Subversion Repositories HelenOS-historic

Rev

Rev 1544 | Rev 1594 | 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.     link_t *cur;
  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.  
  151.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  152.         if (!as->cpu_refcount)
  153.             list_remove(&as->inactive_as_with_asid_link);
  154.         asid_put(as->asid);
  155.     }
  156.     spinlock_unlock(&inactive_as_with_asid_lock);
  157.  
  158.     /*
  159.      * Destroy address space areas of the address space.
  160.      */
  161.     for (cur = as->as_area_btree.leaf_head.next; cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  162.         btree_node_t *node;
  163.         int i;
  164.        
  165.         node = list_get_instance(cur, btree_node_t, leaf_link);
  166.         for (i = 0; i < node->keys; i++)
  167.             as_area_destroy(as, node->key[i]);
  168.     }
  169.  
  170.     btree_destroy(&as->as_area_btree);
  171.     page_table_destroy(as->page_table);
  172.  
  173.     interrupts_restore(ipl);
  174.    
  175.     free(as);
  176. }
  177.  
  178. /** Create address space area of common attributes.
  179.  *
  180.  * The created address space area is added to the target address space.
  181.  *
  182.  * @param as Target address space.
  183.  * @param flags Flags of the area memory.
  184.  * @param size Size of area.
  185.  * @param base Base address of area.
  186.  * @param attrs Attributes of the area.
  187.  * @param backend Address space area backend. NULL if no backend is used.
  188.  * @param backend_data NULL or a pointer to an array holding two void *.
  189.  *
  190.  * @return Address space area on success or NULL on failure.
  191.  */
  192. as_area_t *as_area_create(as_t *as, int flags, size_t size, __address base, int attrs,
  193.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  194. {
  195.     ipl_t ipl;
  196.     as_area_t *a;
  197.    
  198.     if (base % PAGE_SIZE)
  199.         return NULL;
  200.  
  201.     if (!size)
  202.         return NULL;
  203.  
  204.     /* Writeable executable areas are not supported. */
  205.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  206.         return NULL;
  207.    
  208.     ipl = interrupts_disable();
  209.     mutex_lock(&as->lock);
  210.    
  211.     if (!check_area_conflicts(as, base, size, NULL)) {
  212.         mutex_unlock(&as->lock);
  213.         interrupts_restore(ipl);
  214.         return NULL;
  215.     }
  216.    
  217.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  218.  
  219.     mutex_initialize(&a->lock);
  220.    
  221.     a->as = as;
  222.     a->flags = flags;
  223.     a->attributes = attrs;
  224.     a->pages = SIZE2FRAMES(size);
  225.     a->base = base;
  226.     a->sh_info = NULL;
  227.     a->backend = backend;
  228.     if (backend_data)
  229.         a->backend_data = *backend_data;
  230.     else
  231.         memsetb((__address) &a->backend_data, sizeof(a->backend_data), 0);
  232.  
  233.     btree_create(&a->used_space);
  234.    
  235.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  236.  
  237.     mutex_unlock(&as->lock);
  238.     interrupts_restore(ipl);
  239.  
  240.     return a;
  241. }
  242.  
  243. /** Find address space area and change it.
  244.  *
  245.  * @param as Address space.
  246.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  247.  * @param size New size of the virtual memory block starting at address.
  248.  * @param flags Flags influencing the remap operation. Currently unused.
  249.  *
  250.  * @return Zero on success or a value from @ref errno.h otherwise.
  251.  */
  252. int as_area_resize(as_t *as, __address address, size_t size, int flags)
  253. {
  254.     as_area_t *area;
  255.     ipl_t ipl;
  256.     size_t pages;
  257.    
  258.     ipl = interrupts_disable();
  259.     mutex_lock(&as->lock);
  260.    
  261.     /*
  262.      * Locate the area.
  263.      */
  264.     area = find_area_and_lock(as, address);
  265.     if (!area) {
  266.         mutex_unlock(&as->lock);
  267.         interrupts_restore(ipl);
  268.         return ENOENT;
  269.     }
  270.  
  271.     if (area->backend == &phys_backend) {
  272.         /*
  273.          * Remapping of address space areas associated
  274.          * with memory mapped devices is not supported.
  275.          */
  276.         mutex_unlock(&area->lock);
  277.         mutex_unlock(&as->lock);
  278.         interrupts_restore(ipl);
  279.         return ENOTSUP;
  280.     }
  281.     if (area->sh_info) {
  282.         /*
  283.          * Remapping of shared address space areas
  284.          * is not supported.
  285.          */
  286.         mutex_unlock(&area->lock);
  287.         mutex_unlock(&as->lock);
  288.         interrupts_restore(ipl);
  289.         return ENOTSUP;
  290.     }
  291.  
  292.     pages = SIZE2FRAMES((address - area->base) + size);
  293.     if (!pages) {
  294.         /*
  295.          * Zero size address space areas are not allowed.
  296.          */
  297.         mutex_unlock(&area->lock);
  298.         mutex_unlock(&as->lock);
  299.         interrupts_restore(ipl);
  300.         return EPERM;
  301.     }
  302.    
  303.     if (pages < area->pages) {
  304.         bool cond;
  305.         __address start_free = area->base + pages*PAGE_SIZE;
  306.  
  307.         /*
  308.          * Shrinking the area.
  309.          * No need to check for overlaps.
  310.          */
  311.  
  312.         /*
  313.          * Start TLB shootdown sequence.
