Subversion Repositories HelenOS

Rev

Rev 1851 | Rev 1890 | 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 spaces 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. /**
  82.  * Each architecture decides what functions will be used to carry out
  83.  * address space operations such as creating or locking page tables.
  84.  */
  85. as_operations_t *as_operations = NULL;
  86.  
  87. /** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
  88. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  89.  
  90. /**
  91.  * This list contains address spaces that are not active on any
  92.  * processor and that have valid ASID.
  93.  */
  94. LIST_INITIALIZE(inactive_as_with_asid_head);
  95.  
  96. /** Kernel address space. */
  97. as_t *AS_KERNEL = NULL;
  98.  
  99. static int area_flags_to_page_flags(int aflags);
  100. static as_area_t *find_area_and_lock(as_t *as, uintptr_t va);
  101. static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size, as_area_t *avoid_area);
  102. static void sh_info_remove_reference(share_info_t *sh_info);
  103.  
  104. /** Initialize address space subsystem. */
  105. void as_init(void)
  106. {
  107.     as_arch_init();
  108.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  109.     if (!AS_KERNEL)
  110.         panic("can't create kernel address space\n");
  111.    
  112. }
  113.  
  114. /** Create address space.
  115.  *
  116.  * @param flags Flags that influence way in wich the address space is created.
  117.  */
  118. as_t *as_create(int flags)
  119. {
  120.     as_t *as;
  121.  
  122.     as = (as_t *) malloc(sizeof(as_t), 0);
  123.     link_initialize(&as->inactive_as_with_asid_link);
  124.     mutex_initialize(&as->lock);
  125.     btree_create(&as->as_area_btree);
  126.    
  127.     if (flags & FLAG_AS_KERNEL)
  128.         as->asid = ASID_KERNEL;
  129.     else
  130.         as->asid = ASID_INVALID;
  131.    
  132.     as->refcount = 0;
  133.     as->cpu_refcount = 0;
  134.     as->page_table = page_table_create(flags);
  135.  
  136.     return as;
  137. }
  138.  
  139. /** Destroy adress space.
  140.  *
  141.  * When there are no tasks referencing this address space (i.e. its refcount is zero),
  142.  * the address space can be destroyed.
  143.  */
  144. void as_destroy(as_t *as)
  145. {
  146.     ipl_t ipl;
  147.     bool cond;
  148.  
  149.     ASSERT(as->refcount == 0);
  150.    
  151.     /*
  152.      * Since there is no reference to this area,
  153.      * it is safe not to lock its mutex.
  154.      */
  155.     ipl = interrupts_disable();
  156.     spinlock_lock(&inactive_as_with_asid_lock);
  157.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  158.         if (as != AS && as->cpu_refcount == 0)
  159.             list_remove(&as->inactive_as_with_asid_link);
  160.         asid_put(as->asid);
  161.     }
  162.     spinlock_unlock(&inactive_as_with_asid_lock);
  163.  
  164.     /*
  165.      * Destroy address space areas of the address space.
  166.      * The B+tee must be walked carefully because it is
  167.      * also being destroyed.
  168.      */
  169.     for (cond = true; cond; ) {
  170.         btree_node_t *node;
  171.  
  172.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  173.         node = list_get_instance(as->as_area_btree.leaf_head.next, btree_node_t, leaf_link);
  174.  
  175.         if ((cond = node->keys)) {
  176.             as_area_destroy(as, node->key[0]);
  177.         }
  178.     }
  179.  
  180.     btree_destroy(&as->as_area_btree);
  181.     page_table_destroy(as->page_table);
  182.  
  183.     interrupts_restore(ipl);
  184.    
  185.     free(as);
  186. }
  187.  
  188. /** Create address space area of common attributes.
  189.  *
  190.  * The created address space area is added to the target address space.
  191.  *
  192.  * @param as Target address space.
  193.  * @param flags Flags of the area memory.
  194.  * @param size Size of area.
  195.  * @param base Base address of area.
  196.  * @param attrs Attributes of the area.
  197.  * @param backend Address space area backend. NULL if no backend is used.
  198.  * @param backend_data NULL or a pointer to an array holding two void *.
  199.  *
  200.  * @return Address space area on success or NULL on failure.
  201.  */
  202. as_area_t *as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
  203.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  204. {
  205.     ipl_t ipl;
  206.     as_area_t *a;
  207.    
  208.     if (base % PAGE_SIZE)
  209.         return NULL;
  210.  
  211.     if (!size)
  212.         return NULL;
  213.  
  214.     /* Writeable executable areas are not supported. */
  215.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  216.         return NULL;
  217.    
  218.     ipl = interrupts_disable();
  219.     mutex_lock(&as->lock);
  220.    
  221.     if (!check_area_conflicts(as, base, size, NULL)) {
  222.         mutex_unlock(&as->lock);
  223.         interrupts_restore(ipl);
  224.         return NULL;
  225.     }
  226.    
  227.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  228.  
  229.     mutex_initialize(&a->lock);
  230.    
  231.     a->as = as;
  232.     a->flags = flags;
  233.     a->attributes = attrs;
  234.     a->pages = SIZE2FRAMES(size);
  235.     a->base = base;
  236.     a->sh_info = NULL;
  237.     a->backend = backend;
  238.     if (backend_data)
  239.         a->backend_data = *backend_data;
  240.     else
  241.         memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data), 0);
  242.  
  243.     btree_create(&a->used_space);
  244.    
  245.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  246.  
  247.     mutex_unlock(&as->lock);
  248.     interrupts_restore(ipl);
  249.  
  250.     return a;
  251. }
  252.  
  253. /** Find address space area and change it.
  254.  *
  255.  * @param as Address space.
  256.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  257.  * @param size New size of the virtual memory block starting at address.
  258.  * @param flags Flags influencing the remap operation. Currently unused.
  259.  *
  260.  * @return Zero on success or a value from @ref errno.h otherwise.
