Subversion Repositories HelenOS

Rev

Rev 2089 | Rev 2106 | 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 <syscall/copy.h>
  78. #include <arch/interrupt.h>
  79.  
  80. #ifdef CONFIG_VIRT_IDX_DCACHE
  81. #include <arch/mm/cache.h>
  82. #endif /* CONFIG_VIRT_IDX_DCACHE */
  83.  
  84. /**
  85.  * Each architecture decides what functions will be used to carry out
  86.  * address space operations such as creating or locking page tables.
  87.  */
  88. as_operations_t *as_operations = NULL;
  89.  
  90. /**
  91.  * Slab for as_t objects.
  92.  */
  93. static slab_cache_t *as_slab;
  94.  
  95. /**
  96.  * This lock protects inactive_as_with_asid_head list. It must be acquired
  97.  * before as_t mutex.
  98.  */
  99. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  100.  
  101. /**
  102.  * This list contains address spaces that are not active on any
  103.  * processor and that have valid ASID.
  104.  */
  105. LIST_INITIALIZE(inactive_as_with_asid_head);
  106.  
  107. /** Kernel address space. */
  108. as_t *AS_KERNEL = NULL;
  109.  
  110. static int area_flags_to_page_flags(int aflags);
  111. static as_area_t *find_area_and_lock(as_t *as, uintptr_t va);
  112. static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  113.     as_area_t *avoid_area);
  114. static void sh_info_remove_reference(share_info_t *sh_info);
  115.  
  116. static int as_constructor(void *obj, int flags)
  117. {
  118.     as_t *as = (as_t *) obj;
  119.     int rc;
  120.  
  121.     link_initialize(&as->inactive_as_with_asid_link);
  122.     mutex_initialize(&as->lock);   
  123.    
  124.     rc = as_constructor_arch(as, flags);
  125.    
  126.     return rc;
  127. }
  128.  
  129. static int as_destructor(void *obj)
  130. {
  131.     as_t *as = (as_t *) obj;
  132.  
  133.     return as_destructor_arch(as);
  134. }
  135.  
  136. /** Initialize address space subsystem. */
  137. void as_init(void)
  138. {
  139.     as_arch_init();
  140.    
  141.     as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,
  142.         as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);
  143.    
  144.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  145.     if (!AS_KERNEL)
  146.         panic("can't create kernel address space\n");
  147.    
  148. }
  149.  
  150. /** Create address space.
  151.  *
  152.  * @param flags Flags that influence way in wich the address space is created.
  153.  */
  154. as_t *as_create(int flags)
  155. {
  156.     as_t *as;
  157.  
  158.     as = (as_t *) slab_alloc(as_slab, 0);
  159.     (void) as_create_arch(as, 0);
  160.    
  161.     btree_create(&as->as_area_btree);
  162.    
  163.     if (flags & FLAG_AS_KERNEL)
  164.         as->asid = ASID_KERNEL;
  165.     else
  166.         as->asid = ASID_INVALID;
  167.    
  168.     as->refcount = 0;
  169.     as->cpu_refcount = 0;
  170. #ifdef AS_PAGE_TABLE
  171.     as->page_table = page_table_create(flags);
  172. #else
  173.     page_table_create(flags);
  174. #endif
  175.  
  176.     return as;
  177. }
  178.  
  179. /** Destroy adress space.
  180.  *
  181.  * When there are no tasks referencing this address space (i.e. its refcount is
  182.  * zero), the address space can be destroyed.
  183.  */
  184. void as_destroy(as_t *as)
  185. {
  186.     ipl_t ipl;
  187.     bool cond;
  188.  
  189.     ASSERT(as->refcount == 0);
  190.    
  191.     /*
  192.      * Since there is no reference to this area,
  193.      * it is safe not to lock its mutex.
  194.      */
  195.     ipl = interrupts_disable();
  196.     spinlock_lock(&inactive_as_with_asid_lock);
  197.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  198.         if (as != AS && as->cpu_refcount == 0)
  199.             list_remove(&as->inactive_as_with_asid_link);
  200.         asid_put(as->asid);
  201.     }
  202.     spinlock_unlock(&inactive_as_with_asid_lock);
  203.  
  204.     /*
  205.      * Destroy address space areas of the address space.
  206.      * The B+tree must be walked carefully because it is
  207.      * also being destroyed.
  208.      */
  209.     for (cond = true; cond; ) {
  210.         btree_node_t *node;
  211.  
  212.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  213.         node = list_get_instance(as->as_area_btree.leaf_head.next,
  214.             btree_node_t, leaf_link);
  215.  
  216.         if ((cond = node->keys)) {
  217.             as_area_destroy(as, node->key[0]);
  218.         }
  219.     }
  220.  
  221.     btree_destroy(&as->as_area_btree);
  222. #ifdef AS_PAGE_TABLE
  223.     page_table_destroy(as->page_table);
  224. #else
  225.     page_table_destroy(NULL);
  226. #endif
  227.  
  228.     interrupts_restore(ipl);
  229.    
  230.     slab_free(as_slab, as);
  231. }
  232.  
  233. /** Create address space area of common attributes.
  234.  *
  235.  * The created address space area is added to the target address space.
  236.  *
  237.  * @param as Target address space.
  238.  * @param flags Flags of the area memory.
  239.  * @param size Size of area.
  240.  * @param base Base address of area.
  241.  * @param attrs Attributes of the area.
  242.  * @param backend Address space area backend. NULL if no backend is used.
  243.  * @param backend_data NULL or a pointer to an array holding two void *.
  244.  *
  245.  * @return Address space area on success or NULL on failure.
  246.  */
  247. as_area_t *
  248. as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
  249.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  250. {
  251.     ipl_t ipl;
  252.     as_area_t *a;
  253.    
  254.     if (base % PAGE_SIZE)
  255.         return NULL;
  256.  
  257.     if (!size)
  258.         return NULL;
  259.  
  260.     /* Writeable executable areas are not supported. */
  261.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  262.         return NULL;
  263.    
  264.     ipl = interrupts_disable();
  265.     mutex_lock(&as->lock);
  266.    
  267.     if (!check_area_conflicts(as, base, size, NULL)) {
  268.         mutex_unlock(&as->lock);
  269.         interrupts_restore(ipl);
  270.         return NULL;
  271.     }
  272.    
  273.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  274.  
  275.     mutex_initialize(&a->lock);
  276.    
  277.     a->as = as;
  278.     a->flags = flags;
  279.     a->attributes = attrs;
  280.     a->pages = SIZE2FRAMES(size);
  281.     a->base = base;
  282.     a->sh_info = NULL;
  283.     a->backend = backend;
  284.     if (backend_data)
  285.         a->backend_data = *backend_data;
  286.     else
  287.         memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data),
  288.             0);
  289.  
  290.     btree_create(&a->used_space);
  291.    
  292.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  293.  
  294.     mutex_unlock(&as->lock);
  295.     interrupts_restore(ipl);
  296.  
  297.     return a;
  298. }
  299.  
  300. /** Find address space area and change it.
  301.  *
  302.  * @param as Address space.
  303.  * @param address Virtual address belonging to the area to be changed. Must be
  304.  *     page-aligned.