  314.          */
  315.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  316.  
  317.         /*
  318.          * Remove frames belonging to used space starting from
  319.          * the highest addresses downwards until an overlap with
  320.          * the resized address space area is found. Note that this
  321.          * is also the right way to remove part of the used_space
  322.          * B+tree leaf list.
  323.          */    
  324.         for (cond = true; cond;) {
  325.             btree_node_t *node;
  326.        
  327.             ASSERT(!list_empty(&area->used_space.leaf_head));
  328.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  329.             if ((cond = (bool) node->keys)) {
  330.                 __address b = node->key[node->keys - 1];
  331.                 count_t c = (count_t) node->value[node->keys - 1];
  332.                 int i = 0;
  333.            
  334.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  335.                    
  336.                     if (b + c*PAGE_SIZE <= start_free) {
  337.                         /*
  338.                          * The whole interval fits completely
  339.                          * in the resized address space area.
  340.                          */
  341.                         break;
  342.                     }
  343.        
  344.                     /*
  345.                      * Part of the interval corresponding to b and c
  346.                      * overlaps with the resized address space area.
  347.                      */
  348.        
  349.                     cond = false;   /* we are almost done */
  350.                     i = (start_free - b) >> PAGE_WIDTH;
  351.                     if (!used_space_remove(area, start_free, c - i))
  352.                         panic("Could not remove used space.");
  353.                 } else {
  354.                     /*
  355.                      * The interval of used space can be completely removed.
  356.                      */
  357.                     if (!used_space_remove(area, b, c))
  358.                         panic("Could not remove used space.\n");
  359.                 }
  360.            
  361.                 for (; i < c; i++) {
  362.                     pte_t *pte;
  363.            
  364.                     page_table_lock(as, false);
  365.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  366.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  367.                     if (area->backend && area->backend->frame_free) {
  368.                         area->backend->frame_free(area,
  369.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  370.                     }
  371.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  372.                     page_table_unlock(as, false);
  373.                 }
  374.             }
  375.         }
  376.  
  377.         /*
  378.          * Finish TLB shootdown sequence.
  379.          */
  380.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  381.         tlb_shootdown_finalize();
  382.     } else {
  383.         /*
  384.          * Growing the area.
  385.          * Check for overlaps with other address space areas.
  386.          */
  387.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  388.             mutex_unlock(&area->lock);
  389.             mutex_unlock(&as->lock);       
  390.             interrupts_restore(ipl);
  391.             return EADDRNOTAVAIL;
  392.         }
  393.     }
  394.  
  395.     area->pages = pages;
  396.    
  397.     mutex_unlock(&area->lock);
  398.     mutex_unlock(&as->lock);
  399.     interrupts_restore(ipl);
  400.  
  401.     return 0;
  402. }
  403.  
  404. /** Destroy address space area.
  405.  *
  406.  * @param as Address space.
  407.  * @param address Address withing the area to be deleted.
  408.  *
  409.  * @return Zero on success or a value from @ref errno.h on failure.
  410.  */
  411. int as_area_destroy(as_t *as, __address address)
  412. {
  413.     as_area_t *area;
  414.     __address base;
  415.     link_t *cur;
  416.     ipl_t ipl;
  417.  
  418.     ipl = interrupts_disable();
  419.     mutex_lock(&as->lock);
  420.  
  421.     area = find_area_and_lock(as, address);
  422.     if (!area) {
  423.         mutex_unlock(&as->lock);
  424.         interrupts_restore(ipl);
  425.         return ENOENT;
  426.     }
  427.  
  428.     base = area->base;
  429.  
  430.     /*
  431.      * Start TLB shootdown sequence.
  432.      */
  433.     tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
  434.  
  435.     /*
  436.      * Visit only the pages mapped by used_space B+tree.
  437.      */
  438.     for (cur = area->used_space.leaf_head.next; cur != &area->used_space.leaf_head; cur = cur->next) {
  439.         btree_node_t *node;
  440.         int i;
  441.        
  442.         node = list_get_instance(cur, btree_node_t, leaf_link);
  443.         for (i = 0; i < node->keys; i++) {
  444.             __address b = node->key[i];
  445.             count_t j;
  446.             pte_t *pte;
  447.            
  448.             for (j = 0; j < (count_t) node->value[i]; j++) {
  449.                 page_table_lock(as, false);
  450.                 pte = page_mapping_find(as, b + j*PAGE_SIZE);
  451.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  452.                 if (area->backend && area->backend->frame_free) {
  453.                     area->backend->frame_free(area,
  454.                         b + j*PAGE_SIZE, PTE_GET_FRAME(pte));
  455.                 }
  456.                 page_mapping_remove(as, b + j*PAGE_SIZE);
  457.                 page_table_unlock(as, false);
  458.             }
  459.         }
  460.     }
  461.  
  462.     /*
  463.      * Finish TLB shootdown sequence.
  464.      */
  465.     tlb_invalidate_pages(AS->asid, area->base, area->pages);
  466.     tlb_shootdown_finalize();
  467.    
  468.     btree_destroy(&area->used_space);
  469.  
  470.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  471.    
  472.     if (area->sh_info)
  473.         sh_info_remove_reference(area->sh_info);
  474.        
  475.     mutex_unlock(&area->lock);
  476.  
  477.     /*
  478.      * Remove the empty area from address space.
  479.      */
  480.     btree_remove(&AS->as_area_btree, base, NULL);
  481.    
  482.     free(area);
  483.    
