Subversion Repositories HelenOS-historic

Rev

Rev 1413 | Rev 1423 | 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. /** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
  87. SPINLOCK_INITIALIZE(inactive_as_with_asid_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->cpu_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->cpu_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.     bool cond;
  379.  
  380.     ipl = interrupts_disable();
  381.     mutex_lock(&as->lock);
  382.  
  383.     area = find_area_and_lock(as, address);
  384.     if (!area) {
  385.         mutex_unlock(&as->lock);
  386.         interrupts_restore(ipl);
  387.         return ENOENT;
  388.     }
  389.  
  390.     base = area->base;
  391.  
  392.     /*
  393.      * Visit only the pages mapped by used_space B+tree.
  394.      * Note that we must be very careful when walking the tree
  395.      * leaf list and removing used space as the leaf list changes
  396.      * unpredictibly after each remove. The solution is to actually
  397.      * not walk the tree at all, but to remove items from the head
  398.      * of the leaf list until there are some keys left.
  399.      */
  400.     for (cond = true; cond;) {
  401.         btree_node_t *node;
  402.        
  403.         ASSERT(!list_empty(&area->used_space.leaf_head));
  404.         node = list_get_instance(area->used_space.leaf_head.next, btree_node_t, leaf_link);
  405.         if ((cond = (bool) node->keys)) {
  406.             __address b = node->key[0];
  407.             count_t i;
  408.             pte_t *pte;
  409.            
  410.             for (i = 0; i < (count_t) node->value[0]; i++) {
  411.                 page_table_lock(as, false);
  412.                 pte = page_mapping_find(as, b + i*PAGE_SIZE);
  413.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  414.                 if (area->backend && area->backend->backend_frame_free) {
  415.                     area->backend->backend_frame_free(area,
  416.                         b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  417.                 }
  418.                 page_mapping_remove(as, b + i*PAGE_SIZE);
  419.                 page_table_unlock(as, false);
  420.             }
  421.             if (!used_space_remove(area, b, i))
  422.                 panic("Could not remove used space.\n");
  423.         }
  424.     }
  425.     btree_destroy(&area->used_space);
  426.  
  427.     /*
  428.      * Invalidate TLB's.
  429.      */
  430.     tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
  431.     tlb_invalidate_pages(AS->asid, area->base, area->pages);
  432.     tlb_shootdown_finalize();
  433.  
  434.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  435.    
  436.     if (area->sh_info)
  437.         sh_info_remove_reference(area->sh_info);
  438.        
  439.     mutex_unlock(&area->lock);
  440.  
  441.     /*
  442.      * Remove the empty area from address space.
  443.      */
  444.     btree_remove(&AS->as_area_btree, base, NULL);
  445.    
  446.     free(area);
  447.    
  448.     mutex_unlock(&AS->lock);
  449.     interrupts_restore(ipl);
  450.     return 0;
  451. }
  452.  
  453. /** Share address space area with another or the same address space.
  454.  *
  455.  * Address space area of anonymous memory is shared with a new address
  456.  * space area. If the source address space area has not been shared so
  457.  * far, a new sh_info is created and the original mapping is duplicated
  458.  * in its pagemap B+tree. The new address space are simply gets the
  459.  * sh_info of the source area.
  460.  *
  461.  * @param src_as Pointer to source address space
  462.  * @param src_base Base address of the source address space area.
  463.  * @param acc_size Expected size of the source area
  464.  * @param dst_base Target base address
  465.  *
  466.  * @return Zero on success or ENOENT if there is no such task or
  467.  *     if there is no such address space area,
  468.  *     EPERM if there was a problem in accepting the area or
  469.  *     ENOMEM if there was a problem in allocating destination
  470.  *     address space area. ENOTSUP is returned if an attempt
  471.  *     to share non-anonymous address space area is detected.
  472.  */
  473. int as_area_share(as_t *src_as, __address src_base, size_t acc_size,
  474.           __address dst_base)
  475. {
  476.     ipl_t ipl;
  477.     int src_flags;
  478.     size_t src_size;
  479.     as_area_t *src_area, *dst_area;
  480.     share_info_t *sh_info;
  481.     link_t *cur;
  482.  
  483.     ipl = interrupts_disable();
  484.     mutex_lock(&src_as->lock);
  485.     src_area = find_area_and_lock(src_as, src_base);
  486.     if (!src_area) {
  487.         /*
  488.          * Could not find the source address space area.
  489.          */
  490.         mutex_unlock(&src_as->lock);
  491.         interrupts_restore(ipl);
  492.         return ENOENT;
  493.     }
  494.    
  495.     if (!src_area->backend || src_area->backend != &anon_backend) {
  496.         /*
  497.          * As of now, only anonymous address space areas can be shared.
  498.          */
  499.         mutex_unlock(&src_area->lock);
  500.         mutex_unlock(&src_as->lock);
  501.         interrupts_restore(ipl);
  502.         return ENOTSUP;
  503.     }
  504.    
  505.     src_size = src_area->pages * PAGE_SIZE;
  506.     src_flags = src_area->flags;
  507.    
  508.     if (src_size != acc_size) {
  509.         mutex_unlock(&src_area->lock);
  510.         mutex_unlock(&src_as->lock);
  511.         interrupts_restore(ipl);
  512.         return EPERM;
  513.     }
  514.  
