Subversion Repositories HelenOS-historic

Rev

Rev 1468 | Rev 1495 | 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. as_operations_t *as_operations = NULL;
  78.  
  79. /** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
  80. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  81.  
  82. /**
  83.  * This list contains address spaces that are not active on any
  84.  * processor and that have valid ASID.
  85.  */
  86. LIST_INITIALIZE(inactive_as_with_asid_head);
  87.  
  88. /** Kernel address space. */
  89. as_t *AS_KERNEL = NULL;
  90.  
  91. static int area_flags_to_page_flags(int aflags);
  92. static as_area_t *find_area_and_lock(as_t *as, __address va);
  93. static bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area);
  94. static void sh_info_remove_reference(share_info_t *sh_info);
  95.  
  96. /** Initialize address space subsystem. */
  97. void as_init(void)
  98. {
  99.     as_arch_init();
  100.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  101.     if (!AS_KERNEL)
  102.         panic("can't create kernel address space\n");
  103.    
  104. }
  105.  
  106. /** Create address space.
  107.  *
  108.  * @param flags Flags that influence way in wich the address space is created.
  109.  */
  110. as_t *as_create(int flags)
  111. {
  112.     as_t *as;
  113.  
  114.     as = (as_t *) malloc(sizeof(as_t), 0);
  115.     link_initialize(&as->inactive_as_with_asid_link);
  116.     mutex_initialize(&as->lock);
  117.     btree_create(&as->as_area_btree);
  118.    
  119.     if (flags & FLAG_AS_KERNEL)
  120.         as->asid = ASID_KERNEL;
  121.     else
  122.         as->asid = ASID_INVALID;
  123.    
  124.     as->refcount = 0;
  125.     as->cpu_refcount = 0;
  126.     as->page_table = page_table_create(flags);
  127.  
  128.     return as;
  129. }
  130.  
  131. /** Destroy adress space.
  132.  *
  133.  * When there are no tasks referencing this address space (i.e. its refcount is zero),
  134.  * the address space can be destroyed.
  135.  */
  136. void as_destroy(as_t *as)
  137. {
  138.     ipl_t ipl;
  139.     bool cond;
  140.  
  141.     ASSERT(as->refcount == 0);
  142.    
  143.     /*
  144.      * Since there is no reference to this area,
  145.      * it is safe not to lock its mutex.
  146.      */
  147.      
  148.     ipl = interrupts_disable();
  149.     spinlock_lock(&inactive_as_with_asid_lock);
  150.     if (as->asid != ASID_INVALID && as->asid != ASID_KERNEL) {
  151.         list_remove(&as->inactive_as_with_asid_link);
  152.         asid_put(as->asid);
  153.     }
  154.     spinlock_unlock(&inactive_as_with_asid_lock);
  155.  
  156.     /*
  157.      * Destroy address space areas of the address space.
  158.      */
  159.     for (cond = true; cond; ) {
  160.         btree_node_t *node;
  161.        
  162.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  163.         node = list_get_instance(&as->as_area_btree.leaf_head.next, btree_node_t, leaf_link);
  164.         if ((cond = node->keys))
  165.             as_area_destroy(as, node->key[0]);
  166.     }
  167.    
  168.     btree_destroy(&as->as_area_btree);
  169.     page_table_destroy(as->page_table);
  170.  
  171.     interrupts_restore(ipl);
  172.    
  173.     free(as);
  174. }
  175.  
  176. /** Create address space area of common attributes.
  177.  *
  178.  * The created address space area is added to the target address space.
  179.  *
  180.  * @param as Target address space.
  181.  * @param flags Flags of the area memory.
  182.  * @param size Size of area.
  183.  * @param base Base address of area.
  184.  * @param attrs Attributes of the area.
  185.  * @param backend Address space area backend. NULL if no backend is used.
  186.  * @param backend_data NULL or a pointer to an array holding two void *.
  187.  *
  188.  * @return Address space area on success or NULL on failure.
  189.  */
  190. as_area_t *as_area_create(as_t *as, int flags, size_t size, __address base, int attrs,
  191.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  192. {
  193.     ipl_t ipl;
  194.     as_area_t *a;
  195.    
  196.     if (base % PAGE_SIZE)
  197.         return NULL;
  198.  
  199.     if (!size)
  200.         return NULL;
  201.  
  202.     /* Writeable executable areas are not supported. */
  203.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  204.         return NULL;
  205.    
  206.     ipl = interrupts_disable();
  207.     mutex_lock(&as->lock);
  208.    
  209.     if (!check_area_conflicts(as, base, size, NULL)) {
  210.         mutex_unlock(&as->lock);
  211.         interrupts_restore(ipl);
  212.         return NULL;
  213.     }
  214.    
  215.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  216.  
  217.     mutex_initialize(&a->lock);
  218.    
  219.     a->as = as;
  220.     a->flags = flags;
  221.     a->attributes = attrs;
  222.     a->pages = SIZE2FRAMES(size);
  223.     a->base = base;
  224.     a->sh_info = NULL;
  225.     a->backend = backend;
  226.     if (backend_data)
  227.         a->backend_data = *backend_data;
  228.     else
  229.         memsetb((__address) &a->backend_data, sizeof(a->backend_data), 0);
  230.  
  231.     btree_create(&a->used_space);
  232.    
  233.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  234.  
  235.     mutex_unlock(&as->lock);
  236.     interrupts_restore(ipl);
  237.  
  238.     return a;
  239. }
  240.  
  241. /** Find address space area and change it.
  242.  *
  243.  * @param as Address space.
  244.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  245.  * @param size New size of the virtual memory block starting at address.
  246.  * @param flags Flags influencing the remap operation. Currently unused.
  247.  *
  248.  * @return Zero on success or a value from @ref errno.h otherwise.
