Subversion Repositories HelenOS-historic

Rev

Rev 1417 | Rev 1424 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed

  1. /*
  2.  * Copyright (C) 2001-2006 Jakub Jermar
  3.  * All rights reserved.
  4.  *
  5.  * Redistribution and use in source and binary forms, with or without
  6.  * modification, are permitted provided that the following conditions
  7.  * are met:
  8.  *
  9.  * - Redistributions of source code must retain the above copyright
  10.  *   notice, this list of conditions and the following disclaimer.
  11.  * - Redistributions in binary form must reproduce the above copyright
  12.  *   notice, this list of conditions and the following disclaimer in the
  13.  *   documentation and/or other materials provided with the distribution.
  14.  * - The name of the author may not be used to endorse or promote products
  15.  *   derived from this software without specific prior written permission.
  16.  *
  17.  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  18.  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  19.  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  20.  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  21.  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  22.  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  23.  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  24.  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  25.  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  26.  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  27.  */
  28.  
  29. /**
  30.  * @file    as.c
  31.  * @brief   Address space related functions.
  32.  *
  33.  * This file contains address space manipulation functions.
  34.  * Roughly speaking, this is a higher-level client of
  35.  * Virtual Address Translation (VAT) subsystem.
  36.  *
  37.  * Functionality provided by this file allows one to
  38.  * create address space and create, resize and share
  39.  * address space areas.
  40.  *
  41.  * @see page.c
  42.  *
  43.  */
  44.  
  45. #include <mm/as.h>
  46. #include <arch/mm/as.h>
  47. #include <mm/page.h>
  48. #include <mm/frame.h>
  49. #include <mm/slab.h>
  50. #include <mm/tlb.h>
  51. #include <arch/mm/page.h>
  52. #include <genarch/mm/page_pt.h>
  53. #include <genarch/mm/page_ht.h>
  54. #include <mm/asid.h>
  55. #include <arch/mm/asid.h>
  56. #include <synch/spinlock.h>
  57. #include <synch/mutex.h>
  58. #include <adt/list.h>
  59. #include <adt/btree.h>
  60. #include <proc/task.h>
  61. #include <proc/thread.h>
  62. #include <arch/asm.h>
  63. #include <panic.h>
  64. #include <debug.h>
  65. #include <print.h>
  66. #include <memstr.h>
  67. #include <macros.h>
  68. #include <arch.h>
  69. #include <errno.h>
  70. #include <config.h>
  71. #include <align.h>
  72. #include <arch/types.h>
  73. #include <typedefs.h>
  74. #include <syscall/copy.h>
  75. #include <arch/interrupt.h>
  76.  
  77. /** This structure contains information associated with the shared address space area. */
  78. struct share_info {
  79.     mutex_t lock;       /**< This lock must be acquired only when the as_area lock is held. */
  80.     count_t refcount;   /**< This structure can be deallocated if refcount drops to 0. */
  81.     btree_t pagemap;    /**< B+tree containing complete map of anonymous pages of the shared area. */
  82. };
  83.  
  84. as_operations_t *as_operations = NULL;
  85.  
  86. /** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
  87. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  88.  
  89. /**
  90.  * This list contains address spaces that are not active on any
  91.  * processor and that have valid ASID.
  92.  */
  93. LIST_INITIALIZE(inactive_as_with_asid_head);
  94.  
  95. /** Kernel address space. */
  96. as_t *AS_KERNEL = NULL;
  97.  
  98. static int area_flags_to_page_flags(int aflags);
  99. static as_area_t *find_area_and_lock(as_t *as, __address va);
  100. static bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area);
  101. static void sh_info_remove_reference(share_info_t *sh_info);
  102.  
  103. /** Initialize address space subsystem. */
  104. void as_init(void)
  105. {
  106.     as_arch_init();
  107.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  108.     if (!AS_KERNEL)
  109.         panic("can't create kernel address space\n");
  110.    
  111. }
  112.  
  113. /** Create address space.
  114.  *
  115.  * @param flags Flags that influence way in wich the address space is created.
  116.  */
  117. as_t *as_create(int flags)
  118. {
  119.     as_t *as;
  120.  
  121.     as = (as_t *) malloc(sizeof(as_t), 0);
  122.     link_initialize(&as->inactive_as_with_asid_link);
  123.     mutex_initialize(&as->lock);
  124.     btree_create(&as->as_area_btree);
  125.    
  126.     if (flags & FLAG_AS_KERNEL)
  127.         as->asid = ASID_KERNEL;
  128.     else
  129.         as->asid = ASID_INVALID;
  130.    
  131.     as->cpu_refcount = 0;
  132.     as->page_table = page_table_create(flags);
  133.  
  134.     return as;
  135. }
  136.  
  137. /** Free Adress space */
  138. void as_free(as_t *as)
  139. {
  140.     ASSERT(as->cpu_refcount == 0);
  141.  
  142.     /* TODO: free as_areas and other resources held by as */
  143.     /* TODO: free page table */
  144.     free(as);
  145. }
  146.  
  147. /** Create address space area of common attributes.
  148.  *
  149.  * The created address space area is added to the target address space.
  150.  *
  151.  * @param as Target address space.
  152.  * @param flags Flags of the area memory.
  153.  * @param size Size of area.
  154.  * @param base Base address of area.
  155.  * @param attrs Attributes of the area.
  156.  * @param backend Address space area backend. NULL if no backend is used.
  157.  * @param backend_data NULL or a pointer to an array holding two void *.
  158.  *
  159.  * @return Address space area on success or NULL on failure.
  160.  */
  161. as_area_t *as_area_create(as_t *as, int flags, size_t size, __address base, int attrs,
  162.            mem_backend_t *backend, void **backend_data)
  163. {
  164.     ipl_t ipl;
  165.     as_area_t *a;
  166.    
  167.     if (base % PAGE_SIZE)
  168.         return NULL;
  169.  
  170.     if (!size)
  171.         return NULL;
  172.  
  173.     /* Writeable executable areas are not supported. */
  174.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  175.         return NULL;
  176.    
  177.     ipl = interrupts_disable();
  178.     mutex_lock(&as->lock);
  179.    
  180.     if (!check_area_conflicts(as, base, size, NULL)) {
  181.         mutex_unlock(&as->lock);
  182.         interrupts_restore(ipl);
  183.         return NULL;
  184.     }
  185.    
  186.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  187.  
  188.     mutex_initialize(&a->lock);
  189.    
  190.     a->flags = flags;
  191.     a->attributes = attrs;
  192.     a->pages = SIZE2FRAMES(size);
  193.     a->base = base;
  194.     a->sh_info = NULL;
  195.     a->backend = backend;
  196.     if (backend_data) {
  197.         a->backend_data[0] = backend_data[0];
  198.         a->backend_data[1] = backend_data[1];
  199.     }
  200.     btree_create(&a->used_space);
  201.    
