Subversion Repositories HelenOS-historic

Rev

Rev 1702 | Rev 1757 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed

  1. /*
  2.  * Copyright (C) 2001-2006 Jakub Jermar
  3.  * All rights reserved.
  4.  *
  5.  * Redistribution and use in source and binary forms, with or without
  6.  * modification, are permitted provided that the following conditions
  7.  * are met:
  8.  *
  9.  * - Redistributions of source code must retain the above copyright
  10.  *   notice, this list of conditions and the following disclaimer.
  11.  * - Redistributions in binary form must reproduce the above copyright
  12.  *   notice, this list of conditions and the following disclaimer in the
  13.  *   documentation and/or other materials provided with the distribution.
  14.  * - The name of the author may not be used to endorse or promote products
  15.  *   derived from this software without specific prior written permission.
  16.  *
  17.  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  18.  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  19.  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  20.  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  21.  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  22.  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  23.  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  24.  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  25.  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  26.  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  27.  */
  28.  
  29.  /** @addtogroup genericmm
  30.  * @{
  31.  */
  32.  
  33. /**
  34.  * @file
  35.  * @brief   Address space related functions.
  36.  *
  37.  * This file contains address space manipulation functions.
  38.  * Roughly speaking, this is a higher-level client of
  39.  * Virtual Address Translation (VAT) subsystem.
  40.  *
  41.  * Functionality provided by this file allows one to
  42.  * create address space and create, resize and share
  43.  * address space areas.
  44.  *
  45.  * @see page.c
  46.  *
  47.  */
  48.  
  49. #include <mm/as.h>
  50. #include <arch/mm/as.h>
  51. #include <mm/page.h>
  52. #include <mm/frame.h>
  53. #include <mm/slab.h>
  54. #include <mm/tlb.h>
  55. #include <arch/mm/page.h>
  56. #include <genarch/mm/page_pt.h>
  57. #include <genarch/mm/page_ht.h>
  58. #include <mm/asid.h>
  59. #include <arch/mm/asid.h>
  60. #include <synch/spinlock.h>
  61. #include <synch/mutex.h>
  62. #include <adt/list.h>
  63. #include <adt/btree.h>
  64. #include <proc/task.h>
  65. #include <proc/thread.h>
  66. #include <arch/asm.h>
  67. #include <panic.h>
  68. #include <debug.h>
  69. #include <print.h>
  70. #include <memstr.h>
  71. #include <macros.h>
  72. #include <arch.h>
  73. #include <errno.h>
  74. #include <config.h>
  75. #include <align.h>
  76. #include <arch/types.h>
  77. #include <typedefs.h>
  78. #include <syscall/copy.h>
  79. #include <arch/interrupt.h>
  80.  
  81. as_operations_t *as_operations = NULL;
  82.  
  83. /** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
  84. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  85.  
  86. /**
  87.  * This list contains address spaces that are not active on any
  88.  * processor and that have valid ASID.
  89.  */
  90. LIST_INITIALIZE(inactive_as_with_asid_head);
  91.  
  92. /** Kernel address space. */
  93. as_t *AS_KERNEL = NULL;
  94.  
  95. static int area_flags_to_page_flags(int aflags);
  96. static as_area_t *find_area_and_lock(as_t *as, __address va);
  97. static bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area);
  98. static void sh_info_remove_reference(share_info_t *sh_info);
  99.  
  100. /** Initialize address space subsystem. */
  101. void as_init(void)
  102. {
  103.     as_arch_init();
  104.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  105.     if (!AS_KERNEL)
  106.         panic("can't create kernel address space\n");
  107.    
  108. }
  109.  
  110. /** Create address space.
  111.  *
  112.  * @param flags Flags that influence way in wich the address space is created.
  113.  */
  114. as_t *as_create(int flags)
  115. {
  116.     as_t *as;
  117.  
  118.     as = (as_t *) malloc(sizeof(as_t), 0);
  119.     link_initialize(&as->inactive_as_with_asid_link);
  120.     mutex_initialize(&as->lock);
  121.     btree_create(&as->as_area_btree);
  122.    
  123.     if (flags & FLAG_AS_KERNEL)
  124.         as->asid = ASID_KERNEL;
  125.     else
  126.         as->asid = ASID_INVALID;
  127.    
  128.     as->refcount = 0;
  129.     as->cpu_refcount = 0;
  130.     as->page_table = page_table_create(flags);
  131.  
  132.     return as;
  133. }
  134.  
  135. /** Destroy adress space.
  136.  *
  137.  * When there are no tasks referencing this address space (i.e. its refcount is zero),
  138.  * the address space can be destroyed.
  139.  */
  140. void as_destroy(as_t *as)
  141. {
  142.     ipl_t ipl;
  143.     bool cond;
  144.  
  145.     ASSERT(as->refcount == 0);
  146.    
  147.     /*
  148.      * Since there is no reference to this area,
  149.      * it is safe not to lock its mutex.
  150.      */
  151.     ipl = interrupts_disable();
  152.     spinlock_lock(&inactive_as_with_asid_lock);
  153.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  154.         if (as != AS && as->cpu_refcount == 0)
  155.             list_remove(&as->inactive_as_with_asid_link);
  156.         asid_put(as->asid);
  157.     }
  158.     spinlock_unlock(&inactive_as_with_asid_lock);
  159.  
  160.     /*
  161.      * Destroy address space areas of the address space.
  162.      * The B+tee must be walked carefully because it is
  163.      * also being destroyed.
  164.      */
  165.     for (cond = true; cond; ) {
  166.         btree_node_t *node;
  167.  
  168.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  169.         node = list_get_instance(as->as_area_btree.leaf_head.next, btree_node_t, leaf_link);
  170.  
  171.         if ((cond = node->keys)) {
  172.             as_area_destroy(as, node->key[0]);
  173.         }
  174.     }
  175.  
  176.     btree_destroy(&as->as_area_btree);
  177.     page_table_destroy(as->page_table);
  178.  
  179.     interrupts_restore(ipl);
  180.    
  181.     free(as);
  182. }
  183.  
  184. /** Create address space area of common attributes.
  185.  *
  186.  * The created address space area is added to the target address space.
  187.  *
  188.  * @param as Target address space.