  249.  */
  250. int as_area_resize(as_t *as, __address address, size_t size, int flags)
  251. {
  252.     as_area_t *area;
  253.     ipl_t ipl;
  254.     size_t pages;
  255.    
  256.     ipl = interrupts_disable();
  257.     mutex_lock(&as->lock);
  258.    
  259.     /*
  260.      * Locate the area.
  261.      */
  262.     area = find_area_and_lock(as, address);
  263.     if (!area) {
  264.         mutex_unlock(&as->lock);
  265.         interrupts_restore(ipl);
  266.         return ENOENT;
  267.     }
  268.  
  269.     if (area->backend == &phys_backend) {
  270.         /*
  271.          * Remapping of address space areas associated
  272.          * with memory mapped devices is not supported.
  273.          */
  274.         mutex_unlock(&area->lock);
  275.         mutex_unlock(&as->lock);
  276.         interrupts_restore(ipl);
  277.         return ENOTSUP;
  278.     }
  279.     if (area->sh_info) {
  280.         /*
  281.          * Remapping of shared address space areas
  282.          * is not supported.
  283.          */
  284.         mutex_unlock(&area->lock);
  285.         mutex_unlock(&as->lock);
  286.         interrupts_restore(ipl);
  287.         return ENOTSUP;
  288.     }
  289.  
  290.     pages = SIZE2FRAMES((address - area->base) + size);
  291.     if (!pages) {
  292.         /*
  293.          * Zero size address space areas are not allowed.
  294.          */
  295.         mutex_unlock(&area->lock);
  296.         mutex_unlock(&as->lock);
  297.         interrupts_restore(ipl);
  298.         return EPERM;
  299.     }
  300.    
  301.     if (pages < area->pages) {
  302.         bool cond;
  303.         __address start_free = area->base + pages*PAGE_SIZE;
  304.  
  305.         /*
  306.          * Shrinking the area.
  307.          * No need to check for overlaps.
  308.          */
  309.  
  310.         /*
  311.          * Start TLB shootdown sequence.
  312.          */
  313.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  314.  
  315.         /*
  316.          * Remove frames belonging to used space starting from
  317.          * the highest addresses downwards until an overlap with
  318.          * the resized address space area is found. Note that this
  319.          * is also the right way to remove part of the used_space
  320.          * B+tree leaf list.
  321.          */    
  322.         for (cond = true; cond;) {
  323.             btree_node_t *node;
  324.        
  325.             ASSERT(!list_empty(&area->used_space.leaf_head));
  326.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  327.             if ((cond = (bool) node->keys)) {
  328.                 __address b = node->key[node->keys - 1];
  329.                 count_t c = (count_t) node->value[node->keys - 1];
  330.                 int i = 0;
  331.            
  332.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  333.                    
  334.                     if (b + c*PAGE_SIZE <= start_free) {
  335.                         /*
  336.                          * The whole interval fits completely
  337.                          * in the resized address space area.
  338.                          */
  339.                         break;
  340.                     }
  341.        
  342.                     /*
  343.                      * Part of the interval corresponding to b and c
  344.                      * overlaps with the resized address space area.
  345.                      */
  346.        
  347.                     cond = false;   /* we are almost done */
  348.                     i = (start_free - b) >> PAGE_WIDTH;
  349.                     if (!used_space_remove(area, start_free, c - i))
  350.                         panic("Could not remove used space.");
  351.                 } else {
  352.                     /*
  353.                      * The interval of used space can be completely removed.
  354.                      */
  355.                     if (!used_space_remove(area, b, c))
  356.                         panic("Could not remove used space.\n");
  357.                 }
  358.            
  359.                 for (; i < c; i++) {
  360.                     pte_t *pte;
  361.            
  362.                     page_table_lock(as, false);
  363.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  364.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  365.                     if (area->backend && area->backend->frame_free) {
  366.                         area->backend->frame_free(area,
  367.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  368.                     }
  369.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  370.                     page_table_unlock(as, false);
  371.                 }
  372.             }
  373.         }
  374.  
  375.         /*
  376.          * Finish TLB shootdown sequence.
  377.          */
  378.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  379.         tlb_shootdown_finalize();
  380.     } else {
  381.         /*
  382.          * Growing the area.
  383.          * Check for overlaps with other address space areas.
  384.          */
  385.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  386.             mutex_unlock(&area->lock);
  387.             mutex_unlock(&as->lock);       
  388.             interrupts_restore(ipl);
  389.             return EADDRNOTAVAIL;
  390.         }
  391.     }
  392.  
  393.     area->pages = pages;
  394.    
  395.     mutex_unlock(&area->lock);
  396.     mutex_unlock(&as->lock);
  397.     interrupts_restore(ipl);
  398.  
  399.     return 0;
  400. }
  401.  
  402. /** Destroy address space area.
  403.  *
  404.  * @param as Address space.
  405.  * @param address Address withing the area to be deleted.
  406.  *
  407.  * @return Zero on success or a value from @ref errno.h on failure.
  408.  */
  409. int as_area_destroy(as_t *as, __address address)
  410. {
  411.     as_area_t *area;
  412.     __address base;
  413.     ipl_t ipl;
  414.     bool cond;
  415.  
  416.     ipl = interrupts_disable();
  417.     mutex_lock(&as->lock);
  418.  
  419.     area = find_area_and_lock(as, address);
  420.     if (!area) {
  421.         mutex_unlock(&as->lock);
  422.         interrupts_restore(ipl);
  423.         return ENOENT;
  424.     }
  425.  
  426.     base = area->base;
  427.  
  428.     /*
  429.      * Start TLB shootdown sequence.
  430.      */
  431.     tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
  432.  
  433.     /*
  434.      * Visit only the pages mapped by used_space B+tree.
  435.      * Note that we must be very careful when walking the tree
  436.      * leaf list and removing used space as the leaf list changes
  437.      * unpredictibly after each remove. The solution is to actually
  438.      * not walk the tree at all, but to remove items from the head
  439.      * of the leaf list until there are some keys left.