  305.  * @param size New size of the virtual memory block starting at address.
  306.  * @param flags Flags influencing the remap operation. Currently unused.
  307.  *
  308.  * @return Zero on success or a value from @ref errno.h otherwise.
  309.  */
  310. int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
  311. {
  312.     as_area_t *area;
  313.     ipl_t ipl;
  314.     size_t pages;
  315.    
  316.     ipl = interrupts_disable();
  317.     mutex_lock(&as->lock);
  318.    
  319.     /*
  320.      * Locate the area.
  321.      */
  322.     area = find_area_and_lock(as, address);
  323.     if (!area) {
  324.         mutex_unlock(&as->lock);
  325.         interrupts_restore(ipl);
  326.         return ENOENT;
  327.     }
  328.  
  329.     if (area->backend == &phys_backend) {
  330.         /*
  331.          * Remapping of address space areas associated
  332.          * with memory mapped devices is not supported.
  333.          */
  334.         mutex_unlock(&area->lock);
  335.         mutex_unlock(&as->lock);
  336.         interrupts_restore(ipl);
  337.         return ENOTSUP;
  338.     }
  339.     if (area->sh_info) {
  340.         /*
  341.          * Remapping of shared address space areas
  342.          * is not supported.
  343.          */
  344.         mutex_unlock(&area->lock);
  345.         mutex_unlock(&as->lock);
  346.         interrupts_restore(ipl);
  347.         return ENOTSUP;
  348.     }
  349.  
  350.     pages = SIZE2FRAMES((address - area->base) + size);
  351.     if (!pages) {
  352.         /*
  353.          * Zero size address space areas are not allowed.
  354.          */
  355.         mutex_unlock(&area->lock);
  356.         mutex_unlock(&as->lock);
  357.         interrupts_restore(ipl);
  358.         return EPERM;
  359.     }
  360.    
  361.     if (pages < area->pages) {
  362.         bool cond;
  363.         uintptr_t start_free = area->base + pages*PAGE_SIZE;
  364.  
  365.         /*
  366.          * Shrinking the area.
  367.          * No need to check for overlaps.
  368.          */
  369.  
  370.         /*
  371.          * Start TLB shootdown sequence.
  372.          */
  373.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base +
  374.             pages * PAGE_SIZE, area->pages - pages);
  375.  
  376.         /*
  377.          * Remove frames belonging to used space starting from
  378.          * the highest addresses downwards until an overlap with
  379.          * the resized address space area is found. Note that this
  380.          * is also the right way to remove part of the used_space
  381.          * B+tree leaf list.
  382.          */    
  383.         for (cond = true; cond;) {
  384.             btree_node_t *node;
  385.        
  386.             ASSERT(!list_empty(&area->used_space.leaf_head));
  387.             node =
  388.                 list_get_instance(area->used_space.leaf_head.prev,
  389.                 btree_node_t, leaf_link);
  390.             if ((cond = (bool) node->keys)) {
  391.                 uintptr_t b = node->key[node->keys - 1];
  392.                 count_t c =
  393.                     (count_t) node->value[node->keys - 1];
  394.                 int i = 0;
  395.            
  396.                 if (overlaps(b, c * PAGE_SIZE, area->base,
  397.                     pages*PAGE_SIZE)) {
  398.                    
  399.                     if (b + c * PAGE_SIZE <= start_free) {
  400.                         /*
  401.                          * The whole interval fits
  402.                          * completely in the resized
  403.                          * address space area.
  404.                          */
  405.                         break;
  406.                     }
  407.        
  408.                     /*
  409.                      * Part of the interval corresponding
  410.                      * to b and c overlaps with the resized
  411.                      * address space area.
  412.                      */
  413.        
  414.                     cond = false;   /* we are almost done */
  415.                     i = (start_free - b) >> PAGE_WIDTH;
  416.                     if (!used_space_remove(area, start_free,
  417.                         c - i))
  418.                         panic("Could not remove used "
  419.                             "space.\n");
  420.                 } else {
  421.                     /*
  422.                      * The interval of used space can be
  423.                      * completely removed.
  424.                      */
  425.                     if (!used_space_remove(area, b, c))
  426.                         panic("Could not remove used "
  427.                             "space.\n");
  428.                 }
  429.            
  430.                 for (; i < c; i++) {
  431.                     pte_t *pte;
  432.            
  433.                     page_table_lock(as, false);
  434.                     pte = page_mapping_find(as, b +
  435.                         i * PAGE_SIZE);
  436.                     ASSERT(pte && PTE_VALID(pte) &&
  437.                         PTE_PRESENT(pte));
  438.                     if (area->backend &&
  439.                         area->backend->frame_free) {
  440.                         area->backend->frame_free(area,
  441.                             b + i * PAGE_SIZE,
  442.                             PTE_GET_FRAME(pte));
  443.                     }
  444.                     page_mapping_remove(as, b +
  445.                         i * PAGE_SIZE);
  446.                     page_table_unlock(as, false);
  447.                 }
  448.             }
  449.         }
  450.  
  451.         /*
  452.          * Finish TLB shootdown sequence.
  453.          */
  454.         tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,
  455.             area->pages - pages);
  456.         tlb_shootdown_finalize();
  457.        
  458.         /*
  459.          * Invalidate software translation caches (e.g. TSB on sparc64).
  460.          */
  461.         as_invalidate_translation_cache(as, area->base +
  462.             pages * PAGE_SIZE, area->pages - pages);
  463.     } else {
  464.         /*
  465.          * Growing the area.
  466.          * Check for overlaps with other address space areas.
  467.          */
  468.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE,
  469.             area)) {
  470.             mutex_unlock(&area->lock);
  471.             mutex_unlock(&as->lock);       
  472.             interrupts_restore(ipl);
  473.             return EADDRNOTAVAIL;
  474.         }
  475.     }
  476.  
  477.     area->pages = pages;
  478.    
  479.     mutex_unlock(&area->lock);
  480.     mutex_unlock(&as->lock);
  481.     interrupts_restore(ipl);
  482.  
  483.     return 0;
  484. }
  485.  
  486. /** Destroy address space area.
  487.  *
  488.  * @param as Address space.
  489.  * @param address Address withing the area to be deleted.
  490.  *
  491.  * @return Zero on success or a value from @ref errno.h on failure.
  492.  */
  493. int as_area_destroy(as_t *as, uintptr_t address)
  494. {
  495.     as_area_t *area;
  496.     uintptr_t base;
  497.     link_t *cur;
  498.     ipl_t ipl;
  499.  
  500.     ipl = interrupts_disable();
  501.     mutex_lock(&as->lock);
  502.  
  503.     area = find_area_and_lock(as, address);
  504.     if (!area) {
  505.         mutex_unlock(&as->lock);
  506.         interrupts_restore(ipl);
  507.         return ENOENT;
  508.     }
  509.  
  510.     base = area->base;
  511.  
  512.     /*
  513.      * Start TLB shootdown sequence.
  514.      */
  515.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  516.  
  517.     /*
  518.      * Visit only the pages mapped by used_space B+tree.
