fault within the current address space.
  617.  *
  618.  * This is the high-level page fault handler. It decides
  619.  * whether the page fault can be resolved by any backend
  620.  * and if so, it invokes the backend to resolve the page
  621.  * fault.
  622.  *
  623.  * Interrupts are assumed disabled.
  624.  *
  625.  * @param page Faulting page.
  626.  * @param istate Pointer to interrupted state.
  627.  *
  628.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  629.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  630.  */
  631. int as_page_fault(__address page, istate_t *istate)
  632. {
  633.     pte_t *pte;
  634.     as_area_t *area;
  635.    
  636.     if (!THREAD)
  637.         return AS_PF_FAULT;
  638.        
  639.     ASSERT(AS);
  640.  
  641.     mutex_lock(&AS->lock);
  642.     area = find_area_and_lock(AS, page);   
  643.     if (!area) {
  644.         /*
  645.          * No area contained mapping for 'page'.
  646.          * Signal page fault to low-level handler.
  647.          */
  648.         mutex_unlock(&AS->lock);
  649.         goto page_fault;
  650.     }
  651.  
  652.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  653.         /*
  654.          * The address space area is not fully initialized.
  655.          * Avoid possible race by returning error.
  656.          */
  657.         mutex_unlock(&area->lock);
  658.         mutex_unlock(&AS->lock);
  659.         goto page_fault;       
  660.     }
  661.  
  662.     if (!area->backend || !area->backend->backend_page_fault) {
  663.         /*
  664.          * The address space area is not backed by any backend
  665.          * or the backend cannot handle page faults.
  666.          */
  667.         mutex_unlock(&area->lock);
  668.         mutex_unlock(&AS->lock);
  669.         goto page_fault;       
  670.     }
  671.  
  672.     page_table_lock(AS, false);
  673.    
  674.     /*
  675.      * To avoid race condition between two page faults
  676.      * on the same address, we need to make sure
  677.      * the mapping has not been already inserted.
  678.      */
  679.     if ((pte = page_mapping_find(AS, page))) {
  680.         if (PTE_PRESENT(pte)) {
  681.             page_table_unlock(AS, false);
  682.             mutex_unlock(&area->lock);
  683.             mutex_unlock(&AS->lock);
  684.             return AS_PF_OK;
  685.         }
  686.     }
  687.    
  688.     /*
  689.      * Resort to the backend page fault handler.
  690.      */
  691.     if (area->backend->backend_page_fault(area, page) != AS_PF_OK) {
  692.         page_table_unlock(AS, false);
  693.         mutex_unlock(&area->lock);
  694.         mutex_unlock(&AS->lock);
  695.         goto page_fault;
  696.     }
  697.    
  698.     page_table_unlock(AS, false);
  699.     mutex_unlock(&area->lock);
  700.     mutex_unlock(&AS->lock);
  701.     return AS_PF_OK;
  702.  
  703. page_fault:
  704.     if (THREAD->in_copy_from_uspace) {
  705.         THREAD->in_copy_from_uspace = false;
  706.         istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
  707.     } else if (THREAD->in_copy_to_uspace) {
  708.         THREAD->in_copy_to_uspace = false;
  709.         istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
  710.     } else {
  711.         return AS_PF_FAULT;
  712.     }
  713.  
  714.     return AS_PF_DEFER;
  715. }
  716.  
  717. /** Switch address spaces.
  718.  *
  719.  * Note that this function cannot sleep as it is essentially a part of
  720.  * the scheduling. Sleeping here would lead to deadlock on wakeup.
  721.  *
  722.  * @param old Old address space or NULL.
  723.  * @param new New address space.
  724.  */
  725. void as_switch(as_t *old, as_t *new)
  726. {
  727.     ipl_t ipl;
  728.     bool needs_asid = false;
  729.    
  730.     ipl = interrupts_disable();
  731.     spinlock_lock(&as_lock);
  732.  
  733.     /*
  734.      * First, take care of the old address space.