  261.  */
  262. int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
  263. {
  264.     as_area_t *area;
  265.     ipl_t ipl;
  266.     size_t pages;
  267.    
  268.     ipl = interrupts_disable();
  269.     mutex_lock(&as->lock);
  270.    
  271.     /*
  272.      * Locate the area.
  273.      */
  274.     area = find_area_and_lock(as, address);
  275.     if (!area) {
  276.         mutex_unlock(&as->lock);
  277.         interrupts_restore(ipl);
  278.         return ENOENT;
  279.     }
  280.  
  281.     if (area->backend == &phys_backend) {
  282.         /*
  283.          * Remapping of address space areas associated
  284.          * with memory mapped devices is not supported.
  285.          */
  286.         mutex_unlock(&area->lock);
  287.         mutex_unlock(&as->lock);
  288.         interrupts_restore(ipl);
  289.         return ENOTSUP;
  290.     }
  291.     if (area->sh_info) {
  292.         /*
  293.          * Remapping of shared address space areas
  294.          * is not supported.
  295.          */
  296.         mutex_unlock(&area->lock);
  297.         mutex_unlock(&as->lock);
  298.         interrupts_restore(ipl);
  299.         return ENOTSUP;
  300.     }
  301.  
  302.     pages = SIZE2FRAMES((address - area->base) + size);
  303.     if (!pages) {
  304.         /*
  305.          * Zero size address space areas are not allowed.
  306.          */
  307.         mutex_unlock(&area->lock);
  308.         mutex_unlock(&as->lock);
  309.         interrupts_restore(ipl);
  310.         return EPERM;
  311.     }
  312.    
  313.     if (pages < area->pages) {
  314.         bool cond;
  315.         uintptr_t start_free = area->base + pages*PAGE_SIZE;
  316.  
  317.         /*
  318.          * Shrinking the area.
  319.          * No need to check for overlaps.
  320.          */
  321.  
  322.         /*
  323.          * Start TLB shootdown sequence.
  324.          */
  325.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  326.  
  327.         /*
  328.          * Remove frames belonging to used space starting from
  329.          * the highest addresses downwards until an overlap with
  330.          * the resized address space area is found. Note that this
  331.          * is also the right way to remove part of the used_space
  332.          * B+tree leaf list.
  333.          */    
  334.         for (cond = true; cond;) {
  335.             btree_node_t *node;
  336.        
  337.             ASSERT(!list_empty(&area->used_space.leaf_head));
  338.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  339.             if ((cond = (bool) node->keys)) {
  340.                 uintptr_t b = node->key[node->keys - 1];
  341.                 count_t c = (count_t) node->value[node->keys - 1];
  342.                 int i = 0;
  343.            
  344.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  345.                    
  346.                     if (b + c*PAGE_SIZE <= start_free) {
  347.                         /*
  348.                          * The whole interval fits completely
  349.                          * in the resized address space area.
  350.                          */
  351.                         break;
  352.                     }
  353.        
  354.                     /*
  355.                      * Part of the interval corresponding to b and c
  356.                      * overlaps with the resized address space area.
  357.                      */
  358.        
  359.                     cond = false;   /* we are almost done */
  360.                     i = (start_free - b) >> PAGE_WIDTH;
  361.                     if (!used_space_remove(area, start_free, c - i))
  362.                         panic("Could not remove used space.\n");
  363.                 } else {
  364.                     /*
  365.                      * The interval of used space can be completely removed.
  366.                      */
  367.                     if (!used_space_remove(area, b, c))
  368.                         panic("Could not remove used space.\n");
  369.                 }
  370.            
  371.                 for (; i < c; i++) {
  372.                     pte_t *pte;
  373.            
  374.                     page_table_lock(as, false);
  375.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  376.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  377.                     if (area->backend && area->backend->frame_free) {
  378.                         area->backend->frame_free(area,
  379.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  380.                     }
  381.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  382.                     page_table_unlock(as, false);
  383.                 }
  384.             }
  385.         }
  386.  
  387.         /*
  388.          * Finish TLB shootdown sequence.
  389.          */
  390.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  391.         tlb_shootdown_finalize();
  392.        
  393.         /*
  394.          * Invalidate software translation caches (e.g. TSB on sparc64).
  395.          */
  396.         as_invalidate_translation_cache(as, area->base + pages*PAGE_SIZE, area->pages - pages);
  397.     } else {
  398.         /*
  399.          * Growing the area.
  400.          * Check for overlaps with other address space areas.
  401.          */
  402.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  403.             mutex_unlock(&area->lock);
  404.             mutex_unlock(&as->lock);       
  405.             interrupts_restore(ipl);
  406.             return EADDRNOTAVAIL;
  407.         }
  408.     }
  409.  
  410.     area->pages = pages;
  411.    
  412.     mutex_unlock(&area->lock);
  413.     mutex_unlock(&as->lock);
  414.     interrupts_restore(ipl);
  415.  
  416.     return 0;
  417. }
  418.  
  419. /** Destroy address space area.
  420.  *
  421.  * @param as Address space.
  422.  * @param address Address withing the area to be deleted.
  423.  *
  424.  * @return Zero on success or a value from @ref errno.h on failure.
  425.  */
  426. int as_area_destroy(as_t *as, uintptr_t address)
  427. {
  428.     as_area_t *area;
  429.     uintptr_t base;
  430.     link_t *cur;
  431.     ipl_t ipl;
  432.  
  433.     ipl = interrupts_disable();
  434.     mutex_lock(&as->lock);
  435.  
  436.     area = find_area_and_lock(as, address);
  437.     if (!area) {
  438.         mutex_unlock(&as->lock);
  439.         interrupts_restore(ipl);
  440.         return ENOENT;
  441.     }
  442.  
  443.     base = area->base;
  444.  
  445.     /*
  446.      * Start TLB shootdown sequence.
  447.      */
  448.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  449.  
  450.     /*
  451.      * Visit only the pages mapped by used_space B+tree.