  202.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  203.  
  204.     mutex_unlock(&as->lock);
  205.     interrupts_restore(ipl);
  206.  
  207.     return a;
  208. }
  209.  
  210. /** Find address space area and change it.
  211.  *
  212.  * @param as Address space.
  213.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  214.  * @param size New size of the virtual memory block starting at address.
  215.  * @param flags Flags influencing the remap operation. Currently unused.
  216.  *
  217.  * @return Zero on success or a value from @ref errno.h otherwise.
  218.  */
  219. int as_area_resize(as_t *as, __address address, size_t size, int flags)
  220. {
  221.     as_area_t *area;
  222.     ipl_t ipl;
  223.     size_t pages;
  224.    
  225.     ipl = interrupts_disable();
  226.     mutex_lock(&as->lock);
  227.    
  228.     /*
  229.      * Locate the area.
  230.      */
  231.     area = find_area_and_lock(as, address);
  232.     if (!area) {
  233.         mutex_unlock(&as->lock);
  234.         interrupts_restore(ipl);
  235.         return ENOENT;
  236.     }
  237.  
  238.     if (area->flags & AS_AREA_DEVICE) {
  239.         /*
  240.          * Remapping of address space areas associated
  241.          * with memory mapped devices is not supported.
  242.          */
  243.         mutex_unlock(&area->lock);
  244.         mutex_unlock(&as->lock);
  245.         interrupts_restore(ipl);
  246.         return ENOTSUP;
  247.     }
  248.     if (area->sh_info) {
  249.         /*
  250.          * Remapping of shared address space areas
  251.          * is not supported.
  252.          */
  253.         mutex_unlock(&area->lock);
  254.         mutex_unlock(&as->lock);
  255.         interrupts_restore(ipl);
  256.         return ENOTSUP;
  257.     }
  258.  
  259.     pages = SIZE2FRAMES((address - area->base) + size);
  260.     if (!pages) {
  261.         /*
  262.          * Zero size address space areas are not allowed.
  263.          */
  264.         mutex_unlock(&area->lock);
  265.         mutex_unlock(&as->lock);
  266.         interrupts_restore(ipl);
  267.         return EPERM;
  268.     }
  269.    
  270.     if (pages < area->pages) {
  271.         bool cond;
  272.         __address start_free = area->base + pages*PAGE_SIZE;
  273.  
  274.         /*
  275.          * Shrinking the area.
  276.          * No need to check for overlaps.
  277.          */
  278.  
  279.         /*
  280.          * Remove frames belonging to used space starting from
  281.          * the highest addresses downwards until an overlap with
  282.          * the resized address space area is found. Note that this
  283.          * is also the right way to remove part of the used_space
  284.          * B+tree leaf list.
  285.          */    
  286.         for (cond = true; cond;) {
  287.             btree_node_t *node;
  288.        
  289.             ASSERT(!list_empty(&area->used_space.leaf_head));
  290.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  291.             if ((cond = (bool) node->keys)) {
  292.                 __address b = node->key[node->keys - 1];
  293.                 count_t c = (count_t) node->value[node->keys - 1];
  294.                 int i = 0;
  295.            
  296.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  297.                    
  298.                     if (b + c*PAGE_SIZE <= start_free) {
  299.                         /*
  300.                          * The whole interval fits completely
  301.                          * in the resized address space area.
  302.                          */
  303.                         break;
  304.                     }
  305.        
  306.                     /*
  307.                      * Part of the interval corresponding to b and c
  308.                      * overlaps with the resized address space area.
  309.                      */
  310.        
  311.                     cond = false;   /* we are almost done */
  312.                     i = (start_free - b) >> PAGE_WIDTH;
  313.                     if (!used_space_remove(area, start_free, c - i))
  314.                         panic("Could not remove used space.");
  315.                 } else {
  316.                     /*
  317.                      * The interval of used space can be completely removed.
  318.                      */
  319.                     if (!used_space_remove(area, b, c))
  320.                         panic("Could not remove used space.\n");
  321.                 }
  322.            
  323.                 for (; i < c; i++) {
  324.                     pte_t *pte;
  325.            
  326.                     page_table_lock(as, false);
  327.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  328.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  329.                     if (area->backend && area->backend->backend_frame_free) {
  330.                         area->backend->backend_frame_free(area,
  331.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  332.                     }
  333.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  334.                     page_table_unlock(as, false);
  335.                 }
  336.             }
  337.         }
  338.         /*
  339.          * Invalidate TLB's.
  340.          */
  341.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  342.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  343.         tlb_shootdown_finalize();
  344.     } else {
  345.         /*
  346.          * Growing the area.
  347.          * Check for overlaps with other address space areas.
  348.          */
  349.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  350.             mutex_unlock(&area->lock);
  351.             mutex_unlock(&as->lock);       
  352.             interrupts_restore(ipl);
  353.             return EADDRNOTAVAIL;
  354.         }
  355.     }
  356.  
  357.     area->pages = pages;
  358.    
  359.     mutex_unlock(&area->lock);
  360.     mutex_unlock(&as->lock);
  361.     interrupts_restore(ipl);
  362.  
  363.     return 0;
  364. }
  365.  
  366. /** Destroy address space area.
  367.  *
  368.  * @param as Address space.
  369.  * @param address Address withing the area to be deleted.
  370.  *
  371.  * @return Zero on success or a value from @ref errno.h on failure.
  372.  */
  373. int as_area_destroy(as_t *as, __address address)
  374. {
  375.     as_area_t *area;
  376.     __address base;
  377.     ipl_t ipl;
  378.     bool cond;
  379.  
  380.     ipl = interrupts_disable();
  381.     mutex_lock(&as->lock);
  382.  
  383.     area = find_area_and_lock(as, address);
  384.     if (!area) {
  385.         mutex_unlock(&as->lock);
  386.         interrupts_restore(ipl);
  387.         return ENOENT;
  388.     }
  389.  
  390.     base = area->base;
  391.  
  392.     /*
  393.      * Visit only the pages mapped by used_space B+tree.
  394.      * Note that we must be very careful when walking the tree
  395.      * leaf list and removing used space as the leaf list changes
  396.      * unpredictibly after each remove. The solution is to actually
  397.      * not walk the tree at all, but to remove items from the head
  398.      * of the leaf list until there are some keys left.
  399.      */
  400.     for (cond = true; cond;) {
  401.         btree_node_t *node;
  402.        
  403.         ASSERT(!list_empty(&area->used_space.leaf_head));
  404.         node = list_get_instance(area->used_space.leaf_head.next, btree_node_t, leaf_link);
  405.         if ((cond = (bool) node->keys)) {
  406.             __address b = node->key[0];
  407.             count_t i;
  408.             pte_t *pte;
  409.            