  519.      */
  520.     for (cur = area->used_space.leaf_head.next;
  521.         cur != &area->used_space.leaf_head; cur = cur->next) {
  522.         btree_node_t *node;
  523.         int i;
  524.        
  525.         node = list_get_instance(cur, btree_node_t, leaf_link);
  526.         for (i = 0; i < node->keys; i++) {
  527.             uintptr_t b = node->key[i];
  528.             count_t j;
  529.             pte_t *pte;
  530.            
  531.             for (j = 0; j < (count_t) node->value[i]; j++) {
  532.                 page_table_lock(as, false);
  533.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  534.                 ASSERT(pte && PTE_VALID(pte) &&
  535.                     PTE_PRESENT(pte));
  536.                 if (area->backend &&
  537.                     area->backend->frame_free) {
  538.                     area->backend->frame_free(area, b +
  539.                     j * PAGE_SIZE, PTE_GET_FRAME(pte));
  540.                 }
  541.                 page_mapping_remove(as, b + j * PAGE_SIZE);            
  542.                 page_table_unlock(as, false);
  543.             }
  544.         }
  545.     }
  546.  
  547.     /*
  548.      * Finish TLB shootdown sequence.
  549.      */
  550.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  551.     tlb_shootdown_finalize();
  552.    
  553.     /*
  554.      * Invalidate potential software translation caches (e.g. TSB on
  555.      * sparc64).
  556.      */
  557.     as_invalidate_translation_cache(as, area->base, area->pages);
  558.    
  559.     btree_destroy(&area->used_space);
  560.  
  561.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  562.    
  563.     if (area->sh_info)
  564.         sh_info_remove_reference(area->sh_info);
  565.        
  566.     mutex_unlock(&area->lock);
  567.  
  568.     /*
  569.      * Remove the empty area from address space.
  570.      */
  571.     btree_remove(&as->as_area_btree, base, NULL);
  572.    
  573.     free(area);
  574.    
  575.     mutex_unlock(&as->lock);
  576.     interrupts_restore(ipl);
  577.     return 0;
  578. }
  579.  
  580. /** Share address space area with another or the same address space.
  581.  *
  582.  * Address space area mapping is shared with a new address space area.
  583.  * If the source address space area has not been shared so far,
  584.  * a new sh_info is created. The new address space area simply gets the
  585.  * sh_info of the source area. The process of duplicating the
  586.  * mapping is done through the backend share function.
  587.  *
  588.  * @param src_as Pointer to source address space.
  589.  * @param src_base Base address of the source address space area.
  590.  * @param acc_size Expected size of the source area.
  591.  * @param dst_as Pointer to destination address space.
  592.  * @param dst_base Target base address.
  593.  * @param dst_flags_mask Destination address space area flags mask.
  594.  *
  595.  * @return Zero on success or ENOENT if there is no such task or if there is no
  596.  * such address space area, EPERM if there was a problem in accepting the area
  597.  * or ENOMEM if there was a problem in allocating destination address space
  598.  * area. ENOTSUP is returned if the address space area backend does not support
  599.  * sharing or if the kernel detects an attempt to create an illegal address
  600.  * alias.
  601.  */
  602. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  603.           as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  604. {
  605.     ipl_t ipl;
  606.     int src_flags;
  607.     size_t src_size;
  608.     as_area_t *src_area, *dst_area;
  609.     share_info_t *sh_info;
  610.     mem_backend_t *src_backend;
  611.     mem_backend_data_t src_backend_data;
  612.    
  613.     ipl = interrupts_disable();
  614.     mutex_lock(&src_as->lock);
  615.     src_area = find_area_and_lock(src_as, src_base);
  616.     if (!src_area) {
  617.         /*
  618.          * Could not find the source address space area.
  619.          */
  620.         mutex_unlock(&src_as->lock);
  621.         interrupts_restore(ipl);
  622.         return ENOENT;
  623.     }
  624.  
  625.     if (!src_area->backend || !src_area->backend->share) {
  626.         /*
  627.          * There is no backend or the backend does not
  628.          * know how to share the area.
  629.          */
  630.         mutex_unlock(&src_area->lock);
  631.         mutex_unlock(&src_as->lock);
  632.         interrupts_restore(ipl);
  633.         return ENOTSUP;
  634.     }
  635.    
  636.     src_size = src_area->pages * PAGE_SIZE;
  637.     src_flags = src_area->flags;
  638.     src_backend = src_area->backend;
  639.     src_backend_data = src_area->backend_data;
  640.  
  641.     /* Share the cacheable flag from the original mapping */
  642.     if (src_flags & AS_AREA_CACHEABLE)
  643.         dst_flags_mask |= AS_AREA_CACHEABLE;
  644.  
  645.     if (src_size != acc_size ||
  646.         (src_flags & dst_flags_mask) != dst_flags_mask) {
  647.         mutex_unlock(&src_area->lock);
  648.         mutex_unlock(&src_as->lock);
  649.         interrupts_restore(ipl);
  650.         return EPERM;
  651.     }
  652.  
  653. #ifdef CONFIG_VIRT_IDX_DCACHE
  654.     if (!(dst_flags_mask & AS_AREA_EXEC)) {
  655.         if (PAGE_COLOR(src_area->base) != PAGE_COLOR(dst_base)) {
  656.             /*
  657.              * Refuse to create an illegal address alias.
  658.              */
  659.             mutex_unlock(&src_area->lock);
  660.             mutex_unlock(&src_as->lock);
  661.             interrupts_restore(ipl);
  662.             return ENOTSUP;
  663.         }
  664.     }
  665. #endif /* CONFIG_VIRT_IDX_DCACHE */
  666.  
  667.     /*
  668.      * Now we are committed to sharing the area.
  669.      * First, prepare the area for sharing.
  670.      * Then it will be safe to unlock it.
  671.      */
  672.     sh_info = src_area->sh_info;
  673.     if (!sh_info) {
  674.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  675.         mutex_initialize(&sh_info->lock);
  676.         sh_info->refcount = 2;
  677.         btree_create(&sh_info->pagemap);
  678.         src_area->sh_info = sh_info;
  679.     } else {
  680.         mutex_lock(&sh_info->lock);
  681.         sh_info->refcount++;
  682.         mutex_unlock(&sh_info->lock);
  683.     }
  684.  
  685.     src_area->backend->share(src_area);
  686.  
  687.     mutex_unlock(&src_area->lock);
  688.     mutex_unlock(&src_as->lock);
  689.  
  690.     /*
  691.      * Create copy of the source address space area.
  692.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  693.      * attribute set which prevents race condition with
  694.      * preliminary as_page_fault() calls.
  695.      * The flags of the source area are masked against dst_flags_mask
  696.      * to support sharing in less privileged mode.
  697.      */
  698.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  699.         AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  700.     if (!dst_area) {
  701.         /*
  702.          * Destination address space area could not be created.
  703.          */
  704.         sh_info_remove_reference(sh_info);
  705.        
  706.         interrupts_restore(ipl);
  707.         return ENOMEM;
  708.     }
  709.  
  710.     /*
  711.      * Now the destination address space area has been
  712.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  713.      * attribute and set the sh_info.