  484.     mutex_unlock(&AS->lock);
  485.     interrupts_restore(ipl);
  486.     return 0;
  487. }
  488.  
  489. /** Share address space area with another or the same address space.
  490.  *
  491.  * Address space area mapping is shared with a new address space area.
  492.  * If the source address space area has not been shared so far,
  493.  * a new sh_info is created. The new address space area simply gets the
  494.  * sh_info of the source area. The process of duplicating the
  495.  * mapping is done through the backend share function.
  496.  *
  497.  * @param src_as Pointer to source address space.
  498.  * @param src_base Base address of the source address space area.
  499.  * @param acc_size Expected size of the source area.
  500.  * @param dst_as Pointer to destination address space.
  501.  * @param dst_base Target base address.
  502.  * @param dst_flags_mask Destination address space area flags mask.
  503.  *
  504.  * @return Zero on success or ENOENT if there is no such task or
  505.  *     if there is no such address space area,
  506.  *     EPERM if there was a problem in accepting the area or
  507.  *     ENOMEM if there was a problem in allocating destination
  508.  *     address space area. ENOTSUP is returned if an attempt
  509.  *     to share non-anonymous address space area is detected.
  510.  */
  511. int as_area_share(as_t *src_as, __address src_base, size_t acc_size,
  512.           as_t *dst_as, __address dst_base, int dst_flags_mask)
  513. {
  514.     ipl_t ipl;
  515.     int src_flags;
  516.     size_t src_size;
  517.     as_area_t *src_area, *dst_area;
  518.     share_info_t *sh_info;
  519.     mem_backend_t *src_backend;
  520.     mem_backend_data_t src_backend_data;
  521.    
  522.     ipl = interrupts_disable();
  523.     mutex_lock(&src_as->lock);
  524.     src_area = find_area_and_lock(src_as, src_base);
  525.     if (!src_area) {
  526.         /*
  527.          * Could not find the source address space area.
  528.          */
  529.         mutex_unlock(&src_as->lock);
  530.         interrupts_restore(ipl);
  531.         return ENOENT;
  532.     }
  533.    
  534.     if (!src_area->backend || !src_area->backend->share) {
  535.         /*
  536.          * There is now backend or the backend does not
  537.          * know how to share the area.
  538.          */
  539.         mutex_unlock(&src_area->lock);
  540.         mutex_unlock(&src_as->lock);
  541.         interrupts_restore(ipl);
  542.         return ENOTSUP;
  543.     }
  544.    
  545.     src_size = src_area->pages * PAGE_SIZE;
  546.     src_flags = src_area->flags;
  547.     src_backend = src_area->backend;
  548.     src_backend_data = src_area->backend_data;
  549.  
  550.     /* Share the cacheable flag from the original mapping */
  551.     if (src_flags & AS_AREA_CACHEABLE)
  552.         dst_flags_mask |= AS_AREA_CACHEABLE;
  553.  
  554.     if (src_size != acc_size || (src_flags & dst_flags_mask) != dst_flags_mask) {
  555.         mutex_unlock(&src_area->lock);
  556.         mutex_unlock(&src_as->lock);
  557.         interrupts_restore(ipl);
  558.         return EPERM;
  559.     }
  560.  
  561.     /*
  562.      * Now we are committed to sharing the area.
  563.      * First prepare the area for sharing.
  564.      * Then it will be safe to unlock it.
  565.      */
  566.     sh_info = src_area->sh_info;
  567.     if (!sh_info) {
  568.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  569.         mutex_initialize(&sh_info->lock);
  570.         sh_info->refcount = 2;
  571.         btree_create(&sh_info->pagemap);
  572.         src_area->sh_info = sh_info;
  573.     } else {
  574.         mutex_lock(&sh_info->lock);
  575.         sh_info->refcount++;
  576.         mutex_unlock(&sh_info->lock);
  577.     }
  578.  
  579.     src_area->backend->share(src_area);
  580.  
  581.     mutex_unlock(&src_area->lock);
  582.     mutex_unlock(&src_as->lock);
  583.  
  584.     /*
  585.      * Create copy of the source address space area.
  586.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  587.      * attribute set which prevents race condition with
  588.      * preliminary as_page_fault() calls.
  589.      * The flags of the source area are masked against dst_flags_mask
  590.      * to support sharing in less privileged mode.
  591.      */
  592.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  593.                   AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  594.     if (!dst_area) {
  595.         /*
  596.          * Destination address space area could not be created.
  597.          */
  598.         sh_info_remove_reference(sh_info);
  599.        
  600.         interrupts_restore(ipl);
  601.         return ENOMEM;
  602.     }
  603.    
  604.     /*
  605.      * Now the destination address space area has been
  606.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  607.      * attribute and set the sh_info.
  608.      */
  609.     mutex_lock(&dst_area->lock);
  610.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  611.     dst_area->sh_info = sh_info;
  612.     mutex_unlock(&dst_area->lock);
  613.    
  614.     interrupts_restore(ipl);
  615.    
  616.     return 0;
  617. }
  618.  
  619. /** Check access mode for address space area.
  620.  *
  621.  * The address space area must be locked prior to this call.
  622.  *
  623.  * @param area Address space area.
  624.  * @param access Access mode.
  625.  *
  626.  * @return False if access violates area's permissions, true otherwise.
  627.  */
  628. bool as_area_check_access(as_area_t *area, pf_access_t access)
  629. {
  630.     int flagmap[] = {
  631.         [PF_ACCESS_READ] = AS_AREA_READ,
  632.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  633.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  634.     };
  635.  
