Subversion Repositories HelenOS

Rev

Rev 3397 | Rev 3742 | 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 <preemption.h>
  61. #include <synch/spinlock.h>
  62. #include <synch/mutex.h>
  63. #include <adt/list.h>
  64. #include <adt/btree.h>
  65. #include <proc/task.h>
  66. #include <proc/thread.h>
  67. #include <arch/asm.h>
  68. #include <panic.h>
  69. #include <debug.h>
  70. #include <print.h>
  71. #include <memstr.h>
  72. #include <macros.h>
  73. #include <arch.h>
  74. #include <errno.h>
  75. #include <config.h>
  76. #include <align.h>
  77. #include <arch/types.h>
  78. #include <syscall/copy.h>
  79. #include <arch/interrupt.h>
  80.  
  81. #ifdef CONFIG_VIRT_IDX_DCACHE
  82. #include <arch/mm/cache.h>
  83. #endif /* CONFIG_VIRT_IDX_DCACHE */
  84.  
  85. /**
  86.  * Each architecture decides what functions will be used to carry out
  87.  * address space operations such as creating or locking page tables.
  88.  */
  89. as_operations_t *as_operations = NULL;
  90.  
  91. /**
  92.  * Slab for as_t objects.
  93.  */
  94. static slab_cache_t *as_slab;
  95.  
  96. /**
  97.  * This lock serializes access to the ASID subsystem.
  98.  * It protects:
  99.  * - inactive_as_with_asid_head list
  100.  * - as->asid for each as of the as_t type
  101.  * - asids_allocated counter
  102.  */
  103. SPINLOCK_INITIALIZE(asidlock);
  104.  
  105. /**
  106.  * This list contains address spaces that are not active on any
  107.  * processor and that have valid ASID.
  108.  */
  109. LIST_INITIALIZE(inactive_as_with_asid_head);
  110.  
  111. /** Kernel address space. */
  112. as_t *AS_KERNEL = NULL;
  113.  
  114. static int area_flags_to_page_flags(int);
  115. static as_area_t *find_area_and_lock(as_t *, uintptr_t);
  116. static bool check_area_conflicts(as_t *, uintptr_t, size_t, as_area_t *);
  117. static void sh_info_remove_reference(share_info_t *);
  118.  
  119. static int as_constructor(void *obj, int flags)
  120. {
  121.     as_t *as = (as_t *) obj;
  122.     int rc;
  123.  
  124.     link_initialize(&as->inactive_as_with_asid_link);
  125.     mutex_initialize(&as->lock, MUTEX_PASSIVE);
  126.    
  127.     rc = as_constructor_arch(as, flags);
  128.    
  129.     return rc;
  130. }
  131.  
  132. static int as_destructor(void *obj)
  133. {
  134.     as_t *as = (as_t *) obj;
  135.  
  136.     return as_destructor_arch(as);
  137. }
  138.  
  139. /** Initialize address space subsystem. */
  140. void as_init(void)
  141. {
  142.     as_arch_init();
  143.  
  144.     as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,
  145.         as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);
  146.    
  147.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  148.     if (!AS_KERNEL)
  149.         panic("can't create kernel address space\n");
  150.    
  151. }
  152.  
  153. /** Create address space.
  154.  *
  155.  * @param flags     Flags that influence the way in wich the address space
  156.  *          is created.
  157.  */
  158. as_t *as_create(int flags)
  159. {
  160.     as_t *as;
  161.  
  162.     as = (as_t *) slab_alloc(as_slab, 0);
  163.     (void) as_create_arch(as, 0);
  164.    
  165.     btree_create(&as->as_area_btree);
  166.    
  167.     if (flags & FLAG_AS_KERNEL)
  168.         as->asid = ASID_KERNEL;
  169.     else
  170.         as->asid = ASID_INVALID;
  171.    
  172.     atomic_set(&as->refcount, 0);
  173.     as->cpu_refcount = 0;
  174. #ifdef AS_PAGE_TABLE
  175.     as->genarch.page_table = page_table_create(flags);
  176. #else
  177.     page_table_create(flags);
  178. #endif
  179.  
  180.     return as;
  181. }
  182.  
  183. /** Destroy adress space.
  184.  *
  185.  * When there are no tasks referencing this address space (i.e. its refcount is
  186.  * zero), the address space can be destroyed.
  187.  *
  188.  * We know that we don't hold any spinlock.
  189.  *
  190.  * @param as        Address space to be destroyed.
  191.  */
  192. void as_destroy(as_t *as)
  193. {
  194.     ipl_t ipl;
  195.     bool cond;
  196.     DEADLOCK_PROBE_INIT(p_asidlock);
  197.  
  198.     ASSERT(atomic_get(&as->refcount) == 0);
  199.    
  200.     /*
  201.      * Since there is no reference to this area,
  202.      * it is safe not to lock its mutex.
  203.      */
  204.  
  205.     /*
  206.      * We need to avoid deadlock between TLB shootdown and asidlock.
  207.      * We therefore try to take asid conditionally and if we don't succeed,
  208.      * we enable interrupts and try again. This is done while preemption is
  209.      * disabled to prevent nested context switches. We also depend on the
  210.      * fact that so far no spinlocks are held.
  211.      */
  212.     preemption_disable();
  213.     ipl = interrupts_read();
  214. retry:
  215.     interrupts_disable();
  216.     if (!spinlock_trylock(&asidlock)) {
  217.         interrupts_enable();
  218.         DEADLOCK_PROBE(p_asidlock, DEADLOCK_THRESHOLD);
  219.         goto retry;
  220.     }
  221.     preemption_enable();    /* Interrupts disabled, enable preemption */
  222.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  223.         if (as != AS && as->cpu_refcount == 0)
  224.             list_remove(&as->inactive_as_with_asid_link);
  225.         asid_put(as->asid);
  226.     }
  227.     spinlock_unlock(&asidlock);
  228.  
  229.     /*
  230.      * Destroy address space areas of the address space.
  231.      * The B+tree must be walked carefully because it is
  232.      * also being destroyed.
  233.      */
  234.     for (cond = true; cond; ) {
  235.         btree_node_t *node;
  236.  
  237.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  238.         node = list_get_instance(as->as_area_btree.leaf_head.next,
  239.             btree_node_t, leaf_link);
  240.  
  241.         if ((cond = node->keys)) {
  242.             as_area_destroy(as, node->key[0]);
  243.         }
  244.     }
  245.  
  246.     btree_destroy(&as->as_area_btree);
  247. #ifdef AS_PAGE_TABLE
  248.     page_table_destroy(as->genarch.page_table);
  249. #else
  250.     page_table_destroy(NULL);
  251. #endif
  252.  
  253.     interrupts_restore(ipl);
  254.  
  255.     slab_free(as_slab, as);
  256. }
  257.  
  258. /** Create address space area of common attributes.
  259.  *
  260.  * The created address space area is added to the target address space.
  261.  *
  262.  * @param as        Target address space.
  263.  * @param flags     Flags of the area memory.
  264.  * @param size      Size of area.
  265.  * @param base      Base address of area.
  266.  * @param attrs     Attributes of the area.
  267.  * @param backend   Address space area backend. NULL if no backend is used.
  268.  * @param backend_data  NULL or a pointer to an array holding two void *.
  269.  *
  270.  * @return      Address space area on success or NULL on failure.
  271.  */
  272. as_area_t *
  273. as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
  274.     mem_backend_t *backend, mem_backend_data_t *backend_data)
  275. {
  276.     ipl_t ipl;
  277.     as_area_t *a;
  278.    
  279.     if (base % PAGE_SIZE)
  280.         return NULL;
  281.  
  282.     if (!size)
  283.         return NULL;
  284.  
  285.     /* Writeable executable areas are not supported. */
  286.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  287.         return NULL;
  288.    