  452.      */
  453.     for (cur = area->used_space.leaf_head.next; cur != &area->used_space.leaf_head; cur = cur->next) {
  454.         btree_node_t *node;
  455.         int i;
  456.        
  457.         node = list_get_instance(cur, btree_node_t, leaf_link);
  458.         for (i = 0; i < node->keys; i++) {
  459.             uintptr_t b = node->key[i];
  460.             count_t j;
  461.             pte_t *pte;
  462.            
  463.             for (j = 0; j < (count_t) node->value[i]; j++) {
  464.                 page_table_lock(as, false);
  465.                 pte = page_mapping_find(as, b + j*PAGE_SIZE);
  466.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  467.                 if (area->backend && area->backend->frame_free) {
  468.                     area->backend->frame_free(area,
  469.                         b + j*PAGE_SIZE, PTE_GET_FRAME(pte));
  470.                 }
  471.                 page_mapping_remove(as, b + j*PAGE_SIZE);              
  472.                 page_table_unlock(as, false);
  473.             }
  474.         }
  475.     }
  476.  
  477.     /*
  478.      * Finish TLB shootdown sequence.
  479.      */
  480.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  481.     tlb_shootdown_finalize();
  482.    
  483.     /*
  484.      * Invalidate potential software translation caches (e.g. TSB on sparc64).
  485.      */
  486.     as_invalidate_translation_cache(as, area->base, area->pages);
  487.    
  488.     btree_destroy(&area->used_space);
  489.  
  490.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  491.    
  492.     if (area->sh_info)
  493.         sh_info_remove_reference(area->sh_info);
  494.        
  495.     mutex_unlock(&area->lock);
  496.  
  497.     /*
  498.      * Remove the empty area from address space.
  499.      */
  500.     btree_remove(&as->as_area_btree, base, NULL);
  501.    
  502.     free(area);
  503.    
  504.     mutex_unlock(&as->lock);
  505.     interrupts_restore(ipl);
  506.     return 0;
  507. }
  508.  
  509. /** Share address space area with another or the same address space.
  510.  *
  511.  * Address space area mapping is shared with a new address space area.
  512.  * If the source address space area has not been shared so far,
  513.  * a new sh_info is created. The new address space area simply gets the
  514.  * sh_info of the source area. The process of duplicating the
  515.  * mapping is done through the backend share function.
  516.  *
  517.  * @param src_as Pointer to source address space.
  518.  * @param src_base Base address of the source address space area.
  519.  * @param acc_size Expected size of the source area.
  520.  * @param dst_as Pointer to destination address space.
  521.  * @param dst_base Target base address.
  522.  * @param dst_flags_mask Destination address space area flags mask.
  523.  *
  524.  * @return Zero on success or ENOENT if there is no such task or
  525.  *     if there is no such address space area,
  526.  *     EPERM if there was a problem in accepting the area or
  527.  *     ENOMEM if there was a problem in allocating destination
  528.  *     address space area. ENOTSUP is returned if an attempt
  529.  *     to share non-anonymous address space area is detected.
  530.  */
  531. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  532.           as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  533. {
  534.     ipl_t ipl;
  535.     int src_flags;
  536.     size_t src_size;
  537.     as_area_t *src_area, *dst_area;
  538.     share_info_t *sh_info;
  539.     mem_backend_t *src_backend;
  540.     mem_backend_data_t src_backend_data;
  541.    
  542.     ipl = interrupts_disable();
  543.     mutex_lock(&src_as->lock);
  544.     src_area = find_area_and_lock(src_as, src_base);
  545.     if (!src_area) {
  546.         /*
  547.          * Could not find the source address space area.
  548.          */
  549.         mutex_unlock(&src_as->lock);
  550.         interrupts_restore(ipl);
  551.         return ENOENT;
  552.     }
  553.    
  554.     if (!src_area->backend || !src_area->backend->share) {
  555.         /*
  556.          * There is no backend or the backend does not
  557.          * know how to share the area.
  558.          */
  559.         mutex_unlock(&src_area->lock);
  560.         mutex_unlock(&src_as->lock);
  561.         interrupts_restore(ipl);
  562.         return ENOTSUP;
  563.     }
  564.    
  565.     src_size = src_area->pages * PAGE_SIZE;
  566.     src_flags = src_area->flags;
  567.     src_backend = src_area->backend;
  568.     src_backend_data = src_area->backend_data;
  569.  
  570.     /* Share the cacheable flag from the original mapping */
  571.     if (src_flags & AS_AREA_CACHEABLE)
  572.         dst_flags_mask |= AS_AREA_CACHEABLE;
  573.  
  574.     if (src_size != acc_size || (src_flags & dst_flags_mask) != dst_flags_mask) {
  575.         mutex_unlock(&src_area->lock);
  576.         mutex_unlock(&src_as->lock);
  577.         interrupts_restore(ipl);
  578.         return EPERM;
  579.     }
  580.  
  581.     /*
  582.      * Now we are committed to sharing the area.
  583.      * First prepare the area for sharing.
  584.      * Then it will be safe to unlock it.
  585.      */
  586.     sh_info = src_area->sh_info;
  587.     if (!sh_info) {
  588.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  589.         mutex_initialize(&sh_info->lock);
  590.         sh_info->refcount = 2;
  591.         btree_create(&sh_info->pagemap);
  592.         src_area->sh_info = sh_info;
  593.     } else {
  594.         mutex_lock(&sh_info->lock);
  595.         sh_info->refcount++;
  596.         mutex_unlock(&sh_info->lock);
  597.     }
  598.  
  599.     src_area->backend->share(src_area);
  600.  
  601.     mutex_unlock(&src_area->lock);
  602.     mutex_unlock(&src_as->lock);
  603.  
  604.     /*
  605.      * Create copy of the source address space area.
  606.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  607.      * attribute set which prevents race condition with
  608.      * preliminary as_page_fault() calls.
  609.      * The flags of the source area are masked against dst_flags_mask
  610.      * to support sharing in less privileged mode.