  714.      */
  715.     mutex_lock(&dst_as->lock); 
  716.     mutex_lock(&dst_area->lock);
  717.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  718.     dst_area->sh_info = sh_info;
  719.     mutex_unlock(&dst_area->lock);
  720.     mutex_unlock(&dst_as->lock);   
  721.  
  722.     interrupts_restore(ipl);
  723.    
  724.     return 0;
  725. }
  726.  
  727. /** Check access mode for address space area.
  728.  *
  729.  * The address space area must be locked prior to this call.
  730.  *
  731.  * @param area Address space area.
  732.  * @param access Access mode.
  733.  *
  734.  * @return False if access violates area's permissions, true otherwise.
  735.  */
  736. bool as_area_check_access(as_area_t *area, pf_access_t access)
  737. {
  738.     int flagmap[] = {
  739.         [PF_ACCESS_READ] = AS_AREA_READ,
  740.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  741.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  742.     };
  743.  
  744.     if (!(area->flags & flagmap[access]))
  745.         return false;
  746.    
  747.     return true;
  748. }
  749.  
  750. /** Handle page fault within the current address space.
  751.  *
  752.  * This is the high-level page fault handler. It decides
  753.  * whether the page fault can be resolved by any backend
  754.  * and if so, it invokes the backend to resolve the page
  755.  * fault.
  756.  *
  757.  * Interrupts are assumed disabled.
  758.  *
  759.  * @param page Faulting page.
  760.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  761.  * @param istate Pointer to interrupted state.
  762.  *
  763.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  764.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  765.  */
  766. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  767. {
  768.     pte_t *pte;
  769.     as_area_t *area;
  770.    
  771.     if (!THREAD)
  772.         return AS_PF_FAULT;
  773.        
  774.     ASSERT(AS);
  775.  
  776.     mutex_lock(&AS->lock);
  777.     area = find_area_and_lock(AS, page);   
  778.     if (!area) {
  779.         /*
  780.          * No area contained mapping for 'page'.
  781.          * Signal page fault to low-level handler.
  782.          */
  783.         mutex_unlock(&AS->lock);
  784.         goto page_fault;
  785.     }
  786.  
  787.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  788.         /*
  789.          * The address space area is not fully initialized.
  790.          * Avoid possible race by returning error.
  791.          */
  792.         mutex_unlock(&area->lock);
  793.         mutex_unlock(&AS->lock);
  794.         goto page_fault;       
  795.     }
  796.  
  797.     if (!area->backend || !area->backend->page_fault) {
  798.         /*
  799.          * The address space area is not backed by any backend
  800.          * or the backend cannot handle page faults.
  801.          */
  802.         mutex_unlock(&area->lock);
  803.         mutex_unlock(&AS->lock);
  804.         goto page_fault;       
  805.     }
  806.  
  807.     page_table_lock(AS, false);
  808.    
  809.     /*
  810.      * To avoid race condition between two page faults
  811.      * on the same address, we need to make sure
  812.      * the mapping has not been already inserted.
  813.      */
  814.     if ((pte = page_mapping_find(AS, page))) {
  815.         if (PTE_PRESENT(pte)) {
  816.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  817.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  818.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  819.                 page_table_unlock(AS, false);
  820.                 mutex_unlock(&area->lock);
  821.                 mutex_unlock(&AS->lock);
  822.                 return AS_PF_OK;
  823.             }
  824.         }
  825.     }
  826.    
  827.     /*
  828.      * Resort to the backend page fault handler.
  829.      */
  830.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  831.         page_table_unlock(AS, false);
  832.         mutex_unlock(&area->lock);
  833.         mutex_unlock(&AS->lock);
  834.         goto page_fault;
  835.     }
  836.    
  837.     page_table_unlock(AS, false);
  838.     mutex_unlock(&area->lock);
  839.     mutex_unlock(&AS->lock);
  840.     return AS_PF_OK;
  841.  
  842. page_fault:
  843.     if (THREAD->in_copy_from_uspace) {
  844.         THREAD->in_copy_from_uspace = false;
  845.         istate_set_retaddr(istate,
  846.             (uintptr_t) &memcpy_from_uspace_failover_address);
  847.     } else if (THREAD->in_copy_to_uspace) {
  848.         THREAD->in_copy_to_uspace = false;
  849.         istate_set_retaddr(istate,
  850.             (uintptr_t) &memcpy_to_uspace_failover_address);
  851.     } else {
  852.         return AS_PF_FAULT;
  853.     }
  854.  
  855.     return AS_PF_DEFER;
  856. }
  857.  
  858. /** Switch address spaces.
  859.  *
  860.  * Note that this function cannot sleep as it is essentially a part of
  861.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  862.  *
  863.  * @param old Old address space or NULL.
  864.  * @param new New address space.
  865.  */
  866. void as_switch(as_t *old, as_t *replace)
  867. {
  868.     ipl_t ipl;
  869.     bool needs_asid = false;
  870.    
  871.     ipl = interrupts_disable();
  872.     spinlock_lock(&inactive_as_with_asid_lock);
  873.  
  874.     /*
  875.      * First, take care of the old address space.
  876.      */
  877.     if (old) {
  878.         mutex_lock_active(&old->lock);
  879.         ASSERT(old->cpu_refcount);
  880.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  881.             /*
  882.              * The old address space is no longer active on
  883.              * any processor. It can be appended to the
  884.              * list of inactive address spaces with assigned
  885.              * ASID.
  886.              */
  887.              ASSERT(old->asid != ASID_INVALID);
  888.              list_append(&old->inactive_as_with_asid_link,
  889.                  &inactive_as_with_asid_head);
  890.         }
  891.         mutex_unlock(&old->lock);
  892.  
  893.         /*
  894.          * Perform architecture-specific tasks when the address space
  895.          * is being removed from the CPU.
  896.          */
  897.         as_deinstall_arch(old);
  898.     }
  899.  
  900.     /*
  901.      * Second, prepare the new address space.
  902.      */
  903.     mutex_lock_active(&replace->lock);
  904.     if ((replace->cpu_refcount++ == 0) && (replace != AS_KERNEL)) {
  905.         if (replace->asid != ASID_INVALID) {
  906.             list_remove(&replace->inactive_as_with_asid_link);
  907.         } else {
  908.             /*
  909.              * Defer call to asid_get() until replace->lock is released.
  910.              */
  911.             needs_asid = true;
  912.         }
  913.     }
  914.     SET_PTL0_ADDRESS(replace->page_table);
  915.     mutex_unlock(&replace->lock);
  916.  
  917.     if (needs_asid) {
  918.         /*
  919.          * Allocation of new ASID was deferred
  920.          * until now in order to avoid deadlock.
  921.          */
  922.         asid_t asid;
  923.        
  924.         asid = asid_get();
  925.         mutex_lock_active(&replace->lock);
  926.         replace->asid = asid;
  927.         mutex_unlock(&replace->lock);
  928.     }
  929.     spinlock_unlock(&inactive_as_with_asid_lock);
  930.     interrupts_restore(ipl);
  931.    
  932.     /*
  933.      * Perform architecture-specific steps.
  934.      * (e.g. write ASID to hardware register etc.)