  410.             for (i = 0; i < (count_t) node->value[0]; i++) {
  411.                 page_table_lock(as, false);
  412.                 pte = page_mapping_find(as, b + i*PAGE_SIZE);
  413.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  414.                 if (area->backend && area->backend->backend_frame_free) {
  415.                     area->backend->backend_frame_free(area,
  416.                         b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  417.                 }
  418.                 page_mapping_remove(as, b + i*PAGE_SIZE);
  419.                 page_table_unlock(as, false);
  420.             }
  421.             if (!used_space_remove(area, b, i))
  422.                 panic("Could not remove used space.\n");
  423.         }
  424.     }
  425.     btree_destroy(&area->used_space);
  426.  
  427.     /*
  428.      * Invalidate TLB's.
  429.      */
  430.     tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
  431.     tlb_invalidate_pages(AS->asid, area->base, area->pages);
  432.     tlb_shootdown_finalize();
  433.  
  434.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  435.    
  436.     if (area->sh_info)
  437.         sh_info_remove_reference(area->sh_info);
  438.        
  439.     mutex_unlock(&area->lock);
  440.  
  441.     /*
  442.      * Remove the empty area from address space.
  443.      */
  444.     btree_remove(&AS->as_area_btree, base, NULL);
  445.    
  446.     free(area);
  447.    
  448.     mutex_unlock(&AS->lock);
  449.     interrupts_restore(ipl);
  450.     return 0;
  451. }
  452.  
  453. /** Share address space area with another or the same address space.
  454.  *
  455.  * Address space area of anonymous memory is shared with a new address
  456.  * space area. If the source address space area has not been shared so
  457.  * far, a new sh_info is created and the original mapping is duplicated
  458.  * in its pagemap B+tree. The new address space are simply gets the
  459.  * sh_info of the source area.
  460.  *
  461.  * @param src_as Pointer to source address space.
  462.  * @param src_base Base address of the source address space area.
  463.  * @param acc_size Expected size of the source area.
  464.  * @param dst_base Target base address.
  465.  * @param dst_flags_mask Destination address space area flags mask.
  466.  *
  467.  * @return Zero on success or ENOENT if there is no such task or
  468.  *     if there is no such address space area,
  469.  *     EPERM if there was a problem in accepting the area or
  470.  *     ENOMEM if there was a problem in allocating destination
  471.  *     address space area. ENOTSUP is returned if an attempt
  472.  *     to share non-anonymous address space area is detected.
  473.  */
  474. int as_area_share(as_t *src_as, __address src_base, size_t acc_size,
  475.           __address dst_base, int dst_flags_mask)
  476. {
  477.     ipl_t ipl;
  478.     int src_flags;
  479.     size_t src_size;
  480.     as_area_t *src_area, *dst_area;
  481.     share_info_t *sh_info;
  482.     link_t *cur;
  483.  
  484.     ipl = interrupts_disable();
  485.     mutex_lock(&src_as->lock);
  486.     src_area = find_area_and_lock(src_as, src_base);
  487.     if (!src_area) {
  488.         /*
  489.          * Could not find the source address space area.
  490.          */
  491.         mutex_unlock(&src_as->lock);
  492.         interrupts_restore(ipl);
  493.         return ENOENT;
  494.     }
  495.    
  496.     if (!src_area->backend || src_area->backend != &anon_backend) {
  497.         /*
  498.          * As of now, only anonymous address space areas can be shared.
  499.          */
  500.         mutex_unlock(&src_area->lock);
  501.         mutex_unlock(&src_as->lock);
  502.         interrupts_restore(ipl);
  503.         return ENOTSUP;
  504.     }
  505.    
  506.     src_size = src_area->pages * PAGE_SIZE;
  507.     src_flags = src_area->flags;
  508.    
  509.     if (src_size != acc_size) {
  510.         mutex_unlock(&src_area->lock);
  511.         mutex_unlock(&src_as->lock);
  512.         interrupts_restore(ipl);
  513.         return EPERM;
  514.     }
  515.  
  516.     /*
  517.      * Now we are committed to sharing the area.
  518.      * First prepare the area for sharing.
  519.      * Then it will be safe to unlock it.
  520.      */
  521.     sh_info = src_area->sh_info;
  522.     if (!sh_info) {
  523.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  524.         mutex_initialize(&sh_info->lock);
  525.         sh_info->refcount = 2;
  526.         btree_create(&sh_info->pagemap);
  527.         src_area->sh_info = sh_info;
  528.     } else {
  529.         mutex_lock(&sh_info->lock);
  530.         sh_info->refcount++;
  531.         mutex_unlock(&sh_info->lock);
  532.     }
  533.  
  534.     /*
  535.      * Copy used portions of the area to sh_info's page map.
  536.      */
  537.     mutex_lock(&sh_info->lock);
  538.     for (cur = src_area->used_space.leaf_head.next; cur != &src_area->used_space.leaf_head; cur = cur->next) {
  539.         btree_node_t *node;
  540.         int i;
  541.        
  542.         node = list_get_instance(cur, btree_node_t, leaf_link);
  543.         for (i = 0; i < node->keys; i++) {
  544.             __address base = node->key[i];
  545.             count_t count = (count_t) node->value[i];
  546.             int j;
  547.            
  548.             for (j = 0; j < count; j++) {
  549.                 pte_t *pte;
  550.            
  551.                 page_table_lock(src_as, false);
  552.                 pte = page_mapping_find(src_as, base + j*PAGE_SIZE);
  553.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  554.                 btree_insert(&sh_info->pagemap, (base + j*PAGE_SIZE) - src_area->base,
  555.                     (void *) PTE_GET_FRAME(pte), NULL);
  556.                 page_table_unlock(src_as, false);
  557.             }
  558.                
  559.         }
  560.     }
  561.     mutex_unlock(&sh_info->lock);
  562.  
  563.     mutex_unlock(&src_area->lock);
  564.     mutex_unlock(&src_as->lock);
  565.  
  566.     /*
  567.      * Create copy of the source address space area.
  568.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  569.      * attribute set which prevents race condition with
  570.      * preliminary as_page_fault() calls.
  571.      * The flags of the source area are masked against dst_flags_mask
  572.      * to support sharing in less privileged mode.
  573.      */
  574.     dst_area = as_area_create(AS, src_flags & dst_flags_mask, src_size, dst_base,
  575.                   AS_AREA_ATTR_PARTIAL, &anon_backend, NULL);
  576.     if (!dst_area) {
  577.         /*
  578.          * Destination address space area could not be created.
  579.          */
  580.         sh_info_remove_reference(sh_info);
  581.        
  582.         interrupts_restore(ipl);
  583.         return ENOMEM;
  584.     }
  585.    
  586.     /*
  587.      * Now the destination address space area has been
  588.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  589.      * attribute and set the sh_info.
  590.      */
  591.     mutex_lock(&dst_area->lock);
  592.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  593.     dst_area->sh_info = sh_info;
  594.     mutex_unlock(&dst_area->lock);
  595.    
  596.     interrupts_restore(ipl);
  597.    