  440.      */
  441.     for (cond = true; cond;) {
  442.         btree_node_t *node;
  443.        
  444.         ASSERT(!list_empty(&area->used_space.leaf_head));
  445.         node = list_get_instance(area->used_space.leaf_head.next, btree_node_t, leaf_link);
  446.         if ((cond = (bool) node->keys)) {
  447.             __address b = node->key[0];
  448.             count_t i;
  449.             pte_t *pte;
  450.            
  451.             for (i = 0; i < (count_t) node->value[0]; i++) {
  452.                 page_table_lock(as, false);
  453.                 pte = page_mapping_find(as, b + i*PAGE_SIZE);
  454.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  455.                 if (area->backend && area->backend->frame_free) {
  456.                     area->backend->frame_free(area,
  457.                         b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  458.                 }
  459.                 page_mapping_remove(as, b + i*PAGE_SIZE);
  460.                 page_table_unlock(as, false);
  461.             }
  462.         }
  463.     }
  464.  
  465.     /*
  466.      * Finish TLB shootdown sequence.
  467.      */
  468.     tlb_invalidate_pages(AS->asid, area->base, area->pages);
  469.     tlb_shootdown_finalize();
  470.    
  471.     btree_destroy(&area->used_space);
  472.  
  473.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  474.    
  475.     if (area->sh_info)
  476.         sh_info_remove_reference(area->sh_info);
  477.        
  478.     mutex_unlock(&area->lock);
  479.  
  480.     /*
  481.      * Remove the empty area from address space.
  482.      */
  483.     btree_remove(&AS->as_area_btree, base, NULL);
  484.    
  485.     free(area);
  486.    
  487.     mutex_unlock(&AS->lock);
  488.     interrupts_restore(ipl);
  489.     return 0;
  490. }
  491.  
  492. /** Share address space area with another or the same address space.
  493.  *
  494.  * Address space area mapping is shared with a new address space area.
  495.  * If the source address space area has not been shared so far,
  496.  * a new sh_info is created. The new address space area simply gets the
  497.  * sh_info of the source area. The process of duplicating the
  498.  * mapping is done through the backend share function.
  499.  *
  500.  * @param src_as Pointer to source address space.
  501.  * @param src_base Base address of the source address space area.
  502.  * @param acc_size Expected size of the source area.
  503.  * @param dst_as Pointer to destination address space.
  504.  * @param dst_base Target base address.
  505.  * @param dst_flags_mask Destination address space area flags mask.
  506.  *
  507.  * @return Zero on success or ENOENT if there is no such task or
  508.  *     if there is no such address space area,
  509.  *     EPERM if there was a problem in accepting the area or
  510.  *     ENOMEM if there was a problem in allocating destination
  511.  *     address space area. ENOTSUP is returned if an attempt
  512.  *     to share non-anonymous address space area is detected.
  513.  */
  514. int as_area_share(as_t *src_as, __address src_base, size_t acc_size,
  515.           as_t *dst_as, __address dst_base, int dst_flags_mask)
  516. {
  517.     ipl_t ipl;
  518.     int src_flags;
  519.     size_t src_size;
  520.     as_area_t *src_area, *dst_area;
  521.     share_info_t *sh_info;
  522.     mem_backend_t *src_backend;
  523.     mem_backend_data_t src_backend_data;
  524.    
  525.     ipl = interrupts_disable();
  526.     mutex_lock(&src_as->lock);
  527.     src_area = find_area_and_lock(src_as, src_base);
  528.     if (!src_area) {
  529.         /*
  530.          * Could not find the source address space area.
  531.          */
  532.         mutex_unlock(&src_as->lock);
  533.         interrupts_restore(ipl);
  534.         return ENOENT;
  535.     }
  536.    
  537.     if (!src_area->backend || !src_area->backend->share) {
  538.         /*
  539.          * There is now backend or the backend does not
  540.          * know how to share the area.
  541.          */
  542.         mutex_unlock(&src_area->lock);
  543.         mutex_unlock(&src_as->lock);
  544.         interrupts_restore(ipl);
  545.         return ENOTSUP;
  546.     }
  547.    
  548.     src_size = src_area->pages * PAGE_SIZE;
  549.     src_flags = src_area->flags;
  550.     src_backend = src_area->backend;
  551.     src_backend_data = src_area->backend_data;
  552.    
  553.     if (src_size != acc_size || (src_flags & dst_flags_mask) != dst_flags_mask) {
  554.         mutex_unlock(&src_area->lock);
  555.         mutex_unlock(&src_as->lock);
  556.         interrupts_restore(ipl);
  557.         return EPERM;
  558.     }
  559.  
  560.     /*
  561.      * Now we are committed to sharing the area.
  562.      * First prepare the area for sharing.
  563.      * Then it will be safe to unlock it.
  564.      */
  565.     sh_info = src_area->sh_info;
  566.     if (!sh_info) {
  567.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  568.         mutex_initialize(&sh_info->lock);
  569.         sh_info->refcount = 2;
  570.         btree_create(&sh_info->pagemap);
  571.         src_area->sh_info = sh_info;
  572.     } else {
  573.         mutex_lock(&sh_info->lock);
  574.         sh_info->refcount++;
  575.         mutex_unlock(&sh_info->lock);
  576.     }
  577.  
  578.     src_area->backend->share(src_area);
  579.  
  580.     mutex_unlock(&src_area->lock);
  581.     mutex_unlock(&src_as->lock);
  582.  
  583.     /*
  584.      * Create copy of the source address space area.
  585.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  586.      * attribute set which prevents race condition with
  587.      * preliminary as_page_fault() calls.
  588.      * The flags of the source area are masked against dst_flags_mask
  589.      * to support sharing in less privileged mode.