  735.      */
  736.     if (old) {
  737.         mutex_lock_active(&old->lock);
  738.         ASSERT(old->refcount);
  739.         if((--old->refcount == 0) && (old != AS_KERNEL)) {
  740.             /*
  741.              * The old address space is no longer active on
  742.              * any processor. It can be appended to the
  743.              * list of inactive address spaces with assigned
  744.              * ASID.
  745.              */
  746.              ASSERT(old->asid != ASID_INVALID);
  747.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  748.         }
  749.         mutex_unlock(&old->lock);
  750.     }
  751.  
  752.     /*
  753.      * Second, prepare the new address space.
  754.      */
  755.     mutex_lock_active(&new->lock);
  756.     if ((new->refcount++ == 0) && (new != AS_KERNEL)) {
  757.         if (new->asid != ASID_INVALID)
  758.             list_remove(&new->inactive_as_with_asid_link);
  759.         else
  760.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  761.     }
  762.     SET_PTL0_ADDRESS(new->page_table);
  763.     mutex_unlock(&new->lock);
  764.  
  765.     if (needs_asid) {
  766.         /*
  767.          * Allocation of new ASID was deferred
  768.          * until now in order to avoid deadlock.
  769.          */
  770.         asid_t asid;
  771.        
  772.         asid = asid_get();
  773.         mutex_lock_active(&new->lock);
  774.         new->asid = asid;
  775.         mutex_unlock(&new->lock);
  776.     }
  777.     spinlock_unlock(&as_lock);
  778.     interrupts_restore(ipl);
  779.    
  780.     /*
  781.      * Perform architecture-specific steps.
  782.      * (e.g. write ASID to hardware register etc.)
  783.      */
  784.     as_install_arch(new);
  785.    
  786.     AS = new;
  787. }
  788.  
  789. /** Convert address space area flags to page flags.
  790.  *
  791.  * @param aflags Flags of some address space area.
  792.  *
  793.  * @return Flags to be passed to page_mapping_insert().
  794.  */
  795. int area_flags_to_page_flags(int aflags)
  796. {
  797.     int flags;
  798.  
  799.     flags = PAGE_USER | PAGE_PRESENT;
  800.    
  801.     if (aflags & AS_AREA_READ)
  802.         flags |= PAGE_READ;
  803.        
  804.     if (aflags & AS_AREA_WRITE)
  805.         flags |= PAGE_WRITE;
  806.    
  807.     if (aflags & AS_AREA_EXEC)
  808.         flags |= PAGE_EXEC;
  809.    
  810.     if (!(aflags & AS_AREA_DEVICE))
  811.         flags |= PAGE_CACHEABLE;
  812.        
  813.     return flags;
  814. }
  815.  
  816. /** Compute flags for virtual address translation subsytem.
  817.  *
  818.  * The address space area must be locked.
  819.  * Interrupts must be disabled.
  820.  *
  821.  * @param a Address space area.
  822.  *
  823.  * @return Flags to be used in page_mapping_insert().
  824.  */
  825. int as_area_get_flags(as_area_t *a)
  826. {
  827.     return area_flags_to_page_flags(a->flags);
  828. }
  829.  
  830. /** Create page table.
  831.  *
  832.  * Depending on architecture, create either address space
  833.  * private or global page table.
  834.  *
  835.  * @param flags Flags saying whether the page table is for kernel address space.
  836.  *
  837.  * @return First entry of the page table.
  838.  */
  839. pte_t *page_table_create(int flags)
  840. {
  841.         ASSERT(as_operations);
  842.         ASSERT(as_operations->page_table_create);
  843.  
  844.         return as_operations->page_table_create(flags);
  845. }
  846.  
  847. /** Lock page table.
  848.  *
  849.  * This function should be called before any page_mapping_insert(),
  850.  * page_mapping_remove() and page_mapping_find().
  851.  *
  852.  * Locking order is such that address space areas must be locked
  853.  * prior to this call. Address space can be locked prior to this
  854.  * call in which case the lock argument is false.
  855.  *
  856.  * @param as Address space.
  857.  * @param lock If false, do not attempt to lock as->lock.