  611.      */
  612.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  613.                   AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  614.     if (!dst_area) {
  615.         /*
  616.          * Destination address space area could not be created.
  617.          */
  618.         sh_info_remove_reference(sh_info);
  619.        
  620.         interrupts_restore(ipl);
  621.         return ENOMEM;
  622.     }
  623.    
  624.     /*
  625.      * Now the destination address space area has been
  626.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  627.      * attribute and set the sh_info.
  628.      */
  629.     mutex_lock(&dst_area->lock);
  630.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  631.     dst_area->sh_info = sh_info;
  632.     mutex_unlock(&dst_area->lock);
  633.    
  634.     interrupts_restore(ipl);
  635.    
  636.     return 0;
  637. }
  638.  
  639. /** Check access mode for address space area.
  640.  *
  641.  * The address space area must be locked prior to this call.
  642.  *
  643.  * @param area Address space area.
  644.  * @param access Access mode.
  645.  *
  646.  * @return False if access violates area's permissions, true otherwise.
  647.  */
  648. bool as_area_check_access(as_area_t *area, pf_access_t access)
  649. {
  650.     int flagmap[] = {
  651.         [PF_ACCESS_READ] = AS_AREA_READ,
  652.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  653.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  654.     };
  655.  
  656.     if (!(area->flags & flagmap[access]))
  657.         return false;
  658.    
  659.     return true;
  660. }
  661.  
  662. /** Handle page fault within the current address space.
  663.  *
  664.  * This is the high-level page fault handler. It decides
  665.  * whether the page fault can be resolved by any backend
  666.  * and if so, it invokes the backend to resolve the page
  667.  * fault.
  668.  *
  669.  * Interrupts are assumed disabled.
  670.  *
  671.  * @param page Faulting page.
  672.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  673.  * @param istate Pointer to interrupted state.
  674.  *
  675.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  676.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  677.  */
  678. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  679. {
  680.     pte_t *pte;
  681.     as_area_t *area;
  682.    
  683.     if (!THREAD)
  684.         return AS_PF_FAULT;
  685.        
  686.     ASSERT(AS);
  687.  
  688.     mutex_lock(&AS->lock);
  689.     area = find_area_and_lock(AS, page);   
  690.     if (!area) {
  691.         /*
  692.          * No area contained mapping for 'page'.
  693.          * Signal page fault to low-level handler.
  694.          */
  695.         mutex_unlock(&AS->lock);
  696.         goto page_fault;
  697.     }
  698.  
  699.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  700.         /*
  701.          * The address space area is not fully initialized.
  702.          * Avoid possible race by returning error.
  703.          */
  704.         mutex_unlock(&area->lock);
  705.         mutex_unlock(&AS->lock);
  706.         goto page_fault;       
  707.     }
  708.  
  709.     if (!area->backend || !area->backend->page_fault) {
  710.         /*
  711.          * The address space area is not backed by any backend
  712.          * or the backend cannot handle page faults.
  713.          */
  714.         mutex_unlock(&area->lock);
  715.         mutex_unlock(&AS->lock);
  716.         goto page_fault;       
  717.     }
  718.  
  719.     page_table_lock(AS, false);
  720.    
  721.     /*
  722.      * To avoid race condition between two page faults
  723.      * on the same address, we need to make sure
  724.      * the mapping has not been already inserted.
  725.      */
  726.     if ((pte = page_mapping_find(AS, page))) {
  727.         if (PTE_PRESENT(pte)) {
  728.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  729.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  730.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  731.                 page_table_unlock(AS, false);
  732.                 mutex_unlock(&area->lock);
  733.                 mutex_unlock(&AS->lock);
  734.                 return AS_PF_OK;
  735.             }
  736.         }
  737.     }
  738.    
  739.     /*
  740.      * Resort to the backend page fault handler.
  741.      */
  742.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  743.         page_table_unlock(AS, false);
  744.         mutex_unlock(&area->lock);
  745.         mutex_unlock(&AS->lock);
  746.         goto page_fault;
  747.     }
  748.    
  749.     page_table_unlock(AS, false);
  750.     mutex_unlock(&area->lock);
  751.     mutex_unlock(&AS->lock);
  752.     return AS_PF_OK;
  753.  
  754. page_fault:
  755.     if (THREAD->in_copy_from_uspace) {
  756.         THREAD->in_copy_from_uspace = false;
  757.         istate_set_retaddr(istate, (uintptr_t) &memcpy_from_uspace_failover_address);
  758.     } else if (THREAD->in_copy_to_uspace) {
  759.         THREAD->in_copy_to_uspace = false;
  760.         istate_set_retaddr(istate, (uintptr_t) &memcpy_to_uspace_failover_address);
  761.     } else {
  762.         return AS_PF_FAULT;
  763.     }
  764.  
  765.     return AS_PF_DEFER;
  766. }
  767.  
  768. /** Switch address spaces.
  769.  *
  770.  * Note that this function cannot sleep as it is essentially a part of
  771.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  772.  *
  773.  * @param old Old address space or NULL.
  774.  * @param new New address space.
  775.  */
  776. void as_switch(as_t *old, as_t *new)
  777. {
  778.     ipl_t ipl;
  779.     bool needs_asid = false;
  780.    
  781.     ipl = interrupts_disable();
  782.     spinlock_lock(&inactive_as_with_asid_lock);
  783.  
  784.     /*
  785.      * First, take care of the old address space.
  786.      */
  787.     if (old) {
  788.         mutex_lock_active(&old->lock);
  789.         ASSERT(old->cpu_refcount);
  790.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  791.             /*
  792.              * The old address space is no longer active on
  793.              * any processor. It can be appended to the
  794.              * list of inactive address spaces with assigned
  795.              * ASID.
  796.              */
  797.              ASSERT(old->asid != ASID_INVALID);
  798.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  799.         }
  800.         mutex_unlock(&old->lock);
  801.     }
  802.  