  590.      */
  591.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  592.                   AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  593.     if (!dst_area) {
  594.         /*
  595.          * Destination address space area could not be created.
  596.          */
  597.         sh_info_remove_reference(sh_info);
  598.        
  599.         interrupts_restore(ipl);
  600.         return ENOMEM;
  601.     }
  602.    
  603.     /*
  604.      * Now the destination address space area has been
  605.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  606.      * attribute and set the sh_info.
  607.      */
  608.     mutex_lock(&dst_area->lock);
  609.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  610.     dst_area->sh_info = sh_info;
  611.     mutex_unlock(&dst_area->lock);
  612.    
  613.     interrupts_restore(ipl);
  614.    
  615.     return 0;
  616. }
  617.  
  618. /** Check access mode for address space area.
  619.  *
  620.  * The address space area must be locked prior to this call.
  621.  *
  622.  * @param area Address space area.
  623.  * @param access Access mode.
  624.  *
  625.  * @return False if access violates area's permissions, true otherwise.
  626.  */
  627. bool as_area_check_access(as_area_t *area, pf_access_t access)
  628. {
  629.     int flagmap[] = {
  630.         [PF_ACCESS_READ] = AS_AREA_READ,
  631.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  632.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  633.     };
  634.  
  635.     if (!(area->flags & flagmap[access]))
  636.         return false;
  637.    
  638.     return true;
  639. }
  640.  
  641. /** Handle page fault within the current address space.
  642.  *
  643.  * This is the high-level page fault handler. It decides
  644.  * whether the page fault can be resolved by any backend
  645.  * and if so, it invokes the backend to resolve the page
  646.  * fault.
  647.  *
  648.  * Interrupts are assumed disabled.
  649.  *
  650.  * @param page Faulting page.
  651.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  652.  * @param istate Pointer to interrupted state.
  653.  *
  654.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  655.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  656.  */
  657. int as_page_fault(__address page, pf_access_t access, istate_t *istate)
  658. {
  659.     pte_t *pte;
  660.     as_area_t *area;
  661.    
  662.     if (!THREAD)
  663.         return AS_PF_FAULT;
  664.        
  665.     ASSERT(AS);
  666.  
  667.     mutex_lock(&AS->lock);
  668.     area = find_area_and_lock(AS, page);   
  669.     if (!area) {
  670.         /*
  671.          * No area contained mapping for 'page'.
  672.          * Signal page fault to low-level handler.
  673.          */
  674.         mutex_unlock(&AS->lock);
  675.         goto page_fault;
  676.     }
  677.  
  678.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  679.         /*
  680.          * The address space area is not fully initialized.
  681.          * Avoid possible race by returning error.
  682.          */
  683.         mutex_unlock(&area->lock);
  684.         mutex_unlock(&AS->lock);
  685.         goto page_fault;       
  686.     }
  687.  
  688.     if (!area->backend || !area->backend->page_fault) {
  689.         /*
  690.          * The address space area is not backed by any backend
  691.          * or the backend cannot handle page faults.
  692.          */
  693.         mutex_unlock(&area->lock);
  694.         mutex_unlock(&AS->lock);
  695.         goto page_fault;       
  696.     }
  697.  
  698.     page_table_lock(AS, false);
  699.    
  700.     /*
  701.      * To avoid race condition between two page faults
  702.      * on the same address, we need to make sure
  703.      * the mapping has not been already inserted.
  704.      */
  705.     if ((pte = page_mapping_find(AS, page))) {
  706.         if (PTE_PRESENT(pte)) {
  707.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  708.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  709.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  710.                 page_table_unlock(AS, false);
  711.                 mutex_unlock(&area->lock);
  712.                 mutex_unlock(&AS->lock);
  713.                 return AS_PF_OK;
  714.             }
  715.         }
  716.     }
  717.    
  718.     /*
  719.      * Resort to the backend page fault handler.
  720.      */
  721.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  722.         page_table_unlock(AS, false);
  723.         mutex_unlock(&area->lock);
  724.         mutex_unlock(&AS->lock);
  725.         goto page_fault;
  726.     }
  727.    
  728.     page_table_unlock(AS, false);
  729.     mutex_unlock(&area->lock);
  730.     mutex_unlock(&AS->lock);
  731.     return AS_PF_OK;
  732.  
  733. page_fault:
  734.     if (THREAD->in_copy_from_uspace) {
  735.         THREAD->in_copy_from_uspace = false;
  736.         istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
  737.     } else if (THREAD->in_copy_to_uspace) {
  738.         THREAD->in_copy_to_uspace = false;
  739.         istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
  740.     } else {
  741.         return AS_PF_FAULT;
  742.     }
  743.  
  744.     return AS_PF_DEFER;
  745. }
  746.  
  747. /** Switch address spaces.
  748.  *
  749.  * Note that this function cannot sleep as it is essentially a part of
  750.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  751.  *
  752.  * @param old Old address space or NULL.
  753.  * @param new New address space.
  754.  */
  755. void as_switch(as_t *old, as_t *new)
  756. {
  757.     ipl_t ipl;
  758.     bool needs_asid = false;
  759.    
  760.     ipl = interrupts_disable();
  761.     spinlock_lock(&inactive_as_with_asid_lock);
  762.  
  763.     /*
  764.      * First, take care of the old address space.
  765.      */
  766.     if (old) {
  767.         mutex_lock_active(&old->lock);
  768.         ASSERT(old->cpu_refcount);
  769.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  770.             /*
  771.              * The old address space is no longer active on
  772.              * any processor. It can be appended to the
  773.              * list of inactive address spaces with assigned
  774.              * ASID.
  775.              */
  776.              ASSERT(old->asid != ASID_INVALID);
  777.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  778.         }
  779.         mutex_unlock(&old->lock);
  780.     }
  781.  
  782.     /*
  783.      * Second, prepare the new address space.