  858.  */
  859. void page_table_lock(as_t *as, bool lock)
  860. {
  861.     ASSERT(as_operations);
  862.     ASSERT(as_operations->page_table_lock);
  863.  
  864.     as_operations->page_table_lock(as, lock);
  865. }
  866.  
  867. /** Unlock page table.
  868.  *
  869.  * @param as Address space.
  870.  * @param unlock If false, do not attempt to unlock as->lock.
  871.  */
  872. void page_table_unlock(as_t *as, bool unlock)
  873. {
  874.     ASSERT(as_operations);
  875.     ASSERT(as_operations->page_table_unlock);
  876.  
  877.     as_operations->page_table_unlock(as, unlock);
  878. }
  879.  
  880.  
  881. /** Find address space area and lock it.
  882.  *
  883.  * The address space must be locked and interrupts must be disabled.
  884.  *
  885.  * @param as Address space.
  886.  * @param va Virtual address.
  887.  *
  888.  * @return Locked address space area containing va on success or NULL on failure.
  889.  */
  890. as_area_t *find_area_and_lock(as_t *as, __address va)
  891. {
  892.     as_area_t *a;
  893.     btree_node_t *leaf, *lnode;
  894.     int i;
  895.    
  896.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  897.     if (a) {
  898.         /* va is the base address of an address space area */
  899.         mutex_lock(&a->lock);
  900.         return a;
  901.     }
  902.    
  903.     /*
  904.      * Search the leaf node and the righmost record of its left neighbour
  905.      * to find out whether this is a miss or va belongs to an address
  906.      * space area found there.
  907.      */
  908.    
  909.     /* First, search the leaf node itself. */
  910.     for (i = 0; i < leaf->keys; i++) {
  911.         a = (as_area_t *) leaf->value[i];
  912.         mutex_lock(&a->lock);
  913.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  914.             return a;
  915.         }
  916.         mutex_unlock(&a->lock);
  917.     }
  918.  
  919.     /*
  920.      * Second, locate the left neighbour and test its last record.
  921.      * Because of its position in the B+tree, it must have base < va.
  922.      */
  923.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  924.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  925.         mutex_lock(&a->lock);
  926.         if (va < a->base + a->pages * PAGE_SIZE) {
  927.             return a;
  928.         }
  929.         mutex_unlock(&a->lock);
  930.     }
  931.  
  932.     return NULL;
  933. }
  934.  
  935. /** Check area conflicts with other areas.
  936.  *
  937.  * The address space must be locked and interrupts must be disabled.
  938.  *
  939.  * @param as Address space.
  940.  * @param va Starting virtual address of the area being tested.
  941.  * @param size Size of the area being tested.
  942.  * @param avoid_area Do not touch this area.
  943.  *
  944.  * @return True if there is no conflict, false otherwise.
  945.  */
  946. bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area)
  947. {
  948.     as_area_t *a;
  949.     btree_node_t *leaf, *node;
  950.     int i;
  951.    
  952.     /*
  953.      * We don't want any area to have conflicts with NULL page.
  954.      */
  955.     if (overlaps(va, size, NULL, PAGE_SIZE))
  956.         return false;
  957.    
  958.     /*
  959.      * The leaf node is found in O(log n), where n is proportional to
  960.      * the number of address space areas belonging to as.
  961.      * The check for conflicts is then attempted on the rightmost
  962.      * record in the left neighbour, the leftmost record in the right
  963.      * neighbour and all records in the leaf node itself.
  964.      */
  965.    
  966.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  967.         if (a != avoid_area)
  968.             return false;
  969.     }
  970.    
  971.     /* First, check the two border cases. */
  972.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  973.         a = (as_area_t *) node->value[node->keys - 1];
  974.         mutex_lock(&a->lock);
  975.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  976.             mutex_unlock(&a->lock);
  977.             return false;
  978.         }
  979.         mutex_unlock(&a->lock);
  980.     }
  981.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  982.         a = (as_area_t *) node->value[0];
  983.         mutex_lock(&a->lock);
  984.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  985.             mutex_unlock(&a->lock);
  986.             return false;
  987.         }
  988.         mutex_unlock(&a->lock);
  989.     }
  990.    