  636.     if (!(area->flags & flagmap[access]))
  637.         return false;
  638.    
  639.     return true;
  640. }
  641.  
  642. /** Handle page fault within the current address space.
  643.  *
  644.  * This is the high-level page fault handler. It decides
  645.  * whether the page fault can be resolved by any backend
  646.  * and if so, it invokes the backend to resolve the page
  647.  * fault.
  648.  *
  649.  * Interrupts are assumed disabled.
  650.  *
  651.  * @param page Faulting page.
  652.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  653.  * @param istate Pointer to interrupted state.
  654.  *
  655.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  656.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  657.  */
  658. int as_page_fault(__address page, pf_access_t access, istate_t *istate)
  659. {
  660.     pte_t *pte;
  661.     as_area_t *area;
  662.    
  663.     if (!THREAD)
  664.         return AS_PF_FAULT;
  665.        
  666.     ASSERT(AS);
  667.  
  668.     mutex_lock(&AS->lock);
  669.     area = find_area_and_lock(AS, page);   
  670.     if (!area) {
  671.         /*
  672.          * No area contained mapping for 'page'.
  673.          * Signal page fault to low-level handler.
  674.          */
  675.         mutex_unlock(&AS->lock);
  676.         goto page_fault;
  677.     }
  678.  
  679.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  680.         /*
  681.          * The address space area is not fully initialized.
  682.          * Avoid possible race by returning error.
  683.          */
  684.         mutex_unlock(&area->lock);
  685.         mutex_unlock(&AS->lock);
  686.         goto page_fault;       
  687.     }
  688.  
  689.     if (!area->backend || !area->backend->page_fault) {
  690.         /*
  691.          * The address space area is not backed by any backend
  692.          * or the backend cannot handle page faults.
  693.          */
  694.         mutex_unlock(&area->lock);
  695.         mutex_unlock(&AS->lock);
  696.         goto page_fault;       
  697.     }
  698.  
  699.     page_table_lock(AS, false);
  700.    
  701.     /*
  702.      * To avoid race condition between two page faults
  703.      * on the same address, we need to make sure
  704.      * the mapping has not been already inserted.
  705.      */
  706.     if ((pte = page_mapping_find(AS, page))) {
  707.         if (PTE_PRESENT(pte)) {
  708.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  709.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  710.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  711.                 page_table_unlock(AS, false);
  712.                 mutex_unlock(&area->lock);
  713.                 mutex_unlock(&AS->lock);
  714.                 return AS_PF_OK;
  715.             }
  716.         }
  717.     }
  718.    
  719.     /*
  720.      * Resort to the backend page fault handler.
  721.      */
  722.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  723.         page_table_unlock(AS, false);
  724.         mutex_unlock(&area->lock);
  725.         mutex_unlock(&AS->lock);
  726.         goto page_fault;
  727.     }
  728.    
  729.     page_table_unlock(AS, false);
  730.     mutex_unlock(&area->lock);
  731.     mutex_unlock(&AS->lock);
  732.     return AS_PF_OK;
  733.  
  734. page_fault:
  735.     if (THREAD->in_copy_from_uspace) {
  736.         THREAD->in_copy_from_uspace = false;
  737.         istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
  738.     } else if (THREAD->in_copy_to_uspace) {
  739.         THREAD->in_copy_to_uspace = false;
  740.         istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
  741.     } else {
  742.         return AS_PF_FAULT;
  743.     }
  744.  
  745.     return AS_PF_DEFER;
  746. }
  747.  
  748. /** Switch address spaces.
  749.  *
  750.  * Note that this function cannot sleep as it is essentially a part of
  751.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  752.  *
  753.  * @param old Old address space or NULL.
  754.  * @param new New address space.
  755.  */
  756. void as_switch(as_t *old, as_t *new)
  757. {
  758.     ipl_t ipl;
  759.     bool needs_asid = false;
  760.    
  761.     ipl = interrupts_disable();
  762.     spinlock_lock(&inactive_as_with_asid_lock);
  763.  
  764.     /*
  765.      * First, take care of the old address space.
  766.      */
  767.     if (old) {
  768.         mutex_lock_active(&old->lock);
  769.         ASSERT(old->cpu_refcount);
  770.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  771.             /*
  772.              * The old address space is no longer active on
  773.              * any processor. It can be appended to the
  774.              * list of inactive address spaces with assigned
  775.              * ASID.
  776.              */
  777.              ASSERT(old->asid != ASID_INVALID);
  778.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  779.         }
  780.         mutex_unlock(&old->lock);
  781.     }
  782.  
  783.     /*
  784.      * Second, prepare the new address space.
  785.      */
  786.     mutex_lock_active(&new->lock);
  787.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  788.         if (new->asid != ASID_INVALID)
  789.             list_remove(&new->inactive_as_with_asid_link);
  790.         else
  791.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  792.     }
  793.     SET_PTL0_ADDRESS(new->page_table);
  794.     mutex_unlock(&new->lock);
  795.  
  796.     if (needs_asid) {
  797.         /*
  798.          * Allocation of new ASID was deferred
  799.          * until now in order to avoid deadlock.
  800.          */
  801.         asid_t asid;
  802.        
  803.         asid = asid_get();
  804.         mutex_lock_active(&new->lock);
  805.         new->asid = asid;
  806.         mutex_unlock(&new->lock);
  807.     }
  808.     spinlock_unlock(&inactive_as_with_asid_lock);
  809.     interrupts_restore(ipl);
  810.    