  784.      */
  785.     mutex_lock_active(&new->lock);
  786.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  787.         if (new->asid != ASID_INVALID)
  788.             list_remove(&new->inactive_as_with_asid_link);
  789.         else
  790.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  791.     }
  792.     SET_PTL0_ADDRESS(new->page_table);
  793.     mutex_unlock(&new->lock);
  794.  
  795.     if (needs_asid) {
  796.         /*
  797.          * Allocation of new ASID was deferred
  798.          * until now in order to avoid deadlock.
  799.          */
  800.         asid_t asid;
  801.        
  802.         asid = asid_get();
  803.         mutex_lock_active(&new->lock);
  804.         new->asid = asid;
  805.         mutex_unlock(&new->lock);
  806.     }
  807.     spinlock_unlock(&inactive_as_with_asid_lock);
  808.     interrupts_restore(ipl);
  809.    
  810.     /*
  811.      * Perform architecture-specific steps.
  812.      * (e.g. write ASID to hardware register etc.)
  813.      */
  814.     as_install_arch(new);
  815.    
  816.     AS = new;
  817. }
  818.  
  819. /** Convert address space area flags to page flags.
  820.  *
  821.  * @param aflags Flags of some address space area.
  822.  *
  823.  * @return Flags to be passed to page_mapping_insert().
  824.  */
  825. int area_flags_to_page_flags(int aflags)
  826. {
  827.     int flags;
  828.  
  829.     flags = PAGE_USER | PAGE_PRESENT;
  830.    
  831.     if (aflags & AS_AREA_READ)
  832.         flags |= PAGE_READ;
  833.        
  834.     if (aflags & AS_AREA_WRITE)
  835.         flags |= PAGE_WRITE;
  836.    
  837.     if (aflags & AS_AREA_EXEC)
  838.         flags |= PAGE_EXEC;
  839.    
  840.     if (aflags & AS_AREA_CACHEABLE)
  841.         flags |= PAGE_CACHEABLE;
  842.        
  843.     return flags;
  844. }
  845.  
  846. /** Compute flags for virtual address translation subsytem.
  847.  *
  848.  * The address space area must be locked.
  849.  * Interrupts must be disabled.
  850.  *
  851.  * @param a Address space area.
  852.  *
  853.  * @return Flags to be used in page_mapping_insert().
  854.  */
  855. int as_area_get_flags(as_area_t *a)
  856. {
  857.     return area_flags_to_page_flags(a->flags);
  858. }
  859.  
  860. /** Create page table.
  861.  *
  862.  * Depending on architecture, create either address space
  863.  * private or global page table.
  864.  *
  865.  * @param flags Flags saying whether the page table is for kernel address space.
  866.  *
  867.  * @return First entry of the page table.
  868.  */
  869. pte_t *page_table_create(int flags)
  870. {
  871.         ASSERT(as_operations);
  872.         ASSERT(as_operations->page_table_create);
  873.  
  874.         return as_operations->page_table_create(flags);
  875. }
  876.  
  877. /** Destroy page table.
  878.  *
  879.  * Destroy page table in architecture specific way.
  880.  *
  881.  * @param page_table Physical address of PTL0.
  882.  */
  883. void page_table_destroy(pte_t *page_table)
  884. {
  885.         ASSERT(as_operations);
  886.         ASSERT(as_operations->page_table_destroy);
  887.  
  888.         as_operations->page_table_destroy(page_table);
  889. }
  890.  
  891. /** Lock page table.
  892.  *
  893.  * This function should be called before any page_mapping_insert(),
  894.  * page_mapping_remove() and page_mapping_find().
  895.  *
  896.  * Locking order is such that address space areas must be locked
  897.  * prior to this call. Address space can be locked prior to this
  898.  * call in which case the lock argument is false.
  899.  *
  900.  * @param as Address space.
  901.  * @param lock If false, do not attempt to lock as->lock.
  902.  */
  903. void page_table_lock(as_t *as, bool lock)
  904. {
  905.     ASSERT(as_operations);
  906.     ASSERT(as_operations->page_table_lock);
  907.  
  908.     as_operations->page_table_lock(as, lock);
  909. }
  910.  
  911. /** Unlock page table.
  912.  *
  913.  * @param as Address space.
  914.  * @param unlock If false, do not attempt to unlock as->lock.
  915.  */
  916. void page_table_unlock(as_t *as, bool unlock)
  917. {
  918.     ASSERT(as_operations);
  919.     ASSERT(as_operations->page_table_unlock);
  920.  
  921.     as_operations->page_table_unlock(as, unlock);
  922. }
  923.  
  924.  
  925. /** Find address space area and lock it.
  926.  *
  927.  * The address space must be locked and interrupts must be disabled.
  928.  *
  929.  * @param as Address space.
  930.  * @param va Virtual address.
  931.  *
  932.  * @return Locked address space area containing va on success or NULL on failure.
  933.  */
  934. as_area_t *find_area_and_lock(as_t *as, __address va)
  935. {
  936.     as_area_t *a;
  937.     btree_node_t *leaf, *lnode;
  938.     int i;
  939.    
  940.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  941.     if (a) {
  942.         /* va is the base address of an address space area */
  943.         mutex_lock(&a->lock);
  944.         return a;
  945.     }
  946.    
  947.     /*
  948.      * Search the leaf node and the righmost record of its left neighbour
  949.      * to find out whether this is a miss or va belongs to an address
  950.      * space area found there.
  951.      */
  952.    
  953.     /* First, search the leaf node itself. */
  954.     for (i = 0; i < leaf->keys; i++) {
  955.         a = (as_area_t *) leaf->value[i];
  956.         mutex_lock(&a->lock);
  957.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  958.             return a;
  959.         }
  960.         mutex_unlock(&a->lock);
  961.     }
  962.  