  598.     return 0;
  599. }
  600.  
  601. /** Initialize mapping for one page of address space.
  602.  *
  603.  * This functions maps 'page' to 'frame' according
  604.  * to attributes of the address space area to
  605.  * wich 'page' belongs.
  606.  *
  607.  * @param as Target address space.
  608.  * @param page Virtual page within the area.
  609.  * @param frame Physical frame to which page will be mapped.
  610.  */
  611. void as_set_mapping(as_t *as, __address page, __address frame)
  612. {
  613.     as_area_t *area;
  614.     ipl_t ipl;
  615.    
  616.     ipl = interrupts_disable();
  617.     page_table_lock(as, true);
  618.    
  619.     area = find_area_and_lock(as, page);
  620.     if (!area) {
  621.         panic("Page not part of any as_area.\n");
  622.     }
  623.  
  624.     ASSERT(!area->backend);
  625.    
  626.     page_mapping_insert(as, page, frame, as_area_get_flags(area));
  627.     if (!used_space_insert(area, page, 1))
  628.         panic("Could not insert used space.\n");
  629.    
  630.     mutex_unlock(&area->lock);
  631.     page_table_unlock(as, true);
  632.     interrupts_restore(ipl);
  633. }
  634.  
  635. /** Check access mode for address space area.
  636.  *
  637.  * The address space area must be locked prior to this call.
  638.  *
  639.  * @param area Address space area.
  640.  * @param access Access mode.
  641.  *
  642.  * @return False if access violates area's permissions, true otherwise.
  643.  */
  644. bool as_area_check_access(as_area_t *area, pf_access_t access)
  645. {
  646.     int flagmap[] = {
  647.         [PF_ACCESS_READ] = AS_AREA_READ,
  648.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  649.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  650.     };
  651.  
  652.     if (!(area->flags & flagmap[access]))
  653.         return false;
  654.    
  655.     return true;
  656. }
  657.  
  658. /** Handle page fault within the current address space.
  659.  *
  660.  * This is the high-level page fault handler. It decides
  661.  * whether the page fault can be resolved by any backend
  662.  * and if so, it invokes the backend to resolve the page
  663.  * fault.
  664.  *
  665.  * Interrupts are assumed disabled.
  666.  *
  667.  * @param page Faulting page.
  668.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  669.  * @param istate Pointer to interrupted state.
  670.  *
  671.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  672.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  673.  */
  674. int as_page_fault(__address page, pf_access_t access, istate_t *istate)
  675. {
  676.     pte_t *pte;
  677.     as_area_t *area;
  678.    
  679.     if (!THREAD)
  680.         return AS_PF_FAULT;
  681.        
  682.     ASSERT(AS);
  683.  
  684.     mutex_lock(&AS->lock);
  685.     area = find_area_and_lock(AS, page);   
  686.     if (!area) {
  687.         /*
  688.          * No area contained mapping for 'page'.
  689.          * Signal page fault to low-level handler.
  690.          */
  691.         mutex_unlock(&AS->lock);
  692.         goto page_fault;
  693.     }
  694.  
  695.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  696.         /*
  697.          * The address space area is not fully initialized.
  698.          * Avoid possible race by returning error.
  699.          */
  700.         mutex_unlock(&area->lock);
  701.         mutex_unlock(&AS->lock);
  702.         goto page_fault;       
  703.     }
  704.  
  705.     if (!area->backend || !area->backend->backend_page_fault) {
  706.         /*
  707.          * The address space area is not backed by any backend
  708.          * or the backend cannot handle page faults.
  709.          */
  710.         mutex_unlock(&area->lock);
  711.         mutex_unlock(&AS->lock);
  712.         goto page_fault;       
  713.     }
  714.  
  715.     page_table_lock(AS, false);
  716.    
  717.     /*
  718.      * To avoid race condition between two page faults
  719.      * on the same address, we need to make sure
  720.      * the mapping has not been already inserted.
  721.      */
  722.     if ((pte = page_mapping_find(AS, page))) {
  723.         if (PTE_PRESENT(pte)) {
  724.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  725.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  726.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  727.                 page_table_unlock(AS, false);
  728.                 mutex_unlock(&area->lock);
  729.                 mutex_unlock(&AS->lock);
  730.                 return AS_PF_OK;
  731.             }
  732.         }
  733.     }
  734.    
  735.     /*
  736.      * Resort to the backend page fault handler.
  737.      */
  738.     if (area->backend->backend_page_fault(area, page, access) != AS_PF_OK) {
  739.         page_table_unlock(AS, false);
  740.         mutex_unlock(&area->lock);
  741.         mutex_unlock(&AS->lock);
  742.         goto page_fault;
  743.     }
  744.    
  745.     page_table_unlock(AS, false);
  746.     mutex_unlock(&area->lock);
  747.     mutex_unlock(&AS->lock);
  748.     return AS_PF_OK;
  749.  
  750. page_fault:
  751.     if (THREAD->in_copy_from_uspace) {
  752.         THREAD->in_copy_from_uspace = false;
  753.         istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
  754.     } else if (THREAD->in_copy_to_uspace) {
  755.         THREAD->in_copy_to_uspace = false;
  756.         istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
  757.     } else {
  758.         return AS_PF_FAULT;
  759.     }
  760.  
  761.     return AS_PF_DEFER;
  762. }
  763.  
  764. /** Switch address spaces.
  765.  *
  766.  * Note that this function cannot sleep as it is essentially a part of
  767.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  768.  *
  769.  * @param old Old address space or NULL.
  770.  * @param new New address space.
  771.  */
  772. void as_switch(as_t *old, as_t *new)
  773. {
  774.     ipl_t ipl;
  775.     bool needs_asid = false;
  776.    
  777.     ipl = interrupts_disable();
  778.     spinlock_lock(&inactive_as_with_asid_lock);
  779.  
  780.     /*
  781.      * First, take care of the old address space.
  782.      */
  783.     if (old) {
  784.         mutex_lock_active(&old->lock);
  785.         ASSERT(old->cpu_refcount);
  786.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  787.             /*
  788.              * The old address space is no longer active on
  789.              * any processor. It can be appended to the
  790.              * list of inactive address spaces with assigned
  791.              * ASID.
  792.              */
  793.              ASSERT(old->asid != ASID_INVALID);
  794.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  795.         }
  796.         mutex_unlock(&old->lock);
  797.     }
  798.  
  799.     /*
  800.      * Second, prepare the new address space.
  801.      */
  802.     mutex_lock_active(&new->lock);
  803.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  804.         if (new->asid != ASID_INVALID)
  805.             list_remove(&new->inactive_as_with_asid_link);
  806.         else
  807.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  808.     }
  809.     SET_PTL0_ADDRESS(new->page_table);
  810.     mutex_unlock(&new->lock);
  811.  
  812.     if (needs_asid) {
  813.         /*
  814.          * Allocation of new ASID was deferred
  815.          * until now in order to avoid deadlock.