  935.      */
  936.     as_install_arch(replace);
  937.    
  938.     AS = replace;
  939. }
  940.  
  941. /** Convert address space area flags to page flags.
  942.  *
  943.  * @param aflags Flags of some address space area.
  944.  *
  945.  * @return Flags to be passed to page_mapping_insert().
  946.  */
  947. int area_flags_to_page_flags(int aflags)
  948. {
  949.     int flags;
  950.  
  951.     flags = PAGE_USER | PAGE_PRESENT;
  952.    
  953.     if (aflags & AS_AREA_READ)
  954.         flags |= PAGE_READ;
  955.        
  956.     if (aflags & AS_AREA_WRITE)
  957.         flags |= PAGE_WRITE;
  958.    
  959.     if (aflags & AS_AREA_EXEC)
  960.         flags |= PAGE_EXEC;
  961.    
  962.     if (aflags & AS_AREA_CACHEABLE)
  963.         flags |= PAGE_CACHEABLE;
  964.        
  965.     return flags;
  966. }
  967.  
  968. /** Compute flags for virtual address translation subsytem.
  969.  *
  970.  * The address space area must be locked.
  971.  * Interrupts must be disabled.
  972.  *
  973.  * @param a Address space area.
  974.  *
  975.  * @return Flags to be used in page_mapping_insert().
  976.  */
  977. int as_area_get_flags(as_area_t *a)
  978. {
  979.     return area_flags_to_page_flags(a->flags);
  980. }
  981.  
  982. /** Create page table.
  983.  *
  984.  * Depending on architecture, create either address space
  985.  * private or global page table.
  986.  *
  987.  * @param flags Flags saying whether the page table is for kernel address space.
  988.  *
  989.  * @return First entry of the page table.
  990.  */
  991. pte_t *page_table_create(int flags)
  992. {
  993.         ASSERT(as_operations);
  994.         ASSERT(as_operations->page_table_create);
  995.  
  996.         return as_operations->page_table_create(flags);
  997. }
  998.  
  999. /** Destroy page table.
  1000.  *
  1001.  * Destroy page table in architecture specific way.
  1002.  *
  1003.  * @param page_table Physical address of PTL0.
  1004.  */
  1005. void page_table_destroy(pte_t *page_table)
  1006. {
  1007.         ASSERT(as_operations);
  1008.         ASSERT(as_operations->page_table_destroy);
  1009.  
  1010.         as_operations->page_table_destroy(page_table);
  1011. }
  1012.  
  1013. /** Lock page table.
  1014.  *
  1015.  * This function should be called before any page_mapping_insert(),
  1016.  * page_mapping_remove() and page_mapping_find().
  1017.  *
  1018.  * Locking order is such that address space areas must be locked
  1019.  * prior to this call. Address space can be locked prior to this
  1020.  * call in which case the lock argument is false.
  1021.  *
  1022.  * @param as Address space.
  1023.  * @param lock If false, do not attempt to lock as->lock.
  1024.  */
  1025. void page_table_lock(as_t *as, bool lock)
  1026. {
  1027.     ASSERT(as_operations);
  1028.     ASSERT(as_operations->page_table_lock);
  1029.  
  1030.     as_operations->page_table_lock(as, lock);
  1031. }
  1032.  
  1033. /** Unlock page table.
  1034.  *
  1035.  * @param as Address space.
  1036.  * @param unlock If false, do not attempt to unlock as->lock.
  1037.  */
  1038. void page_table_unlock(as_t *as, bool unlock)
  1039. {
  1040.     ASSERT(as_operations);
  1041.     ASSERT(as_operations->page_table_unlock);
  1042.  
  1043.     as_operations->page_table_unlock(as, unlock);
  1044. }
  1045.  
  1046.  
  1047. /** Find address space area and lock it.
  1048.  *
  1049.  * The address space must be locked and interrupts must be disabled.
  1050.  *
  1051.  * @param as Address space.
  1052.  * @param va Virtual address.
  1053.  *
  1054.  * @return Locked address space area containing va on success or NULL on
  1055.  *     failure.
  1056.  */
  1057. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  1058. {
  1059.     as_area_t *a;
  1060.     btree_node_t *leaf, *lnode;
  1061.     int i;
  1062.    
  1063.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  1064.     if (a) {
  1065.         /* va is the base address of an address space area */
  1066.         mutex_lock(&a->lock);
  1067.         return a;
  1068.     }
  1069.    
  1070.     /*
  1071.      * Search the leaf node and the righmost record of its left neighbour
  1072.      * to find out whether this is a miss or va belongs to an address
  1073.      * space area found there.
  1074.      */
  1075.    
  1076.     /* First, search the leaf node itself. */
  1077.     for (i = 0; i < leaf->keys; i++) {
  1078.         a = (as_area_t *) leaf->value[i];
  1079.         mutex_lock(&a->lock);
  1080.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  1081.             return a;
  1082.         }
  1083.         mutex_unlock(&a->lock);
  1084.     }
  1085.  
  1086.     /*
  1087.      * Second, locate the left neighbour and test its last record.
  1088.      * Because of its position in the B+tree, it must have base < va.
  1089.      */
  1090.     lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf);
  1091.     if (lnode) {
  1092.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  1093.         mutex_lock(&a->lock);
  1094.         if (va < a->base + a->pages * PAGE_SIZE) {
  1095.             return a;
  1096.         }
  1097.         mutex_unlock(&a->lock);
  1098.     }
  1099.  
  1100.     return NULL;
  1101. }
  1102.  
  1103. /** Check area conflicts with other areas.
  1104.  *
  1105.  * The address space must be locked and interrupts must be disabled.
  1106.  *
  1107.  * @param as Address space.
  1108.  * @param va Starting virtual address of the area being tested.
  1109.  * @param size Size of the area being tested.
  1110.  * @param avoid_area Do not touch this area.
  1111.  *
  1112.  * @return True if there is no conflict, false otherwise.
  1113.  */
  1114. bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  1115.               as_area_t *avoid_area)
  1116. {
  1117.     as_area_t *a;
  1118.     btree_node_t *leaf, *node;
  1119.     int i;
  1120.    
  1121.     /*
  1122.      * We don't want any area to have conflicts with NULL page.
  1123.      */
  1124.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1125.         return false;
  1126.    
  1127.     /*
  1128.      * The leaf node is found in O(log n), where n is proportional to
  1129.      * the number of address space areas belonging to as.
  1130.      * The check for conflicts is then attempted on the rightmost
  1131.      * record in the left neighbour, the leftmost record in the right
  1132.      * neighbour and all records in the leaf node itself.
  1133.      */
  1134.    
  1135.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1136.         if (a != avoid_area)
  1137.             return false;
  1138.     }
  1139.    
  1140.     /* First, check the two border cases. */
  1141.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1142.         a = (as_area_t *) node->value[node->keys - 1];
  1143.         mutex_lock(&a->lock);
  1144.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1145.             mutex_unlock(&a->lock);
  1146.             return false;
  1147.         }
  1148.         mutex_unlock(&a->lock);
  1149.     }
  1150.     node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf);
  1151.     if (node) {
  1152.         a = (as_area_t *) node->value[0];
  1153.         mutex_lock(&a->lock);
  1154.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1155.             mutex_unlock(&a->lock);
  1156.             return false;
  1157.         }
  1158.         mutex_unlock(&a->lock);
  1159.     }
  1160.    