  189.  * @param flags Flags of the area memory.
  190.  * @param size Size of area.
  191.  * @param base Base address of area.
  192.  * @param attrs Attributes of the area.
  193.  * @param backend Address space area backend. NULL if no backend is used.
  194.  * @param backend_data NULL or a pointer to an array holding two void *.
  195.  *
  196.  * @return Address space area on success or NULL on failure.
  197.  */
  198. as_area_t *as_area_create(as_t *as, int flags, size_t size, __address base, int attrs,
  199.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  200. {
  201.     ipl_t ipl;
  202.     as_area_t *a;
  203.    
  204.     if (base % PAGE_SIZE)
  205.         return NULL;
  206.  
  207.     if (!size)
  208.         return NULL;
  209.  
  210.     /* Writeable executable areas are not supported. */
  211.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  212.         return NULL;
  213.    
  214.     ipl = interrupts_disable();
  215.     mutex_lock(&as->lock);
  216.    
  217.     if (!check_area_conflicts(as, base, size, NULL)) {
  218.         mutex_unlock(&as->lock);
  219.         interrupts_restore(ipl);
  220.         return NULL;
  221.     }
  222.    
  223.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  224.  
  225.     mutex_initialize(&a->lock);
  226.    
  227.     a->as = as;
  228.     a->flags = flags;
  229.     a->attributes = attrs;
  230.     a->pages = SIZE2FRAMES(size);
  231.     a->base = base;
  232.     a->sh_info = NULL;
  233.     a->backend = backend;
  234.     if (backend_data)
  235.         a->backend_data = *backend_data;
  236.     else
  237.         memsetb((__address) &a->backend_data, sizeof(a->backend_data), 0);
  238.  
  239.     btree_create(&a->used_space);
  240.    
  241.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  242.  
  243.     mutex_unlock(&as->lock);
  244.     interrupts_restore(ipl);
  245.  
  246.     return a;
  247. }
  248.  
  249. /** Find address space area and change it.
  250.  *
  251.  * @param as Address space.
  252.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  253.  * @param size New size of the virtual memory block starting at address.
  254.  * @param flags Flags influencing the remap operation. Currently unused.
  255.  *
  256.  * @return Zero on success or a value from @ref errno.h otherwise.
  257.  */
  258. int as_area_resize(as_t *as, __address address, size_t size, int flags)
  259. {
  260.     as_area_t *area;
  261.     ipl_t ipl;
  262.     size_t pages;
  263.    
  264.     ipl = interrupts_disable();
  265.     mutex_lock(&as->lock);
  266.    
  267.     /*
  268.      * Locate the area.
  269.      */
  270.     area = find_area_and_lock(as, address);
  271.     if (!area) {
  272.         mutex_unlock(&as->lock);
  273.         interrupts_restore(ipl);
  274.         return ENOENT;
  275.     }
  276.  
  277.     if (area->backend == &phys_backend) {
  278.         /*
  279.          * Remapping of address space areas associated
  280.          * with memory mapped devices is not supported.
  281.          */
  282.         mutex_unlock(&area->lock);
  283.         mutex_unlock(&as->lock);
  284.         interrupts_restore(ipl);
  285.         return ENOTSUP;
  286.     }
  287.     if (area->sh_info) {
  288.         /*
  289.          * Remapping of shared address space areas
  290.          * is not supported.
  291.          */
  292.         mutex_unlock(&area->lock);
  293.         mutex_unlock(&as->lock);
  294.         interrupts_restore(ipl);
  295.         return ENOTSUP;
  296.     }
  297.  
  298.     pages = SIZE2FRAMES((address - area->base) + size);
  299.     if (!pages) {
  300.         /*
  301.          * Zero size address space areas are not allowed.
  302.          */
  303.         mutex_unlock(&area->lock);
  304.         mutex_unlock(&as->lock);
  305.         interrupts_restore(ipl);
  306.         return EPERM;
  307.     }
  308.    
  309.     if (pages < area->pages) {
  310.         bool cond;
  311.         __address start_free = area->base + pages*PAGE_SIZE;
  312.  
  313.         /*
  314.          * Shrinking the area.
  315.          * No need to check for overlaps.
  316.          */
  317.  
  318.         /*
  319.          * Start TLB shootdown sequence.
  320.          */
  321.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  322.  
  323.         /*
  324.          * Remove frames belonging to used space starting from
  325.          * the highest addresses downwards until an overlap with
  326.          * the resized address space area is found. Note that this
  327.          * is also the right way to remove part of the used_space
  328.          * B+tree leaf list.
  329.          */    
  330.         for (cond = true; cond;) {
  331.             btree_node_t *node;
  332.        
  333.             ASSERT(!list_empty(&area->used_space.leaf_head));
  334.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  335.             if ((cond = (bool) node->keys)) {
  336.                 __address b = node->key[node->keys - 1];
  337.                 count_t c = (count_t) node->value[node->keys - 1];
  338.                 int i = 0;
  339.            
  340.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  341.                    
  342.                     if (b + c*PAGE_SIZE <= start_free) {
  343.                         /*
  344.                          * The whole interval fits completely
  345.                          * in the resized address space area.
  346.                          */
  347.                         break;
  348.                     }
  349.        
  350.                     /*
  351.                      * Part of the interval corresponding to b and c
  352.                      * overlaps with the resized address space area.
  353.                      */
  354.        
  355.                     cond = false;   /* we are almost done */
  356.                     i = (start_free - b) >> PAGE_WIDTH;
  357.                     if (!used_space_remove(area, start_free, c - i))
  358.                         panic("Could not remove used space.");
  359.                 } else {
  360.                     /*
  361.                      * The interval of used space can be completely removed.
  362.                      */
  363.                     if (!used_space_remove(area, b, c))
  364.                         panic("Could not remove used space.\n");
  365.                 }
  366.            
  367.                 for (; i < c; i++) {
  368.                     pte_t *pte;
  369.            
  370.                     page_table_lock(as, false);
  371.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  372.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  373.                     if (area->backend && area->backend->frame_free) {
  374.                         area->backend->frame_free(area,
  375.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  376.                     }
  377.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  378.                     page_table_unlock(as, false);
  379.                 }
  380.             }
  381.         }
  382.  