  816.          */
  817.         asid_t asid;
  818.        
  819.         asid = asid_get();
  820.         mutex_lock_active(&new->lock);
  821.         new->asid = asid;
  822.         mutex_unlock(&new->lock);
  823.     }
  824.     spinlock_unlock(&inactive_as_with_asid_lock);
  825.     interrupts_restore(ipl);
  826.    
  827.     /*
  828.      * Perform architecture-specific steps.
  829.      * (e.g. write ASID to hardware register etc.)
  830.      */
  831.     as_install_arch(new);
  832.    
  833.     AS = new;
  834. }
  835.  
  836. /** Convert address space area flags to page flags.
  837.  *
  838.  * @param aflags Flags of some address space area.
  839.  *
  840.  * @return Flags to be passed to page_mapping_insert().
  841.  */
  842. int area_flags_to_page_flags(int aflags)
  843. {
  844.     int flags;
  845.  
  846.     flags = PAGE_USER | PAGE_PRESENT;
  847.    
  848.     if (aflags & AS_AREA_READ)
  849.         flags |= PAGE_READ;
  850.        
  851.     if (aflags & AS_AREA_WRITE)
  852.         flags |= PAGE_WRITE;
  853.    
  854.     if (aflags & AS_AREA_EXEC)
  855.         flags |= PAGE_EXEC;
  856.    
  857.     if (!(aflags & AS_AREA_DEVICE))
  858.         flags |= PAGE_CACHEABLE;
  859.        
  860.     return flags;
  861. }
  862.  
  863. /** Compute flags for virtual address translation subsytem.
  864.  *
  865.  * The address space area must be locked.
  866.  * Interrupts must be disabled.
  867.  *
  868.  * @param a Address space area.
  869.  *
  870.  * @return Flags to be used in page_mapping_insert().
  871.  */
  872. int as_area_get_flags(as_area_t *a)
  873. {
  874.     return area_flags_to_page_flags(a->flags);
  875. }
  876.  
  877. /** Create page table.
  878.  *
  879.  * Depending on architecture, create either address space
  880.  * private or global page table.
  881.  *
  882.  * @param flags Flags saying whether the page table is for kernel address space.
  883.  *
  884.  * @return First entry of the page table.
  885.  */
  886. pte_t *page_table_create(int flags)
  887. {
  888.         ASSERT(as_operations);
  889.         ASSERT(as_operations->page_table_create);
  890.  
  891.         return as_operations->page_table_create(flags);
  892. }
  893.  
  894. /** Lock page table.
  895.  *
  896.  * This function should be called before any page_mapping_insert(),
  897.  * page_mapping_remove() and page_mapping_find().
  898.  *
  899.  * Locking order is such that address space areas must be locked
  900.  * prior to this call. Address space can be locked prior to this
  901.  * call in which case the lock argument is false.
  902.  *
  903.  * @param as Address space.
  904.  * @param lock If false, do not attempt to lock as->lock.
  905.  */
  906. void page_table_lock(as_t *as, bool lock)
  907. {
  908.     ASSERT(as_operations);
  909.     ASSERT(as_operations->page_table_lock);
  910.  
  911.     as_operations->page_table_lock(as, lock);
  912. }
  913.  
  914. /** Unlock page table.
  915.  *
  916.  * @param as Address space.
  917.  * @param unlock If false, do not attempt to unlock as->lock.
  918.  */
  919. void page_table_unlock(as_t *as, bool unlock)
  920. {
  921.     ASSERT(as_operations);
  922.     ASSERT(as_operations->page_table_unlock);
  923.  
  924.     as_operations->page_table_unlock(as, unlock);
  925. }
  926.  
  927.  
  928. /** Find address space area and lock it.
  929.  *
  930.  * The address space must be locked and interrupts must be disabled.
  931.  *
  932.  * @param as Address space.
  933.  * @param va Virtual address.
  934.  *
  935.  * @return Locked address space area containing va on success or NULL on failure.
  936.  */
  937. as_area_t *find_area_and_lock(as_t *as, __address va)
  938. {
  939.     as_area_t *a;
  940.     btree_node_t *leaf, *lnode;
  941.     int i;
  942.    
  943.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  944.     if (a) {
  945.         /* va is the base address of an address space area */
  946.         mutex_lock(&a->lock);
  947.         return a;
  948.     }
  949.    
  950.     /*
  951.      * Search the leaf node and the righmost record of its left neighbour
  952.      * to find out whether this is a miss or va belongs to an address
  953.      * space area found there.
  954.      */
  955.    
  956.     /* First, search the leaf node itself. */
  957.     for (i = 0; i < leaf->keys; i++) {
  958.         a = (as_area_t *) leaf->value[i];
  959.         mutex_lock(&a->lock);
  960.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  961.             return a;
  962.         }
  963.         mutex_unlock(&a->lock);
  964.     }
  965.  
  966.     /*
  967.      * Second, locate the left neighbour and test its last record.
  968.      * Because of its position in the B+tree, it must have base < va.
  969.      */
  970.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  971.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  972.         mutex_lock(&a->lock);
  973.         if (va < a->base + a->pages * PAGE_SIZE) {
  974.             return a;
  975.         }
  976.         mutex_unlock(&a->lock);
  977.     }
  978.  
  979.     return NULL;
  980. }
  981.  
  982. /** Check area conflicts with other areas.
  983.  *
  984.  * The address space must be locked and interrupts must be disabled.
  985.  *
  986.  * @param as Address space.
  987.  * @param va Starting virtual address of the area being tested.
  988.  * @param size Size of the area being tested.
  989.  * @param avoid_area Do not touch this area.
  990.  *
  991.  * @return True if there is no conflict, false otherwise.
  992.  */
  993. bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area)
  994. {
  995.     as_area_t *a;
  996.     btree_node_t *leaf, *node;
  997.     int i;
  998.    
  999.     /*
  1000.      * We don't want any area to have conflicts with NULL page.
  1001.      */
  1002.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1003.         return false;
  1004.    
  1005.     /*
  1006.      * The leaf node is found in O(log n), where n is proportional to
  1007.      * the number of address space areas belonging to as.
  1008.      * The check for conflicts is then attempted on the rightmost
  1009.      * record in the left neighbour, the leftmost record in the right
  1010.      * neighbour and all records in the leaf node itself.
  1011.      */
  1012.    
  1013.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1014.         if (a != avoid_area)
  1015.             return false;
  1016.     }
  1017.    
  1018.     /* First, check the two border cases. */
  1019.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1020.         a = (as_area_t *) node->value[node->keys - 1];
  1021.         mutex_lock(&a->lock);
  1022.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1023.             mutex_unlock(&a->lock);
  1024.             return false;
  1025.         }
  1026.         mutex_unlock(&a->lock);
  1027.     }
  1028.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1029.         a = (as_area_t *) node->value[0];
  1030.         mutex_lock(&a->lock);
  1031.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1032.             mutex_unlock(&a->lock);
  1033.             return false;
  1034.         }
  1035.         mutex_unlock(&a->lock);
  1036.     }
  1037.    