  803.     /*
  804.      * Second, prepare the new address space.
  805.      */
  806.     mutex_lock_active(&new->lock);
  807.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  808.         if (new->asid != ASID_INVALID)
  809.             list_remove(&new->inactive_as_with_asid_link);
  810.         else
  811.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  812.     }
  813.     SET_PTL0_ADDRESS(new->page_table);
  814.     mutex_unlock(&new->lock);
  815.  
  816.     if (needs_asid) {
  817.         /*
  818.          * Allocation of new ASID was deferred
  819.          * until now in order to avoid deadlock.
  820.          */
  821.         asid_t asid;
  822.        
  823.         asid = asid_get();
  824.         mutex_lock_active(&new->lock);
  825.         new->asid = asid;
  826.         mutex_unlock(&new->lock);
  827.     }
  828.     spinlock_unlock(&inactive_as_with_asid_lock);
  829.     interrupts_restore(ipl);
  830.    
  831.     /*
  832.      * Perform architecture-specific steps.
  833.      * (e.g. write ASID to hardware register etc.)
  834.      */
  835.     as_install_arch(new);
  836.    
  837.     AS = new;
  838. }
  839.  
  840. /** Convert address space area flags to page flags.
  841.  *
  842.  * @param aflags Flags of some address space area.
  843.  *
  844.  * @return Flags to be passed to page_mapping_insert().
  845.  */
  846. int area_flags_to_page_flags(int aflags)
  847. {
  848.     int flags;
  849.  
  850.     flags = PAGE_USER | PAGE_PRESENT;
  851.    
  852.     if (aflags & AS_AREA_READ)
  853.         flags |= PAGE_READ;
  854.        
  855.     if (aflags & AS_AREA_WRITE)
  856.         flags |= PAGE_WRITE;
  857.    
  858.     if (aflags & AS_AREA_EXEC)
  859.         flags |= PAGE_EXEC;
  860.    
  861.     if (aflags & AS_AREA_CACHEABLE)
  862.         flags |= PAGE_CACHEABLE;
  863.        
  864.     return flags;
  865. }
  866.  
  867. /** Compute flags for virtual address translation subsytem.
  868.  *
  869.  * The address space area must be locked.
  870.  * Interrupts must be disabled.
  871.  *
  872.  * @param a Address space area.
  873.  *
  874.  * @return Flags to be used in page_mapping_insert().
  875.  */
  876. int as_area_get_flags(as_area_t *a)
  877. {
  878.     return area_flags_to_page_flags(a->flags);
  879. }
  880.  
  881. /** Create page table.
  882.  *
  883.  * Depending on architecture, create either address space
  884.  * private or global page table.
  885.  *
  886.  * @param flags Flags saying whether the page table is for kernel address space.
  887.  *
  888.  * @return First entry of the page table.
  889.  */
  890. pte_t *page_table_create(int flags)
  891. {
  892.         ASSERT(as_operations);
  893.         ASSERT(as_operations->page_table_create);
  894.  
  895.         return as_operations->page_table_create(flags);
  896. }
  897.  
  898. /** Destroy page table.
  899.  *
  900.  * Destroy page table in architecture specific way.
  901.  *
  902.  * @param page_table Physical address of PTL0.
  903.  */
  904. void page_table_destroy(pte_t *page_table)
  905. {
  906.         ASSERT(as_operations);
  907.         ASSERT(as_operations->page_table_destroy);
  908.  
  909.         as_operations->page_table_destroy(page_table);
  910. }
  911.  
  912. /** Lock page table.
  913.  *
  914.  * This function should be called before any page_mapping_insert(),
  915.  * page_mapping_remove() and page_mapping_find().
  916.  *
  917.  * Locking order is such that address space areas must be locked
  918.  * prior to this call. Address space can be locked prior to this
  919.  * call in which case the lock argument is false.
  920.  *
  921.  * @param as Address space.
  922.  * @param lock If false, do not attempt to lock as->lock.
  923.  */
  924. void page_table_lock(as_t *as, bool lock)
  925. {
  926.     ASSERT(as_operations);
  927.     ASSERT(as_operations->page_table_lock);
  928.  
  929.     as_operations->page_table_lock(as, lock);
  930. }
  931.  
  932. /** Unlock page table.
  933.  *
  934.  * @param as Address space.
  935.  * @param unlock If false, do not attempt to unlock as->lock.
  936.  */
  937. void page_table_unlock(as_t *as, bool unlock)
  938. {
  939.     ASSERT(as_operations);
  940.     ASSERT(as_operations->page_table_unlock);
  941.  
  942.     as_operations->page_table_unlock(as, unlock);
  943. }
  944.  
  945.  
  946. /** Find address space area and lock it.
  947.  *
  948.  * The address space must be locked and interrupts must be disabled.
  949.  *
  950.  * @param as Address space.
  951.  * @param va Virtual address.
  952.  *
  953.  * @return Locked address space area containing va on success or NULL on failure.
  954.  */
  955. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  956. {
  957.     as_area_t *a;
  958.     btree_node_t *leaf, *lnode;
  959.     int i;
  960.    
  961.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  962.     if (a) {
  963.         /* va is the base address of an address space area */
  964.         mutex_lock(&a->lock);
  965.         return a;
  966.     }
  967.    
  968.     /*
  969.      * Search the leaf node and the righmost record of its left neighbour
  970.      * to find out whether this is a miss or va belongs to an address
  971.      * space area found there.
  972.      */
  973.    
  974.     /* First, search the leaf node itself. */
  975.     for (i = 0; i < leaf->keys; i++) {
  976.         a = (as_area_t *) leaf->value[i];
  977.         mutex_lock(&a->lock);
  978.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  979.             return a;
  980.         }
  981.         mutex_unlock(&a->lock);
  982.     }
  983.  
  984.     /*
  985.      * Second, locate the left neighbour and test its last record.
  986.      * Because of its position in the B+tree, it must have base < va.