  289.     ipl = interrupts_disable();
  290.     mutex_lock(&as->lock);
  291.    
  292.     if (!check_area_conflicts(as, base, size, NULL)) {
  293.         mutex_unlock(&as->lock);
  294.         interrupts_restore(ipl);
  295.         return NULL;
  296.     }
  297.    
  298.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  299.  
  300.     mutex_initialize(&a->lock, MUTEX_PASSIVE);
  301.    
  302.     a->as = as;
  303.     a->flags = flags;
  304.     a->attributes = attrs;
  305.     a->pages = SIZE2FRAMES(size);
  306.     a->base = base;
  307.     a->sh_info = NULL;
  308.     a->backend = backend;
  309.     if (backend_data)
  310.         a->backend_data = *backend_data;
  311.     else
  312.         memsetb(&a->backend_data, sizeof(a->backend_data), 0);
  313.  
  314.     btree_create(&a->used_space);
  315.    
  316.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  317.  
  318.     mutex_unlock(&as->lock);
  319.     interrupts_restore(ipl);
  320.  
  321.     return a;
  322. }
  323.  
  324. /** Find address space area and change it.
  325.  *
  326.  * @param as        Address space.
  327.  * @param address   Virtual address belonging to the area to be changed.
  328.  *          Must be page-aligned.
  329.  * @param size      New size of the virtual memory block starting at
  330.  *          address.
  331.  * @param flags     Flags influencing the remap operation. Currently unused.
  332.  *
  333.  * @return      Zero on success or a value from @ref errno.h otherwise.
  334.  */
  335. int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
  336. {
  337.     as_area_t *area;
  338.     ipl_t ipl;
  339.     size_t pages;
  340.    
  341.     ipl = interrupts_disable();
  342.     mutex_lock(&as->lock);
  343.    
  344.     /*
  345.      * Locate the area.
  346.      */
  347.     area = find_area_and_lock(as, address);
  348.     if (!area) {
  349.         mutex_unlock(&as->lock);
  350.         interrupts_restore(ipl);
  351.         return ENOENT;
  352.     }
  353.  
  354.     if (area->backend == &phys_backend) {
  355.         /*
  356.          * Remapping of address space areas associated
  357.          * with memory mapped devices is not supported.
  358.          */
  359.         mutex_unlock(&area->lock);
  360.         mutex_unlock(&as->lock);
  361.         interrupts_restore(ipl);
  362.         return ENOTSUP;
  363.     }
  364.     if (area->sh_info) {
  365.         /*
  366.          * Remapping of shared address space areas
  367.          * is not supported.
  368.          */
  369.         mutex_unlock(&area->lock);
  370.         mutex_unlock(&as->lock);
  371.         interrupts_restore(ipl);
  372.         return ENOTSUP;
  373.     }
  374.  
  375.     pages = SIZE2FRAMES((address - area->base) + size);
  376.     if (!pages) {
  377.         /*
  378.          * Zero size address space areas are not allowed.
  379.          */
  380.         mutex_unlock(&area->lock);
  381.         mutex_unlock(&as->lock);
  382.         interrupts_restore(ipl);
  383.         return EPERM;
  384.     }
  385.    
  386.     if (pages < area->pages) {
  387.         bool cond;
  388.         uintptr_t start_free = area->base + pages * PAGE_SIZE;
  389.  
  390.         /*
  391.          * Shrinking the area.
  392.          * No need to check for overlaps.
  393.          */
  394.  
  395.         /*
  396.          * Start TLB shootdown sequence.
  397.          */
  398.         tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base +
  399.             pages * PAGE_SIZE, area->pages - pages);
  400.  
  401.         /*
  402.          * Remove frames belonging to used space starting from
  403.          * the highest addresses downwards until an overlap with
  404.          * the resized address space area is found. Note that this
  405.          * is also the right way to remove part of the used_space
  406.          * B+tree leaf list.
  407.          */    
  408.         for (cond = true; cond;) {
  409.             btree_node_t *node;
  410.        
  411.             ASSERT(!list_empty(&area->used_space.leaf_head));
  412.             node =
  413.                 list_get_instance(area->used_space.leaf_head.prev,
  414.                 btree_node_t, leaf_link);
  415.             if ((cond = (bool) node->keys)) {
  416.                 uintptr_t b = node->key[node->keys - 1];
  417.                 count_t c =
  418.                     (count_t) node->value[node->keys - 1];
  419.                 unsigned int i = 0;
  420.            
  421.                 if (overlaps(b, c * PAGE_SIZE, area->base,
  422.                     pages * PAGE_SIZE)) {
  423.                    
  424.                     if (b + c * PAGE_SIZE <= start_free) {
  425.                         /*
  426.                          * The whole interval fits
  427.                          * completely in the resized
  428.                          * address space area.
  429.                          */
  430.                         break;
  431.                     }
  432.        
  433.                     /*
  434.                      * Part of the interval corresponding
  435.                      * to b and c overlaps with the resized
  436.                      * address space area.
  437.                      */
  438.        
  439.                     cond = false;   /* we are almost done */
  440.                     i = (start_free - b) >> PAGE_WIDTH;
  441.                     if (!used_space_remove(area, start_free,
  442.                         c - i))
  443.                         panic("Could not remove used "
  444.                             "space.\n");
  445.                 } else {
  446.                     /*
  447.                      * The interval of used space can be
  448.                      * completely removed.
  449.                      */
  450.                     if (!used_space_remove(area, b, c))
  451.                         panic("Could not remove used "
  452.                             "space.\n");
  453.                 }
  454.            
  455.                 for (; i < c; i++) {
  456.                     pte_t *pte;
  457.            
  458.                     page_table_lock(as, false);
  459.                     pte = page_mapping_find(as, b +
  460.                         i * PAGE_SIZE);
  461.                     ASSERT(pte && PTE_VALID(pte) &&
  462.                         PTE_PRESENT(pte));
  463.                     if (area->backend &&
  464.                         area->backend->frame_free) {
  465.                         area->backend->frame_free(area,
  466.                             b + i * PAGE_SIZE,
  467.                             PTE_GET_FRAME(pte));
  468.                     }
  469.                     page_mapping_remove(as, b +
  470.                         i * PAGE_SIZE);
  471.                     page_table_unlock(as, false);
  472.                 }
  473.             }
  474.         }
  475.  
  476.         /*
  477.          * Finish TLB shootdown sequence.
  478.          */
  479.  
  480.         tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,
  481.             area->pages - pages);
  482.         /*
  483.          * Invalidate software translation caches (e.g. TSB on sparc64).
  484.          */
  485.         as_invalidate_translation_cache(as, area->base +
  486.             pages * PAGE_SIZE, area->pages - pages);
  487.         tlb_shootdown_finalize();
  488.        
  489.     } else {
  490.         /*
  491.          * Growing the area.
  492.          * Check for overlaps with other address space areas.
  493.          */
  494.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE,
  495.             area)) {
  496.             mutex_unlock(&area->lock);
  497.             mutex_unlock(&as->lock);       
  498.             interrupts_restore(ipl);
  499.             return EADDRNOTAVAIL;
  500.         }
  501.     }
  502.  
  503.     area->pages = pages;
  504.    
  505.     mutex_unlock(&area->lock);
  506.     mutex_unlock(&as->lock);
  507.     interrupts_restore(ipl);
  508.  
  509.     return 0;
  510. }
  511.  
  512. /** Destroy address space area.
  513.  *
  514.  * @param as        Address space.
  515.  * @param address   Address within the area to be deleted.
  516.  *
  517.  * @return      Zero on success or a value from @ref errno.h on failure.
  518.  */
  519. int as_area_destroy(as_t *as, uintptr_t address)
  520. {
  521.     as_area_t *area;
  522.     uintptr_t base;
  523.     link_t *cur;
  524.     ipl_t ipl;
  525.  