  1038.     /* Second, check the leaf node. */
  1039.     for (i = 0; i < leaf->keys; i++) {
  1040.         a = (as_area_t *) leaf->value[i];
  1041.    
  1042.         if (a == avoid_area)
  1043.             continue;
  1044.    
  1045.         mutex_lock(&a->lock);
  1046.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1047.             mutex_unlock(&a->lock);
  1048.             return false;
  1049.         }
  1050.         mutex_unlock(&a->lock);
  1051.     }
  1052.  
  1053.     /*
  1054.      * So far, the area does not conflict with other areas.
  1055.      * Check if it doesn't conflict with kernel address space.
  1056.      */  
  1057.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1058.         return !overlaps(va, size,
  1059.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1060.     }
  1061.  
  1062.     return true;
  1063. }
  1064.  
  1065. /** Return size of the address space area with given base.  */
  1066. size_t as_get_size(__address base)
  1067. {
  1068.     ipl_t ipl;
  1069.     as_area_t *src_area;
  1070.     size_t size;
  1071.  
  1072.     ipl = interrupts_disable();
  1073.     src_area = find_area_and_lock(AS, base);
  1074.     if (src_area){
  1075.         size = src_area->pages * PAGE_SIZE;
  1076.         mutex_unlock(&src_area->lock);
  1077.     } else {
  1078.         size = 0;
  1079.     }
  1080.     interrupts_restore(ipl);
  1081.     return size;
  1082. }
  1083.  
  1084. /** Mark portion of address space area as used.
  1085.  *
  1086.  * The address space area must be already locked.
  1087.  *
  1088.  * @param a Address space area.
  1089.  * @param page First page to be marked.
  1090.  * @param count Number of page to be marked.
  1091.  *
  1092.  * @return 0 on failure and 1 on success.
  1093.  */
  1094. int used_space_insert(as_area_t *a, __address page, count_t count)
  1095. {
  1096.     btree_node_t *leaf, *node;
  1097.     count_t pages;
  1098.     int i;
  1099.  
  1100.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1101.     ASSERT(count);
  1102.  
  1103.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1104.     if (pages) {
  1105.         /*
  1106.          * We hit the beginning of some used space.
  1107.          */
  1108.         return 0;
  1109.     }
  1110.  
  1111.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1112.     if (node) {
  1113.         __address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1114.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1115.        
  1116.         /*
  1117.          * Examine the possibility that the interval fits
  1118.          * somewhere between the rightmost interval of
  1119.          * the left neigbour and the first interval of the leaf.
  1120.          */
  1121.          
  1122.         if (page >= right_pg) {
  1123.             /* Do nothing. */
  1124.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1125.             /* The interval intersects with the left interval. */
  1126.             return 0;
  1127.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1128.             /* The interval intersects with the right interval. */
  1129.             return 0;          
  1130.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1131.             /* The interval can be added by merging the two already present intervals. */
  1132.             node->value[node->keys - 1] += count + right_cnt;
  1133.             btree_remove(&a->used_space, right_pg, leaf);
  1134.             return 1;
  1135.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1136.             /* The interval can be added by simply growing the left interval. */
  1137.             node->value[node->keys - 1] += count;
  1138.             return 1;
  1139.         } else if (page + count*PAGE_SIZE == right_pg) {
  1140.             /*
  1141.              * The interval can be addded by simply moving base of the right
  1142.              * interval down and increasing its size accordingly.
  1143.              */
  1144.             leaf->value[0] += count;
  1145.             leaf->key[0] = page;
  1146.             return 1;
  1147.         } else {
  1148.             /*
  1149.              * The interval is between both neigbouring intervals,
  1150.              * but cannot be merged with any of them.
  1151.              */
  1152.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1153.             return 1;
  1154.         }
  1155.     } else if (page < leaf->key[0]) {
  1156.         __address right_pg = leaf->key[0];
  1157.         count_t right_cnt = (count_t) leaf->value[0];
  1158.    
  1159.         /*
  1160.          * Investigate the border case in which the left neighbour does not
  1161.          * exist but the interval fits from the left.
  1162.          */
  1163.          
  1164.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1165.             /* The interval intersects with the right interval. */
  1166.             return 0;
  1167.         } else if (page + count*PAGE_SIZE == right_pg) {
  1168.             /*
  1169.              * The interval can be added by moving the base of the right interval down
  1170.              * and increasing its size accordingly.
  1171.              */
  1172.             leaf->key[0] = page;
  1173.             leaf->value[0] += count;
  1174.             return 1;
  1175.         } else {
  1176.             /*
  1177.              * The interval doesn't adjoin with the right interval.
  1178.              * It must be added individually.
  1179.              */
  1180.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1181.             return 1;
  1182.         }
  1183.     }
  1184.  
  1185.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1186.     if (node) {
  1187.         __address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1188.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1189.        
  1190.         /*
  1191.          * Examine the possibility that the interval fits
  1192.          * somewhere between the leftmost interval of
  1193.          * the right neigbour and the last interval of the leaf.
  1194.          */
  1195.  
  1196.         if (page < left_pg) {
  1197.             /* Do nothing. */
  1198.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1199.             /* The interval intersects with the left interval. */
  1200.             return 0;
  1201.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1202.             /* The interval intersects with the right interval. */
  1203.             return 0;          
  1204.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1205.             /* The interval can be added by merging the two already present intervals. */
  1206.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1207.             btree_remove(&a->used_space, right_pg, node);
  1208.             return 1;
  1209.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1210.             /* The interval can be added by simply growing the left interval. */
  1211.             leaf->value[leaf->keys - 1] +=  count;
  1212.             return 1;
  1213.         } else if (page + count*PAGE_SIZE == right_pg) {
  1214.             /*
  1215.              * The interval can be addded by simply moving base of the right
  1216.              * interval down and increasing its size accordingly.
  1217.              */
  1218.             node->value[0] += count;
  1219.             node->key[0] = page;
  1220.             return 1;
  1221.         } else {
  1222.             /*
  1223.              * The interval is between both neigbouring intervals,
  1224.              * but cannot be merged with any of them.
  1225.              */
  1226.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1227.             return 1;
  1228.         }
  1229.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1230.         __address left_pg = leaf->key[leaf->keys - 1];
  1231.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1232.    
  1233.         /*
  1234.          * Investigate the border case in which the right neighbour does not
  1235.          * exist but the interval fits from the right.
  1236.          */
  1237.          
  1238.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1239.             /* The interval intersects with the left interval. */
  1240.             return 0;
  1241.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1242.             /* The interval can be added by growing the left interval. */
  1243.             leaf->value[leaf->keys - 1] += count;
  1244.             return 1;
  1245.         } else {
  1246.             /*
  1247.              * The interval doesn't adjoin with the left interval.