  991.     /* Second, check the leaf node. */
  992.     for (i = 0; i < leaf->keys; i++) {
  993.         a = (as_area_t *) leaf->value[i];
  994.    
  995.         if (a == avoid_area)
  996.             continue;
  997.    
  998.         mutex_lock(&a->lock);
  999.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1000.             mutex_unlock(&a->lock);
  1001.             return false;
  1002.         }
  1003.         mutex_unlock(&a->lock);
  1004.     }
  1005.  
  1006.     /*
  1007.      * So far, the area does not conflict with other areas.
  1008.      * Check if it doesn't conflict with kernel address space.
  1009.      */  
  1010.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1011.         return !overlaps(va, size,
  1012.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1013.     }
  1014.  
  1015.     return true;
  1016. }
  1017.  
  1018. /** Return size of the address space area with given base.  */
  1019. size_t as_get_size(__address base)
  1020. {
  1021.     ipl_t ipl;
  1022.     as_area_t *src_area;
  1023.     size_t size;
  1024.  
  1025.     ipl = interrupts_disable();
  1026.     src_area = find_area_and_lock(AS, base);
  1027.     if (src_area){
  1028.         size = src_area->pages * PAGE_SIZE;
  1029.         mutex_unlock(&src_area->lock);
  1030.     } else {
  1031.         size = 0;
  1032.     }
  1033.     interrupts_restore(ipl);
  1034.     return size;
  1035. }
  1036.  
  1037. /** Mark portion of address space area as used.
  1038.  *
  1039.  * The address space area must be already locked.
  1040.  *
  1041.  * @param a Address space area.
  1042.  * @param page First page to be marked.
  1043.  * @param count Number of page to be marked.
  1044.  *
  1045.  * @return 0 on failure and 1 on success.
  1046.  */
  1047. int used_space_insert(as_area_t *a, __address page, count_t count)
  1048. {
  1049.     btree_node_t *leaf, *node;
  1050.     count_t pages;
  1051.     int i;
  1052.  
  1053.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1054.     ASSERT(count);
  1055.  
  1056.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1057.     if (pages) {
  1058.         /*
  1059.          * We hit the beginning of some used space.
  1060.          */
  1061.         return 0;
  1062.     }
  1063.  
  1064.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1065.     if (node) {
  1066.         __address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1067.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1068.        
  1069.         /*
  1070.          * Examine the possibility that the interval fits
  1071.          * somewhere between the rightmost interval of
  1072.          * the left neigbour and the first interval of the leaf.
  1073.          */
  1074.          
  1075.         if (page >= right_pg) {
  1076.             /* Do nothing. */
  1077.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1078.             /* The interval intersects with the left interval. */
  1079.             return 0;
  1080.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1081.             /* The interval intersects with the right interval. */
  1082.             return 0;          
  1083.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1084.             /* The interval can be added by merging the two already present intervals. */
  1085.             node->value[node->keys - 1] += count + right_cnt;
  1086.             btree_remove(&a->used_space, right_pg, leaf);
  1087.             return 1;
  1088.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1089.             /* The interval can be added by simply growing the left interval. */
  1090.             node->value[node->keys - 1] += count;
  1091.             return 1;
  1092.         } else if (page + count*PAGE_SIZE == right_pg) {
  1093.             /*
  1094.              * The interval can be addded by simply moving base of the right
  1095.              * interval down and increasing its size accordingly.
  1096.              */
  1097.             leaf->value[0] += count;
  1098.             leaf->key[0] = page;
  1099.             return 1;
  1100.         } else {
  1101.             /*
  1102.              * The interval is between both neigbouring intervals,
  1103.              * but cannot be merged with any of them.
  1104.              */
  1105.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1106.             return 1;
  1107.         }
  1108.     } else if (page < leaf->key[0]) {
  1109.         __address right_pg = leaf->key[0];
  1110.         count_t right_cnt = (count_t) leaf->value[0];
  1111.    
  1112.         /*
  1113.          * Investigate the border case in which the left neighbour does not
  1114.          * exist but the interval fits from the left.