  526.     ipl = interrupts_disable();
  527.     mutex_lock(&as->lock);
  528.  
  529.     area = find_area_and_lock(as, address);
  530.     if (!area) {
  531.         mutex_unlock(&as->lock);
  532.         interrupts_restore(ipl);
  533.         return ENOENT;
  534.     }
  535.  
  536.     base = area->base;
  537.  
  538.     /*
  539.      * Start TLB shootdown sequence.
  540.      */
  541.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  542.  
  543.     /*
  544.      * Visit only the pages mapped by used_space B+tree.
  545.      */
  546.     for (cur = area->used_space.leaf_head.next;
  547.         cur != &area->used_space.leaf_head; cur = cur->next) {
  548.         btree_node_t *node;
  549.         unsigned int i;
  550.        
  551.         node = list_get_instance(cur, btree_node_t, leaf_link);
  552.         for (i = 0; i < node->keys; i++) {
  553.             uintptr_t b = node->key[i];
  554.             count_t j;
  555.             pte_t *pte;
  556.            
  557.             for (j = 0; j < (count_t) node->value[i]; j++) {
  558.                 page_table_lock(as, false);
  559.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  560.                 ASSERT(pte && PTE_VALID(pte) &&
  561.                     PTE_PRESENT(pte));
  562.                 if (area->backend &&
  563.                     area->backend->frame_free) {
  564.                     area->backend->frame_free(area, b +
  565.                         j * PAGE_SIZE, PTE_GET_FRAME(pte));
  566.                 }
  567.                 page_mapping_remove(as, b + j * PAGE_SIZE);            
  568.                 page_table_unlock(as, false);
  569.             }
  570.         }
  571.     }
  572.  
  573.     /*
  574.      * Finish TLB shootdown sequence.
  575.      */
  576.  
  577.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  578.     /*
  579.      * Invalidate potential software translation caches (e.g. TSB on
  580.      * sparc64).
  581.      */
  582.     as_invalidate_translation_cache(as, area->base, area->pages);
  583.     tlb_shootdown_finalize();
  584.    
  585.     btree_destroy(&area->used_space);
  586.  
  587.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  588.    
  589.     if (area->sh_info)
  590.         sh_info_remove_reference(area->sh_info);
  591.        
  592.     mutex_unlock(&area->lock);
  593.  
  594.     /*
  595.      * Remove the empty area from address space.
  596.      */
  597.     btree_remove(&as->as_area_btree, base, NULL);
  598.    
  599.     free(area);
  600.    
  601.     mutex_unlock(&as->lock);
  602.     interrupts_restore(ipl);
  603.     return 0;
  604. }
  605.  
  606. /** Share address space area with another or the same address space.
  607.  *
  608.  * Address space area mapping is shared with a new address space area.
  609.  * If the source address space area has not been shared so far,
  610.  * a new sh_info is created. The new address space area simply gets the
  611.  * sh_info of the source area. The process of duplicating the
  612.  * mapping is done through the backend share function.
  613.  *
  614.  * @param src_as    Pointer to source address space.
  615.  * @param src_base  Base address of the source address space area.
  616.  * @param acc_size  Expected size of the source area.
  617.  * @param dst_as    Pointer to destination address space.
  618.  * @param dst_base  Target base address.
  619.  * @param dst_flags_mask Destination address space area flags mask.
  620.  *
  621.  * @return      Zero on success or ENOENT if there is no such task or if
  622.  *          there is no such address space area, EPERM if there was
  623.  *          a problem in accepting the area or ENOMEM if there was a
  624.  *          problem in allocating destination address space area.
  625.  *          ENOTSUP is returned if the address space area backend
  626.  *          does not support sharing.
  627.  */
  628. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  629.     as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  630. {
  631.     ipl_t ipl;
  632.     int src_flags;
  633.     size_t src_size;
  634.     as_area_t *src_area, *dst_area;
  635.     share_info_t *sh_info;
  636.     mem_backend_t *src_backend;
  637.     mem_backend_data_t src_backend_data;
  638.    
  639.     ipl = interrupts_disable();
  640.     mutex_lock(&src_as->lock);
  641.     src_area = find_area_and_lock(src_as, src_base);
  642.     if (!src_area) {
  643.         /*
  644.          * Could not find the source address space area.
  645.          */
  646.         mutex_unlock(&src_as->lock);
  647.         interrupts_restore(ipl);
  648.         return ENOENT;
  649.     }
  650.  
  651.     if (!src_area->backend || !src_area->backend->share) {
  652.         /*
  653.          * There is no backend or the backend does not
  654.          * know how to share the area.
  655.          */
  656.         mutex_unlock(&src_area->lock);
  657.         mutex_unlock(&src_as->lock);
  658.         interrupts_restore(ipl);
  659.         return ENOTSUP;
  660.     }
  661.    
  662.     src_size = src_area->pages * PAGE_SIZE;
  663.     src_flags = src_area->flags;
  664.     src_backend = src_area->backend;
  665.     src_backend_data = src_area->backend_data;
  666.  
  667.     /* Share the cacheable flag from the original mapping */
  668.     if (src_flags & AS_AREA_CACHEABLE)
  669.         dst_flags_mask |= AS_AREA_CACHEABLE;
  670.  
  671.     if (src_size != acc_size ||
  672.         (src_flags & dst_flags_mask) != dst_flags_mask) {
  673.         mutex_unlock(&src_area->lock);
  674.         mutex_unlock(&src_as->lock);
  675.         interrupts_restore(ipl);
  676.         return EPERM;
  677.     }
  678.  
  679.     /*
  680.      * Now we are committed to sharing the area.
  681.      * First, prepare the area for sharing.
  682.      * Then it will be safe to unlock it.
  683.      */
  684.     sh_info = src_area->sh_info;
  685.     if (!sh_info) {
  686.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  687.         mutex_initialize(&sh_info->lock, MUTEX_PASSIVE);
  688.         sh_info->refcount = 2;
  689.         btree_create(&sh_info->pagemap);
  690.         src_area->sh_info = sh_info;
  691.         /*
  692.          * Call the backend to setup sharing.
  693.          */
  694.         src_area->backend->share(src_area);
  695.     } else {
  696.         mutex_lock(&sh_info->lock);
  697.         sh_info->refcount++;
  698.         mutex_unlock(&sh_info->lock);
  699.     }
  700.  
  701.     mutex_unlock(&src_area->lock);
  702.     mutex_unlock(&src_as->lock);
  703.  
  704.     /*
  705.      * Create copy of the source address space area.
  706.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  707.      * attribute set which prevents race condition with
  708.      * preliminary as_page_fault() calls.
  709.      * The flags of the source area are masked against dst_flags_mask
  710.      * to support sharing in less privileged mode.
  711.      */
  712.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  713.         AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  714.     if (!dst_area) {
  715.         /*
  716.          * Destination address space area could not be created.
  717.          */
  718.         sh_info_remove_reference(sh_info);
  719.        
  720.         interrupts_restore(ipl);
  721.         return ENOMEM;
  722.     }
  723.  
  724.     /*
  725.      * Now the destination address space area has been
  726.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  727.      * attribute and set the sh_info.
  728.      */
  729.     mutex_lock(&dst_as->lock); 
  730.     mutex_lock(&dst_area->lock);
  731.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  732.     dst_area->sh_info = sh_info;
  733.     mutex_unlock(&dst_area->lock);
  734.     mutex_unlock(&dst_as->lock);   
  735.  
  736.     interrupts_restore(ipl);
  737.    
  738.     return 0;
  739. }
  740.  
  741. /** Check access mode for address space area.
  742.  *
  743.  * The address space area must be locked prior to this call.
  744.  *
  745.  * @param area      Address space area.
  746.  * @param access    Access mode.
  747.  *
  748.  * @return      False if access violates area's permissions, true
  749.  *          otherwise.