  1248.              * It must be added individually.
  1249.              */
  1250.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1251.             return 1;
  1252.         }
  1253.     }
  1254.    
  1255.     /*
  1256.      * Note that if the algorithm made it thus far, the interval can fit only
  1257.      * between two other intervals of the leaf. The two border cases were already
  1258.      * resolved.
  1259.      */
  1260.     for (i = 1; i < leaf->keys; i++) {
  1261.         if (page < leaf->key[i]) {
  1262.             __address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1263.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1264.  
  1265.             /*
  1266.              * The interval fits between left_pg and right_pg.
  1267.              */
  1268.  
  1269.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1270.                 /* The interval intersects with the left interval. */
  1271.                 return 0;
  1272.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1273.                 /* The interval intersects with the right interval. */
  1274.                 return 0;          
  1275.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1276.                 /* The interval can be added by merging the two already present intervals. */
  1277.                 leaf->value[i - 1] += count + right_cnt;
  1278.                 btree_remove(&a->used_space, right_pg, leaf);
  1279.                 return 1;
  1280.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1281.                 /* The interval can be added by simply growing the left interval. */
  1282.                 leaf->value[i - 1] += count;
  1283.                 return 1;
  1284.             } else if (page + count*PAGE_SIZE == right_pg) {
  1285.                 /*
  1286.                      * The interval can be addded by simply moving base of the right
  1287.                  * interval down and increasing its size accordingly.
  1288.                  */
  1289.                 leaf->value[i] += count;
  1290.                 leaf->key[i] = page;
  1291.                 return 1;
  1292.             } else {
  1293.                 /*
  1294.                  * The interval is between both neigbouring intervals,
  1295.                  * but cannot be merged with any of them.
  1296.                  */
  1297.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1298.                 return 1;
  1299.             }
  1300.         }
  1301.     }
  1302.  
  1303.     panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
  1304. }
  1305.  
  1306. /** Mark portion of address space area as unused.
  1307.  *
  1308.  * The address space area must be already locked.
  1309.  *
  1310.  * @param a Address space area.
  1311.  * @param page First page to be marked.
  1312.  * @param count Number of page to be marked.
  1313.  *
  1314.  * @return 0 on failure and 1 on success.
  1315.  */
  1316. int used_space_remove(as_area_t *a, __address page, count_t count)
  1317. {
  1318.     btree_node_t *leaf, *node;
  1319.     count_t pages;
  1320.     int i;
  1321.  
  1322.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1323.     ASSERT(count);
  1324.  
  1325.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1326.     if (pages) {
  1327.         /*
  1328.          * We are lucky, page is the beginning of some interval.
  1329.          */
  1330.         if (count > pages) {
  1331.             return 0;
  1332.         } else if (count == pages) {
  1333.             btree_remove(&a->used_space, page, leaf);
  1334.             return 1;
  1335.         } else {
  1336.             /*
  1337.              * Find the respective interval.
  1338.              * Decrease its size and relocate its start address.
  1339.              */
  1340.             for (i = 0; i < leaf->keys; i++) {
  1341.                 if (leaf->key[i] == page) {
  1342.                     leaf->key[i] += count*PAGE_SIZE;
  1343.                     leaf->value[i] -= count;
  1344.                     return 1;
  1345.                 }
  1346.             }
  1347.             goto error;
  1348.         }
  1349.     }
  1350.  
  1351.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1352.     if (node && page < leaf->key[0]) {
  1353.         __address left_pg = node->key[node->keys - 1];
  1354.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1355.  
  1356.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1357.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1358.                 /*
  1359.                  * The interval is contained in the rightmost interval
  1360.                  * of the left neighbour and can be removed by
  1361.                  * updating the size of the bigger interval.
  1362.                  */
  1363.                 node->value[node->keys - 1] -= count;
  1364.                 return 1;
  1365.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1366.                 count_t new_cnt;
  1367.                
  1368.                 /*
  1369.                  * The interval is contained in the rightmost interval
  1370.                  * of the left neighbour but its removal requires
  1371.                  * both updating the size of the original interval and
  1372.                  * also inserting a new interval.
  1373.                  */
  1374.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1375.                 node->value[node->keys - 1] -= count + new_cnt;
  1376.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1377.                 return 1;
  1378.             }
  1379.         }
  1380.         return 0;
  1381.     } else if (page < leaf->key[0]) {
  1382.         return 0;
  1383.     }
  1384.    
  1385.     if (page > leaf->key[leaf->keys - 1]) {
  1386.         __address left_pg = leaf->key[leaf->keys - 1];
  1387.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1388.  
  1389.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1390.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1391.                 /*
  1392.                  * The interval is contained in the rightmost interval
  1393.                  * of the leaf and can be removed by updating the size
  1394.                  * of the bigger interval.
  1395.                  */
  1396.                 leaf->value[leaf->keys - 1] -= count;
  1397.                 return 1;
  1398.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1399.                 count_t new_cnt;
  1400.                
  1401.                 /*
  1402.                  * The interval is contained in the rightmost interval
  1403.                  * of the leaf but its removal requires both updating
  1404.                  * the size of the original interval and
  1405.                  * also inserting a new interval.
  1406.                  */
  1407.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1408.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1409.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1410.                 return 1;
  1411.             }
  1412.         }
  1413.         return 0;
  1414.     }  
  1415.    
  1416.     /*
  1417.      * The border cases have been already resolved.
  1418.      * Now the interval can be only between intervals of the leaf.
  1419.      */
  1420.     for (i = 1; i < leaf->keys - 1; i++) {
  1421.         if (page < leaf->key[i]) {
  1422.             __address left_pg = leaf->key[i - 1];
  1423.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1424.  
  1425.             /*
  1426.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1427.              */
  1428.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1429.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1430.                     /*
  1431.                     * The interval is contained in the interval (i - 1)
  1432.                      * of the leaf and can be removed by updating the size
  1433.                      * of the bigger interval.
  1434.                      */
  1435.                     leaf->value[i - 1] -= count;
  1436.                     return 1;
  1437.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1438.                     count_t new_cnt;
  1439.                
  1440.                     /*
  1441.                      * The interval is contained in the interval (i - 1)
  1442.                      * of the leaf but its removal requires both updating
  1443.                      * the size of the original interval and
  1444.                      * also inserting a new interval.
  1445.                      */
  1446.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1447.                     leaf->value[i - 1] -= count + new_cnt;
  1448.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1449.                     return 1;
  1450.                 }
  1451.             }
  1452.             return 0;
  1453.         }
  1454.     }
  1455.  
  1456. error:
  1457.     panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
  1458. }
  1459.  
  1460. /** Remove reference to address space area share info.
  1461.  *
  1462.  * If the reference count drops to 0, the sh_info is deallocated.
  1463.  *
  1464.  * @param sh_info Pointer to address space area share info.