  1115.          */
  1116.          
  1117.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1118.             /* The interval intersects with the right interval. */
  1119.             return 0;
  1120.         } else if (page + count*PAGE_SIZE == right_pg) {
  1121.             /*
  1122.              * The interval can be added by moving the base of the right interval down
  1123.              * and increasing its size accordingly.
  1124.              */
  1125.             leaf->key[0] = page;
  1126.             leaf->value[0] += count;
  1127.             return 1;
  1128.         } else {
  1129.             /*
  1130.              * The interval doesn't adjoin with the right interval.
  1131.              * It must be added individually.
  1132.              */
  1133.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1134.             return 1;
  1135.         }
  1136.     }
  1137.  
  1138.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1139.     if (node) {
  1140.         __address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1141.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1142.        
  1143.         /*
  1144.          * Examine the possibility that the interval fits
  1145.          * somewhere between the leftmost interval of
  1146.          * the right neigbour and the last interval of the leaf.
  1147.          */
  1148.  
  1149.         if (page < left_pg) {
  1150.             /* Do nothing. */
  1151.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1152.             /* The interval intersects with the left interval. */
  1153.             return 0;
  1154.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1155.             /* The interval intersects with the right interval. */
  1156.             return 0;          
  1157.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1158.             /* The interval can be added by merging the two already present intervals. */
  1159.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1160.             btree_remove(&a->used_space, right_pg, node);
  1161.             return 1;
  1162.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1163.             /* The interval can be added by simply growing the left interval. */
  1164.             leaf->value[leaf->keys - 1] +=  count;
  1165.             return 1;
  1166.         } else if (page + count*PAGE_SIZE == right_pg) {
  1167.             /*
  1168.              * The interval can be addded by simply moving base of the right
  1169.              * interval down and increasing its size accordingly.
  1170.              */
  1171.             node->value[0] += count;
  1172.             node->key[0] = page;
  1173.             return 1;
  1174.         } else {
  1175.             /*
  1176.              * The interval is between both neigbouring intervals,
  1177.              * but cannot be merged with any of them.
  1178.              */
  1179.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1180.             return 1;
  1181.         }
  1182.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1183.         __address left_pg = leaf->key[leaf->keys - 1];
  1184.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1185.    
  1186.         /*
  1187.          * Investigate the border case in which the right neighbour does not
  1188.          * exist but the interval fits from the right.
  1189.          */
  1190.          
  1191.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1192.             /* The interval intersects with the left interval. */
  1193.             return 0;
  1194.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1195.             /* The interval can be added by growing the left interval. */
  1196.             leaf->value[leaf->keys - 1] += count;
  1197.             return 1;
  1198.         } else {
  1199.             /*
  1200.              * The interval doesn't adjoin with the left interval.
  1201.              * It must be added individually.
  1202.              */
  1203.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1204.             return 1;
  1205.         }
  1206.     }
  1207.    
  1208.     /*
  1209.      * Note that if the algorithm made it thus far, the interval can fit only
  1210.      * between two other intervals of the leaf. The two border cases were already
  1211.      * resolved.
  1212.      */
  1213.     for (i = 1; i < leaf->keys; i++) {
  1214.         if (page < leaf->key[i]) {
  1215.             __address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1216.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1217.  
  1218.             /*
  1219.              * The interval fits between left_pg and right_pg.
  1220.              */
  1221.  
  1222.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1223.                 /* The interval intersects with the left interval. */
  1224.                 return 0;
  1225.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1226.                 /* The interval intersects with the right interval. */
  1227.                 return 0;          
  1228.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1229.                 /* The interval can be added by merging the two already present intervals. */
  1230.                 leaf->value[i - 1] += count + right_cnt;
  1231.                 btree_remove(&a->used_space, right_pg, leaf);
  1232.                 return 1;
  1233.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1234.                 /* The interval can be added by simply growing the left interval. */
  1235.                 leaf->value[i - 1] += count;
  1236.                 return 1;
  1237.             } else if (page + count*PAGE_SIZE == right_pg) {
  1238.                 /*
  1239.                      * The interval can be addded by simply moving base of the right
  1240.                  * interval down and increasing its size accordingly.