  750.  */
  751. bool as_area_check_access(as_area_t *area, pf_access_t access)
  752. {
  753.     int flagmap[] = {
  754.         [PF_ACCESS_READ] = AS_AREA_READ,
  755.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  756.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  757.     };
  758.  
  759.     if (!(area->flags & flagmap[access]))
  760.         return false;
  761.    
  762.     return true;
  763. }
  764.  
  765. /** Change adress space area flags.
  766.  *
  767.  * The idea is to have the same data, but with a different access mode.
  768.  * This is needed e.g. for writing code into memory and then executing it.
  769.  * In order for this to work properly, this may copy the data
  770.  * into private anonymous memory (unless it's already there).
  771.  *
  772.  * @param as        Address space.
  773.  * @param flags     Flags of the area memory.
  774.  * @param address   Address withing the area to be changed.
  775.  *
  776.  * @return      Zero on success or a value from @ref errno.h on failure.
  777.  */
  778. int as_area_change_flags(as_t *as, int flags, uintptr_t address)
  779. {
  780.     as_area_t *area;
  781.     uintptr_t base;
  782.     link_t *cur;
  783.     ipl_t ipl;
  784.     int page_flags;
  785.     uintptr_t *old_frame;
  786.     index_t frame_idx;
  787.     count_t used_pages;
  788.  
  789.     /* Flags for the new memory mapping */
  790.     page_flags = area_flags_to_page_flags(flags);
  791.  
  792.     ipl = interrupts_disable();
  793.     mutex_lock(&as->lock);
  794.  
  795.     area = find_area_and_lock(as, address);
  796.     if (!area) {
  797.         mutex_unlock(&as->lock);
  798.         interrupts_restore(ipl);
  799.         return ENOENT;
  800.     }
  801.  
  802.     if (area->sh_info || area->backend != &anon_backend) {
  803.         /* Copying shared areas not supported yet */
  804.         /* Copying non-anonymous memory not supported yet */
  805.         mutex_unlock(&area->lock);
  806.         mutex_unlock(&as->lock);
  807.         interrupts_restore(ipl);
  808.         return ENOTSUP;
  809.     }
  810.  
  811.     base = area->base;
  812.  
  813.     /*
  814.      * Compute total number of used pages in the used_space B+tree
  815.      */
  816.     used_pages = 0;
  817.  
  818.     for (cur = area->used_space.leaf_head.next;
  819.         cur != &area->used_space.leaf_head; cur = cur->next) {
  820.         btree_node_t *node;
  821.         unsigned int i;
  822.        
  823.         node = list_get_instance(cur, btree_node_t, leaf_link);
  824.         for (i = 0; i < node->keys; i++) {
  825.             used_pages += (count_t) node->value[i];
  826.         }
  827.     }
  828.  
  829.     /* An array for storing frame numbers */
  830.     old_frame = malloc(used_pages * sizeof(uintptr_t), 0);
  831.  
  832.     /*
  833.      * Start TLB shootdown sequence.
  834.      */
  835.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  836.  
  837.     /*
  838.      * Remove used pages from page tables and remember their frame
  839.      * numbers.
  840.      */
  841.     frame_idx = 0;
  842.  
  843.     for (cur = area->used_space.leaf_head.next;
  844.         cur != &area->used_space.leaf_head; cur = cur->next) {
  845.         btree_node_t *node;
  846.         unsigned int i;
  847.        
  848.         node = list_get_instance(cur, btree_node_t, leaf_link);
  849.         for (i = 0; i < node->keys; i++) {
  850.             uintptr_t b = node->key[i];
  851.             count_t j;
  852.             pte_t *pte;
  853.            
  854.             for (j = 0; j < (count_t) node->value[i]; j++) {
  855.                 page_table_lock(as, false);
  856.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  857.                 ASSERT(pte && PTE_VALID(pte) &&
  858.                     PTE_PRESENT(pte));
  859.                 old_frame[frame_idx++] = PTE_GET_FRAME(pte);
  860.  
  861.                 /* Remove old mapping */
  862.                 page_mapping_remove(as, b + j * PAGE_SIZE);
  863.                 page_table_unlock(as, false);
  864.             }
  865.         }
  866.     }
  867.  
  868.     /*
  869.      * Finish TLB shootdown sequence.
  870.      */
  871.  
  872.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  873.     /*
  874.      * Invalidate potential software translation caches (e.g. TSB on
  875.      * sparc64).
  876.      */
  877.     as_invalidate_translation_cache(as, area->base, area->pages);
  878.     tlb_shootdown_finalize();
  879.  
  880.     /*
  881.      * Set the new flags.
  882.      */
  883.     area->flags = flags;
  884.  
  885.     /*
  886.      * Map pages back in with new flags. This step is kept separate
  887.      * so that the memory area could not be accesed with both the old and
  888.      * the new flags at once.
  889.      */
  890.     frame_idx = 0;
  891.  
  892.     for (cur = area->used_space.leaf_head.next;
  893.         cur != &area->used_space.leaf_head; cur = cur->next) {
  894.         btree_node_t *node;
  895.         unsigned int i;
  896.        
  897.         node = list_get_instance(cur, btree_node_t, leaf_link);
  898.         for (i = 0; i < node->keys; i++) {
  899.             uintptr_t b = node->key[i];
  900.             count_t j;
  901.            
  902.             for (j = 0; j < (count_t) node->value[i]; j++) {
  903.                 page_table_lock(as, false);
  904.  
  905.                 /* Insert the new mapping */
  906.                 page_mapping_insert(as, b + j * PAGE_SIZE,
  907.                     old_frame[frame_idx++], page_flags);
  908.  
  909.                 page_table_unlock(as, false);
  910.             }
  911.         }
  912.     }
  913.  
  914.     free(old_frame);
  915.  
  916.     mutex_unlock(&area->lock);
  917.     mutex_unlock(&as->lock);
  918.     interrupts_restore(ipl);
  919.  
  920.     return 0;
  921. }
  922.  
  923.  
  924. /** Handle page fault within the current address space.
  925.  *
  926.  * This is the high-level page fault handler. It decides whether the page fault
  927.  * can be resolved by any backend and if so, it invokes the backend to resolve
  928.  * the page fault.
  929.  *
  930.  * Interrupts are assumed disabled.
  931.  *
  932.  * @param page      Faulting page.
  933.  * @param access    Access mode that caused the page fault (i.e.
  934.  *          read/write/exec).
  935.  * @param istate    Pointer to the interrupted state.
  936.  *
  937.  * @return      AS_PF_FAULT on page fault, AS_PF_OK on success or
  938.  *          AS_PF_DEFER if the fault was caused by copy_to_uspace()
  939.  *          or copy_from_uspace().
  940.  */
  941. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  942. {
  943.     pte_t *pte;
  944.     as_area_t *area;
  945.    
  946.     if (!THREAD)
  947.         return AS_PF_FAULT;
  948.        
  949.     ASSERT(AS);
  950.  
  951.     mutex_lock(&AS->lock);
  952.     area = find_area_and_lock(AS, page);   
  953.     if (!area) {
  954.         /*
  955.          * No area contained mapping for 'page'.
  956.          * Signal page fault to low-level handler.
  957.          */
  958.         mutex_unlock(&AS->lock);
  959.         goto page_fault;
  960.     }
  961.  
  962.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  963.         /*
  964.          * The address space area is not fully initialized.
  965.          * Avoid possible race by returning error.
  966.          */
  967.         mutex_unlock(&area->lock);
  968.         mutex_unlock(&AS->lock);
  969.         goto page_fault;       
  970.     }
  971.  
  972.     if (!area->backend || !area->backend->page_fault) {
  973.         /*
  974.          * The address space area is not backed by any backend
  975.          * or the backend cannot handle page faults.