  811.     /*
  812.      * Perform architecture-specific steps.
  813.      * (e.g. write ASID to hardware register etc.)
  814.      */
  815.     as_install_arch(new);
  816.    
  817.     AS = new;
  818. }
  819.  
  820. /** Convert address space area flags to page flags.
  821.  *
  822.  * @param aflags Flags of some address space area.
  823.  *
  824.  * @return Flags to be passed to page_mapping_insert().
  825.  */
  826. int area_flags_to_page_flags(int aflags)
  827. {
  828.     int flags;
  829.  
  830.     flags = PAGE_USER | PAGE_PRESENT;
  831.    
  832.     if (aflags & AS_AREA_READ)
  833.         flags |= PAGE_READ;
  834.        
  835.     if (aflags & AS_AREA_WRITE)
  836.         flags |= PAGE_WRITE;
  837.    
  838.     if (aflags & AS_AREA_EXEC)
  839.         flags |= PAGE_EXEC;
  840.    
  841.     if (aflags & AS_AREA_CACHEABLE)
  842.         flags |= PAGE_CACHEABLE;
  843.        
  844.     return flags;
  845. }
  846.  
  847. /** Compute flags for virtual address translation subsytem.
  848.  *
  849.  * The address space area must be locked.
  850.  * Interrupts must be disabled.
  851.  *
  852.  * @param a Address space area.
  853.  *
  854.  * @return Flags to be used in page_mapping_insert().
  855.  */
  856. int as_area_get_flags(as_area_t *a)
  857. {
  858.     return area_flags_to_page_flags(a->flags);
  859. }
  860.  
  861. /** Create page table.
  862.  *
  863.  * Depending on architecture, create either address space
  864.  * private or global page table.
  865.  *
  866.  * @param flags Flags saying whether the page table is for kernel address space.
  867.  *
  868.  * @return First entry of the page table.
  869.  */
  870. pte_t *page_table_create(int flags)
  871. {
  872.         ASSERT(as_operations);
  873.         ASSERT(as_operations->page_table_create);
  874.  
  875.         return as_operations->page_table_create(flags);
  876. }
  877.  
  878. /** Destroy page table.
  879.  *
  880.  * Destroy page table in architecture specific way.
  881.  *
  882.  * @param page_table Physical address of PTL0.
  883.  */
  884. void page_table_destroy(pte_t *page_table)
  885. {
  886.         ASSERT(as_operations);
  887.         ASSERT(as_operations->page_table_destroy);
  888.  
  889.         as_operations->page_table_destroy(page_table);
  890. }
  891.  
  892. /** Lock page table.
  893.  *
  894.  * This function should be called before any page_mapping_insert(),
  895.  * page_mapping_remove() and page_mapping_find().
  896.  *
  897.  * Locking order is such that address space areas must be locked
  898.  * prior to this call. Address space can be locked prior to this
  899.  * call in which case the lock argument is false.
  900.  *
  901.  * @param as Address space.
  902.  * @param lock If false, do not attempt to lock as->lock.
  903.  */
  904. void page_table_lock(as_t *as, bool lock)
  905. {
  906.     ASSERT(as_operations);
  907.     ASSERT(as_operations->page_table_lock);
  908.  
  909.     as_operations->page_table_lock(as, lock);
  910. }
  911.  
  912. /** Unlock page table.
  913.  *
  914.  * @param as Address space.
  915.  * @param unlock If false, do not attempt to unlock as->lock.
  916.  */
  917. void page_table_unlock(as_t *as, bool unlock)
  918. {
  919.     ASSERT(as_operations);
  920.     ASSERT(as_operations->page_table_unlock);
  921.  
  922.     as_operations->page_table_unlock(as, unlock);
  923. }
  924.  
  925.  
  926. /** Find address space area and lock it.
  927.  *
  928.  * The address space must be locked and interrupts must be disabled.
  929.  *
  930.  * @param as Address space.
  931.  * @param va Virtual address.
  932.  *
  933.  * @return Locked address space area containing va on success or NULL on failure.
  934.  */
  935. as_area_t *find_area_and_lock(as_t *as, __address va)
  936. {
  937.     as_area_t *a;
  938.     btree_node_t *leaf, *lnode;
  939.     int i;
  940.    
  941.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  942.     if (a) {
  943.         /* va is the base address of an address space area */
  944.         mutex_lock(&a->lock);
  945.         return a;
  946.     }
  947.    
  948.     /*
  949.      * Search the leaf node and the righmost record of its left neighbour
  950.      * to find out whether this is a miss or va belongs to an address
  951.      * space area found there.
  952.      */
  953.    
  954.     /* First, search the leaf node itself. */
  955.     for (i = 0; i < leaf->keys; i++) {
  956.         a = (as_area_t *) leaf->value[i];
  957.         mutex_lock(&a->lock);
  958.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  959.             return a;
  960.         }
  961.         mutex_unlock(&a->lock);
  962.     }
  963.  
  964.     /*
  965.      * Second, locate the left neighbour and test its last record.
  966.      * Because of its position in the B+tree, it must have base < va.
  967.      */
  968.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  969.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  970.         mutex_lock(&a->lock);
  971.         if (va < a->base + a->pages * PAGE_SIZE) {
  972.             return a;
  973.         }
  974.         mutex_unlock(&a->lock);
  975.     }
  976.  
  977.     return NULL;
  978. }
  979.  