  383.         /*
  384.          * Finish TLB shootdown sequence.
  385.          */
  386.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  387.         tlb_shootdown_finalize();
  388.     } else {
  389.         /*
  390.          * Growing the area.
  391.          * Check for overlaps with other address space areas.
  392.          */
  393.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  394.             mutex_unlock(&area->lock);
  395.             mutex_unlock(&as->lock);       
  396.             interrupts_restore(ipl);
  397.             return EADDRNOTAVAIL;
  398.         }
  399.     }
  400.  
  401.     area->pages = pages;
  402.    
  403.     mutex_unlock(&area->lock);
  404.     mutex_unlock(&as->lock);
  405.     interrupts_restore(ipl);
  406.  
  407.     return 0;
  408. }
  409.  
  410. /** Destroy address space area.
  411.  *
  412.  * @param as Address space.
  413.  * @param address Address withing the area to be deleted.
  414.  *
  415.  * @return Zero on success or a value from @ref errno.h on failure.
  416.  */
  417. int as_area_destroy(as_t *as, __address address)
  418. {
  419.     as_area_t *area;
  420.     __address base;
  421.     link_t *cur;
  422.     ipl_t ipl;
  423.  
  424.     ipl = interrupts_disable();
  425.     mutex_lock(&as->lock);
  426.  
  427.     area = find_area_and_lock(as, address);
  428.     if (!area) {
  429.         mutex_unlock(&as->lock);
  430.         interrupts_restore(ipl);
  431.         return ENOENT;
  432.     }
  433.  
  434.     base = area->base;
  435.  
  436.     /*
  437.      * Start TLB shootdown sequence.
  438.      */
  439.     tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
  440.  
  441.     /*
  442.      * Visit only the pages mapped by used_space B+tree.
  443.      */
  444.     for (cur = area->used_space.leaf_head.next; cur != &area->used_space.leaf_head; cur = cur->next) {
  445.         btree_node_t *node;
  446.         int i;
  447.        
  448.         node = list_get_instance(cur, btree_node_t, leaf_link);
  449.         for (i = 0; i < node->keys; i++) {
  450.             __address b = node->key[i];
  451.             count_t j;
  452.             pte_t *pte;
  453.            
  454.             for (j = 0; j < (count_t) node->value[i]; j++) {
  455.                 page_table_lock(as, false);
  456.                 pte = page_mapping_find(as, b + j*PAGE_SIZE);
  457.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  458.                 if (area->backend && area->backend->frame_free) {
  459.                     area->backend->frame_free(area,
  460.                         b + j*PAGE_SIZE, PTE_GET_FRAME(pte));
  461.                 }
  462.                 page_mapping_remove(as, b + j*PAGE_SIZE);
  463.                 page_table_unlock(as, false);
  464.             }
  465.         }
  466.     }
  467.  
  468.     /*
  469.      * Finish TLB shootdown sequence.
  470.      */
  471.     tlb_invalidate_pages(AS->asid, area->base, area->pages);
  472.     tlb_shootdown_finalize();
  473.    
  474.     btree_destroy(&area->used_space);
  475.  
  476.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  477.    
  478.     if (area->sh_info)
  479.         sh_info_remove_reference(area->sh_info);
  480.        
  481.     mutex_unlock(&area->lock);
  482.  
  483.     /*
  484.      * Remove the empty area from address space.
  485.      */
  486.     btree_remove(&AS->as_area_btree, base, NULL);
  487.    
  488.     free(area);
  489.    
  490.     mutex_unlock(&AS->lock);
  491.     interrupts_restore(ipl);
  492.     return 0;
  493. }
  494.  
  495. /** Share address space area with another or the same address space.
  496.  *
  497.  * Address space area mapping is shared with a new address space area.
  498.  * If the source address space area has not been shared so far,
  499.  * a new sh_info is created. The new address space area simply gets the
  500.  * sh_info of the source area. The process of duplicating the
  501.  * mapping is done through the backend share function.
  502.  *
  503.  * @param src_as Pointer to source address space.
  504.  * @param src_base Base address of the source address space area.
  505.  * @param acc_size Expected size of the source area.
  506.  * @param dst_as Pointer to destination address space.
  507.  * @param dst_base Target base address.
  508.  * @param dst_flags_mask Destination address space area flags mask.
  509.  *
  510.  * @return Zero on success or ENOENT if there is no such task or
  511.  *     if there is no such address space area,
  512.  *     EPERM if there was a problem in accepting the area or
  513.  *     ENOMEM if there was a problem in allocating destination
  514.  *     address space area. ENOTSUP is returned if an attempt
  515.  *     to share non-anonymous address space area is detected.
  516.  */
  517. int as_area_share(as_t *src_as, __address src_base, size_t acc_size,
  518.           as_t *dst_as, __address dst_base, int dst_flags_mask)
  519. {
  520.     ipl_t ipl;
  521.     int src_flags;
  522.     size_t src_size;
  523.     as_area_t *src_area, *dst_area;
  524.     share_info_t *sh_info;
  525.     mem_backend_t *src_backend;
  526.     mem_backend_data_t src_backend_data;
  527.    
  528.     ipl = interrupts_disable();
  529.     mutex_lock(&src_as->lock);
  530.     src_area = find_area_and_lock(src_as, src_base);
  531.     if (!src_area) {
  532.         /*
  533.          * Could not find the source address space area.
  534.          */
  535.         mutex_unlock(&src_as->lock);
  536.         interrupts_restore(ipl);
  537.         return ENOENT;
  538.     }
  539.    
  540.     if (!src_area->backend || !src_area->backend->share) {
  541.         /*
  542.          * There is now backend or the backend does not
  543.          * know how to share the area.
  544.          */
  545.         mutex_unlock(&src_area->lock);
  546.         mutex_unlock(&src_as->lock);
  547.         interrupts_restore(ipl);
  548.         return ENOTSUP;
  549.     }
  550.    
  551.     src_size = src_area->pages * PAGE_SIZE;
  552.     src_flags = src_area->flags;
  553.     src_backend = src_area->backend;
  554.     src_backend_data = src_area->backend_data;
  555.  