  515.     /*
  516.      * Now we are committed to sharing the area.
  517.      * First prepare the area for sharing.
  518.      * Then it will be safe to unlock it.
  519.      */
  520.     sh_info = src_area->sh_info;
  521.     if (!sh_info) {
  522.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  523.         mutex_initialize(&sh_info->lock);
  524.         sh_info->refcount = 2;
  525.         btree_create(&sh_info->pagemap);
  526.         src_area->sh_info = sh_info;
  527.     } else {
  528.         mutex_lock(&sh_info->lock);
  529.         sh_info->refcount++;
  530.         mutex_unlock(&sh_info->lock);
  531.     }
  532.  
  533.     /*
  534.      * Copy used portions of the area to sh_info's page map.
  535.      */
  536.     mutex_lock(&sh_info->lock);
  537.     for (cur = src_area->used_space.leaf_head.next; cur != &src_area->used_space.leaf_head; cur = cur->next) {
  538.         btree_node_t *node;
  539.         int i;
  540.        
  541.         node = list_get_instance(cur, btree_node_t, leaf_link);
  542.         for (i = 0; i < node->keys; i++) {
  543.             __address base = node->key[i];
  544.             count_t count = (count_t) node->value[i];
  545.             int j;
  546.            
  547.             for (j = 0; j < count; j++) {
  548.                 pte_t *pte;
  549.            
  550.                 page_table_lock(src_as, false);
  551.                 pte = page_mapping_find(src_as, base + j*PAGE_SIZE);
  552.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  553.                 btree_insert(&sh_info->pagemap, (base + j*PAGE_SIZE) - src_area->base,
  554.                     (void *) PTE_GET_FRAME(pte), NULL);
  555.                 page_table_unlock(src_as, false);
  556.             }
  557.                
  558.         }
  559.     }
  560.     mutex_unlock(&sh_info->lock);
  561.  
  562.     mutex_unlock(&src_area->lock);
  563.     mutex_unlock(&src_as->lock);
  564.  
  565.     /*
  566.      * Create copy of the source address space area.
  567.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  568.      * attribute set which prevents race condition with
  569.      * preliminary as_page_fault() calls.
  570.      */
  571.     dst_area = as_area_create(AS, src_flags, src_size, dst_base, AS_AREA_ATTR_PARTIAL, &anon_backend, NULL);
  572.     if (!dst_area) {
  573.         /*
  574.          * Destination address space area could not be created.
  575.          */
  576.         sh_info_remove_reference(sh_info);
  577.        
  578.         interrupts_restore(ipl);
  579.         return ENOMEM;
  580.     }
  581.    
  582.     /*
  583.      * Now the destination address space area has been
  584.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  585.      * attribute and set the sh_info.
  586.      */
  587.     mutex_lock(&dst_area->lock);
  588.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  589.     dst_area->sh_info = sh_info;
  590.     mutex_unlock(&dst_area->lock);
  591.    
  592.     interrupts_restore(ipl);
  593.    
  594.     return 0;
  595. }
  596.  
  597. /** Initialize mapping for one page of address space.
  598.  *
  599.  * This functions maps 'page' to 'frame' according
  600.  * to attributes of the address space area to
  601.  * wich 'page' belongs.
  602.  *
  603.  * @param as Target address space.
  604.  * @param page Virtual page within the area.
  605.  * @param frame Physical frame to which page will be mapped.
  606.  */
  607. void as_set_mapping(as_t *as, __address page, __address frame)
  608. {
  609.     as_area_t *area;
  610.     ipl_t ipl;
  611.    
  612.     ipl = interrupts_disable();
  613.     page_table_lock(as, true);
  614.    
  615.     area = find_area_and_lock(as, page);
  616.     if (!area) {
  617.         panic("Page not part of any as_area.\n");
  618.     }
  619.  
  620.     ASSERT(!area->backend);
  621.    
  622.     page_mapping_insert(as, page, frame, as_area_get_flags(area));
  623.     if (!used_space_insert(area, page, 1))
  624.         panic("Could not insert used space.\n");
  625.    
  626.     mutex_unlock(&area->lock);
  627.     page_table_unlock(as, true);
  628.     interrupts_restore(ipl);
  629. }
  630.  
  631. /** Handle page fault within the current address space.
  632.  *
  633.  * This is the high-level page fault handler. It decides
  634.  * whether the page fault can be resolved by any backend
  635.  * and if so, it invokes the backend to resolve the page
  636.  * fault.
  637.  *
  638.  * Interrupts are assumed disabled.
  639.  *
  640.  * @param page Faulting page.
  641.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  642.  * @param istate Pointer to interrupted state.
  643.  *
  644.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  645.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  646.  */
  647. int as_page_fault(__address page, pf_access_t access, istate_t *istate)
  648. {
  649.     pte_t *pte;
  650.     as_area_t *area;
  651.    
  652.     if (!THREAD)
  653.         return AS_PF_FAULT;
  654.        
  655.     ASSERT(AS);
  656.  
  657.     mutex_lock(&AS->lock);
  658.     area = find_area_and_lock(AS, page);   
  659.     if (!area) {
  660.         /*
  661.          * No area contained mapping for 'page'.