  987.      */
  988.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  989.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  990.         mutex_lock(&a->lock);
  991.         if (va < a->base + a->pages * PAGE_SIZE) {
  992.             return a;
  993.         }
  994.         mutex_unlock(&a->lock);
  995.     }
  996.  
  997.     return NULL;
  998. }
  999.  
  1000. /** Check area conflicts with other areas.
  1001.  *
  1002.  * The address space must be locked and interrupts must be disabled.
  1003.  *
  1004.  * @param as Address space.
  1005.  * @param va Starting virtual address of the area being tested.
  1006.  * @param size Size of the area being tested.
  1007.  * @param avoid_area Do not touch this area.
  1008.  *
  1009.  * @return True if there is no conflict, false otherwise.
  1010.  */
  1011. bool check_area_conflicts(as_t *as, uintptr_t va, size_t size, as_area_t *avoid_area)
  1012. {
  1013.     as_area_t *a;
  1014.     btree_node_t *leaf, *node;
  1015.     int i;
  1016.    
  1017.     /*
  1018.      * We don't want any area to have conflicts with NULL page.
  1019.      */
  1020.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1021.         return false;
  1022.    
  1023.     /*
  1024.      * The leaf node is found in O(log n), where n is proportional to
  1025.      * the number of address space areas belonging to as.
  1026.      * The check for conflicts is then attempted on the rightmost
  1027.      * record in the left neighbour, the leftmost record in the right
  1028.      * neighbour and all records in the leaf node itself.
  1029.      */
  1030.    
  1031.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1032.         if (a != avoid_area)
  1033.             return false;
  1034.     }
  1035.    
  1036.     /* First, check the two border cases. */
  1037.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1038.         a = (as_area_t *) node->value[node->keys - 1];
  1039.         mutex_lock(&a->lock);
  1040.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1041.             mutex_unlock(&a->lock);
  1042.             return false;
  1043.         }
  1044.         mutex_unlock(&a->lock);
  1045.     }
  1046.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1047.         a = (as_area_t *) node->value[0];
  1048.         mutex_lock(&a->lock);
  1049.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1050.             mutex_unlock(&a->lock);
  1051.             return false;
  1052.         }
  1053.         mutex_unlock(&a->lock);
  1054.     }
  1055.    
  1056.     /* Second, check the leaf node. */
  1057.     for (i = 0; i < leaf->keys; i++) {
  1058.         a = (as_area_t *) leaf->value[i];
  1059.    
  1060.         if (a == avoid_area)
  1061.             continue;
  1062.    
  1063.         mutex_lock(&a->lock);
  1064.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1065.             mutex_unlock(&a->lock);
  1066.             return false;
  1067.         }
  1068.         mutex_unlock(&a->lock);
  1069.     }
  1070.  
  1071.     /*
  1072.      * So far, the area does not conflict with other areas.
  1073.      * Check if it doesn't conflict with kernel address space.
  1074.      */  
  1075.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1076.         return !overlaps(va, size,
  1077.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1078.     }
  1079.  
  1080.     return true;
  1081. }
  1082.  
  1083. /** Return size of the address space area with given base.  */
  1084. size_t as_get_size(uintptr_t base)
  1085. {
  1086.     ipl_t ipl;
  1087.     as_area_t *src_area;
  1088.     size_t size;
  1089.  
  1090.     ipl = interrupts_disable();
  1091.     src_area = find_area_and_lock(AS, base);
  1092.     if (src_area){
  1093.         size = src_area->pages * PAGE_SIZE;
  1094.         mutex_unlock(&src_area->lock);
  1095.     } else {
  1096.         size = 0;
  1097.     }
  1098.     interrupts_restore(ipl);
  1099.     return size;
  1100. }
  1101.  
  1102. /** Mark portion of address space area as used.
  1103.  *
  1104.  * The address space area must be already locked.
  1105.  *
  1106.  * @param a Address space area.
  1107.  * @param page First page to be marked.
  1108.  * @param count Number of page to be marked.
  1109.  *
  1110.  * @return 0 on failure and 1 on success.
  1111.  */
  1112. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1113. {
  1114.     btree_node_t *leaf, *node;
  1115.     count_t pages;
  1116.     int i;
  1117.  
  1118.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1119.     ASSERT(count);
  1120.  
  1121.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1122.     if (pages) {
  1123.         /*
  1124.          * We hit the beginning of some used space.
  1125.          */
  1126.         return 0;
  1127.     }
  1128.  
  1129.     if (!leaf->keys) {
  1130.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1131.         return 1;
  1132.     }
  1133.  
  1134.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1135.     if (node) {
  1136.         uintptr_t left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1137.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1138.        
  1139.         /*
  1140.          * Examine the possibility that the interval fits
  1141.          * somewhere between the rightmost interval of
  1142.          * the left neigbour and the first interval of the leaf.
  1143.          */
  1144.          
  1145.         if (page >= right_pg) {
  1146.             /* Do nothing. */
  1147.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1148.             /* The interval intersects with the left interval. */
  1149.             return 0;
  1150.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1151.             /* The interval intersects with the right interval. */
  1152.             return 0;          
  1153.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1154.             /* The interval can be added by merging the two already present intervals. */
  1155.             node->value[node->keys - 1] += count + right_cnt;
  1156.             btree_remove(&a->used_space, right_pg, leaf);
  1157.             return 1;
  1158.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1159.             /* The interval can be added by simply growing the left interval. */
  1160.             node->value[node->keys - 1] += count;
  1161.             return 1;
  1162.         } else if (page + count*PAGE_SIZE == right_pg) {
  1163.             /*
  1164.              * The interval can be addded by simply moving base of the right
  1165.              * interval down and increasing its size accordingly.
  1166.              */
  1167.             leaf->value[0] += count;
  1168.             leaf->key[0] = page;
  1169.             return 1;
  1170.         } else {
  1171.             /*
  1172.              * The interval is between both neigbouring intervals,
  1173.              * but cannot be merged with any of them.