  980. /** Check area conflicts with other areas.
  981.  *
  982.  * The address space must be locked and interrupts must be disabled.
  983.  *
  984.  * @param as Address space.
  985.  * @param va Starting virtual address of the area being tested.
  986.  * @param size Size of the area being tested.
  987.  * @param avoid_area Do not touch this area.
  988.  *
  989.  * @return True if there is no conflict, false otherwise.
  990.  */
  991. bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area)
  992. {
  993.     as_area_t *a;
  994.     btree_node_t *leaf, *node;
  995.     int i;
  996.    
  997.     /*
  998.      * We don't want any area to have conflicts with NULL page.
  999.      */
  1000.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1001.         return false;
  1002.    
  1003.     /*
  1004.      * The leaf node is found in O(log n), where n is proportional to
  1005.      * the number of address space areas belonging to as.
  1006.      * The check for conflicts is then attempted on the rightmost
  1007.      * record in the left neighbour, the leftmost record in the right
  1008.      * neighbour and all records in the leaf node itself.
  1009.      */
  1010.    
  1011.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1012.         if (a != avoid_area)
  1013.             return false;
  1014.     }
  1015.    
  1016.     /* First, check the two border cases. */
  1017.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1018.         a = (as_area_t *) node->value[node->keys - 1];
  1019.         mutex_lock(&a->lock);
  1020.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1021.             mutex_unlock(&a->lock);
  1022.             return false;
  1023.         }
  1024.         mutex_unlock(&a->lock);
  1025.     }
  1026.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1027.         a = (as_area_t *) node->value[0];
  1028.         mutex_lock(&a->lock);
  1029.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1030.             mutex_unlock(&a->lock);
  1031.             return false;
  1032.         }
  1033.         mutex_unlock(&a->lock);
  1034.     }
  1035.    
  1036.     /* Second, check the leaf node. */
  1037.     for (i = 0; i < leaf->keys; i++) {
  1038.         a = (as_area_t *) leaf->value[i];
  1039.    
  1040.         if (a == avoid_area)
  1041.             continue;
  1042.    
  1043.         mutex_lock(&a->lock);
  1044.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1045.             mutex_unlock(&a->lock);
  1046.             return false;
  1047.         }
  1048.         mutex_unlock(&a->lock);
  1049.     }
  1050.  
  1051.     /*
  1052.      * So far, the area does not conflict with other areas.
  1053.      * Check if it doesn't conflict with kernel address space.
  1054.      */  
  1055.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1056.         return !overlaps(va, size,
  1057.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1058.     }
  1059.  
  1060.     return true;
  1061. }
  1062.  
  1063. /** Return size of the address space area with given base.  */
  1064. size_t as_get_size(__address base)
  1065. {
  1066.     ipl_t ipl;
  1067.     as_area_t *src_area;
  1068.     size_t size;
  1069.  
  1070.     ipl = interrupts_disable();
  1071.     src_area = find_area_and_lock(AS, base);
  1072.     if (src_area){
  1073.         size = src_area->pages * PAGE_SIZE;
  1074.         mutex_unlock(&src_area->lock);
  1075.     } else {
  1076.         size = 0;
  1077.     }
  1078.     interrupts_restore(ipl);
  1079.     return size;
  1080. }
  1081.  
  1082. /** Mark portion of address space area as used.
  1083.  *
  1084.  * The address space area must be already locked.
  1085.  *
  1086.  * @param a Address space area.
  1087.  * @param page First page to be marked.
  1088.  * @param count Number of page to be marked.
  1089.  *
  1090.  * @return 0 on failure and 1 on success.
  1091.  */
  1092. int used_space_insert(as_area_t *a, __address page, count_t count)
  1093. {
  1094.     btree_node_t *leaf, *node;
  1095.     count_t pages;
  1096.     int i;
  1097.  
  1098.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1099.     ASSERT(count);
  1100.  
  1101.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1102.     if (pages) {
  1103.         /*
  1104.          * We hit the beginning of some used space.
  1105.          */
  1106.         return 0;
  1107.     }
  1108.  
  1109.     if (!leaf->keys) {
  1110.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1111.         return 1;
  1112.     }
  1113.  
  1114.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1115.     if (node) {
  1116.         __address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1117.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1118.        
  1119.         /*
  1120.          * Examine the possibility that the interval fits
  1121.          * somewhere between the rightmost interval of
  1122.          * the left neigbour and the first interval of the leaf.
  1123.          */
  1124.          
  1125.         if (page >= right_pg) {
  1126.             /* Do nothing. */
  1127.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1128.             /* The interval intersects with the left interval. */
  1129.             return 0;
  1130.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1131.             /* The interval intersects with the right interval. */
  1132.             return 0;          
  1133.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1134.             /* The interval can be added by merging the two already present intervals. */
  1135.             node->value[node->keys - 1] += count + right_cnt;
  1136.             btree_remove(&a->used_space, right_pg, leaf);
  1137.             return 1;
  1138.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1139.             /* The interval can be added by simply growing the left interval. */
  1140.             node->value[node->keys - 1] += count;
  1141.             return 1;
  1142.         } else if (page + count*PAGE_SIZE == right_pg) {
  1143.             /*
  1144.              * The interval can be addded by simply moving base of the right
  1145.              * interval down and increasing its size accordingly.
  1146.              */
  1147.             leaf->value[0] += count;
  1148.             leaf->key[0] = page;
  1149.             return 1;
  1150.         } else {
  1151.             /*
  1152.              * The interval is between both neigbouring intervals,
  1153.              * but cannot be merged with any of them.