  963.     /*
  964.      * Second, locate the left neighbour and test its last record.
  965.      * Because of its position in the B+tree, it must have base < va.
  966.      */
  967.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  968.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  969.         mutex_lock(&a->lock);
  970.         if (va < a->base + a->pages * PAGE_SIZE) {
  971.             return a;
  972.         }
  973.         mutex_unlock(&a->lock);
  974.     }
  975.  
  976.     return NULL;
  977. }
  978.  
  979. /** Check area conflicts with other areas.
  980.  *
  981.  * The address space must be locked and interrupts must be disabled.
  982.  *
  983.  * @param as Address space.
  984.  * @param va Starting virtual address of the area being tested.
  985.  * @param size Size of the area being tested.
  986.  * @param avoid_area Do not touch this area.
  987.  *
  988.  * @return True if there is no conflict, false otherwise.
  989.  */
  990. bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area)
  991. {
  992.     as_area_t *a;
  993.     btree_node_t *leaf, *node;
  994.     int i;
  995.    
  996.     /*
  997.      * We don't want any area to have conflicts with NULL page.
  998.      */
  999.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1000.         return false;
  1001.    
  1002.     /*
  1003.      * The leaf node is found in O(log n), where n is proportional to
  1004.      * the number of address space areas belonging to as.
  1005.      * The check for conflicts is then attempted on the rightmost
  1006.      * record in the left neighbour, the leftmost record in the right
  1007.      * neighbour and all records in the leaf node itself.
  1008.      */
  1009.    
  1010.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1011.         if (a != avoid_area)
  1012.             return false;
  1013.     }
  1014.    
  1015.     /* First, check the two border cases. */
  1016.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1017.         a = (as_area_t *) node->value[node->keys - 1];
  1018.         mutex_lock(&a->lock);
  1019.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1020.             mutex_unlock(&a->lock);
  1021.             return false;
  1022.         }
  1023.         mutex_unlock(&a->lock);
  1024.     }
  1025.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1026.         a = (as_area_t *) node->value[0];
  1027.         mutex_lock(&a->lock);
  1028.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1029.             mutex_unlock(&a->lock);
  1030.             return false;
  1031.         }
  1032.         mutex_unlock(&a->lock);
  1033.     }
  1034.    
  1035.     /* Second, check the leaf node. */
  1036.     for (i = 0; i < leaf->keys; i++) {
  1037.         a = (as_area_t *) leaf->value[i];
  1038.    
  1039.         if (a == avoid_area)
  1040.             continue;
  1041.    
  1042.         mutex_lock(&a->lock);
  1043.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1044.             mutex_unlock(&a->lock);
  1045.             return false;
  1046.         }
  1047.         mutex_unlock(&a->lock);
  1048.     }
  1049.  
  1050.     /*
  1051.      * So far, the area does not conflict with other areas.
  1052.      * Check if it doesn't conflict with kernel address space.
  1053.      */  
  1054.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1055.         return !overlaps(va, size,
  1056.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1057.     }
  1058.  
  1059.     return true;
  1060. }
  1061.  
  1062. /** Return size of the address space area with given base.  */
  1063. size_t as_get_size(__address base)
  1064. {
  1065.     ipl_t ipl;
  1066.     as_area_t *src_area;
  1067.     size_t size;
  1068.  
  1069.     ipl = interrupts_disable();
  1070.     src_area = find_area_and_lock(AS, base);
  1071.     if (src_area){
  1072.         size = src_area->pages * PAGE_SIZE;
  1073.         mutex_unlock(&src_area->lock);
  1074.     } else {
  1075.         size = 0;
  1076.     }
  1077.     interrupts_restore(ipl);
  1078.     return size;
  1079. }
  1080.  
  1081. /** Mark portion of address space area as used.
  1082.  *
  1083.  * The address space area must be already locked.
  1084.  *
  1085.  * @param a Address space area.
  1086.  * @param page First page to be marked.
  1087.  * @param count Number of page to be marked.
  1088.  *
  1089.  * @return 0 on failure and 1 on success.
  1090.  */
  1091. int used_space_insert(as_area_t *a, __address page, count_t count)
  1092. {
  1093.     btree_node_t *leaf, *node;
  1094.     count_t pages;
  1095.     int i;
  1096.  
  1097.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1098.     ASSERT(count);
  1099.  
  1100.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1101.     if (pages) {
  1102.         /*
  1103.          * We hit the beginning of some used space.
  1104.          */
  1105.         return 0;
  1106.     }
  1107.  
  1108.     if (!leaf->keys) {
  1109.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1110.         return 1;
  1111.     }
  1112.  
  1113.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1114.     if (node) {
  1115.         __address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1116.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1117.        
  1118.         /*
  1119.          * Examine the possibility that the interval fits
  1120.          * somewhere between the rightmost interval of
  1121.          * the left neigbour and the first interval of the leaf.
  1122.          */
  1123.          
  1124.         if (page >= right_pg) {
  1125.             /* Do nothing. */
  1126.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1127.             /* The interval intersects with the left interval. */
  1128.             return 0;
  1129.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1130.             /* The interval intersects with the right interval. */
  1131.             return 0;          
  1132.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1133.             /* The interval can be added by merging the two already present intervals. */
  1134.             node->value[node->keys - 1] += count + right_cnt;
  1135.             btree_remove(&a->used_space, right_pg, leaf);
  1136.             return 1;
  1137.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1138.             /* The interval can be added by simply growing the left interval. */
  1139.             node->value[node->keys - 1] += count;
  1140.             return 1;
  1141.         } else if (page + count*PAGE_SIZE == right_pg) {
  1142.             /*
  1143.              * The interval can be addded by simply moving base of the right
  1144.              * interval down and increasing its size accordingly.