  1241.                  */
  1242.                 leaf->value[i] += count;
  1243.                 leaf->key[i] = page;
  1244.                 return 1;
  1245.             } else {
  1246.                 /*
  1247.                  * The interval is between both neigbouring intervals,
  1248.                  * but cannot be merged with any of them.
  1249.                  */
  1250.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1251.                 return 1;
  1252.             }
  1253.         }
  1254.     }
  1255.  
  1256.     panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
  1257. }
  1258.  
  1259. /** Mark portion of address space area as unused.
  1260.  *
  1261.  * The address space area must be already locked.
  1262.  *
  1263.  * @param a Address space area.
  1264.  * @param page First page to be marked.
  1265.  * @param count Number of page to be marked.
  1266.  *
  1267.  * @return 0 on failure and 1 on success.
  1268.  */
  1269. int used_space_remove(as_area_t *a, __address page, count_t count)
  1270. {
  1271.     btree_node_t *leaf, *node;
  1272.     count_t pages;
  1273.     int i;
  1274.  
  1275.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1276.     ASSERT(count);
  1277.  
  1278.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1279.     if (pages) {
  1280.         /*
  1281.          * We are lucky, page is the beginning of some interval.
  1282.          */
  1283.         if (count > pages) {
  1284.             return 0;
  1285.         } else if (count == pages) {
  1286.             btree_remove(&a->used_space, page, leaf);
  1287.             return 1;
  1288.         } else {
  1289.             /*
  1290.              * Find the respective interval.
  1291.              * Decrease its size and relocate its start address.
  1292.              */
  1293.             for (i = 0; i < leaf->keys; i++) {
  1294.                 if (leaf->key[i] == page) {
  1295.                     leaf->key[i] += count*PAGE_SIZE;
  1296.                     leaf->value[i] -= count;
  1297.                     return 1;
  1298.                 }
  1299.             }
  1300.             goto error;
  1301.         }
  1302.     }
  1303.  
  1304.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1305.     if (node && page < leaf->key[0]) {
  1306.         __address left_pg = node->key[node->keys - 1];
  1307.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1308.  
  1309.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1310.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1311.                 /*
  1312.                  * The interval is contained in the rightmost interval
  1313.                  * of the left neighbour and can be removed by
  1314.                  * updating the size of the bigger interval.
  1315.                  */
  1316.                 node->value[node->keys - 1] -= count;
  1317.                 return 1;
  1318.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1319.                 count_t new_cnt;
  1320.                
  1321.                 /*
  1322.                  * The interval is contained in the rightmost interval
  1323.                  * of the left neighbour but its removal requires
  1324.                  * both updating the size of the original interval and
  1325.                  * also inserting a new interval.
  1326.                  */
  1327.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1328.                 node->value[node->keys - 1] -= count + new_cnt;
  1329.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1330.                 return 1;
  1331.             }
  1332.         }
  1333.         return 0;
  1334.     } else if (page < leaf->key[0]) {
  1335.         return 0;
  1336.     }
  1337.    
  1338.     if (page > leaf->key[leaf->keys - 1]) {
  1339.         __address left_pg = leaf->key[leaf->keys - 1];
  1340.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1341.  
  1342.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1343.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1344.                 /*
  1345.                  * The interval is contained in the rightmost interval
  1346.                  * of the leaf and can be removed by updating the size
  1347.                  * of the bigger interval.
  1348.                  */
  1349.                 leaf->value[leaf->keys - 1] -= count;
  1350.                 return 1;
  1351.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1352.                 count_t new_cnt;
  1353.                
  1354.                 /*
  1355.                  * The interval is contained in the rightmost interval
  1356.                  * of the leaf but its removal requires both updating
  1357.                  * the size of the original interval and
  1358.                  * also inserting a new interval.
  1359.                  */
  1360.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1361.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1362.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1363.                 return 1;
  1364.             }
  1365.         }
  1366.         return 0;
  1367.     }  
  1368.    
  1369.     /*
  1370.      * The border cases have been already resolved.