  556.     /* Share the cacheable flag from the original mapping */
  557.     if (src_flags & AS_AREA_CACHEABLE)
  558.         dst_flags_mask |= AS_AREA_CACHEABLE;
  559.  
  560.     if (src_size != acc_size || (src_flags & dst_flags_mask) != dst_flags_mask) {
  561.         mutex_unlock(&src_area->lock);
  562.         mutex_unlock(&src_as->lock);
  563.         interrupts_restore(ipl);
  564.         return EPERM;
  565.     }
  566.  
  567.     /*
  568.      * Now we are committed to sharing the area.
  569.      * First prepare the area for sharing.
  570.      * Then it will be safe to unlock it.
  571.      */
  572.     sh_info = src_area->sh_info;
  573.     if (!sh_info) {
  574.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  575.         mutex_initialize(&sh_info->lock);
  576.         sh_info->refcount = 2;
  577.         btree_create(&sh_info->pagemap);
  578.         src_area->sh_info = sh_info;
  579.     } else {
  580.         mutex_lock(&sh_info->lock);
  581.         sh_info->refcount++;
  582.         mutex_unlock(&sh_info->lock);
  583.     }
  584.  
  585.     src_area->backend->share(src_area);
  586.  
  587.     mutex_unlock(&src_area->lock);
  588.     mutex_unlock(&src_as->lock);
  589.  
  590.     /*
  591.      * Create copy of the source address space area.
  592.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  593.      * attribute set which prevents race condition with
  594.      * preliminary as_page_fault() calls.
  595.      * The flags of the source area are masked against dst_flags_mask
  596.      * to support sharing in less privileged mode.
  597.      */
  598.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  599.                   AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  600.     if (!dst_area) {
  601.         /*
  602.          * Destination address space area could not be created.
  603.          */
  604.         sh_info_remove_reference(sh_info);
  605.        
  606.         interrupts_restore(ipl);
  607.         return ENOMEM;
  608.     }
  609.    
  610.     /*
  611.      * Now the destination address space area has been
  612.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  613.      * attribute and set the sh_info.
  614.      */
  615.     mutex_lock(&dst_area->lock);
  616.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  617.     dst_area->sh_info = sh_info;
  618.     mutex_unlock(&dst_area->lock);
  619.    
  620.     interrupts_restore(ipl);
  621.    
  622.     return 0;
  623. }
  624.  
  625. /** Check access mode for address space area.
  626.  *
  627.  * The address space area must be locked prior to this call.
  628.  *
  629.  * @param area Address space area.
  630.  * @param access Access mode.
  631.  *
  632.  * @return False if access violates area's permissions, true otherwise.
  633.  */
  634. bool as_area_check_access(as_area_t *area, pf_access_t access)
  635. {
  636.     int flagmap[] = {
  637.         [PF_ACCESS_READ] = AS_AREA_READ,
  638.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  639.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  640.     };
  641.  
  642.     if (!(area->flags & flagmap[access]))
  643.         return false;
  644.    
  645.     return true;
  646. }
  647.  
  648. /** Handle page fault within the current address space.
  649.  *
  650.  * This is the high-level page fault handler. It decides
  651.  * whether the page fault can be resolved by any backend
  652.  * and if so, it invokes the backend to resolve the page
  653.  * fault.
  654.  *
  655.  * Interrupts are assumed disabled.
  656.  *
  657.  * @param page Faulting page.
  658.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  659.  * @param istate Pointer to interrupted state.
  660.  *
  661.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  662.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  663.  */
  664. int as_page_fault(__address page, pf_access_t access, istate_t *istate)
  665. {
  666.     pte_t *pte;
  667.     as_area_t *area;
  668.    
  669.     if (!THREAD)
  670.         return AS_PF_FAULT;
  671.        
  672.     ASSERT(AS);
  673.  
  674.     mutex_lock(&AS->lock);
  675.     area = find_area_and_lock(AS, page);   
  676.     if (!area) {
  677.         /*
  678.          * No area contained mapping for 'page'.
  679.          * Signal page fault to low-level handler.
  680.          */
  681.         mutex_unlock(&AS->lock);
  682.         goto page_fault;
  683.     }
  684.  
  685.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  686.         /*
  687.          * The address space area is not fully initialized.
  688.          * Avoid possible race by returning error.
  689.          */
  690.         mutex_unlock(&area->lock);
  691.         mutex_unlock(&AS->lock);
  692.         goto page_fault;       
  693.     }
  694.  
  695.     if (!area->backend || !area->backend->page_fault) {
  696.         /*
  697.          * The address space area is not backed by any backend
  698.          * or the backend cannot handle page faults.
  699.          */
  700.         mutex_unlock(&area->lock);
  701.         mutex_unlock(&AS->lock);
  702.         goto page_fault;       
  703.     }
  704.  
  705.     page_table_lock(AS, false);
  706.    
  707.     /*
  708.      * To avoid race condition between two page faults
  709.      * on the same address, we need to make sure
  710.      * the mapping has not been already inserted.
  711.      */
  712.     if ((pte = page_mapping_find(AS, page))) {
  713.         if (PTE_PRESENT(pte)) {
  714.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  715.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  716.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  717.                 page_table_unlock(AS, false);
  718.                 mutex_unlock(&area->lock);
  719.                 mutex_unlock(&AS->lock);
  720.                 return AS_PF_OK;
  721.             }
  722.         }
  723.     }
  724.    
  725.     /*
  726.      * Resort to the backend page fault handler.
  727.      */
  728.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  729.         page_table_unlock(AS, false);
  730.         mutex_unlock(&area->lock);
  731.         mutex_unlock(&AS->lock);
  732.         goto page_fault;
  733.     }
  734.    
  735.     page_table_unlock(AS, false);
  736.     mutex_unlock(&area->lock);
  737.     mutex_unlock(&AS->lock);
  738.     return AS_PF_OK;
  739.  
  740. page_fault:
  741.     if (THREAD->in_copy_from_uspace) {
  742.         THREAD->in_copy_from_uspace = false;
  743.         istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
  744.     } else if (THREAD->in_copy_to_uspace) {
  745.         THREAD->in_copy_to_uspace = false;
  746.         istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
  747.     } else {
  748.         return AS_PF_FAULT;
  749.     }
  750.  