  976.          */
  977.         mutex_unlock(&area->lock);
  978.         mutex_unlock(&AS->lock);
  979.         goto page_fault;       
  980.     }
  981.  
  982.     page_table_lock(AS, false);
  983.    
  984.     /*
  985.      * To avoid race condition between two page faults on the same address,
  986.      * we need to make sure the mapping has not been already inserted.
  987.      */
  988.     if ((pte = page_mapping_find(AS, page))) {
  989.         if (PTE_PRESENT(pte)) {
  990.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  991.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  992.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  993.                 page_table_unlock(AS, false);
  994.                 mutex_unlock(&area->lock);
  995.                 mutex_unlock(&AS->lock);
  996.                 return AS_PF_OK;
  997.             }
  998.         }
  999.     }
  1000.    
  1001.     /*
  1002.      * Resort to the backend page fault handler.
  1003.      */
  1004.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  1005.         page_table_unlock(AS, false);
  1006.         mutex_unlock(&area->lock);
  1007.         mutex_unlock(&AS->lock);
  1008.         goto page_fault;
  1009.     }
  1010.    
  1011.     page_table_unlock(AS, false);
  1012.     mutex_unlock(&area->lock);
  1013.     mutex_unlock(&AS->lock);
  1014.     return AS_PF_OK;
  1015.  
  1016. page_fault:
  1017.     if (THREAD->in_copy_from_uspace) {
  1018.         THREAD->in_copy_from_uspace = false;
  1019.         istate_set_retaddr(istate,
  1020.             (uintptr_t) &memcpy_from_uspace_failover_address);
  1021.     } else if (THREAD->in_copy_to_uspace) {
  1022.         THREAD->in_copy_to_uspace = false;
  1023.         istate_set_retaddr(istate,
  1024.             (uintptr_t) &memcpy_to_uspace_failover_address);
  1025.     } else {
  1026.         return AS_PF_FAULT;
  1027.     }
  1028.  
  1029.     return AS_PF_DEFER;
  1030. }
  1031.  
  1032. /** Switch address spaces.
  1033.  *
  1034.  * Note that this function cannot sleep as it is essentially a part of
  1035.  * scheduling. Sleeping here would lead to deadlock on wakeup. Another
  1036.  * thing which is forbidden in this context is locking the address space.
  1037.  *
  1038.  * When this function is enetered, no spinlocks may be held.
  1039.  *
  1040.  * @param old       Old address space or NULL.
  1041.  * @param new       New address space.
  1042.  */
  1043. void as_switch(as_t *old_as, as_t *new_as)
  1044. {
  1045.     DEADLOCK_PROBE_INIT(p_asidlock);
  1046.     preemption_disable();
  1047. retry:
  1048.     (void) interrupts_disable();
  1049.     if (!spinlock_trylock(&asidlock)) {
  1050.         /*
  1051.          * Avoid deadlock with TLB shootdown.
  1052.          * We can enable interrupts here because
  1053.          * preemption is disabled. We should not be
  1054.          * holding any other lock.
  1055.          */
  1056.         (void) interrupts_enable();
  1057.         DEADLOCK_PROBE(p_asidlock, DEADLOCK_THRESHOLD);
  1058.         goto retry;
  1059.     }
  1060.     preemption_enable();
  1061.  
  1062.     /*
  1063.      * First, take care of the old address space.
  1064.      */
  1065.     if (old_as) {
  1066.         ASSERT(old_as->cpu_refcount);
  1067.         if((--old_as->cpu_refcount == 0) && (old_as != AS_KERNEL)) {
  1068.             /*
  1069.              * The old address space is no longer active on
  1070.              * any processor. It can be appended to the
  1071.              * list of inactive address spaces with assigned
  1072.              * ASID.
  1073.              */
  1074.             ASSERT(old_as->asid != ASID_INVALID);
  1075.             list_append(&old_as->inactive_as_with_asid_link,
  1076.                 &inactive_as_with_asid_head);
  1077.         }
  1078.  
  1079.         /*
  1080.          * Perform architecture-specific tasks when the address space
  1081.          * is being removed from the CPU.
  1082.          */
  1083.         as_deinstall_arch(old_as);
  1084.     }
  1085.  
  1086.     /*
  1087.      * Second, prepare the new address space.
  1088.      */
  1089.     if ((new_as->cpu_refcount++ == 0) && (new_as != AS_KERNEL)) {
  1090.         if (new_as->asid != ASID_INVALID)
  1091.             list_remove(&new_as->inactive_as_with_asid_link);
  1092.         else
  1093.             new_as->asid = asid_get();
  1094.     }
  1095. #ifdef AS_PAGE_TABLE
  1096.     SET_PTL0_ADDRESS(new_as->genarch.page_table);
  1097. #endif
  1098.    
  1099.     /*
  1100.      * Perform architecture-specific steps.
  1101.      * (e.g. write ASID to hardware register etc.)
  1102.      */
  1103.     as_install_arch(new_as);
  1104.  
  1105.     spinlock_unlock(&asidlock);
  1106.    
  1107.     AS = new_as;
  1108. }
  1109.  
  1110. /** Convert address space area flags to page flags.
  1111.  *
  1112.  * @param aflags    Flags of some address space area.
  1113.  *
  1114.  * @return      Flags to be passed to page_mapping_insert().
  1115.  */
  1116. int area_flags_to_page_flags(int aflags)
  1117. {
  1118.     int flags;
  1119.  
  1120.     flags = PAGE_USER | PAGE_PRESENT;
  1121.    
  1122.     if (aflags & AS_AREA_READ)
  1123.         flags |= PAGE_READ;
  1124.        
  1125.     if (aflags & AS_AREA_WRITE)
  1126.         flags |= PAGE_WRITE;
  1127.    
  1128.     if (aflags & AS_AREA_EXEC)
  1129.         flags |= PAGE_EXEC;
  1130.    
  1131.     if (aflags & AS_AREA_CACHEABLE)
  1132.         flags |= PAGE_CACHEABLE;
  1133.        
  1134.     return flags;
  1135. }
  1136.  
  1137. /** Compute flags for virtual address translation subsytem.
  1138.  *
  1139.  * The address space area must be locked.
  1140.  * Interrupts must be disabled.
  1141.  *
  1142.  * @param a     Address space area.
  1143.  *
  1144.  * @return      Flags to be used in page_mapping_insert().
  1145.  */
  1146. int as_area_get_flags(as_area_t *a)
  1147. {
  1148.     return area_flags_to_page_flags(a->flags);
  1149. }
  1150.  
  1151. /** Create page table.
  1152.  *
  1153.  * Depending on architecture, create either address space private or global page
  1154.  * table.
  1155.  *
  1156.  * @param flags     Flags saying whether the page table is for the kernel
  1157.  *          address space.
  1158.  *
  1159.  * @return      First entry of the page table.
  1160.  */
  1161. pte_t *page_table_create(int flags)
  1162. {
  1163.     ASSERT(as_operations);
  1164.     ASSERT(as_operations->page_table_create);
  1165.    
  1166.     return as_operations->page_table_create(flags);
  1167. }
  1168.  
  1169. /** Destroy page table.
  1170.  *
  1171.  * Destroy page table in architecture specific way.
  1172.  *
  1173.  * @param page_table    Physical address of PTL0.
  1174.  */
  1175. void page_table_destroy(pte_t *page_table)
  1176. {
  1177.     ASSERT(as_operations);
  1178.     ASSERT(as_operations->page_table_destroy);
  1179.    
  1180.     as_operations->page_table_destroy(page_table);
  1181. }
  1182.  
  1183. /** Lock page table.
  1184.  *
  1185.  * This function should be called before any page_mapping_insert(),
  1186.  * page_mapping_remove() and page_mapping_find().
  1187.  *
  1188.  * Locking order is such that address space areas must be locked
  1189.  * prior to this call. Address space can be locked prior to this
  1190.  * call in which case the lock argument is false.