  1154.              */
  1155.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1156.             return 1;
  1157.         }
  1158.     } else if (page < leaf->key[0]) {
  1159.         __address right_pg = leaf->key[0];
  1160.         count_t right_cnt = (count_t) leaf->value[0];
  1161.    
  1162.         /*
  1163.          * Investigate the border case in which the left neighbour does not
  1164.          * exist but the interval fits from the left.
  1165.          */
  1166.          
  1167.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1168.             /* The interval intersects with the right interval. */
  1169.             return 0;
  1170.         } else if (page + count*PAGE_SIZE == right_pg) {
  1171.             /*
  1172.              * The interval can be added by moving the base of the right interval down
  1173.              * and increasing its size accordingly.
  1174.              */
  1175.             leaf->key[0] = page;
  1176.             leaf->value[0] += count;
  1177.             return 1;
  1178.         } else {
  1179.             /*
  1180.              * The interval doesn't adjoin with the right interval.
  1181.              * It must be added individually.
  1182.              */
  1183.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1184.             return 1;
  1185.         }
  1186.     }
  1187.  
  1188.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1189.     if (node) {
  1190.         __address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1191.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1192.        
  1193.         /*
  1194.          * Examine the possibility that the interval fits
  1195.          * somewhere between the leftmost interval of
  1196.          * the right neigbour and the last interval of the leaf.
  1197.          */
  1198.  
  1199.         if (page < left_pg) {
  1200.             /* Do nothing. */
  1201.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1202.             /* The interval intersects with the left interval. */
  1203.             return 0;
  1204.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1205.             /* The interval intersects with the right interval. */
  1206.             return 0;          
  1207.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1208.             /* The interval can be added by merging the two already present intervals. */
  1209.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1210.             btree_remove(&a->used_space, right_pg, node);
  1211.             return 1;
  1212.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1213.             /* The interval can be added by simply growing the left interval. */
  1214.             leaf->value[leaf->keys - 1] +=  count;
  1215.             return 1;
  1216.         } else if (page + count*PAGE_SIZE == right_pg) {
  1217.             /*
  1218.              * The interval can be addded by simply moving base of the right
  1219.              * interval down and increasing its size accordingly.
  1220.              */
  1221.             node->value[0] += count;
  1222.             node->key[0] = page;
  1223.             return 1;
  1224.         } else {
  1225.             /*
  1226.              * The interval is between both neigbouring intervals,
  1227.              * but cannot be merged with any of them.
  1228.              */
  1229.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1230.             return 1;
  1231.         }
  1232.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1233.         __address left_pg = leaf->key[leaf->keys - 1];
  1234.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1235.    
  1236.         /*
  1237.          * Investigate the border case in which the right neighbour does not
  1238.          * exist but the interval fits from the right.
  1239.          */
  1240.          
  1241.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1242.             /* The interval intersects with the left interval. */
  1243.             return 0;
  1244.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1245.             /* The interval can be added by growing the left interval. */
  1246.             leaf->value[leaf->keys - 1] += count;
  1247.             return 1;
  1248.         } else {
  1249.             /*
  1250.              * The interval doesn't adjoin with the left interval.
  1251.              * It must be added individually.
  1252.              */
  1253.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1254.             return 1;
  1255.         }
  1256.     }
  1257.    
  1258.     /*
  1259.      * Note that if the algorithm made it thus far, the interval can fit only
  1260.      * between two other intervals of the leaf. The two border cases were already
  1261.      * resolved.
  1262.      */
  1263.     for (i = 1; i < leaf->keys; i++) {
  1264.         if (page < leaf->key[i]) {
  1265.             __address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1266.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1267.  
  1268.             /*
  1269.              * The interval fits between left_pg and right_pg.
  1270.              */
  1271.  
  1272.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1273.                 /* The interval intersects with the left interval. */
  1274.                 return 0;
  1275.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1276.                 /* The interval intersects with the right interval. */
  1277.                 return 0;          
  1278.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1279.                 /* The interval can be added by merging the two already present intervals. */
  1280.                 leaf->value[i - 1] += count + right_cnt;
  1281.                 btree_remove(&a->used_space, right_pg, leaf);
  1282.                 return 1;
  1283.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1284.                 /* The interval can be added by simply growing the left interval. */
  1285.                 leaf->value[i - 1] += count;
  1286.                 return 1;
  1287.             } else if (page + count*PAGE_SIZE == right_pg) {
  1288.                 /*
  1289.                      * The interval can be addded by simply moving base of the right
  1290.                  * interval down and increasing its size accordingly.
  1291.                  */
  1292.                 leaf->value[i] += count;
  1293.                 leaf->key[i] = page;
  1294.                 return 1;
  1295.             } else {
  1296.                 /*
  1297.                  * The interval is between both neigbouring intervals,
  1298.                  * but cannot be merged with any of them.
  1299.                  */
  1300.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1301.                 return 1;
  1302.             }
  1303.         }
  1304.     }
  1305.  
  1306.     panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
  1307. }
  1308.  
  1309. /** Mark portion of address space area as unused.
  1310.  *
  1311.  * The address space area must be already locked.
  1312.  *
  1313.  * @param a Address space area.
  1314.  * @param page First page to be marked.
  1315.  * @param count Number of page to be marked.
  1316.  *
  1317.  * @return 0 on failure and 1 on success.