  662.          * Signal page fault to low-level handler.
  663.          */
  664.         mutex_unlock(&AS->lock);
  665.         goto page_fault;
  666.     }
  667.  
  668.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  669.         /*
  670.          * The address space area is not fully initialized.
  671.          * Avoid possible race by returning error.
  672.          */
  673.         mutex_unlock(&area->lock);
  674.         mutex_unlock(&AS->lock);
  675.         goto page_fault;       
  676.     }
  677.  
  678.     if (!area->backend || !area->backend->backend_page_fault) {
  679.         /*
  680.          * The address space area is not backed by any backend
  681.          * or the backend cannot handle page faults.
  682.          */
  683.         mutex_unlock(&area->lock);
  684.         mutex_unlock(&AS->lock);
  685.         goto page_fault;       
  686.     }
  687.  
  688.     page_table_lock(AS, false);
  689.    
  690.     /*
  691.      * To avoid race condition between two page faults
  692.      * on the same address, we need to make sure
  693.      * the mapping has not been already inserted.
  694.      */
  695.     if ((pte = page_mapping_find(AS, page))) {
  696.         if (PTE_PRESENT(pte)) {
  697.             page_table_unlock(AS, false);
  698.             mutex_unlock(&area->lock);
  699.             mutex_unlock(&AS->lock);
  700.             return AS_PF_OK;
  701.         }
  702.     }
  703.    
  704.     /*
  705.      * Resort to the backend page fault handler.
  706.      */
  707.     if (area->backend->backend_page_fault(area, page, access) != AS_PF_OK) {
  708.         page_table_unlock(AS, false);
  709.         mutex_unlock(&area->lock);
  710.         mutex_unlock(&AS->lock);
  711.         goto page_fault;
  712.     }
  713.    
  714.     page_table_unlock(AS, false);
  715.     mutex_unlock(&area->lock);
  716.     mutex_unlock(&AS->lock);
  717.     return AS_PF_OK;
  718.  
  719. page_fault:
  720.     if (THREAD->in_copy_from_uspace) {
  721.         THREAD->in_copy_from_uspace = false;
  722.         istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
  723.     } else if (THREAD->in_copy_to_uspace) {
  724.         THREAD->in_copy_to_uspace = false;
  725.         istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
  726.     } else {
  727.         return AS_PF_FAULT;
  728.     }
  729.  
  730.     return AS_PF_DEFER;
  731. }
  732.  
  733. /** Switch address spaces.
  734.  *
  735.  * Note that this function cannot sleep as it is essentially a part of
  736.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  737.  *
  738.  * @param old Old address space or NULL.
  739.  * @param new New address space.
  740.  */
  741. void as_switch(as_t *old, as_t *new)
  742. {
  743.     ipl_t ipl;
  744.     bool needs_asid = false;
  745.    
  746.     ipl = interrupts_disable();
  747.     spinlock_lock(&inactive_as_with_asid_lock);
  748.  
  749.     /*
  750.      * First, take care of the old address space.
  751.      */
  752.     if (old) {
  753.         mutex_lock_active(&old->lock);
  754.         ASSERT(old->cpu_refcount);
  755.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  756.             /*
  757.              * The old address space is no longer active on
  758.              * any processor. It can be appended to the
  759.              * list of inactive address spaces with assigned
  760.              * ASID.
  761.              */
  762.              ASSERT(old->asid != ASID_INVALID);
  763.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  764.         }
  765.         mutex_unlock(&old->lock);
  766.     }
  767.  
  768.     /*
  769.      * Second, prepare the new address space.
  770.      */
  771.     mutex_lock_active(&new->lock);
  772.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  773.         if (new->asid != ASID_INVALID)
  774.             list_remove(&new->inactive_as_with_asid_link);
  775.         else
  776.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  777.     }
  778.     SET_PTL0_ADDRESS(new->page_table);
  779.     mutex_unlock(&new->lock);
  780.  
  781.     if (needs_asid) {
  782.         /*
  783.          * Allocation of new ASID was deferred
  784.          * until now in order to avoid deadlock.
  785.          */
  786.         asid_t asid;
  787.        
  788.         asid = asid_get();
  789.         mutex_lock_active(&new->lock);
  790.         new->asid = asid;
  791.         mutex_unlock(&new->lock);
  792.     }
  793.     spinlock_unlock(&inactive_as_with_asid_lock);
  794.     interrupts_restore(ipl);
  795.    
  796.     /*
  797.      * Perform architecture-specific steps.
  798.      * (e.g. write ASID to hardware register etc.)
  799.      */
  800.     as_install_arch(new);
  801.    
  802.     AS = new;
  803. }
  804.  
  805. /** Convert address space area flags to page flags.
  806.  *
  807.  * @param aflags Flags of some address space area.
  808.  *
  809.  * @return Flags to be passed to page_mapping_insert().
  810.  */
  811. int area_flags_to_page_flags(int aflags)
  812. {
  813.     int flags;
  814.  
  815.     flags = PAGE_USER | PAGE_PRESENT;
  816.    
  817.     if (aflags & AS_AREA_READ)
  818.         flags |= PAGE_READ;
  819.        
  820.     if (aflags & AS_AREA_WRITE)
  821.         flags |= PAGE_WRITE;
  822.    