  1145.              */
  1146.             leaf->value[0] += count;
  1147.             leaf->key[0] = page;
  1148.             return 1;
  1149.         } else {
  1150.             /*
  1151.              * The interval is between both neigbouring intervals,
  1152.              * but cannot be merged with any of them.
  1153.              */
  1154.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1155.             return 1;
  1156.         }
  1157.     } else if (page < leaf->key[0]) {
  1158.         __address right_pg = leaf->key[0];
  1159.         count_t right_cnt = (count_t) leaf->value[0];
  1160.    
  1161.         /*
  1162.          * Investigate the border case in which the left neighbour does not
  1163.          * exist but the interval fits from the left.
  1164.          */
  1165.          
  1166.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1167.             /* The interval intersects with the right interval. */
  1168.             return 0;
  1169.         } else if (page + count*PAGE_SIZE == right_pg) {
  1170.             /*
  1171.              * The interval can be added by moving the base of the right interval down
  1172.              * and increasing its size accordingly.
  1173.              */
  1174.             leaf->key[0] = page;
  1175.             leaf->value[0] += count;
  1176.             return 1;
  1177.         } else {
  1178.             /*
  1179.              * The interval doesn't adjoin with the right interval.
  1180.              * It must be added individually.
  1181.              */
  1182.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1183.             return 1;
  1184.         }
  1185.     }
  1186.  
  1187.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1188.     if (node) {
  1189.         __address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1190.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1191.        
  1192.         /*
  1193.          * Examine the possibility that the interval fits
  1194.          * somewhere between the leftmost interval of
  1195.          * the right neigbour and the last interval of the leaf.
  1196.          */
  1197.  
  1198.         if (page < left_pg) {
  1199.             /* Do nothing. */
  1200.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1201.             /* The interval intersects with the left interval. */
  1202.             return 0;
  1203.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1204.             /* The interval intersects with the right interval. */
  1205.             return 0;          
  1206.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1207.             /* The interval can be added by merging the two already present intervals. */
  1208.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1209.             btree_remove(&a->used_space, right_pg, node);
  1210.             return 1;
  1211.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1212.             /* The interval can be added by simply growing the left interval. */
  1213.             leaf->value[leaf->keys - 1] +=  count;
  1214.             return 1;
  1215.         } else if (page + count*PAGE_SIZE == right_pg) {
  1216.             /*
  1217.              * The interval can be addded by simply moving base of the right
  1218.              * interval down and increasing its size accordingly.
  1219.              */
  1220.             node->value[0] += count;
  1221.             node->key[0] = page;
  1222.             return 1;
  1223.         } else {
  1224.             /*
  1225.              * The interval is between both neigbouring intervals,
  1226.              * but cannot be merged with any of them.
  1227.              */
  1228.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1229.             return 1;
  1230.         }
  1231.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1232.         __address left_pg = leaf->key[leaf->keys - 1];
  1233.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1234.    
  1235.         /*
  1236.          * Investigate the border case in which the right neighbour does not
  1237.          * exist but the interval fits from the right.
  1238.          */
  1239.          
  1240.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1241.             /* The interval intersects with the left interval. */
  1242.             return 0;
  1243.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1244.             /* The interval can be added by growing the left interval. */
  1245.             leaf->value[leaf->keys - 1] += count;
  1246.             return 1;
  1247.         } else {
  1248.             /*
  1249.              * The interval doesn't adjoin with the left interval.
  1250.              * It must be added individually.
  1251.              */
  1252.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1253.             return 1;
  1254.         }
  1255.     }
  1256.    
  1257.     /*
  1258.      * Note that if the algorithm made it thus far, the interval can fit only
  1259.      * between two other intervals of the leaf. The two border cases were already
  1260.      * resolved.
  1261.      */
  1262.     for (i = 1; i < leaf->keys; i++) {
  1263.         if (page < leaf->key[i]) {
  1264.             __address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1265.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1266.  
  1267.             /*
  1268.              * The interval fits between left_pg and right_pg.
  1269.              */
  1270.  
  1271.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1272.                 /* The interval intersects with the left interval. */
  1273.                 return 0;
  1274.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1275.                 /* The interval intersects with the right interval. */
  1276.                 return 0;          
  1277.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1278.                 /* The interval can be added by merging the two already present intervals. */
  1279.                 leaf->value[i - 1] += count + right_cnt;
  1280.                 btree_remove(&a->used_space, right_pg, leaf);
  1281.                 return 1;
  1282.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1283.                 /* The interval can be added by simply growing the left interval. */
  1284.                 leaf->value[i - 1] += count;
  1285.                 return 1;
  1286.             } else if (page + count*PAGE_SIZE == right_pg) {
  1287.                 /*
  1288.                      * The interval can be addded by simply moving base of the right
  1289.                  * interval down and increasing its size accordingly.
  1290.                  */
  1291.                 leaf->value[i] += count;
  1292.                 leaf->key[i] = page;
  1293.                 return 1;
  1294.             } else {
  1295.                 /*
  1296.                  * The interval is between both neigbouring intervals,
  1297.                  * but cannot be merged with any of them.
  1298.                  */
  1299.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1300.                 return 1;
  1301.             }
  1302.         }
  1303.     }
  1304.  
  1305.     panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
  1306. }
  1307.  
  1308. /** Mark portion of address space area as unused.
  1309.  *
  1310.  * The address space area must be already locked.
  1311.  *
  1312.  * @param a Address space area.
  1313.  * @param page First page to be marked.
  1314.  * @param count Number of page to be marked.
  1315.  *
  1316.  * @return 0 on failure and 1 on success.
  1317.  */
  1318. int used_space_remove(as_area_t *a, __address page, count_t count)
  1319. {
  1320.     btree_node_t *leaf, *node;
  1321.     count_t pages;
  1322.     int i;
  1323.  
  1324.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1325.     ASSERT(count);
  1326.  
  1327.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1328.     if (pages) {
  1329.         /*
  1330.          * We are lucky, page is the beginning of some interval.