  751.     return AS_PF_DEFER;
  752. }
  753.  
  754. /** Switch address spaces.
  755.  *
  756.  * Note that this function cannot sleep as it is essentially a part of
  757.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  758.  *
  759.  * @param old Old address space or NULL.
  760.  * @param new New address space.
  761.  */
  762. void as_switch(as_t *old, as_t *new)
  763. {
  764.     ipl_t ipl;
  765.     bool needs_asid = false;
  766.    
  767.     ipl = interrupts_disable();
  768.     spinlock_lock(&inactive_as_with_asid_lock);
  769.  
  770.     /*
  771.      * First, take care of the old address space.
  772.      */
  773.     if (old) {
  774.         mutex_lock_active(&old->lock);
  775.         ASSERT(old->cpu_refcount);
  776.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  777.             /*
  778.              * The old address space is no longer active on
  779.              * any processor. It can be appended to the
  780.              * list of inactive address spaces with assigned
  781.              * ASID.
  782.              */
  783.              ASSERT(old->asid != ASID_INVALID);
  784.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  785.         }
  786.         mutex_unlock(&old->lock);
  787.     }
  788.  
  789.     /*
  790.      * Second, prepare the new address space.
  791.      */
  792.     mutex_lock_active(&new->lock);
  793.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  794.         if (new->asid != ASID_INVALID)
  795.             list_remove(&new->inactive_as_with_asid_link);
  796.         else
  797.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  798.     }
  799.     SET_PTL0_ADDRESS(new->page_table);
  800.     mutex_unlock(&new->lock);
  801.  
  802.     if (needs_asid) {
  803.         /*
  804.          * Allocation of new ASID was deferred
  805.          * until now in order to avoid deadlock.
  806.          */
  807.         asid_t asid;
  808.        
  809.         asid = asid_get();
  810.         mutex_lock_active(&new->lock);
  811.         new->asid = asid;
  812.         mutex_unlock(&new->lock);
  813.     }
  814.     spinlock_unlock(&inactive_as_with_asid_lock);
  815.     interrupts_restore(ipl);
  816.    
  817.     /*
  818.      * Perform architecture-specific steps.
  819.      * (e.g. write ASID to hardware register etc.)
  820.      */
  821.     as_install_arch(new);
  822.    
  823.     AS = new;
  824. }
  825.  
  826. /** Convert address space area flags to page flags.
  827.  *
  828.  * @param aflags Flags of some address space area.
  829.  *
  830.  * @return Flags to be passed to page_mapping_insert().
  831.  */
  832. int area_flags_to_page_flags(int aflags)
  833. {
  834.     int flags;
  835.  
  836.     flags = PAGE_USER | PAGE_PRESENT;
  837.    
  838.     if (aflags & AS_AREA_READ)
  839.         flags |= PAGE_READ;
  840.        
  841.     if (aflags & AS_AREA_WRITE)
  842.         flags |= PAGE_WRITE;
  843.    
  844.     if (aflags & AS_AREA_EXEC)
  845.         flags |= PAGE_EXEC;
  846.    
  847.     if (aflags & AS_AREA_CACHEABLE)
  848.         flags |= PAGE_CACHEABLE;
  849.        
  850.     return flags;
  851. }
  852.  
  853. /** Compute flags for virtual address translation subsytem.
  854.  *
  855.  * The address space area must be locked.
  856.  * Interrupts must be disabled.
  857.  *
  858.  * @param a Address space area.
  859.  *
  860.  * @return Flags to be used in page_mapping_insert().
  861.  */
  862. int as_area_get_flags(as_area_t *a)
  863. {
  864.     return area_flags_to_page_flags(a->flags);
  865. }
  866.  
  867. /** Create page table.
  868.  *
  869.  * Depending on architecture, create either address space
  870.  * private or global page table.
  871.  *
  872.  * @param flags Flags saying whether the page table is for kernel address space.
  873.  *
  874.  * @return First entry of the page table.
  875.  */
  876. pte_t *page_table_create(int flags)
  877. {
  878.         ASSERT(as_operations);
  879.         ASSERT(as_operations->page_table_create);
  880.  
  881.         return as_operations->page_table_create(flags);
  882. }
  883.  
  884. /** Destroy page table.
  885.  *
  886.  * Destroy page table in architecture specific way.
  887.  *
  888.  * @param page_table Physical address of PTL0.
  889.  */
  890. void page_table_destroy(pte_t *page_table)
  891. {
  892.         ASSERT(as_operations);
  893.         ASSERT(as_operations->page_table_destroy);
  894.  
  895.         as_operations->page_table_destroy(page_table);
  896. }
  897.  
  898. /** Lock page table.
  899.  *
  900.  * This function should be called before any page_mapping_insert(),
  901.  * page_mapping_remove() and page_mapping_find().
  902.  *
  903.  * Locking order is such that address space areas must be locked
  904.  * prior to this call. Address space can be locked prior to this
  905.  * call in which case the lock argument is false.
  906.  *
  907.  * @param as Address space.
  908.  * @param lock If false, do not attempt to lock as->lock.
  909.  */
  910. void page_table_lock(as_t *as, bool lock)
  911. {
  912.     ASSERT(as_operations);
  913.     ASSERT(as_operations->page_table_lock);
  914.  
  915.     as_operations->page_table_lock(as, lock);
  916. }
  917.  
  918. /** Unlock page table.
  919.  *
  920.  * @param as Address space.
  921.  * @param unlock If false, do not attempt to unlock as->lock.
  922.  */
  923. void page_table_unlock(as_t *as, bool unlock)
  924. {
  925.     ASSERT(as_operations);
  926.     ASSERT(as_operations->page_table_unlock);
  927.  
  928.     as_operations->page_table_unlock(as, unlock);
  929. }
  930.  
  931.  
  932. /** Find address space area and lock it.
  933.  *
  934.  * The address space must be locked and interrupts must be disabled.
  935.  *
  936.  * @param as Address space.
  937.  * @param va Virtual address.
  938.  *
  939.  * @return Locked address space area containing va on success or NULL on failure.
  940.  */
  941. as_area_t *find_area_and_lock(as_t *as, __address va)
  942. {
  943.     as_area_t *a;
  944.     btree_node_t *leaf, *lnode;
  945.     int i;
  946.    