  1318.  */
  1319. int used_space_remove(as_area_t *a, __address page, count_t count)
  1320. {
  1321.     btree_node_t *leaf, *node;
  1322.     count_t pages;
  1323.     int i;
  1324.  
  1325.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1326.     ASSERT(count);
  1327.  
  1328.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1329.     if (pages) {
  1330.         /*
  1331.          * We are lucky, page is the beginning of some interval.
  1332.          */
  1333.         if (count > pages) {
  1334.             return 0;
  1335.         } else if (count == pages) {
  1336.             btree_remove(&a->used_space, page, leaf);
  1337.             return 1;
  1338.         } else {
  1339.             /*
  1340.              * Find the respective interval.
  1341.              * Decrease its size and relocate its start address.
  1342.              */
  1343.             for (i = 0; i < leaf->keys; i++) {
  1344.                 if (leaf->key[i] == page) {
  1345.                     leaf->key[i] += count*PAGE_SIZE;
  1346.                     leaf->value[i] -= count;
  1347.                     return 1;
  1348.                 }
  1349.             }
  1350.             goto error;
  1351.         }
  1352.     }
  1353.  
  1354.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1355.     if (node && page < leaf->key[0]) {
  1356.         __address left_pg = node->key[node->keys - 1];
  1357.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1358.  
  1359.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1360.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1361.                 /*
  1362.                  * The interval is contained in the rightmost interval
  1363.                  * of the left neighbour and can be removed by
  1364.                  * updating the size of the bigger interval.
  1365.                  */
  1366.                 node->value[node->keys - 1] -= count;
  1367.                 return 1;
  1368.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1369.                 count_t new_cnt;
  1370.                
  1371.                 /*
  1372.                  * The interval is contained in the rightmost interval
  1373.                  * of the left neighbour but its removal requires
  1374.                  * both updating the size of the original interval and
  1375.                  * also inserting a new interval.
  1376.                  */
  1377.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1378.                 node->value[node->keys - 1] -= count + new_cnt;
  1379.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1380.                 return 1;
  1381.             }
  1382.         }
  1383.         return 0;
  1384.     } else if (page < leaf->key[0]) {
  1385.         return 0;
  1386.     }
  1387.    
  1388.     if (page > leaf->key[leaf->keys - 1]) {
  1389.         __address left_pg = leaf->key[leaf->keys - 1];
  1390.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1391.  
  1392.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1393.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1394.                 /*
  1395.                  * The interval is contained in the rightmost interval
  1396.                  * of the leaf and can be removed by updating the size
  1397.                  * of the bigger interval.
  1398.                  */
  1399.                 leaf->value[leaf->keys - 1] -= count;
  1400.                 return 1;
  1401.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1402.                 count_t new_cnt;
  1403.                
  1404.                 /*
  1405.                  * The interval is contained in the rightmost interval
  1406.                  * of the leaf but its removal requires both updating
  1407.                  * the size of the original interval and
  1408.                  * also inserting a new interval.
  1409.                  */
  1410.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1411.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1412.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1413.                 return 1;
  1414.             }
  1415.         }
  1416.         return 0;
  1417.     }  
  1418.    
  1419.     /*
  1420.      * The border cases have been already resolved.
  1421.      * Now the interval can be only between intervals of the leaf.
  1422.      */
  1423.     for (i = 1; i < leaf->keys - 1; i++) {
  1424.         if (page < leaf->key[i]) {
  1425.             __address left_pg = leaf->key[i - 1];
  1426.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1427.  
  1428.             /*
  1429.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1430.              */
  1431.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1432.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1433.                     /*
  1434.                     * The interval is contained in the interval (i - 1)
  1435.                      * of the leaf and can be removed by updating the size
  1436.                      * of the bigger interval.
  1437.                      */
  1438.                     leaf->value[i - 1] -= count;
  1439.                     return 1;
  1440.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1441.                     count_t new_cnt;
  1442.                
  1443.                     /*
  1444.                      * The interval is contained in the interval (i - 1)
  1445.                      * of the leaf but its removal requires both updating
  1446.                      * the size of the original interval and
  1447.                      * also inserting a new interval.
  1448.                      */
  1449.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1450.                     leaf->value[i - 1] -= count + new_cnt;
  1451.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1452.                     return 1;
  1453.                 }
  1454.             }
  1455.             return 0;
  1456.         }
  1457.     }
  1458.  
  1459. error:
  1460.     panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
  1461. }
  1462.  
  1463. /** Remove reference to address space area share info.
  1464.  *
  1465.  * If the reference count drops to 0, the sh_info is deallocated.
  1466.  *
  1467.  * @param sh_info Pointer to address space area share info.
  1468.  */
  1469. void sh_info_remove_reference(share_info_t *sh_info)
  1470. {
  1471.     bool dealloc = false;
  1472.  
  1473.     mutex_lock(&sh_info->lock);
  1474.     ASSERT(sh_info->refcount);
  1475.     if (--sh_info->refcount == 0) {
  1476.         dealloc = true;
  1477.         link_t *cur;
  1478.        
  1479.         /*
  1480.          * Now walk carefully the pagemap B+tree and free/remove
  1481.          * reference from all frames found there.
  1482.          */
  1483.         for (cur = sh_info->pagemap.leaf_head.next; cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1484.             btree_node_t *node;
  1485.             int i;
  1486.            
  1487.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1488.             for (i = 0; i < node->keys; i++)
  1489.                 frame_free(ADDR2PFN((__address) node->value[i]));
  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.