  1191.  *
  1192.  * @param as        Address space.
  1193.  * @param lock      If false, do not attempt to lock as->lock.
  1194.  */
  1195. void page_table_lock(as_t *as, bool lock)
  1196. {
  1197.     ASSERT(as_operations);
  1198.     ASSERT(as_operations->page_table_lock);
  1199.    
  1200.     as_operations->page_table_lock(as, lock);
  1201. }
  1202.  
  1203. /** Unlock page table.
  1204.  *
  1205.  * @param as        Address space.
  1206.  * @param unlock    If false, do not attempt to unlock as->lock.
  1207.  */
  1208. void page_table_unlock(as_t *as, bool unlock)
  1209. {
  1210.     ASSERT(as_operations);
  1211.     ASSERT(as_operations->page_table_unlock);
  1212.    
  1213.     as_operations->page_table_unlock(as, unlock);
  1214. }
  1215.  
  1216.  
  1217. /** Find address space area and lock it.
  1218.  *
  1219.  * The address space must be locked and interrupts must be disabled.
  1220.  *
  1221.  * @param as        Address space.
  1222.  * @param va        Virtual address.
  1223.  *
  1224.  * @return      Locked address space area containing va on success or
  1225.  *          NULL on failure.
  1226.  */
  1227. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  1228. {
  1229.     as_area_t *a;
  1230.     btree_node_t *leaf, *lnode;
  1231.     unsigned int i;
  1232.    
  1233.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  1234.     if (a) {
  1235.         /* va is the base address of an address space area */
  1236.         mutex_lock(&a->lock);
  1237.         return a;
  1238.     }
  1239.    
  1240.     /*
  1241.      * Search the leaf node and the righmost record of its left neighbour
  1242.      * to find out whether this is a miss or va belongs to an address
  1243.      * space area found there.
  1244.      */
  1245.    
  1246.     /* First, search the leaf node itself. */
  1247.     for (i = 0; i < leaf->keys; i++) {
  1248.         a = (as_area_t *) leaf->value[i];
  1249.         mutex_lock(&a->lock);
  1250.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  1251.             return a;
  1252.         }
  1253.         mutex_unlock(&a->lock);
  1254.     }
  1255.  
  1256.     /*
  1257.      * Second, locate the left neighbour and test its last record.
  1258.      * Because of its position in the B+tree, it must have base < va.
  1259.      */
  1260.     lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf);
  1261.     if (lnode) {
  1262.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  1263.         mutex_lock(&a->lock);
  1264.         if (va < a->base + a->pages * PAGE_SIZE) {
  1265.             return a;
  1266.         }
  1267.         mutex_unlock(&a->lock);
  1268.     }
  1269.  
  1270.     return NULL;
  1271. }
  1272.  
  1273. /** Check area conflicts with other areas.
  1274.  *
  1275.  * The address space must be locked and interrupts must be disabled.
  1276.  *
  1277.  * @param as        Address space.
  1278.  * @param va        Starting virtual address of the area being tested.
  1279.  * @param size      Size of the area being tested.
  1280.  * @param avoid_area    Do not touch this area.
  1281.  *
  1282.  * @return      True if there is no conflict, false otherwise.
  1283.  */
  1284. bool
  1285. check_area_conflicts(as_t *as, uintptr_t va, size_t size, as_area_t *avoid_area)
  1286. {
  1287.     as_area_t *a;
  1288.     btree_node_t *leaf, *node;
  1289.     unsigned int i;
  1290.    
  1291.     /*
  1292.      * We don't want any area to have conflicts with NULL page.
  1293.      */
  1294.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1295.         return false;
  1296.    
  1297.     /*
  1298.      * The leaf node is found in O(log n), where n is proportional to
  1299.      * the number of address space areas belonging to as.
  1300.      * The check for conflicts is then attempted on the rightmost
  1301.      * record in the left neighbour, the leftmost record in the right
  1302.      * neighbour and all records in the leaf node itself.
  1303.      */
  1304.    
  1305.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1306.         if (a != avoid_area)
  1307.             return false;
  1308.     }
  1309.    
  1310.     /* First, check the two border cases. */
  1311.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1312.         a = (as_area_t *) node->value[node->keys - 1];
  1313.         mutex_lock(&a->lock);
  1314.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1315.             mutex_unlock(&a->lock);
  1316.             return false;
  1317.         }
  1318.         mutex_unlock(&a->lock);
  1319.     }
  1320.     node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf);
  1321.     if (node) {
  1322.         a = (as_area_t *) node->value[0];
  1323.         mutex_lock(&a->lock);
  1324.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1325.             mutex_unlock(&a->lock);
  1326.             return false;
  1327.         }
  1328.         mutex_unlock(&a->lock);
  1329.     }
  1330.    
  1331.     /* Second, check the leaf node. */
  1332.     for (i = 0; i < leaf->keys; i++) {
  1333.         a = (as_area_t *) leaf->value[i];
  1334.    
  1335.         if (a == avoid_area)
  1336.             continue;
  1337.    
  1338.         mutex_lock(&a->lock);
  1339.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1340.             mutex_unlock(&a->lock);
  1341.             return false;
  1342.         }
  1343.         mutex_unlock(&a->lock);
  1344.     }
  1345.  
  1346.     /*
  1347.      * So far, the area does not conflict with other areas.
  1348.      * Check if it doesn't conflict with kernel address space.
  1349.      */  
  1350.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1351.         return !overlaps(va, size,
  1352.             KERNEL_ADDRESS_SPACE_START,
  1353.             KERNEL_ADDRESS_SPACE_END - KERNEL_ADDRESS_SPACE_START);
  1354.     }
  1355.  
  1356.     return true;
  1357. }
  1358.  
  1359. /** Return size of the address space area with given base.
  1360.  *
  1361.  * @param base      Arbitrary address insede the address space area.
  1362.  *
  1363.  * @return      Size of the address space area in bytes or zero if it
  1364.  *          does not exist.
  1365.  */
  1366. size_t as_area_get_size(uintptr_t base)
  1367. {
  1368.     ipl_t ipl;
  1369.     as_area_t *src_area;
  1370.     size_t size;
  1371.  
  1372.     ipl = interrupts_disable();
  1373.     src_area = find_area_and_lock(AS, base);
  1374.     if (src_area) {
  1375.         size = src_area->pages * PAGE_SIZE;
  1376.         mutex_unlock(&src_area->lock);
  1377.     } else {
  1378.         size = 0;
  1379.     }
  1380.     interrupts_restore(ipl);
  1381.     return size;
  1382. }
  1383.  
  1384. /** Mark portion of address space area as used.
  1385.  *
  1386.  * The address space area must be already locked.
  1387.  *
  1388.  * @param a     Address space area.
  1389.  * @param page      First page to be marked.
  1390.  * @param count     Number of page to be marked.
  1391.  *
  1392.  * @return      Zero on failure and non-zero on success.
  1393.  */
  1394. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1395. {
  1396.     btree_node_t *leaf, *node;
  1397.     count_t pages;
  1398.     unsigned int i;
  1399.  
  1400.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1401.     ASSERT(count);
  1402.  
  1403.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1404.     if (pages) {
  1405.         /*
  1406.          * We hit the beginning of some used space.
  1407.          */
  1408.         return 0;
  1409.     }
  1410.  
  1411.     if (!leaf->keys) {
  1412.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1413.         return 1;
  1414.     }
  1415.  
  1416.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1417.     if (node) {
  1418.         uintptr_t left_pg = node->key[node->keys - 1];
  1419.         uintptr_t right_pg = leaf->key[0];
  1420.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1421.         count_t right_cnt = (count_t) leaf->value[0];
  1422.        
  1423.         /*
  1424.          * Examine the possibility that the interval fits
  1425.          * somewhere between the rightmost interval of
  1426.          * the left neigbour and the first interval of the leaf.