  1161.     /* Second, check the leaf node. */
  1162.     for (i = 0; i < leaf->keys; i++) {
  1163.         a = (as_area_t *) leaf->value[i];
  1164.    
  1165.         if (a == avoid_area)
  1166.             continue;
  1167.    
  1168.         mutex_lock(&a->lock);
  1169.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1170.             mutex_unlock(&a->lock);
  1171.             return false;
  1172.         }
  1173.         mutex_unlock(&a->lock);
  1174.     }
  1175.  
  1176.     /*
  1177.      * So far, the area does not conflict with other areas.
  1178.      * Check if it doesn't conflict with kernel address space.
  1179.      */  
  1180.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1181.         return !overlaps(va, size,
  1182.             KERNEL_ADDRESS_SPACE_START,
  1183.             KERNEL_ADDRESS_SPACE_END - KERNEL_ADDRESS_SPACE_START);
  1184.     }
  1185.  
  1186.     return true;
  1187. }
  1188.  
  1189. /** Return size of the address space area with given base.  */
  1190. size_t as_get_size(uintptr_t base)
  1191. {
  1192.     ipl_t ipl;
  1193.     as_area_t *src_area;
  1194.     size_t size;
  1195.  
  1196.     ipl = interrupts_disable();
  1197.     src_area = find_area_and_lock(AS, base);
  1198.     if (src_area){
  1199.         size = src_area->pages * PAGE_SIZE;
  1200.         mutex_unlock(&src_area->lock);
  1201.     } else {
  1202.         size = 0;
  1203.     }
  1204.     interrupts_restore(ipl);
  1205.     return size;
  1206. }
  1207.  
  1208. /** Mark portion of address space area as used.
  1209.  *
  1210.  * The address space area must be already locked.
  1211.  *
  1212.  * @param a Address space area.
  1213.  * @param page First page to be marked.
  1214.  * @param count Number of page to be marked.
  1215.  *
  1216.  * @return 0 on failure and 1 on success.
  1217.  */
  1218. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1219. {
  1220.     btree_node_t *leaf, *node;
  1221.     count_t pages;
  1222.     int i;
  1223.  
  1224.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1225.     ASSERT(count);
  1226.  
  1227.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1228.     if (pages) {
  1229.         /*
  1230.          * We hit the beginning of some used space.
  1231.          */
  1232.         return 0;
  1233.     }
  1234.  
  1235.     if (!leaf->keys) {
  1236.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1237.         return 1;
  1238.     }
  1239.  
  1240.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1241.     if (node) {
  1242.         uintptr_t left_pg = node->key[node->keys - 1];
  1243.         uintptr_t right_pg = leaf->key[0];
  1244.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1245.         count_t right_cnt = (count_t) leaf->value[0];
  1246.        
  1247.         /*
  1248.          * Examine the possibility that the interval fits
  1249.          * somewhere between the rightmost interval of
  1250.          * the left neigbour and the first interval of the leaf.
  1251.          */
  1252.          
  1253.         if (page >= right_pg) {
  1254.             /* Do nothing. */
  1255.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1256.             left_cnt * PAGE_SIZE)) {
  1257.             /* The interval intersects with the left interval. */
  1258.             return 0;
  1259.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1260.             right_cnt * PAGE_SIZE)) {
  1261.             /* The interval intersects with the right interval. */
  1262.             return 0;          
  1263.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1264.             (page + count * PAGE_SIZE == right_pg)) {
  1265.             /*
  1266.              * The interval can be added by merging the two already
  1267.              * present intervals.
  1268.              */
  1269.             node->value[node->keys - 1] += count + right_cnt;
  1270.             btree_remove(&a->used_space, right_pg, leaf);
  1271.             return 1;
  1272.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1273.             /*
  1274.              * The interval can be added by simply growing the left
  1275.              * interval.
  1276.              */
  1277.             node->value[node->keys - 1] += count;
  1278.             return 1;
  1279.         } else if (page + count * PAGE_SIZE == right_pg) {
  1280.             /*
  1281.              * The interval can be addded by simply moving base of
  1282.              * the right interval down and increasing its size
  1283.              * accordingly.
  1284.              */
  1285.             leaf->value[0] += count;
  1286.             leaf->key[0] = page;
  1287.             return 1;
  1288.         } else {
  1289.             /*
  1290.              * The interval is between both neigbouring intervals,
  1291.              * but cannot be merged with any of them.
  1292.              */
  1293.             btree_insert(&a->used_space, page, (void *) count,
  1294.                 leaf);
  1295.             return 1;
  1296.         }
  1297.     } else if (page < leaf->key[0]) {
  1298.         uintptr_t right_pg = leaf->key[0];
  1299.         count_t right_cnt = (count_t) leaf->value[0];
  1300.    
  1301.         /*
  1302.          * Investigate the border case in which the left neighbour does
  1303.          * not exist but the interval fits from the left.
  1304.          */
  1305.          
  1306.         if (overlaps(page, count * PAGE_SIZE, right_pg,
  1307.             right_cnt * PAGE_SIZE)) {
  1308.             /* The interval intersects with the right interval. */
  1309.             return 0;
  1310.         } else if (page + count * PAGE_SIZE == right_pg) {
  1311.             /*
  1312.              * The interval can be added by moving the base of the
  1313.              * right interval down and increasing its size
  1314.              * accordingly.
  1315.              */
  1316.             leaf->key[0] = page;
  1317.             leaf->value[0] += count;
  1318.             return 1;
  1319.         } else {
  1320.             /*
  1321.              * The interval doesn't adjoin with the right interval.
  1322.              * It must be added individually.
  1323.              */
  1324.             btree_insert(&a->used_space, page, (void *) count,
  1325.                 leaf);
  1326.             return 1;
  1327.         }
  1328.     }
  1329.  
  1330.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1331.     if (node) {
  1332.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1333.         uintptr_t right_pg = node->key[0];
  1334.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1335.         count_t right_cnt = (count_t) node->value[0];
  1336.        
  1337.         /*
  1338.          * Examine the possibility that the interval fits
  1339.          * somewhere between the leftmost interval of
  1340.          * the right neigbour and the last interval of the leaf.
  1341.          */
  1342.  
  1343.         if (page < left_pg) {
  1344.             /* Do nothing. */
  1345.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1346.             left_cnt * PAGE_SIZE)) {
  1347.             /* The interval intersects with the left interval. */
  1348.             return 0;
  1349.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1350.             right_cnt * PAGE_SIZE)) {
  1351.             /* The interval intersects with the right interval. */
  1352.             return 0;          
  1353.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1354.             (page + count * PAGE_SIZE == right_pg)) {
  1355.             /*
  1356.              * The interval can be added by merging the two already
  1357.              * present intervals.
  1358.              * */
  1359.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1360.             btree_remove(&a->used_space, right_pg, node);
  1361.             return 1;
  1362.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1363.             /*
  1364.              * The interval can be added by simply growing the left
  1365.              * interval.