  1465.  */
  1466. void sh_info_remove_reference(share_info_t *sh_info)
  1467. {
  1468.     bool dealloc = false;
  1469.  
  1470.     mutex_lock(&sh_info->lock);
  1471.     ASSERT(sh_info->refcount);
  1472.     if (--sh_info->refcount == 0) {
  1473.         dealloc = true;
  1474.         bool cond;
  1475.        
  1476.         /*
  1477.          * Now walk carefully the pagemap B+tree and free/remove
  1478.          * reference from all frames found there.
  1479.          */
  1480.         for (cond = true; cond;) {
  1481.             btree_node_t *node;
  1482.            
  1483.             ASSERT(!list_empty(&sh_info->pagemap.leaf_head));
  1484.             node = list_get_instance(sh_info->pagemap.leaf_head.next, btree_node_t, leaf_link);
  1485.             if ((cond = node->keys)) {
  1486.                 frame_free(ADDR2PFN((__address) node->value[0]));
  1487.                 btree_remove(&sh_info->pagemap, node->key[0], node);
  1488.             }
  1489.         }
  1490.        
  1491.     }
  1492.     mutex_unlock(&sh_info->lock);
  1493.    
  1494.     if (dealloc) {
  1495.         btree_destroy(&sh_info->pagemap);
  1496.         free(sh_info);
  1497.     }
  1498. }
  1499.  
  1500. static int anon_page_fault(as_area_t *area, __address addr, pf_access_t access);
  1501. static void anon_frame_free(as_area_t *area, __address page, __address frame);
  1502.  
  1503. /*
  1504.  * Anonymous memory backend.
  1505.  */
  1506. mem_backend_t anon_backend = {
  1507.     .backend_page_fault = anon_page_fault,
  1508.     .backend_frame_free = anon_frame_free
  1509. };
  1510.  
  1511. /** Service a page fault in the anonymous memory address space area.
  1512.  *
  1513.  * The address space area and page tables must be already locked.
  1514.  *
  1515.  * @param area Pointer to the address space area.
  1516.  * @param addr Faulting virtual address.
  1517.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  1518.  *
  1519.  * @return AS_PF_FAULT on failure (i.e. page fault) or AS_PF_OK on success (i.e. serviced).
  1520.  */
  1521. int anon_page_fault(as_area_t *area, __address addr, pf_access_t access)
  1522. {
  1523.     __address frame;
  1524.  
  1525.     if (!as_area_check_access(area, access))
  1526.         return AS_PF_FAULT;
  1527.  
  1528.     if (area->sh_info) {
  1529.         btree_node_t *leaf;
  1530.        
  1531.         /*
  1532.          * The area is shared, chances are that the mapping can be found
  1533.          * in the pagemap of the address space area share info structure.
  1534.          * In the case that the pagemap does not contain the respective
  1535.          * mapping, a new frame is allocated and the mapping is created.
  1536.          */
  1537.         mutex_lock(&area->sh_info->lock);
  1538.         frame = (__address) btree_search(&area->sh_info->pagemap,
  1539.             ALIGN_DOWN(addr, PAGE_SIZE) - area->base, &leaf);
  1540.         if (!frame) {
  1541.             bool allocate = true;
  1542.             int i;
  1543.            
  1544.             /*
  1545.              * Zero can be returned as a valid frame address.
  1546.              * Just a small workaround.
  1547.              */
  1548.             for (i = 0; i < leaf->keys; i++) {
  1549.                 if (leaf->key[i] == ALIGN_DOWN(addr, PAGE_SIZE)) {
  1550.                     allocate = false;
  1551.                     break;
  1552.                 }
  1553.             }
  1554.             if (allocate) {
  1555.                 frame = PFN2ADDR(frame_alloc(ONE_FRAME, 0));
  1556.                 memsetb(PA2KA(frame), FRAME_SIZE, 0);
  1557.                
  1558.                 /*
  1559.                  * Insert the address of the newly allocated frame to the pagemap.
  1560.                  */
  1561.                 btree_insert(&area->sh_info->pagemap, ALIGN_DOWN(addr, PAGE_SIZE) - area->base, (void *) frame, leaf);
  1562.             }
  1563.         }
  1564.         mutex_unlock(&area->sh_info->lock);
  1565.     } else {
  1566.  
  1567.         /*
  1568.          * In general, there can be several reasons that
  1569.          * can have caused this fault.
  1570.          *
  1571.          * - non-existent mapping: the area is an anonymous
  1572.          *   area (e.g. heap or stack) and so far has not been
  1573.          *   allocated a frame for the faulting page
  1574.          *
  1575.          * - non-present mapping: another possibility,
  1576.          *   currently not implemented, would be frame
  1577.          *   reuse; when this becomes a possibility,
  1578.          *   do not forget to distinguish between
  1579.          *   the different causes
  1580.          */
  1581.         frame = PFN2ADDR(frame_alloc(ONE_FRAME, 0));
  1582.         memsetb(PA2KA(frame), FRAME_SIZE, 0);
  1583.     }
  1584.    
  1585.     /*
  1586.      * Map 'page' to 'frame'.
  1587.      * Note that TLB shootdown is not attempted as only new information is being
  1588.      * inserted into page tables.
  1589.      */
  1590.     page_mapping_insert(AS, addr, frame, as_area_get_flags(area));
  1591.     if (!used_space_insert(area, ALIGN_DOWN(addr, PAGE_SIZE), 1))
  1592.         panic("Could not insert used space.\n");
  1593.        
  1594.     return AS_PF_OK;
  1595. }
  1596.  
  1597. /** Free a frame that is backed by the anonymous memory backend.
  1598.  *
  1599.  * The address space area and page tables must be already locked.
  1600.  *
  1601.  * @param area Ignored.
  1602.  * @param page Ignored.
  1603.  * @param frame Frame to be released.
  1604.  */
  1605. void anon_frame_free(as_area_t *area, __address page, __address frame)
  1606. {
  1607.     frame_free(ADDR2PFN(frame));
  1608. }
  1609.  
  1610. /*
  1611.  * Address space related syscalls.
  1612.  */
  1613.  
  1614. /** Wrapper for as_area_create(). */
  1615. __native sys_as_area_create(__address address, size_t size, int flags)
  1616. {
  1617.     if (as_area_create(AS, flags, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1618.         return (__native) address;
  1619.     else
  1620.         return (__native) -1;
  1621. }
  1622.  
  1623. /** Wrapper for as_area_resize. */
  1624. __native sys_as_area_resize(__address address, size_t size, int flags)
  1625. {
  1626.     return (__native) as_area_resize(AS, address, size, 0);
  1627. }
  1628.  
  1629. /** Wrapper for as_area_destroy. */
  1630. __native sys_as_area_destroy(__address address)
  1631. {
  1632.     return (__native) as_area_destroy(AS, address);
  1633. }
  1634.