  1174.              */
  1175.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1176.             return 1;
  1177.         }
  1178.     } else if (page < leaf->key[0]) {
  1179.         uintptr_t right_pg = leaf->key[0];
  1180.         count_t right_cnt = (count_t) leaf->value[0];
  1181.    
  1182.         /*
  1183.          * Investigate the border case in which the left neighbour does not
  1184.          * exist but the interval fits from the left.
  1185.          */
  1186.          
  1187.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1188.             /* The interval intersects with the right interval. */
  1189.             return 0;
  1190.         } else if (page + count*PAGE_SIZE == right_pg) {
  1191.             /*
  1192.              * The interval can be added by moving the base of the right interval down
  1193.              * and increasing its size accordingly.
  1194.              */
  1195.             leaf->key[0] = page;
  1196.             leaf->value[0] += count;
  1197.             return 1;
  1198.         } else {
  1199.             /*
  1200.              * The interval doesn't adjoin with the right interval.
  1201.              * It must be added individually.
  1202.              */
  1203.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1204.             return 1;
  1205.         }
  1206.     }
  1207.  
  1208.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1209.     if (node) {
  1210.         uintptr_t left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1211.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1212.        
  1213.         /*
  1214.          * Examine the possibility that the interval fits
  1215.          * somewhere between the leftmost interval of
  1216.          * the right neigbour and the last interval of the leaf.
  1217.          */
  1218.  
  1219.         if (page < left_pg) {
  1220.             /* Do nothing. */
  1221.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1222.             /* The interval intersects with the left interval. */
  1223.             return 0;
  1224.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1225.             /* The interval intersects with the right interval. */
  1226.             return 0;          
  1227.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1228.             /* The interval can be added by merging the two already present intervals. */
  1229.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1230.             btree_remove(&a->used_space, right_pg, node);
  1231.             return 1;
  1232.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1233.             /* The interval can be added by simply growing the left interval. */
  1234.             leaf->value[leaf->keys - 1] +=  count;
  1235.             return 1;
  1236.         } else if (page + count*PAGE_SIZE == right_pg) {
  1237.             /*
  1238.              * The interval can be addded by simply moving base of the right
  1239.              * interval down and increasing its size accordingly.
  1240.              */
  1241.             node->value[0] += count;
  1242.             node->key[0] = page;
  1243.             return 1;
  1244.         } else {
  1245.             /*
  1246.              * The interval is between both neigbouring intervals,
  1247.              * but cannot be merged with any of them.
  1248.              */
  1249.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1250.             return 1;
  1251.         }
  1252.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1253.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1254.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1255.    
  1256.         /*
  1257.          * Investigate the border case in which the right neighbour does not
  1258.          * exist but the interval fits from the right.
  1259.          */
  1260.          
  1261.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1262.             /* The interval intersects with the left interval. */
  1263.             return 0;
  1264.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1265.             /* The interval can be added by growing the left interval. */
  1266.             leaf->value[leaf->keys - 1] += count;
  1267.             return 1;
  1268.         } else {
  1269.             /*
  1270.              * The interval doesn't adjoin with the left interval.
  1271.              * It must be added individually.
  1272.              */
  1273.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1274.             return 1;
  1275.         }
  1276.     }
  1277.    
  1278.     /*
  1279.      * Note that if the algorithm made it thus far, the interval can fit only
  1280.      * between two other intervals of the leaf. The two border cases were already
  1281.      * resolved.
  1282.      */
  1283.     for (i = 1; i < leaf->keys; i++) {
  1284.         if (page < leaf->key[i]) {
  1285.             uintptr_t left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1286.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1287.  
  1288.             /*
  1289.              * The interval fits between left_pg and right_pg.
  1290.              */
  1291.  
  1292.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1293.                 /* The interval intersects with the left interval. */
  1294.                 return 0;
  1295.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1296.                 /* The interval intersects with the right interval. */
  1297.                 return 0;          
  1298.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1299.                 /* The interval can be added by merging the two already present intervals. */
  1300.                 leaf->value[i - 1] += count + right_cnt;
  1301.                 btree_remove(&a->used_space, right_pg, leaf);
  1302.                 return 1;
  1303.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1304.                 /* The interval can be added by simply growing the left interval. */
  1305.                 leaf->value[i - 1] += count;
  1306.                 return 1;
  1307.             } else if (page + count*PAGE_SIZE == right_pg) {
  1308.                 /*
  1309.                      * The interval can be addded by simply moving base of the right
  1310.                  * interval down and increasing its size accordingly.
  1311.                  */
  1312.                 leaf->value[i] += count;
  1313.                 leaf->key[i] = page;
  1314.                 return 1;
  1315.             } else {
  1316.                 /*
  1317.                  * The interval is between both neigbouring intervals,
  1318.                  * but cannot be merged with any of them.
  1319.                  */
  1320.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1321.                 return 1;
  1322.             }
  1323.         }
  1324.     }
  1325.  
  1326.     panic("Inconsistency detected while adding %d pages of used space at %p.\n", count, page);
  1327. }
  1328.  
  1329. /** Mark portion of address space area as unused.
  1330.  *
  1331.  * The address space area must be already locked.
  1332.  *
  1333.  * @param a Address space area.
  1334.  * @param page First page to be marked.
  1335.  * @param count Number of page to be marked.
  1336.  *
  1337.  * @return 0 on failure and 1 on success.
  1338.  */
  1339. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1340. {
  1341.     btree_node_t *leaf, *node;
  1342.     count_t pages;
  1343.     int i;
  1344.  
  1345.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1346.     ASSERT(count);
  1347.  
  1348.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1349.     if (pages) {
  1350.         /*
  1351.          * We are lucky, page is the beginning of some interval.
  1352.          */
  1353.         if (count > pages) {
  1354.             return 0;
  1355.         } else if (count == pages) {
  1356.             btree_remove(&a->used_space, page, leaf);
  1357.             return 1;
  1358.         } else {
  1359.             /*
  1360.              * Find the respective interval.