  823.     if (aflags & AS_AREA_EXEC)
  824.         flags |= PAGE_EXEC;
  825.    
  826.     if (!(aflags & AS_AREA_DEVICE))
  827.         flags |= PAGE_CACHEABLE;
  828.        
  829.     return flags;
  830. }
  831.  
  832. /** Compute flags for virtual address translation subsytem.
  833.  *
  834.  * The address space area must be locked.
  835.  * Interrupts must be disabled.
  836.  *
  837.  * @param a Address space area.
  838.  *
  839.  * @return Flags to be used in page_mapping_insert().
  840.  */
  841. int as_area_get_flags(as_area_t *a)
  842. {
  843.     return area_flags_to_page_flags(a->flags);
  844. }
  845.  
  846. /** Create page table.
  847.  *
  848.  * Depending on architecture, create either address space
  849.  * private or global page table.
  850.  *
  851.  * @param flags Flags saying whether the page table is for kernel address space.
  852.  *
  853.  * @return First entry of the page table.
  854.  */
  855. pte_t *page_table_create(int flags)
  856. {
  857.         ASSERT(as_operations);
  858.         ASSERT(as_operations->page_table_create);
  859.  
  860.         return as_operations->page_table_create(flags);
  861. }
  862.  
  863. /** Lock page table.
  864.  *
  865.  * This function should be called before any page_mapping_insert(),
  866.  * page_mapping_remove() and page_mapping_find().
  867.  *
  868.  * Locking order is such that address space areas must be locked
  869.  * prior to this call. Address space can be locked prior to this
  870.  * call in which case the lock argument is false.
  871.  *
  872.  * @param as Address space.
  873.  * @param lock If false, do not attempt to lock as->lock.
  874.  */
  875. void page_table_lock(as_t *as, bool lock)
  876. {
  877.     ASSERT(as_operations);
  878.     ASSERT(as_operations->page_table_lock);
  879.  
  880.     as_operations->page_table_lock(as, lock);
  881. }
  882.  
  883. /** Unlock page table.
  884.  *
  885.  * @param as Address space.
  886.  * @param unlock If false, do not attempt to unlock as->lock.
  887.  */
  888. void page_table_unlock(as_t *as, bool unlock)
  889. {
  890.     ASSERT(as_operations);
  891.     ASSERT(as_operations->page_table_unlock);
  892.  
  893.     as_operations->page_table_unlock(as, unlock);
  894. }
  895.  
  896.  
  897. /** Find address space area and lock it.
  898.  *
  899.  * The address space must be locked and interrupts must be disabled.
  900.  *
  901.  * @param as Address space.
  902.  * @param va Virtual address.
  903.  *
  904.  * @return Locked address space area containing va on success or NULL on failure.
  905.  */
  906. as_area_t *find_area_and_lock(as_t *as, __address va)
  907. {
  908.     as_area_t *a;
  909.     btree_node_t *leaf, *lnode;
  910.     int i;
  911.    
  912.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  913.     if (a) {
  914.         /* va is the base address of an address space area */
  915.         mutex_lock(&a->lock);
  916.         return a;
  917.     }
  918.    
  919.     /*
  920.      * Search the leaf node and the righmost record of its left neighbour
  921.      * to find out whether this is a miss or va belongs to an address
  922.      * space area found there.
  923.      */
  924.    
  925.     /* First, search the leaf node itself. */
  926.     for (i = 0; i < leaf->keys; i++) {
  927.         a = (as_area_t *) leaf->value[i];
  928.         mutex_lock(&a->lock);
  929.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  930.             return a;
  931.         }
  932.         mutex_unlock(&a->lock);
  933.     }
  934.  
  935.     /*
  936.      * Second, locate the left neighbour and test its last record.
  937.      * Because of its position in the B+tree, it must have base < va.
  938.      */
  939.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  940.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  941.         mutex_lock(&a->lock);
  942.         if (va < a->base + a->pages * PAGE_SIZE) {
  943.             return a;
  944.         }
  945.         mutex_unlock(&a->lock);
  946.     }
  947.  
  948.     return NULL;
  949. }
  950.  
  951. /** Check area conflicts with other areas.
  952.  *
  953.  * The address space must be locked and interrupts must be disabled.
  954.  *
  955.  * @param as Address space.
  956.  * @param va Starting virtual address of the area being tested.
  957.  * @param size Size of the area being tested.
  958.  * @param avoid_area Do not touch this area.
  959.  *
  960.  * @return True if there is no conflict, false otherwise.
  961.  */
  962. bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area)
  963. {
  964.     as_area_t *a;
  965.     btree_node_t *leaf, *node;
  966.     int i;
  967.    
  968.     /*
  969.      * We don't want any area to have conflicts with NULL page.
  970.      */
  971.     if (overlaps(va, size, NULL, PAGE_SIZE))
  972.         return false;
  973.    
  974.     /*
  975.      * The leaf node is found in O(log n), where n is proportional to
  976.      * the number of address space areas belonging to as.
  977.      * The check for conflicts is then attempted on the rightmost
  978.      * record in the left neighbour, the leftmost record in the right
  979.      * neighbour and all records in the leaf node itself.
  980.      */
  981.    
  982.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  983.         if (a != avoid_area)
  984.             return false;
  985.     }
  986.    