  1371.      * Now the interval can be only between intervals of the leaf.
  1372.      */
  1373.     for (i = 1; i < leaf->keys - 1; i++) {
  1374.         if (page < leaf->key[i]) {
  1375.             __address left_pg = leaf->key[i - 1];
  1376.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1377.  
  1378.             /*
  1379.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1380.              */
  1381.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1382.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1383.                     /*
  1384.                     * The interval is contained in the interval (i - 1)
  1385.                      * of the leaf and can be removed by updating the size
  1386.                      * of the bigger interval.
  1387.                      */
  1388.                     leaf->value[i - 1] -= count;
  1389.                     return 1;
  1390.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1391.                     count_t new_cnt;
  1392.                
  1393.                     /*
  1394.                      * The interval is contained in the interval (i - 1)
  1395.                      * of the leaf but its removal requires both updating
  1396.                      * the size of the original interval and
  1397.                      * also inserting a new interval.
  1398.                      */
  1399.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1400.                     leaf->value[i - 1] -= count + new_cnt;
  1401.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1402.                     return 1;
  1403.                 }
  1404.             }
  1405.             return 0;
  1406.         }
  1407.     }
  1408.  
  1409. error:
  1410.     panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
  1411. }
  1412.  
  1413. /** Remove reference to address space area share info.
  1414.  *
  1415.  * If the reference count drops to 0, the sh_info is deallocated.
  1416.  *
  1417.  * @param sh_info Pointer to address space area share info.
  1418.  */
  1419. void sh_info_remove_reference(share_info_t *sh_info)
  1420. {
  1421.     bool dealloc = false;
  1422.  
  1423.     mutex_lock(&sh_info->lock);
  1424.     ASSERT(sh_info->refcount);
  1425.     if (--sh_info->refcount == 0) {
  1426.         dealloc = true;
  1427.         bool cond;
  1428.        
  1429.         /*
  1430.          * Now walk carefully the pagemap B+tree and free/remove
  1431.          * reference from all frames found there.
  1432.          */
  1433.         for (cond = true; cond;) {
  1434.             btree_node_t *node;
  1435.            
  1436.             ASSERT(!list_empty(&sh_info->pagemap.leaf_head));
  1437.             node = list_get_instance(sh_info->pagemap.leaf_head.next, btree_node_t, leaf_link);
  1438.             if ((cond = node->keys)) {
  1439.                 frame_free(ADDR2PFN((__address) node->value[0]));
  1440.                 btree_remove(&sh_info->pagemap, node->key[0], node);
  1441.             }
  1442.         }
  1443.        
  1444.     }
  1445.     mutex_unlock(&sh_info->lock);
  1446.    
  1447.     if (dealloc) {
  1448.         btree_destroy(&sh_info->pagemap);
  1449.         free(sh_info);
  1450.     }
  1451. }
  1452.  
  1453. static int anon_page_fault(as_area_t *area, __address addr);
  1454. static void anon_frame_free(as_area_t *area, __address page, __address frame);
  1455.  
  1456. /*
  1457.  * Anonymous memory backend.
  1458.  */
  1459. mem_backend_t anon_backend = {
  1460.     .backend_page_fault = anon_page_fault,
  1461.     .backend_frame_free = anon_frame_free
  1462. };
  1463.  
  1464. /** Service a page fault in the anonymous memory address space area.
  1465.  *
  1466.  * The address space area and page tables must be already locked.
  1467.  *
  1468.  * @param area Pointer to the address space area.
  1469.  * @param addr Faulting virtual address.
  1470.  *
  1471.  * @return AS_PF_FAULT on failure (i.e. page fault) or AS_PF_OK on success (i.e. serviced).
  1472.  */
  1473. int anon_page_fault(as_area_t *area, __address addr)
  1474. {
  1475.     __address frame;
  1476.  
  1477.     if (area->sh_info) {
  1478.         btree_node_t *leaf;
  1479.        
  1480.         /*
  1481.          * The area is shared, chances are that the mapping can be found
  1482.          * in the pagemap of the address space area share info structure.