  1366.              * */
  1367.             leaf->value[leaf->keys - 1] +=  count;
  1368.             return 1;
  1369.         } else if (page + count * PAGE_SIZE == right_pg) {
  1370.             /*
  1371.              * The interval can be addded by simply moving base of
  1372.              * the right interval down and increasing its size
  1373.              * accordingly.
  1374.              */
  1375.             node->value[0] += count;
  1376.             node->key[0] = page;
  1377.             return 1;
  1378.         } else {
  1379.             /*
  1380.              * The interval is between both neigbouring intervals,
  1381.              * but cannot be merged with any of them.
  1382.              */
  1383.             btree_insert(&a->used_space, page, (void *) count,
  1384.                 leaf);
  1385.             return 1;
  1386.         }
  1387.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1388.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1389.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1390.    
  1391.         /*
  1392.          * Investigate the border case in which the right neighbour
  1393.          * does not exist but the interval fits from the right.
  1394.          */
  1395.          
  1396.         if (overlaps(page, count * PAGE_SIZE, left_pg,
  1397.             left_cnt * PAGE_SIZE)) {
  1398.             /* The interval intersects with the left interval. */
  1399.             return 0;
  1400.         } else if (left_pg + left_cnt * PAGE_SIZE == page) {
  1401.             /*
  1402.              * The interval can be added by growing the left
  1403.              * interval.
  1404.              */
  1405.             leaf->value[leaf->keys - 1] += count;
  1406.             return 1;
  1407.         } else {
  1408.             /*
  1409.              * The interval doesn't adjoin with the left interval.
  1410.              * It must be added individually.
  1411.              */
  1412.             btree_insert(&a->used_space, page, (void *) count,
  1413.                 leaf);
  1414.             return 1;
  1415.         }
  1416.     }
  1417.    
  1418.     /*
  1419.      * Note that if the algorithm made it thus far, the interval can fit
  1420.      * only between two other intervals of the leaf. The two border cases
  1421.      * were already resolved.
  1422.      */
  1423.     for (i = 1; i < leaf->keys; i++) {
  1424.         if (page < leaf->key[i]) {
  1425.             uintptr_t left_pg = leaf->key[i - 1];
  1426.             uintptr_t right_pg = leaf->key[i];
  1427.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1428.             count_t right_cnt = (count_t) leaf->value[i];
  1429.  
  1430.             /*
  1431.              * The interval fits between left_pg and right_pg.
  1432.              */
  1433.  
  1434.             if (overlaps(page, count * PAGE_SIZE, left_pg,
  1435.                 left_cnt * PAGE_SIZE)) {
  1436.                 /*
  1437.                  * The interval intersects with the left
  1438.                  * interval.
  1439.                  */
  1440.                 return 0;
  1441.             } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1442.                 right_cnt * PAGE_SIZE)) {
  1443.                 /*
  1444.                  * The interval intersects with the right
  1445.                  * interval.
  1446.                  */
  1447.                 return 0;          
  1448.             } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1449.                 (page + count * PAGE_SIZE == right_pg)) {
  1450.                 /*
  1451.                  * The interval can be added by merging the two
  1452.                  * already present intervals.
  1453.                  */
  1454.                 leaf->value[i - 1] += count + right_cnt;
  1455.                 btree_remove(&a->used_space, right_pg, leaf);
  1456.                 return 1;
  1457.             } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1458.                 /*
  1459.                  * The interval can be added by simply growing
  1460.                  * the left interval.
  1461.                  */
  1462.                 leaf->value[i - 1] += count;
  1463.                 return 1;
  1464.             } else if (page + count * PAGE_SIZE == right_pg) {
  1465.                 /*
  1466.                      * The interval can be addded by simply moving
  1467.                  * base of the right interval down and
  1468.                  * increasing its size accordingly.
  1469.                  */
  1470.                 leaf->value[i] += count;
  1471.                 leaf->key[i] = page;
  1472.                 return 1;
  1473.             } else {
  1474.                 /*
  1475.                  * The interval is between both neigbouring
  1476.                  * intervals, but cannot be merged with any of
  1477.                  * them.
  1478.                  */
  1479.                 btree_insert(&a->used_space, page,
  1480.                     (void *) count, leaf);
  1481.                 return 1;
  1482.             }
  1483.         }
  1484.     }
  1485.  
  1486.     panic("Inconsistency detected while adding %d pages of used space at "
  1487.         "%p.\n", count, page);
  1488. }
  1489.  
  1490. /** Mark portion of address space area as unused.
  1491.  *
  1492.  * The address space area must be already locked.
  1493.  *
  1494.  * @param a Address space area.
  1495.  * @param page First page to be marked.
  1496.  * @param count Number of page to be marked.
  1497.  *
  1498.  * @return 0 on failure and 1 on success.
  1499.  */
  1500. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1501. {
  1502.     btree_node_t *leaf, *node;
  1503.     count_t pages;
  1504.     int i;
  1505.  
  1506.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1507.     ASSERT(count);
  1508.  
  1509.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1510.     if (pages) {
  1511.         /*
  1512.          * We are lucky, page is the beginning of some interval.
  1513.          */
  1514.         if (count > pages) {
  1515.             return 0;
  1516.         } else if (count == pages) {
  1517.             btree_remove(&a->used_space, page, leaf);
  1518.             return 1;
  1519.         } else {
  1520.             /*
  1521.              * Find the respective interval.
  1522.              * Decrease its size and relocate its start address.
  1523.              */
  1524.             for (i = 0; i < leaf->keys; i++) {
  1525.                 if (leaf->key[i] == page) {
  1526.                     leaf->key[i] += count * PAGE_SIZE;
  1527.                     leaf->value[i] -= count;
  1528.                     return 1;
  1529.                 }
  1530.             }
  1531.             goto error;
  1532.         }
  1533.     }
  1534.  
  1535.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1536.     if (node && page < leaf->key[0]) {
  1537.         uintptr_t left_pg = node->key[node->keys - 1];
  1538.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1539.  
  1540.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1541.             count * PAGE_SIZE)) {
  1542.             if (page + count * PAGE_SIZE ==
  1543.                 left_pg + left_cnt * PAGE_SIZE) {
  1544.                 /*
  1545.                  * The interval is contained in the rightmost
  1546.                  * interval of the left neighbour and can be
  1547.                  * removed by updating the size of the bigger
  1548.                  * interval.
  1549.                  */
  1550.                 node->value[node->keys - 1] -= count;
  1551.                 return 1;
  1552.             } else if (page + count * PAGE_SIZE <
  1553.                 left_pg + left_cnt*PAGE_SIZE) {
  1554.                 count_t new_cnt;
  1555.                
  1556.                 /*
  1557.                  * The interval is contained in the rightmost
  1558.                  * interval of the left neighbour but its
  1559.                  * removal requires both updating the size of
  1560.                  * the original interval and also inserting a
  1561.                  * new interval.