  947.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  948.     if (a) {
  949.         /* va is the base address of an address space area */
  950.         mutex_lock(&a->lock);
  951.         return a;
  952.     }
  953.    
  954.     /*
  955.      * Search the leaf node and the righmost record of its left neighbour
  956.      * to find out whether this is a miss or va belongs to an address
  957.      * space area found there.
  958.      */
  959.    
  960.     /* First, search the leaf node itself. */
  961.     for (i = 0; i < leaf->keys; i++) {
  962.         a = (as_area_t *) leaf->value[i];
  963.         mutex_lock(&a->lock);
  964.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  965.             return a;
  966.         }
  967.         mutex_unlock(&a->lock);
  968.     }
  969.  
  970.     /*
  971.      * Second, locate the left neighbour and test its last record.
  972.      * Because of its position in the B+tree, it must have base < va.
  973.      */
  974.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  975.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  976.         mutex_lock(&a->lock);
  977.         if (va < a->base + a->pages * PAGE_SIZE) {
  978.             return a;
  979.         }
  980.         mutex_unlock(&a->lock);
  981.     }
  982.  
  983.     return NULL;
  984. }
  985.  
  986. /** Check area conflicts with other areas.
  987.  *
  988.  * The address space must be locked and interrupts must be disabled.
  989.  *
  990.  * @param as Address space.
  991.  * @param va Starting virtual address of the area being tested.
  992.  * @param size Size of the area being tested.
  993.  * @param avoid_area Do not touch this area.
  994.  *
  995.  * @return True if there is no conflict, false otherwise.
  996.  */
  997. bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area)
  998. {
  999.     as_area_t *a;
  1000.     btree_node_t *leaf, *node;
  1001.     int i;
  1002.    
  1003.     /*
  1004.      * We don't want any area to have conflicts with NULL page.
  1005.      */
  1006.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1007.         return false;
  1008.    
  1009.     /*
  1010.      * The leaf node is found in O(log n), where n is proportional to
  1011.      * the number of address space areas belonging to as.
  1012.      * The check for conflicts is then attempted on the rightmost
  1013.      * record in the left neighbour, the leftmost record in the right
  1014.      * neighbour and all records in the leaf node itself.
  1015.      */
  1016.    
  1017.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1018.         if (a != avoid_area)
  1019.             return false;
  1020.     }
  1021.    
  1022.     /* First, check the two border cases. */
  1023.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1024.         a = (as_area_t *) node->value[node->keys - 1];
  1025.         mutex_lock(&a->lock);
  1026.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1027.             mutex_unlock(&a->lock);
  1028.             return false;
  1029.         }
  1030.         mutex_unlock(&a->lock);
  1031.     }
  1032.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1033.         a = (as_area_t *) node->value[0];
  1034.         mutex_lock(&a->lock);
  1035.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1036.             mutex_unlock(&a->lock);
  1037.             return false;
  1038.         }
  1039.         mutex_unlock(&a->lock);
  1040.     }
  1041.    
  1042.     /* Second, check the leaf node. */
  1043.     for (i = 0; i < leaf->keys; i++) {
  1044.         a = (as_area_t *) leaf->value[i];
  1045.    
  1046.         if (a == avoid_area)
  1047.             continue;
  1048.    
  1049.         mutex_lock(&a->lock);
  1050.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1051.             mutex_unlock(&a->lock);
  1052.             return false;
  1053.         }
  1054.         mutex_unlock(&a->lock);
  1055.     }
  1056.  
  1057.     /*
  1058.      * So far, the area does not conflict with other areas.
  1059.      * Check if it doesn't conflict with kernel address space.
  1060.      */  
  1061.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1062.         return !overlaps(va, size,
  1063.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1064.     }
  1065.  
  1066.     return true;
  1067. }
  1068.  
  1069. /** Return size of the address space area with given base.  */
  1070. size_t as_get_size(__address base)
  1071. {
  1072.     ipl_t ipl;
  1073.     as_area_t *src_area;
  1074.     size_t size;
  1075.  
  1076.     ipl = interrupts_disable();
  1077.     src_area = find_area_and_lock(AS, base);
  1078.     if (src_area){
  1079.         size = src_area->pages * PAGE_SIZE;
  1080.         mutex_unlock(&src_area->lock);
  1081.     } else {
  1082.         size = 0;
  1083.     }
  1084.     interrupts_restore(ipl);
  1085.     return size;
  1086. }
  1087.  
  1088. /** Mark portion of address space area as used.
  1089.  *
  1090.  * The address space area must be already locked.
  1091.  *
  1092.  * @param a Address space area.
  1093.  * @param page First page to be marked.
  1094.  * @param count Number of page to be marked.
  1095.  *
  1096.  * @return 0 on failure and 1 on success.
  1097.  */
  1098. int used_space_insert(as_area_t *a, __address page, count_t count)
  1099. {
  1100.     btree_node_t *leaf, *node;
  1101.     count_t pages;
  1102.     int i;
  1103.  
  1104.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1105.     ASSERT(count);
  1106.  
  1107.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1108.     if (pages) {
  1109.         /*
  1110.          * We hit the beginning of some used space.
  1111.          */
  1112.         return 0;
  1113.     }
  1114.  
  1115.     if (!leaf->keys) {
  1116.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1117.         return 1;
  1118.     }
  1119.  
  1120.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1121.     if (node) {
  1122.         __address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1123.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1124.        
  1125.         /*
  1126.          * Examine the possibility that the interval fits
  1127.          * somewhere between the rightmost interval of
  1128.          * the left neigbour and the first interval of the leaf.
  1129.          */
  1130.          
  1131.         if (page >= right_pg) {
  1132.             /* Do nothing. */
  1133.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1134.             /* The interval intersects with the left interval. */
  1135.             return 0;
  1136.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1137.             /* The interval intersects with the right interval. */
  1138.             return 0;          
  1139.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1140.             /* The interval can be added by merging the two already present intervals. */
  1141.             node->value[node->keys - 1] += count + right_cnt;
  1142.             btree_remove(&a->used_space, right_pg, leaf);
  1143.             return 1;
  1144.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1145.             /* The interval can be added by simply growing the left interval. */
  1146.             node->value[node->keys - 1] += count;
  1147.             return 1;
  1148.         } else if (page + count*PAGE_SIZE == right_pg) {
  1149.             /*
  1150.              * The interval can be addded by simply moving base of the right
  1151.              * interval down and increasing its size accordingly.