  987.     /* First, check the two border cases. */
  988.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  989.         a = (as_area_t *) node->value[node->keys - 1];
  990.         mutex_lock(&a->lock);
  991.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  992.             mutex_unlock(&a->lock);
  993.             return false;
  994.         }
  995.         mutex_unlock(&a->lock);
  996.     }
  997.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  998.         a = (as_area_t *) node->value[0];
  999.         mutex_lock(&a->lock);
  1000.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1001.             mutex_unlock(&a->lock);
  1002.             return false;
  1003.         }
  1004.         mutex_unlock(&a->lock);
  1005.     }
  1006.    
  1007.     /* Second, check the leaf node. */
  1008.     for (i = 0; i < leaf->keys; i++) {
  1009.         a = (as_area_t *) leaf->value[i];
  1010.    
  1011.         if (a == avoid_area)
  1012.             continue;
  1013.    
  1014.         mutex_lock(&a->lock);
  1015.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1016.             mutex_unlock(&a->lock);
  1017.             return false;
  1018.         }
  1019.         mutex_unlock(&a->lock);
  1020.     }
  1021.  
  1022.     /*
  1023.      * So far, the area does not conflict with other areas.
  1024.      * Check if it doesn't conflict with kernel address space.
  1025.      */  
  1026.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1027.         return !overlaps(va, size,
  1028.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1029.     }
  1030.  
  1031.     return true;
  1032. }
  1033.  
  1034. /** Return size of the address space area with given base.  */
  1035. size_t as_get_size(__address base)
  1036. {
  1037.     ipl_t ipl;
  1038.     as_area_t *src_area;
  1039.     size_t size;
  1040.  
  1041.     ipl = interrupts_disable();
  1042.     src_area = find_area_and_lock(AS, base);
  1043.     if (src_area){
  1044.         size = src_area->pages * PAGE_SIZE;
  1045.         mutex_unlock(&src_area->lock);
  1046.     } else {
  1047.         size = 0;
  1048.     }
  1049.     interrupts_restore(ipl);
  1050.     return size;
  1051. }
  1052.  
  1053. /** Mark portion of address space area as used.
  1054.  *
  1055.  * The address space area must be already locked.
  1056.  *
  1057.  * @param a Address space area.
  1058.  * @param page First page to be marked.
  1059.  * @param count Number of page to be marked.
  1060.  *
  1061.  * @return 0 on failure and 1 on success.
  1062.  */
  1063. int used_space_insert(as_area_t *a, __address page, count_t count)
  1064. {
  1065.     btree_node_t *leaf, *node;
  1066.     count_t pages;
  1067.     int i;
  1068.  
  1069.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1070.     ASSERT(count);
  1071.  
  1072.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1073.     if (pages) {
  1074.         /*
  1075.          * We hit the beginning of some used space.
  1076.          */
  1077.         return 0;
  1078.     }
  1079.  
  1080.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1081.     if (node) {
  1082.         __address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1083.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1084.        
  1085.         /*
  1086.          * Examine the possibility that the interval fits
  1087.          * somewhere between the rightmost interval of
  1088.          * the left neigbour and the first interval of the leaf.
  1089.          */
  1090.          
  1091.         if (page >= right_pg) {
  1092.             /* Do nothing. */
  1093.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1094.             /* The interval intersects with the left interval. */
  1095.             return 0;
  1096.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1097.             /* The interval intersects with the right interval. */
  1098.             return 0;          
  1099.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1100.             /* The interval can be added by merging the two already present intervals. */
  1101.             node->value[node->keys - 1] += count + right_cnt;
  1102.             btree_remove(&a->used_space, right_pg, leaf);
  1103.             return 1;
  1104.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1105.             /* The interval can be added by simply growing the left interval. */
  1106.             node->value[node->keys - 1] += count;
  1107.             return 1;
  1108.         } else if (page + count*PAGE_SIZE == right_pg) {
  1109.             /*
  1110.              * The interval can be addded by simply moving base of the right
  1111.              * interval down and increasing its size accordingly.
  1112.              */
  1113.             leaf->value[0] += count;
  1114.             leaf->key[0] = page;
  1115.             return 1;
  1116.         } else {
  1117.             /*
  1118.              * The interval is between both neigbouring intervals,
  1119.              * but cannot be merged with any of them.
  1120.              */
  1121.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1122.             return 1;
  1123.         }
  1124.     } else if (page < leaf->key[0]) {
  1125.         __address right_pg = leaf->key[0];
  1126.         count_t right_cnt = (count_t) leaf->value[0];
  1127.    
  1128.         /*
  1129.          * Investigate the border case in which the left neighbour does not
  1130.          * exist but the interval fits from the left.
  1131.          */
  1132.          
  1133.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1134.             /* The interval intersects with the right interval. */
  1135.             return 0;
  1136.         } else if (page + count*PAGE_SIZE == right_pg) {
  1137.             /*
  1138.              * The interval can be added by moving the base of the right interval down
  1139.              * and increasing its size accordingly.
  1140.              */
  1141.             leaf->key[0] = page;
  1142.             leaf->value[0] += count;
  1143.             return 1;
  1144.         } else {
  1145.             /*
  1146.              * The interval doesn't adjoin with the right interval.
  1147.              * It must be added individually.