  1562.                  */
  1563.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1564.                     (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1565.                 node->value[node->keys - 1] -= count + new_cnt;
  1566.                 btree_insert(&a->used_space, page +
  1567.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1568.                 return 1;
  1569.             }
  1570.         }
  1571.         return 0;
  1572.     } else if (page < leaf->key[0]) {
  1573.         return 0;
  1574.     }
  1575.    
  1576.     if (page > leaf->key[leaf->keys - 1]) {
  1577.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1578.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1579.  
  1580.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1581.             count * PAGE_SIZE)) {
  1582.             if (page + count * PAGE_SIZE ==
  1583.                 left_pg + left_cnt * PAGE_SIZE) {
  1584.                 /*
  1585.                  * The interval is contained in the rightmost
  1586.                  * interval of the leaf and can be removed by
  1587.                  * updating the size of the bigger interval.
  1588.                  */
  1589.                 leaf->value[leaf->keys - 1] -= count;
  1590.                 return 1;
  1591.             } else if (page + count * PAGE_SIZE < left_pg +
  1592.                 left_cnt * PAGE_SIZE) {
  1593.                 count_t new_cnt;
  1594.                
  1595.                 /*
  1596.                  * The interval is contained in the rightmost
  1597.                  * interval of the leaf but its removal
  1598.                  * requires both updating the size of the
  1599.                  * original interval and also inserting a new
  1600.                  * interval.
  1601.                  */
  1602.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1603.                     (page + count * PAGE_SIZE)) >> PAGE_WIDTH;
  1604.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1605.                 btree_insert(&a->used_space, page +
  1606.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1607.                 return 1;
  1608.             }
  1609.         }
  1610.         return 0;
  1611.     }  
  1612.    
  1613.     /*
  1614.      * The border cases have been already resolved.
  1615.      * Now the interval can be only between intervals of the leaf.
  1616.      */
  1617.     for (i = 1; i < leaf->keys - 1; i++) {
  1618.         if (page < leaf->key[i]) {
  1619.             uintptr_t left_pg = leaf->key[i - 1];
  1620.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1621.  
  1622.             /*
  1623.              * Now the interval is between intervals corresponding
  1624.              * to (i - 1) and i.
  1625.              */
  1626.             if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1627.                 count * PAGE_SIZE)) {
  1628.                 if (page + count * PAGE_SIZE ==
  1629.                     left_pg + left_cnt*PAGE_SIZE) {
  1630.                     /*
  1631.                      * The interval is contained in the
  1632.                      * interval (i - 1) of the leaf and can
  1633.                      * be removed by updating the size of
  1634.                      * the bigger interval.
  1635.                      */
  1636.                     leaf->value[i - 1] -= count;
  1637.                     return 1;
  1638.                 } else if (page + count * PAGE_SIZE <
  1639.                     left_pg + left_cnt * PAGE_SIZE) {
  1640.                     count_t new_cnt;
  1641.                
  1642.                     /*
  1643.                      * The interval is contained in the
  1644.                      * interval (i - 1) of the leaf but its
  1645.                      * removal requires both updating the
  1646.                      * size of the original interval and
  1647.                      * also inserting a new interval.
  1648.                      */
  1649.                     new_cnt = ((left_pg +
  1650.                         left_cnt * PAGE_SIZE) -
  1651.                         (page + count * PAGE_SIZE)) >>
  1652.                         PAGE_WIDTH;
  1653.                     leaf->value[i - 1] -= count + new_cnt;
  1654.                     btree_insert(&a->used_space, page +
  1655.                         count * PAGE_SIZE, (void *) new_cnt,
  1656.                         leaf);
  1657.                     return 1;
  1658.                 }
  1659.             }
  1660.             return 0;
  1661.         }
  1662.     }
  1663.  
  1664. error:
  1665.     panic("Inconsistency detected while removing %d pages of used space "
  1666.         "from %p.\n", count, page);
  1667. }
  1668.  
  1669. /** Remove reference to address space area share info.
  1670.  *
  1671.  * If the reference count drops to 0, the sh_info is deallocated.
  1672.  *
  1673.  * @param sh_info Pointer to address space area share info.
  1674.  */
  1675. void sh_info_remove_reference(share_info_t *sh_info)
  1676. {
  1677.     bool dealloc = false;
  1678.  
  1679.     mutex_lock(&sh_info->lock);
  1680.     ASSERT(sh_info->refcount);
  1681.     if (--sh_info->refcount == 0) {
  1682.         dealloc = true;
  1683.         link_t *cur;
  1684.        
  1685.         /*
  1686.          * Now walk carefully the pagemap B+tree and free/remove
  1687.          * reference from all frames found there.
  1688.          */
  1689.         for (cur = sh_info->pagemap.leaf_head.next;
  1690.             cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1691.             btree_node_t *node;
  1692.             int i;
  1693.            
  1694.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1695.             for (i = 0; i < node->keys; i++)
  1696.                 frame_free((uintptr_t) node->value[i]);
  1697.         }
  1698.        
  1699.     }
  1700.     mutex_unlock(&sh_info->lock);
  1701.    
  1702.     if (dealloc) {
  1703.         btree_destroy(&sh_info->pagemap);
  1704.         free(sh_info);
  1705.     }
  1706. }
  1707.  
  1708. /*
  1709.  * Address space related syscalls.
  1710.  */
  1711.  
  1712. /** Wrapper for as_area_create(). */
  1713. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1714. {
  1715.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address,
  1716.         AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1717.         return (unative_t) address;
  1718.     else
  1719.         return (unative_t) -1;
  1720. }
  1721.  
  1722. /** Wrapper for as_area_resize(). */
  1723. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1724. {
  1725.     return (unative_t) as_area_resize(AS, address, size, 0);
  1726. }
  1727.  
  1728. /** Wrapper for as_area_destroy(). */
  1729. unative_t sys_as_area_destroy(uintptr_t address)
  1730. {
  1731.     return (unative_t) as_area_destroy(AS, address);
  1732. }
  1733.  
  1734. /** Print out information about address space.
  1735.  *
  1736.  * @param as Address space.
  1737.  */
  1738. void as_print(as_t *as)
  1739. {
  1740.     ipl_t ipl;
  1741.    
  1742.     ipl = interrupts_disable();
  1743.     mutex_lock(&as->lock);
  1744.    
  1745.     /* print out info about address space areas */
  1746.     link_t *cur;
  1747.     for (cur = as->as_area_btree.leaf_head.next;
  1748.         cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  1749.         btree_node_t *node;
  1750.        
  1751.         node = list_get_instance(cur, btree_node_t, leaf_link);
  1752.        
  1753.         int i;
  1754.         for (i = 0; i < node->keys; i++) {
  1755.             as_area_t *area = node->value[i];
  1756.        
  1757.             mutex_lock(&area->lock);
  1758.             printf("as_area: %p, base=%p, pages=%d (%p - %p)\n",
  1759.                 area, area->base, area->pages, area->base,
  1760.                 area->base + area->pages*PAGE_SIZE);
  1761.             mutex_unlock(&area->lock);
  1762.         }
  1763.     }
  1764.    
  1765.     mutex_unlock(&as->lock);
  1766.     interrupts_restore(ipl);
  1767. }
  1768.  
  1769. /** @}
  1770.  */
  1771.