  1152.              */
  1153.             leaf->value[0] += count;
  1154.             leaf->key[0] = page;
  1155.             return 1;
  1156.         } else {
  1157.             /*
  1158.              * The interval is between both neigbouring intervals,
  1159.              * but cannot be merged with any of them.
  1160.              */
  1161.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1162.             return 1;
  1163.         }
  1164.     } else if (page < leaf->key[0]) {
  1165.         __address right_pg = leaf->key[0];
  1166.         count_t right_cnt = (count_t) leaf->value[0];
  1167.    
  1168.         /*
  1169.          * Investigate the border case in which the left neighbour does not
  1170.          * exist but the interval fits from the left.
  1171.          */
  1172.          
  1173.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1174.             /* The interval intersects with the right interval. */
  1175.             return 0;
  1176.         } else if (page + count*PAGE_SIZE == right_pg) {
  1177.             /*
  1178.              * The interval can be added by moving the base of the right interval down
  1179.              * and increasing its size accordingly.
  1180.              */
  1181.             leaf->key[0] = page;
  1182.             leaf->value[0] += count;
  1183.             return 1;
  1184.         } else {
  1185.             /*
  1186.              * The interval doesn't adjoin with the right interval.
  1187.              * It must be added individually.
  1188.              */
  1189.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1190.             return 1;
  1191.         }
  1192.     }
  1193.  
  1194.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1195.     if (node) {
  1196.         __address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1197.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1198.        
  1199.         /*
  1200.          * Examine the possibility that the interval fits
  1201.          * somewhere between the leftmost interval of
  1202.          * the right neigbour and the last interval of the leaf.
  1203.          */
  1204.  
  1205.         if (page < left_pg) {
  1206.             /* Do nothing. */
  1207.         } else 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 (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1211.             /* The interval intersects with the right interval. */
  1212.             return 0;          
  1213.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1214.             /* The interval can be added by merging the two already present intervals. */
  1215.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1216.             btree_remove(&a->used_space, right_pg, node);
  1217.             return 1;
  1218.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1219.             /* The interval can be added by simply growing the left interval. */
  1220.             leaf->value[leaf->keys - 1] +=  count;
  1221.             return 1;
  1222.         } else if (page + count*PAGE_SIZE == right_pg) {
  1223.             /*
  1224.              * The interval can be addded by simply moving base of the right
  1225.              * interval down and increasing its size accordingly.
  1226.              */
  1227.             node->value[0] += count;
  1228.             node->key[0] = page;
  1229.             return 1;
  1230.         } else {
  1231.             /*
  1232.              * The interval is between both neigbouring intervals,
  1233.              * but cannot be merged with any of them.
  1234.              */
  1235.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1236.             return 1;
  1237.         }
  1238.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1239.         __address left_pg = leaf->key[leaf->keys - 1];
  1240.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1241.    
  1242.         /*
  1243.          * Investigate the border case in which the right neighbour does not
  1244.          * exist but the interval fits from the right.
  1245.          */
  1246.          
  1247.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1248.             /* The interval intersects with the left interval. */
  1249.             return 0;
  1250.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1251.             /* The interval can be added by growing the left interval. */
  1252.             leaf->value[leaf->keys - 1] += count;
  1253.             return 1;
  1254.         } else {
  1255.             /*
  1256.              * The interval doesn't adjoin with the left interval.
  1257.              * It must be added individually.
  1258.              */
  1259.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1260.             return 1;
  1261.         }
  1262.     }
  1263.    
  1264.     /*
  1265.      * Note that if the algorithm made it thus far, the interval can fit only
  1266.      * between two other intervals of the leaf. The two border cases were already
  1267.      * resolved.
  1268.      */
  1269.     for (i = 1; i < leaf->keys; i++) {
  1270.         if (page < leaf->key[i]) {
  1271.             __address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1272.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1273.  
  1274.             /*
  1275.              * The interval fits between left_pg and right_pg.
  1276.              */
  1277.  
  1278.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1279.                 /* The interval intersects with the left interval. */
  1280.                 return 0;
  1281.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1282.                 /* The interval intersects with the right interval. */
  1283.                 return 0;          
  1284.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1285.                 /* The interval can be added by merging the two already present intervals. */
  1286.                 leaf->value[i - 1] += count + right_cnt;
  1287.                 btree_remove(&a->used_space, right_pg, leaf);
  1288.                 return 1;
  1289.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1290.                 /* The interval can be added by simply growing the left interval. */
  1291.                 leaf->value[i - 1] += count;
  1292.                 return 1;
  1293.             } else if (page + count*PAGE_SIZE == right_pg) {
  1294.                 /*
  1295.                      * The interval can be addded by simply moving base of the right
  1296.                  * interval down and increasing its size accordingly.
  1297.                  */
  1298.                 leaf->value[i] += count;
  1299.                 leaf->key[i] = page;
  1300.                 return 1;
  1301.             } else {
  1302.                 /*
  1303.                  * The interval is between both neigbouring intervals,
  1304.                  * but cannot be merged with any of them.
  1305.                  */
  1306.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1307.                 return 1;
  1308.             }
  1309.         }
  1310.     }
  1311.  
  1312.     panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
  1313. }
  1314.  
  1315. /** Mark portion of address space area as unused.
  1316.  *
  1317.  * The address space area must be already locked.
  1318.  *
  1319.  * @param a Address space area.
  1320.  * @param page First page to be marked.
  1321.  * @param count Number of page to be marked.
  1322.  *
  1323.  * @return 0 on failure and 1 on success.
  1324.  */
  1325. int used_space_remove(as_area_t *a, __address page, count_t count)
  1326. {
  1327.     btree_node_t *leaf, *node;
  1328.     count_t pages;
  1329.     int i;
  1330.  
  1331.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1332.     ASSERT(count);
  1333.  
  1334.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1335.     if (pages) {
  1336.         /*
  1337.          * We are lucky, page is the beginning of some interval.