  1483.          * In the case that the pagemap does not contain the respective
  1484.          * mapping, a new frame is allocated and the mapping is created.
  1485.          */
  1486.         mutex_lock(&area->sh_info->lock);
  1487.         frame = (__address) btree_search(&area->sh_info->pagemap, ALIGN_DOWN(addr, PAGE_SIZE), &leaf);
  1488.         if (!frame) {
  1489.             bool allocate = true;
  1490.             int i;
  1491.            
  1492.             /*
  1493.              * Zero can be returned as a valid frame address.
  1494.              * Just a small workaround.
  1495.              */
  1496.             for (i = 0; i < leaf->keys; i++) {
  1497.                 if (leaf->key[i] == ALIGN_DOWN(addr, PAGE_SIZE)) {
  1498.                     allocate = false;
  1499.                     break;
  1500.                 }
  1501.             }
  1502.             if (allocate) {
  1503.                 frame = PFN2ADDR(frame_alloc(ONE_FRAME, 0));
  1504.                 memsetb(PA2KA(frame), FRAME_SIZE, 0);
  1505.                
  1506.                 /*
  1507.                  * Insert the address of the newly allocated frame to the pagemap.
  1508.                  */
  1509.                 btree_insert(&area->sh_info->pagemap, ALIGN_DOWN(addr, PAGE_SIZE), (void *) frame, leaf);
  1510.             }
  1511.         }
  1512.         mutex_unlock(&area->sh_info->lock);
  1513.     } else {
  1514.  
  1515.         /*
  1516.          * In general, there can be several reasons that
  1517.          * can have caused this fault.
  1518.          *
  1519.          * - non-existent mapping: the area is an anonymous
  1520.          *   area (e.g. heap or stack) and so far has not been
  1521.          *   allocated a frame for the faulting page
  1522.          *
  1523.          * - non-present mapping: another possibility,
  1524.          *   currently not implemented, would be frame
  1525.          *   reuse; when this becomes a possibility,
  1526.          *   do not forget to distinguish between
  1527.          *   the different causes
  1528.          */
  1529.         frame = PFN2ADDR(frame_alloc(ONE_FRAME, 0));
  1530.         memsetb(PA2KA(frame), FRAME_SIZE, 0);
  1531.     }
  1532.    
  1533.     /*
  1534.      * Map 'page' to 'frame'.
  1535.      * Note that TLB shootdown is not attempted as only new information is being
  1536.      * inserted into page tables.
  1537.      */
  1538.     page_mapping_insert(AS, addr, frame, as_area_get_flags(area));
  1539.     if (!used_space_insert(area, ALIGN_DOWN(addr, PAGE_SIZE), 1))
  1540.         panic("Could not insert used space.\n");
  1541.        
  1542.     return AS_PF_OK;
  1543. }
  1544.  
  1545. /** Free a frame that is backed by the anonymous memory backend.
  1546.  *
  1547.  * The address space area and page tables must be already locked.
  1548.  *
  1549.  * @param area Ignored.
  1550.  * @param page Ignored.
  1551.  * @param frame Frame to be released.
  1552.  */
  1553. void anon_frame_free(as_area_t *area, __address page, __address frame)
  1554. {
  1555.     frame_free(ADDR2PFN(frame));
  1556. }
  1557.  
  1558. /*
  1559.  * Address space related syscalls.
  1560.  */
  1561.  
  1562. /** Wrapper for as_area_create(). */
  1563. __native sys_as_area_create(__address address, size_t size, int flags)
  1564. {
  1565.     if (as_area_create(AS, flags, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1566.         return (__native) address;
  1567.     else
  1568.         return (__native) -1;
  1569. }
  1570.  
  1571. /** Wrapper for as_area_resize. */
  1572. __native sys_as_area_resize(__address address, size_t size, int flags)
  1573. {
  1574.     return (__native) as_area_resize(AS, address, size, 0);
  1575. }
  1576.  
  1577. /** Wrapper for as_area_destroy. */
  1578. __native sys_as_area_destroy(__address address)
  1579. {
  1580.     return (__native) as_area_destroy(AS, address);
  1581. }
  1582.