  1427.          */
  1428.          
  1429.         if (page >= right_pg) {
  1430.             /* Do nothing. */
  1431.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1432.             left_cnt * PAGE_SIZE)) {
  1433.             /* The interval intersects with the left interval. */
  1434.             return 0;
  1435.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1436.             right_cnt * PAGE_SIZE)) {
  1437.             /* The interval intersects with the right interval. */
  1438.             return 0;          
  1439.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1440.             (page + count * PAGE_SIZE == right_pg)) {
  1441.             /*
  1442.              * The interval can be added by merging the two already
  1443.              * present intervals.
  1444.              */
  1445.             node->value[node->keys - 1] += count + right_cnt;
  1446.             btree_remove(&a->used_space, right_pg, leaf);
  1447.             return 1;
  1448.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1449.             /*
  1450.              * The interval can be added by simply growing the left
  1451.              * interval.
  1452.              */
  1453.             node->value[node->keys - 1] += count;
  1454.             return 1;
  1455.         } else if (page + count * PAGE_SIZE == right_pg) {
  1456.             /*
  1457.              * The interval can be addded by simply moving base of
  1458.              * the right interval down and increasing its size
  1459.              * accordingly.
  1460.              */
  1461.             leaf->value[0] += count;
  1462.             leaf->key[0] = page;
  1463.             return 1;
  1464.         } else {
  1465.             /*
  1466.              * The interval is between both neigbouring intervals,
  1467.              * but cannot be merged with any of them.
  1468.              */
  1469.             btree_insert(&a->used_space, page, (void *) count,
  1470.                 leaf);
  1471.             return 1;
  1472.         }
  1473.     } else if (page < leaf->key[0]) {
  1474.         uintptr_t right_pg = leaf->key[0];
  1475.         count_t right_cnt = (count_t) leaf->value[0];
  1476.    
  1477.         /*
  1478.          * Investigate the border case in which the left neighbour does
  1479.          * not exist but the interval fits from the left.
  1480.          */
  1481.          
  1482.         if (overlaps(page, count * PAGE_SIZE, right_pg,
  1483.             right_cnt * PAGE_SIZE)) {
  1484.             /* The interval intersects with the right interval. */
  1485.             return 0;
  1486.         } else if (page + count * PAGE_SIZE == right_pg) {
  1487.             /*
  1488.              * The interval can be added by moving the base of the
  1489.              * right interval down and increasing its size
  1490.              * accordingly.
  1491.              */
  1492.             leaf->key[0] = page;
  1493.             leaf->value[0] += count;
  1494.             return 1;
  1495.         } else {
  1496.             /*
  1497.              * The interval doesn't adjoin with the right interval.
  1498.              * It must be added individually.
  1499.              */
  1500.             btree_insert(&a->used_space, page, (void *) count,
  1501.                 leaf);
  1502.             return 1;
  1503.         }
  1504.     }
  1505.  
  1506.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1507.     if (node) {
  1508.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1509.         uintptr_t right_pg = node->key[0];
  1510.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1511.         count_t right_cnt = (count_t) node->value[0];
  1512.        
  1513.         /*
  1514.          * Examine the possibility that the interval fits
  1515.          * somewhere between the leftmost interval of
  1516.          * the right neigbour and the last interval of the leaf.
  1517.          */
  1518.  
  1519.         if (page < left_pg) {
  1520.             /* Do nothing. */
  1521.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1522.             left_cnt * PAGE_SIZE)) {
  1523.             /* The interval intersects with the left interval. */
  1524.             return 0;
  1525.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1526.             right_cnt * PAGE_SIZE)) {
  1527.             /* The interval intersects with the right interval. */
  1528.             return 0;          
  1529.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1530.             (page + count * PAGE_SIZE == right_pg)) {
  1531.             /*
  1532.              * The interval can be added by merging the two already
  1533.              * present intervals.
  1534.              * */
  1535.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1536.             btree_remove(&a->used_space, right_pg, node);
  1537.             return 1;
  1538.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1539.             /*
  1540.              * The interval can be added by simply growing the left
  1541.              * interval.
  1542.              * */
  1543.             leaf->value[leaf->keys - 1] +=  count;
  1544.             return 1;
  1545.         } else if (page + count * PAGE_SIZE == right_pg) {
  1546.             /*
  1547.              * The interval can be addded by simply moving base of
  1548.              * the right interval down and increasing its size
  1549.              * accordingly.
  1550.              */
  1551.             node->value[0] += count;
  1552.             node->key[0] = page;
  1553.             return 1;
  1554.         } else {
  1555.             /*
  1556.              * The interval is between both neigbouring intervals,
  1557.              * but cannot be merged with any of them.
  1558.              */
  1559.             btree_insert(&a->used_space, page, (void *) count,
  1560.                 leaf);
  1561.             return 1;
  1562.         }
  1563.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1564.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1565.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1566.    
  1567.         /*
  1568.          * Investigate the border case in which the right neighbour
  1569.          * does not exist but the interval fits from the right.
  1570.          */
  1571.          
  1572.         if (overlaps(page, count * PAGE_SIZE, left_pg,
  1573.             left_cnt * PAGE_SIZE)) {
  1574.             /* The interval intersects with the left interval. */
  1575.             return 0;
  1576.         } else if (left_pg + left_cnt * PAGE_SIZE == page) {
  1577.             /*
  1578.              * The interval can be added by growing the left
  1579.              * interval.
  1580.              */
  1581.             leaf->value[leaf->keys - 1] += count;
  1582.             return 1;
  1583.         } else {
  1584.             /*
  1585.              * The interval doesn't adjoin with the left interval.
  1586.              * It must be added individually.
  1587.              */
  1588.             btree_insert(&a->used_space, page, (void *) count,
  1589.                 leaf);
  1590.             return 1;
  1591.         }
  1592.     }
  1593.    
  1594.     /*
  1595.      * Note that if the algorithm made it thus far, the interval can fit
  1596.      * only between two other intervals of the leaf. The two border cases
  1597.      * were already resolved.
  1598.      */
  1599.     for (i = 1; i < leaf->keys; i++) {
  1600.         if (page < leaf->key[i]) {
  1601.             uintptr_t left_pg = leaf->key[i - 1];
  1602.             uintptr_t right_pg = leaf->key[i];
  1603.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1604.             count_t right_cnt = (count_t) leaf->value[i];
  1605.  
  1606.             /*
  1607.              * The interval fits between left_pg and right_pg.
  1608.              */
  1609.  
  1610.             if (overlaps(page, count * PAGE_SIZE, left_pg,
  1611.                 left_cnt * PAGE_SIZE)) {
  1612.                 /*
  1613.                  * The interval intersects with the left
  1614.                  * interval.
  1615.                  */
  1616.                 return 0;
  1617.             } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1618.                 right_cnt * PAGE_SIZE)) {
  1619.                 /*
  1620.                  * The interval intersects with the right
  1621.                  * interval.
  1622.                  */
  1623.                 return 0;          
  1624.             } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1625.                 (page + count * PAGE_SIZE == right_pg)) {
  1626.                 /*
  1627.                  * The interval can be added by merging the two
  1628.                  * already present intervals.
  1629.                  */
  1630.                 leaf->value[i - 1] += count + right_cnt;
  1631.                 btree_remove(&a->used_space, right_pg, leaf);
  1632.                 return 1;
  1633.             } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1634.                 /*
  1635.                  * The interval can be added by simply growing
  1636.                  * the left interval.
  1637.                  */
  1638.                 leaf->value[i - 1] += count;
  1639.                 return 1;
  1640.             } else if (page + count * PAGE_SIZE == right_pg) {
  1641.                 /*
  1642.                      * The interval can be addded by simply moving
  1643.                  * base of the right interval down and
  1644.                  * increasing its size accordingly.