  1338.          */
  1339.         if (count > pages) {
  1340.             return 0;
  1341.         } else if (count == pages) {
  1342.             btree_remove(&a->used_space, page, leaf);
  1343.             return 1;
  1344.         } else {
  1345.             /*
  1346.              * Find the respective interval.
  1347.              * Decrease its size and relocate its start address.
  1348.              */
  1349.             for (i = 0; i < leaf->keys; i++) {
  1350.                 if (leaf->key[i] == page) {
  1351.                     leaf->key[i] += count*PAGE_SIZE;
  1352.                     leaf->value[i] -= count;
  1353.                     return 1;
  1354.                 }
  1355.             }
  1356.             goto error;
  1357.         }
  1358.     }
  1359.  
  1360.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1361.     if (node && page < leaf->key[0]) {
  1362.         __address left_pg = node->key[node->keys - 1];
  1363.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1364.  
  1365.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1366.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1367.                 /*
  1368.                  * The interval is contained in the rightmost interval
  1369.                  * of the left neighbour and can be removed by
  1370.                  * updating the size of the bigger interval.
  1371.                  */
  1372.                 node->value[node->keys - 1] -= count;
  1373.                 return 1;
  1374.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1375.                 count_t new_cnt;
  1376.                
  1377.                 /*
  1378.                  * The interval is contained in the rightmost interval
  1379.                  * of the left neighbour but its removal requires
  1380.                  * both updating the size of the original interval and
  1381.                  * also inserting a new interval.
  1382.                  */
  1383.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1384.                 node->value[node->keys - 1] -= count + new_cnt;
  1385.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1386.                 return 1;
  1387.             }
  1388.         }
  1389.         return 0;
  1390.     } else if (page < leaf->key[0]) {
  1391.         return 0;
  1392.     }
  1393.    
  1394.     if (page > leaf->key[leaf->keys - 1]) {
  1395.         __address left_pg = leaf->key[leaf->keys - 1];
  1396.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1397.  
  1398.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1399.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1400.                 /*
  1401.                  * The interval is contained in the rightmost interval
  1402.                  * of the leaf and can be removed by updating the size
  1403.                  * of the bigger interval.
  1404.                  */
  1405.                 leaf->value[leaf->keys - 1] -= count;
  1406.                 return 1;
  1407.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1408.                 count_t new_cnt;
  1409.                
  1410.                 /*
  1411.                  * The interval is contained in the rightmost interval
  1412.                  * of the leaf but its removal requires both updating
  1413.                  * the size of the original interval and
  1414.                  * also inserting a new interval.
  1415.                  */
  1416.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1417.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1418.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1419.                 return 1;
  1420.             }
  1421.         }
  1422.         return 0;
  1423.     }  
  1424.    
  1425.     /*
  1426.      * The border cases have been already resolved.
  1427.      * Now the interval can be only between intervals of the leaf.
  1428.      */
  1429.     for (i = 1; i < leaf->keys - 1; i++) {
  1430.         if (page < leaf->key[i]) {
  1431.             __address left_pg = leaf->key[i - 1];
  1432.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1433.  
  1434.             /*
  1435.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1436.              */
  1437.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1438.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1439.                     /*
  1440.                     * The interval is contained in the interval (i - 1)
  1441.                      * of the leaf and can be removed by updating the size
  1442.                      * of the bigger interval.
  1443.                      */
  1444.                     leaf->value[i - 1] -= count;
  1445.                     return 1;
  1446.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1447.                     count_t new_cnt;
  1448.                
  1449.                     /*
  1450.                      * The interval is contained in the interval (i - 1)
  1451.                      * of the leaf but its removal requires both updating
  1452.                      * the size of the original interval and
  1453.                      * also inserting a new interval.
  1454.                      */
  1455.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1456.                     leaf->value[i - 1] -= count + new_cnt;
  1457.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1458.                     return 1;
  1459.                 }
  1460.             }
  1461.             return 0;
  1462.         }
  1463.     }
  1464.  
  1465. error:
  1466.     panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
  1467. }
  1468.  
  1469. /** Remove reference to address space area share info.
  1470.  *
  1471.  * If the reference count drops to 0, the sh_info is deallocated.
  1472.  *
  1473.  * @param sh_info Pointer to address space area share info.
  1474.  */
  1475. void sh_info_remove_reference(share_info_t *sh_info)
  1476. {
  1477.     bool dealloc = false;
  1478.  
  1479.     mutex_lock(&sh_info->lock);
  1480.     ASSERT(sh_info->refcount);
  1481.     if (--sh_info->refcount == 0) {
  1482.         dealloc = true;
  1483.         link_t *cur;
  1484.        
  1485.         /*
  1486.          * Now walk carefully the pagemap B+tree and free/remove
  1487.          * reference from all frames found there.
  1488.          */
  1489.         for (cur = sh_info->pagemap.leaf_head.next; cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1490.             btree_node_t *node;
  1491.             int i;
  1492.            
  1493.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1494.             for (i = 0; i < node->keys; i++)
  1495.                 frame_free(ADDR2PFN((__address) node->value[i]));
  1496.         }
  1497.        
  1498.     }
  1499.     mutex_unlock(&sh_info->lock);
  1500.    
  1501.     if (dealloc) {
  1502.         btree_destroy(&sh_info->pagemap);
  1503.         free(sh_info);
  1504.     }
  1505. }
  1506.  
  1507. /*
  1508.  * Address space related syscalls.
  1509.  */
  1510.  
  1511. /** Wrapper for as_area_create(). */
  1512. __native sys_as_area_create(__address address, size_t size, int flags)
  1513. {
  1514.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1515.         return (__native) address;
  1516.     else
  1517.         return (__native) -1;
  1518. }
  1519.  
  1520. /** Wrapper for as_area_resize. */
  1521. __native sys_as_area_resize(__address address, size_t size, int flags)
  1522. {
  1523.     return (__native) as_area_resize(AS, address, size, 0);
  1524. }
  1525.  
  1526. /** Wrapper for as_area_destroy. */
  1527. __native sys_as_area_destroy(__address address)
  1528. {
  1529.     return (__native) as_area_destroy(AS, address);
  1530. }
  1531.  
  1532.  /** @}
  1533.  */
  1534.  
  1535.