  1148.              */
  1149.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1150.             return 1;
  1151.         }
  1152.     }
  1153.  
  1154.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1155.     if (node) {
  1156.         __address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1157.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1158.        
  1159.         /*
  1160.          * Examine the possibility that the interval fits
  1161.          * somewhere between the leftmost interval of
  1162.          * the right neigbour and the last interval of the leaf.
  1163.          */
  1164.  
  1165.         if (page < left_pg) {
  1166.             /* Do nothing. */
  1167.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1168.             /* The interval intersects with the left interval. */
  1169.             return 0;
  1170.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1171.             /* The interval intersects with the right interval. */
  1172.             return 0;          
  1173.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1174.             /* The interval can be added by merging the two already present intervals. */
  1175.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1176.             btree_remove(&a->used_space, right_pg, node);
  1177.             return 1;
  1178.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1179.             /* The interval can be added by simply growing the left interval. */
  1180.             leaf->value[leaf->keys - 1] +=  count;
  1181.             return 1;
  1182.         } else if (page + count*PAGE_SIZE == right_pg) {
  1183.             /*
  1184.              * The interval can be addded by simply moving base of the right
  1185.              * interval down and increasing its size accordingly.
  1186.              */
  1187.             node->value[0] += count;
  1188.             node->key[0] = page;
  1189.             return 1;
  1190.         } else {
  1191.             /*
  1192.              * The interval is between both neigbouring intervals,
  1193.              * but cannot be merged with any of them.
  1194.              */
  1195.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1196.             return 1;
  1197.         }
  1198.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1199.         __address left_pg = leaf->key[leaf->keys - 1];
  1200.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1201.    
  1202.         /*
  1203.          * Investigate the border case in which the right neighbour does not
  1204.          * exist but the interval fits from the right.
  1205.          */
  1206.          
  1207.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1208.             /* The interval intersects with the left interval. */
  1209.             return 0;
  1210.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1211.             /* The interval can be added by growing the left interval. */
  1212.             leaf->value[leaf->keys - 1] += count;
  1213.             return 1;
  1214.         } else {
  1215.             /*
  1216.              * The interval doesn't adjoin with the left interval.
  1217.              * It must be added individually.
  1218.              */
  1219.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1220.             return 1;
  1221.         }
  1222.     }
  1223.    
  1224.     /*
  1225.      * Note that if the algorithm made it thus far, the interval can fit only
  1226.      * between two other intervals of the leaf. The two border cases were already
  1227.      * resolved.
  1228.      */
  1229.     for (i = 1; i < leaf->keys; i++) {
  1230.         if (page < leaf->key[i]) {
  1231.             __address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1232.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1233.  
  1234.             /*
  1235.              * The interval fits between left_pg and right_pg.
  1236.              */
  1237.  
  1238.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1239.                 /* The interval intersects with the left interval. */
  1240.                 return 0;
  1241.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1242.                 /* The interval intersects with the right interval. */
  1243.                 return 0;          
  1244.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1245.                 /* The interval can be added by merging the two already present intervals. */
  1246.                 leaf->value[i - 1] += count + right_cnt;
  1247.                 btree_remove(&a->used_space, right_pg, leaf);
  1248.                 return 1;
  1249.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1250.                 /* The interval can be added by simply growing the left interval. */
  1251.                 leaf->value[i - 1] += count;
  1252.                 return 1;
  1253.             } else if (page + count*PAGE_SIZE == right_pg) {
  1254.                 /*
  1255.                      * The interval can be addded by simply moving base of the right
  1256.                  * interval down and increasing its size accordingly.
  1257.                  */
  1258.                 leaf->value[i] += count;
  1259.                 leaf->key[i] = page;
  1260.                 return 1;
  1261.             } else {
  1262.                 /*
  1263.                  * The interval is between both neigbouring intervals,
  1264.                  * but cannot be merged with any of them.
  1265.                  */
  1266.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1267.                 return 1;
  1268.             }
  1269.         }
  1270.     }
  1271.  
  1272.     panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
  1273. }
  1274.  
  1275. /** Mark portion of address space area as unused.
  1276.  *
  1277.  * The address space area must be already locked.
  1278.  *
  1279.  * @param a Address space area.
  1280.  * @param page First page to be marked.
  1281.  * @param count Number of page to be marked.
  1282.  *
  1283.  * @return 0 on failure and 1 on success.
  1284.  */
  1285. int used_space_remove(as_area_t *a, __address page, count_t count)
  1286. {
  1287.     btree_node_t *leaf, *node;
  1288.     count_t pages;
  1289.     int i;
  1290.  
  1291.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1292.     ASSERT(count);
  1293.  
  1294.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1295.     if (pages) {
  1296.         /*
  1297.          * We are lucky, page is the beginning of some interval.
  1298.          */
  1299.         if (count > pages) {
  1300.             return 0;
  1301.         } else if (count == pages) {
  1302.             btree_remove(&a->used_space, page, leaf);
  1303.             return 1;
  1304.         } else {
  1305.             /*
  1306.              * Find the respective interval.
  1307.              * Decrease its size and relocate its start address.