  1331.          */
  1332.         if (count > pages) {
  1333.             return 0;
  1334.         } else if (count == pages) {
  1335.             btree_remove(&a->used_space, page, leaf);
  1336.             return 1;
  1337.         } else {
  1338.             /*
  1339.              * Find the respective interval.
  1340.              * Decrease its size and relocate its start address.
  1341.              */
  1342.             for (i = 0; i < leaf->keys; i++) {
  1343.                 if (leaf->key[i] == page) {
  1344.                     leaf->key[i] += count*PAGE_SIZE;
  1345.                     leaf->value[i] -= count;
  1346.                     return 1;
  1347.                 }
  1348.             }
  1349.             goto error;
  1350.         }
  1351.     }
  1352.  
  1353.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1354.     if (node && page < leaf->key[0]) {
  1355.         __address left_pg = node->key[node->keys - 1];
  1356.         count_t left_cnt = (count_t) node->value[node->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 left neighbour and can be removed by
  1363.                  * updating the size of the bigger interval.
  1364.                  */
  1365.                 node->value[node->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 left neighbour but its removal requires
  1373.                  * both updating 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.                 node->value[node->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.     } else if (page < leaf->key[0]) {
  1384.         return 0;
  1385.     }
  1386.    
  1387.     if (page > leaf->key[leaf->keys - 1]) {
  1388.         __address left_pg = leaf->key[leaf->keys - 1];
  1389.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1390.  
  1391.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1392.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1393.                 /*
  1394.                  * The interval is contained in the rightmost interval
  1395.                  * of the leaf and can be removed by updating the size
  1396.                  * of the bigger interval.
  1397.                  */
  1398.                 leaf->value[leaf->keys - 1] -= count;
  1399.                 return 1;
  1400.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1401.                 count_t new_cnt;
  1402.                
  1403.                 /*
  1404.                  * The interval is contained in the rightmost interval
  1405.                  * of the leaf but its removal requires both updating
  1406.                  * the size of the original interval and
  1407.                  * also inserting a new interval.
  1408.                  */
  1409.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1410.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1411.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1412.                 return 1;
  1413.             }
  1414.         }
  1415.         return 0;
  1416.     }  
  1417.    
  1418.     /*
  1419.      * The border cases have been already resolved.
  1420.      * Now the interval can be only between intervals of the leaf.
  1421.      */
  1422.     for (i = 1; i < leaf->keys - 1; i++) {
  1423.         if (page < leaf->key[i]) {
  1424.             __address left_pg = leaf->key[i - 1];
  1425.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1426.  
  1427.             /*
  1428.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1429.              */
  1430.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1431.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1432.                     /*
  1433.                     * The interval is contained in the interval (i - 1)
  1434.                      * of the leaf and can be removed by updating the size
  1435.                      * of the bigger interval.
  1436.                      */
  1437.                     leaf->value[i - 1] -= count;
  1438.                     return 1;
  1439.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1440.                     count_t new_cnt;
  1441.                
  1442.                     /*
  1443.                      * The interval is contained in the interval (i - 1)
  1444.                      * of the leaf but its removal requires both updating
  1445.                      * the size of the original interval and
  1446.                      * also inserting a new interval.
  1447.                      */
  1448.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1449.                     leaf->value[i - 1] -= count + new_cnt;
  1450.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1451.                     return 1;
  1452.                 }
  1453.             }
  1454.             return 0;
  1455.         }
  1456.     }
  1457.  
  1458. error:
  1459.     panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
  1460. }
  1461.  
  1462. /** Remove reference to address space area share info.
  1463.  *
  1464.  * If the reference count drops to 0, the sh_info is deallocated.
  1465.  *
  1466.  * @param sh_info Pointer to address space area share info.
  1467.  */
  1468. void sh_info_remove_reference(share_info_t *sh_info)
  1469. {
  1470.     bool dealloc = false;
  1471.  
  1472.     mutex_lock(&sh_info->lock);
  1473.     ASSERT(sh_info->refcount);
  1474.     if (--sh_info->refcount == 0) {
  1475.         dealloc = true;
  1476.         bool cond;
  1477.        
  1478.         /*
  1479.          * Now walk carefully the pagemap B+tree and free/remove
  1480.          * reference from all frames found there.
  1481.          */
  1482.         for (cond = true; cond;) {
  1483.             btree_node_t *node;
  1484.            
  1485.             ASSERT(!list_empty(&sh_info->pagemap.leaf_head));
  1486.             node = list_get_instance(sh_info->pagemap.leaf_head.next, btree_node_t, leaf_link);
  1487.             if ((cond = node->keys)) {
  1488.                 frame_free(ADDR2PFN((__address) node->value[0]));
  1489.             }
  1490.         }
  1491.        
  1492.     }
  1493.     mutex_unlock(&sh_info->lock);
  1494.    
  1495.     if (dealloc) {
  1496.         btree_destroy(&sh_info->pagemap);
  1497.         free(sh_info);
  1498.     }
  1499. }
  1500.  
  1501. /*
  1502.  * Address space related syscalls.
  1503.  */
  1504.  
  1505. /** Wrapper for as_area_create(). */
  1506. __native sys_as_area_create(__address address, size_t size, int flags)
  1507. {
  1508.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1509.         return (__native) address;
  1510.     else
  1511.         return (__native) -1;
  1512. }
  1513.  
  1514. /** Wrapper for as_area_resize. */
  1515. __native sys_as_area_resize(__address address, size_t size, int flags)
  1516. {
  1517.     return (__native) as_area_resize(AS, address, size, 0);
  1518. }
  1519.  
  1520. /** Wrapper for as_area_destroy. */
  1521. __native sys_as_area_destroy(__address address)
  1522. {
  1523.     return (__native) as_area_destroy(AS, address);
  1524. }
  1525.