  1361.              * Decrease its size and relocate its start address.
  1362.              */
  1363.             for (i = 0; i < leaf->keys; i++) {
  1364.                 if (leaf->key[i] == page) {
  1365.                     leaf->key[i] += count*PAGE_SIZE;
  1366.                     leaf->value[i] -= count;
  1367.                     return 1;
  1368.                 }
  1369.             }
  1370.             goto error;
  1371.         }
  1372.     }
  1373.  
  1374.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1375.     if (node && page < leaf->key[0]) {
  1376.         uintptr_t left_pg = node->key[node->keys - 1];
  1377.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1378.  
  1379.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1380.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1381.                 /*
  1382.                  * The interval is contained in the rightmost interval
  1383.                  * of the left neighbour and can be removed by
  1384.                  * updating the size of the bigger interval.
  1385.                  */
  1386.                 node->value[node->keys - 1] -= count;
  1387.                 return 1;
  1388.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1389.                 count_t new_cnt;
  1390.                
  1391.                 /*
  1392.                  * The interval is contained in the rightmost interval
  1393.                  * of the left neighbour but its removal requires
  1394.                  * both updating the size of the original interval and
  1395.                  * also inserting a new interval.
  1396.                  */
  1397.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1398.                 node->value[node->keys - 1] -= count + new_cnt;
  1399.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1400.                 return 1;
  1401.             }
  1402.         }
  1403.         return 0;
  1404.     } else if (page < leaf->key[0]) {
  1405.         return 0;
  1406.     }
  1407.    
  1408.     if (page > leaf->key[leaf->keys - 1]) {
  1409.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1410.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1411.  
  1412.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1413.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1414.                 /*
  1415.                  * The interval is contained in the rightmost interval
  1416.                  * of the leaf and can be removed by updating the size
  1417.                  * of the bigger interval.
  1418.                  */
  1419.                 leaf->value[leaf->keys - 1] -= count;
  1420.                 return 1;
  1421.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1422.                 count_t new_cnt;
  1423.                
  1424.                 /*
  1425.                  * The interval is contained in the rightmost interval
  1426.                  * of the leaf but its removal requires both updating
  1427.                  * the size of the original interval and
  1428.                  * also inserting a new interval.
  1429.                  */
  1430.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1431.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1432.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1433.                 return 1;
  1434.             }
  1435.         }
  1436.         return 0;
  1437.     }  
  1438.    
  1439.     /*
  1440.      * The border cases have been already resolved.
  1441.      * Now the interval can be only between intervals of the leaf.
  1442.      */
  1443.     for (i = 1; i < leaf->keys - 1; i++) {
  1444.         if (page < leaf->key[i]) {
  1445.             uintptr_t left_pg = leaf->key[i - 1];
  1446.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1447.  
  1448.             /*
  1449.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1450.              */
  1451.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1452.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1453.                     /*
  1454.                     * The interval is contained in the interval (i - 1)
  1455.                      * of the leaf and can be removed by updating the size
  1456.                      * of the bigger interval.
  1457.                      */
  1458.                     leaf->value[i - 1] -= count;
  1459.                     return 1;
  1460.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1461.                     count_t new_cnt;
  1462.                
  1463.                     /*
  1464.                      * The interval is contained in the interval (i - 1)
  1465.                      * of the leaf but its removal requires both updating
  1466.                      * the size of the original interval and
  1467.                      * also inserting a new interval.
  1468.                      */
  1469.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1470.                     leaf->value[i - 1] -= count + new_cnt;
  1471.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1472.                     return 1;
  1473.                 }
  1474.             }
  1475.             return 0;
  1476.         }
  1477.     }
  1478.  
  1479. error:
  1480.     panic("Inconsistency detected while removing %d pages of used space from %p.\n", count, page);
  1481. }
  1482.  
  1483. /** Remove reference to address space area share info.
  1484.  *
  1485.  * If the reference count drops to 0, the sh_info is deallocated.
  1486.  *
  1487.  * @param sh_info Pointer to address space area share info.
  1488.  */
  1489. void sh_info_remove_reference(share_info_t *sh_info)
  1490. {
  1491.     bool dealloc = false;
  1492.  
  1493.     mutex_lock(&sh_info->lock);
  1494.     ASSERT(sh_info->refcount);
  1495.     if (--sh_info->refcount == 0) {
  1496.         dealloc = true;
  1497.         link_t *cur;
  1498.        
  1499.         /*
  1500.          * Now walk carefully the pagemap B+tree and free/remove
  1501.          * reference from all frames found there.
  1502.          */
  1503.         for (cur = sh_info->pagemap.leaf_head.next; cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1504.             btree_node_t *node;
  1505.             int i;
  1506.            
  1507.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1508.             for (i = 0; i < node->keys; i++)
  1509.                 frame_free((uintptr_t) node->value[i]);
  1510.         }
  1511.        
  1512.     }
  1513.     mutex_unlock(&sh_info->lock);
  1514.    
  1515.     if (dealloc) {
  1516.         btree_destroy(&sh_info->pagemap);
  1517.         free(sh_info);
  1518.     }
  1519. }
  1520.  
  1521. /*
  1522.  * Address space related syscalls.
  1523.  */
  1524.  
  1525. /** Wrapper for as_area_create(). */
  1526. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1527. {
  1528.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1529.         return (unative_t) address;
  1530.     else
  1531.         return (unative_t) -1;
  1532. }
  1533.  
  1534. /** Wrapper for as_area_resize(). */
  1535. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1536. {
  1537.     return (unative_t) as_area_resize(AS, address, size, 0);
  1538. }
  1539.  
  1540. /** Wrapper for as_area_destroy(). */
  1541. unative_t sys_as_area_destroy(uintptr_t address)
  1542. {
  1543.     return (unative_t) as_area_destroy(AS, address);
  1544. }
  1545.  
  1546. /** @}
  1547.  */
  1548.