  1308.              */
  1309.             for (i = 0; i < leaf->keys; i++) {
  1310.                 if (leaf->key[i] == page) {
  1311.                     leaf->key[i] += count*PAGE_SIZE;
  1312.                     leaf->value[i] -= count;
  1313.                     return 1;
  1314.                 }
  1315.             }
  1316.             goto error;
  1317.         }
  1318.     }
  1319.  
  1320.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1321.     if (node && page < leaf->key[0]) {
  1322.         __address left_pg = node->key[node->keys - 1];
  1323.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1324.  
  1325.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1326.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1327.                 /*
  1328.                  * The interval is contained in the rightmost interval
  1329.                  * of the left neighbour and can be removed by
  1330.                  * updating the size of the bigger interval.
  1331.                  */
  1332.                 node->value[node->keys - 1] -= count;
  1333.                 return 1;
  1334.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1335.                 count_t new_cnt;
  1336.                
  1337.                 /*
  1338.                  * The interval is contained in the rightmost interval
  1339.                  * of the left neighbour but its removal requires
  1340.                  * both updating the size of the original interval and
  1341.                  * also inserting a new interval.
  1342.                  */
  1343.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1344.                 node->value[node->keys - 1] -= count + new_cnt;
  1345.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1346.                 return 1;
  1347.             }
  1348.         }
  1349.         return 0;
  1350.     } else if (page < leaf->key[0]) {
  1351.         return 0;
  1352.     }
  1353.    
  1354.     if (page > leaf->key[leaf->keys - 1]) {
  1355.         __address left_pg = leaf->key[leaf->keys - 1];
  1356.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1357.  
  1358.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1359.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1360.                 /*
  1361.                  * The interval is contained in the rightmost interval
  1362.                  * of the leaf and can be removed by updating the size
  1363.                  * of the bigger interval.
  1364.                  */
  1365.                 leaf->value[leaf->keys - 1] -= count;
  1366.                 return 1;
  1367.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1368.                 count_t new_cnt;
  1369.                
  1370.                 /*
  1371.                  * The interval is contained in the rightmost interval
  1372.                  * of the leaf but its removal requires both updating
  1373.                  * the size of the original interval and
  1374.                  * also inserting a new interval.
  1375.                  */
  1376.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1377.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1378.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1379.                 return 1;
  1380.             }
  1381.         }
  1382.         return 0;
  1383.     }  
  1384.    
  1385.     /*
  1386.      * The border cases have been already resolved.
  1387.      * Now the interval can be only between intervals of the leaf.
  1388.      */
  1389.     for (i = 1; i < leaf->keys - 1; i++) {
  1390.         if (page < leaf->key[i]) {
  1391.             __address left_pg = leaf->key[i - 1];
  1392.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1393.  
  1394.             /*
  1395.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1396.              */
  1397.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1398.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1399.                     /*
  1400.                     * The interval is contained in the interval (i - 1)
  1401.                      * of the leaf and can be removed by updating the size
  1402.                      * of the bigger interval.
  1403.                      */
  1404.                     leaf->value[i - 1] -= count;
  1405.                     return 1;
  1406.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1407.                     count_t new_cnt;
  1408.                
  1409.                     /*
  1410.                      * The interval is contained in the interval (i - 1)
  1411.                      * of the leaf but its removal requires both updating
  1412.                      * the size of the original interval and
  1413.                      * also inserting a new interval.
  1414.                      */
  1415.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1416.                     leaf->value[i - 1] -= count + new_cnt;
  1417.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1418.                     return 1;
  1419.                 }
  1420.             }
  1421.             return 0;
  1422.         }
  1423.     }
  1424.  
  1425. error:
  1426.     panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
  1427. }
  1428.  
  1429. /** Remove reference to address space area share info.
  1430.  *
  1431.  * If the reference count drops to 0, the sh_info is deallocated.
  1432.  *
  1433.  * @param sh_info Pointer to address space area share info.
  1434.  */
  1435. void sh_info_remove_reference(share_info_t *sh_info)
  1436. {
  1437.     bool dealloc = false;
  1438.  
  1439.     mutex_lock(&sh_info->lock);
  1440.     ASSERT(sh_info->refcount);
  1441.     if (--sh_info->refcount == 0) {
  1442.         dealloc = true;
  1443.         bool cond;
  1444.        
  1445.         /*
  1446.          * Now walk carefully the pagemap B+tree and free/remove
  1447.          * reference from all frames found there.
  1448.          */
  1449.         for (cond = true; cond;) {
  1450.             btree_node_t *node;
  1451.            
  1452.             ASSERT(!list_empty(&sh_info->pagemap.leaf_head));
  1453.             node = list_get_instance(sh_info->pagemap.leaf_head.next, btree_node_t, leaf_link);
  1454.             if ((cond = node->keys)) {
  1455.                 frame_free(ADDR2PFN((__address) node->value[0]));
  1456.                 btree_remove(&sh_info->pagemap, node->key[0], node);
  1457.             }
  1458.         }
  1459.        
  1460.     }
  1461.     mutex_unlock(&sh_info->lock);
  1462.    
  1463.     if (dealloc) {
  1464.         btree_destroy(&sh_info->pagemap);
  1465.         free(sh_info);
  1466.     }
  1467. }
  1468.  
  1469. static int anon_page_fault(as_area_t *area, __address addr, pf_access_t access);
  1470. static void anon_frame_free(as_area_t *area, __address page, __address frame);
  1471.  
  1472. /*
  1473.  * Anonymous memory backend.
  1474.  */
  1475. mem_backend_t anon_backend = {
  1476.     .backend_page_fault = anon_page_fault,
  1477.     .backend_frame_free = anon_frame_free
  1478. };
  1479.  
  1480. /** Service a page fault in the anonymous memory address space area.
  1481.  *
  1482.  * The address space area and page tables must be already locked.
  1483.  *
  1484.  * @param area Pointer to the address space area.
  1485.  * @param addr Faulting virtual address.
  1486.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  1487.  *
  1488.  * @return AS_PF_FAULT on failure (i.e. page fault) or AS_PF_OK on success (i.e. serviced).
  1489.  */
  1490. int anon_page_fault(as_area_t *area, __address addr, pf_access_t access)
  1491. {
  1492.     __address frame;
  1493.  
  1494.     if (area->sh_info) {
  1495.         btree_node_t *leaf;
  1496.        
  1497.         /*
  1498.          * The area is shared, chances are that the mapping can be found
  1499.          * in the pagemap of the address space area share info structure.
  1500.          * In the case that the pagemap does not contain the respective
  1501.          * mapping, a new frame is allocated and the mapping is created.
  1502.          */
  1503.         mutex_lock(&area->sh_info->lock);
  1504.         frame = (__address) btree_search(&area->sh_info->pagemap,
  1505.             ALIGN_DOWN(addr, PAGE_SIZE) - area->base, &leaf);
  1506.         if (!frame) {
  1507.             bool allocate = true;
  1508.             int i;
  1509.            
  1510.             /*
  1511.              * Zero can be returned as a valid frame address.
  1512.              * Just a small workaround.
  1513.              */
  1514.             for (i = 0; i < leaf->keys; i++) {
  1515.                 if (leaf->key[i] == ALIGN_DOWN(addr, PAGE_SIZE)) {
  1516.                     allocate = false;
  1517.                     break;
  1518.                 }
  1519.             }
  1520.             if (allocate) {
  1521.                 frame = PFN2ADDR(frame_alloc(ONE_FRAME, 0));
  1522.                 memsetb(PA2KA(frame), FRAME_SIZE, 0);
  1523.                
  1524.                 /*
  1525.                  * Insert the address of the newly allocated frame to the pagemap.
  1526.                  */
  1527.                 btree_insert(&area->sh_info->pagemap, ALIGN_DOWN(addr, PAGE_SIZE) - area->base, (void *) frame, leaf);
  1528.             }
  1529.         }
  1530.         mutex_unlock(&area->sh_info->lock);
  1531.     } else {
  1532.  
  1533.         /*
  1534.          * In general, there can be several reasons that
  1535.          * can have caused this fault.
  1536.          *
  1537.          * - non-existent mapping: the area is an anonymous
  1538.          *   area (e.g. heap or stack) and so far has not been
  1539.          *   allocated a frame for the faulting page
  1540.          *
  1541.          * - non-present mapping: another possibility,
  1542.          *   currently not implemented, would be frame
  1543.          *   reuse; when this becomes a possibility,
  1544.          *   do not forget to distinguish between
  1545.          *   the different causes
  1546.          */
  1547.         frame = PFN2ADDR(frame_alloc(ONE_FRAME, 0));
  1548.         memsetb(PA2KA(frame), FRAME_SIZE, 0);
  1549.     }
  1550.    
  1551.     /*
  1552.      * Map 'page' to 'frame'.
  1553.      * Note that TLB shootdown is not attempted as only new information is being
  1554.      * inserted into page tables.
  1555.      */
  1556.     page_mapping_insert(AS, addr, frame, as_area_get_flags(area));
  1557.     if (!used_space_insert(area, ALIGN_DOWN(addr, PAGE_SIZE), 1))
  1558.         panic("Could not insert used space.\n");
  1559.        
  1560.     return AS_PF_OK;
  1561. }
  1562.  
  1563. /** Free a frame that is backed by the anonymous memory backend.
  1564.  *
  1565.  * The address space area and page tables must be already locked.
  1566.  *
  1567.  * @param area Ignored.
  1568.  * @param page Ignored.
  1569.  * @param frame Frame to be released.
  1570.  */
  1571. void anon_frame_free(as_area_t *area, __address page, __address frame)
  1572. {
  1573.     frame_free(ADDR2PFN(frame));
  1574. }
  1575.  
  1576. /*
  1577.  * Address space related syscalls.
  1578.  */
  1579.  
  1580. /** Wrapper for as_area_create(). */
  1581. __native sys_as_area_create(__address address, size_t size, int flags)
  1582. {
  1583.     if (as_area_create(AS, flags, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1584.         return (__native) address;
  1585.     else
  1586.         return (__native) -1;
  1587. }
  1588.  
  1589. /** Wrapper for as_area_resize. */
  1590. __native sys_as_area_resize(__address address, size_t size, int flags)
  1591. {
  1592.     return (__native) as_area_resize(AS, address, size, 0);
  1593. }
  1594.  
  1595. /** Wrapper for as_area_destroy. */
  1596. __native sys_as_area_destroy(__address address)
  1597. {
  1598.     return (__native) as_area_destroy(AS, address);
  1599. }
  1600.