  1645.                  */
  1646.                 leaf->value[i] += count;
  1647.                 leaf->key[i] = page;
  1648.                 return 1;
  1649.             } else {
  1650.                 /*
  1651.                  * The interval is between both neigbouring
  1652.                  * intervals, but cannot be merged with any of
  1653.                  * them.
  1654.                  */
  1655.                 btree_insert(&a->used_space, page,
  1656.                     (void *) count, leaf);
  1657.                 return 1;
  1658.             }
  1659.         }
  1660.     }
  1661.  
  1662.     panic("Inconsistency detected while adding %" PRIc " pages of used "
  1663.         "space at %p.\n", count, page);
  1664. }
  1665.  
  1666. /** Mark portion of address space area as unused.
  1667.  *
  1668.  * The address space area must be already locked.
  1669.  *
  1670.  * @param a     Address space area.
  1671.  * @param page      First page to be marked.
  1672.  * @param count     Number of page to be marked.
  1673.  *
  1674.  * @return      Zero on failure and non-zero on success.
  1675.  */
  1676. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1677. {
  1678.     btree_node_t *leaf, *node;
  1679.     count_t pages;
  1680.     unsigned int i;
  1681.  
  1682.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1683.     ASSERT(count);
  1684.  
  1685.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1686.     if (pages) {
  1687.         /*
  1688.          * We are lucky, page is the beginning of some interval.
  1689.          */
  1690.         if (count > pages) {
  1691.             return 0;
  1692.         } else if (count == pages) {
  1693.             btree_remove(&a->used_space, page, leaf);
  1694.             return 1;
  1695.         } else {
  1696.             /*
  1697.              * Find the respective interval.
  1698.              * Decrease its size and relocate its start address.
  1699.              */
  1700.             for (i = 0; i < leaf->keys; i++) {
  1701.                 if (leaf->key[i] == page) {
  1702.                     leaf->key[i] += count * PAGE_SIZE;
  1703.                     leaf->value[i] -= count;
  1704.                     return 1;
  1705.                 }
  1706.             }
  1707.             goto error;
  1708.         }
  1709.     }
  1710.  
  1711.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1712.     if (node && page < leaf->key[0]) {
  1713.         uintptr_t left_pg = node->key[node->keys - 1];
  1714.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1715.  
  1716.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1717.             count * PAGE_SIZE)) {
  1718.             if (page + count * PAGE_SIZE ==
  1719.                 left_pg + left_cnt * PAGE_SIZE) {
  1720.                 /*
  1721.                  * The interval is contained in the rightmost
  1722.                  * interval of the left neighbour and can be
  1723.                  * removed by updating the size of the bigger
  1724.                  * interval.
  1725.                  */
  1726.                 node->value[node->keys - 1] -= count;
  1727.                 return 1;
  1728.             } else if (page + count * PAGE_SIZE <
  1729.                 left_pg + left_cnt*PAGE_SIZE) {
  1730.                 count_t new_cnt;
  1731.                
  1732.                 /*
  1733.                  * The interval is contained in the rightmost
  1734.                  * interval of the left neighbour but its
  1735.                  * removal requires both updating the size of
  1736.                  * the original interval and also inserting a
  1737.                  * new interval.
  1738.                  */
  1739.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1740.                     (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1741.                 node->value[node->keys - 1] -= count + new_cnt;
  1742.                 btree_insert(&a->used_space, page +
  1743.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1744.                 return 1;
  1745.             }
  1746.         }
  1747.         return 0;
  1748.     } else if (page < leaf->key[0]) {
  1749.         return 0;
  1750.     }
  1751.    
  1752.     if (page > leaf->key[leaf->keys - 1]) {
  1753.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1754.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1755.  
  1756.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1757.             count * PAGE_SIZE)) {
  1758.             if (page + count * PAGE_SIZE ==
  1759.                 left_pg + left_cnt * PAGE_SIZE) {
  1760.                 /*
  1761.                  * The interval is contained in the rightmost
  1762.                  * interval of the leaf and can be removed by
  1763.                  * updating the size of the bigger interval.
  1764.                  */
  1765.                 leaf->value[leaf->keys - 1] -= count;
  1766.                 return 1;
  1767.             } else if (page + count * PAGE_SIZE < left_pg +
  1768.                 left_cnt * PAGE_SIZE) {
  1769.                 count_t new_cnt;
  1770.                
  1771.                 /*
  1772.                  * The interval is contained in the rightmost
  1773.                  * interval of the leaf but its removal
  1774.                  * requires both updating the size of the
  1775.                  * original interval and also inserting a new
  1776.                  * interval.
  1777.                  */
  1778.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1779.                     (page + count * PAGE_SIZE)) >> PAGE_WIDTH;
  1780.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1781.                 btree_insert(&a->used_space, page +
  1782.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1783.                 return 1;
  1784.             }
  1785.         }
  1786.         return 0;
  1787.     }  
  1788.    
  1789.     /*
  1790.      * The border cases have been already resolved.
  1791.      * Now the interval can be only between intervals of the leaf.
  1792.      */
  1793.     for (i = 1; i < leaf->keys - 1; i++) {
  1794.         if (page < leaf->key[i]) {
  1795.             uintptr_t left_pg = leaf->key[i - 1];
  1796.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1797.  
  1798.             /*
  1799.              * Now the interval is between intervals corresponding
  1800.              * to (i - 1) and i.
  1801.              */
  1802.             if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1803.                 count * PAGE_SIZE)) {
  1804.                 if (page + count * PAGE_SIZE ==
  1805.                     left_pg + left_cnt*PAGE_SIZE) {
  1806.                     /*
  1807.                      * The interval is contained in the
  1808.                      * interval (i - 1) of the leaf and can
  1809.                      * be removed by updating the size of
  1810.                      * the bigger interval.
  1811.                      */
  1812.                     leaf->value[i - 1] -= count;
  1813.                     return 1;
  1814.                 } else if (page + count * PAGE_SIZE <
  1815.                     left_pg + left_cnt * PAGE_SIZE) {
  1816.                     count_t new_cnt;
  1817.                
  1818.                     /*
  1819.                      * The interval is contained in the
  1820.                      * interval (i - 1) of the leaf but its
  1821.                      * removal requires both updating the
  1822.                      * size of the original interval and
  1823.                      * also inserting a new interval.
  1824.                      */
  1825.                     new_cnt = ((left_pg +
  1826.                         left_cnt * PAGE_SIZE) -
  1827.                         (page + count * PAGE_SIZE)) >>
  1828.                         PAGE_WIDTH;
  1829.                     leaf->value[i - 1] -= count + new_cnt;
  1830.                     btree_insert(&a->used_space, page +
  1831.                         count * PAGE_SIZE, (void *) new_cnt,
  1832.                         leaf);
  1833.                     return 1;
  1834.                 }
  1835.             }
  1836.             return 0;
  1837.         }
  1838.     }
  1839.  
  1840. error:
  1841.     panic("Inconsistency detected while removing %" PRIc " pages of used "
  1842.         "space from %p.\n", count, page);
  1843. }
  1844.  
  1845. /** Remove reference to address space area share info.
  1846.  *
  1847.  * If the reference count drops to 0, the sh_info is deallocated.
  1848.  *
  1849.  * @param sh_info   Pointer to address space area share info.
  1850.  */
  1851. void sh_info_remove_reference(share_info_t *sh_info)
  1852. {
  1853.     bool dealloc = false;
  1854.  
  1855.     mutex_lock(&sh_info->lock);
  1856.     ASSERT(sh_info->refcount);
  1857.     if (--sh_info->refcount == 0) {
  1858.         dealloc = true;
  1859.         link_t *cur;
  1860.        
  1861.         /*
  1862.          * Now walk carefully the pagemap B+tree and free/remove
  1863.          * reference from all frames found there.
  1864.          */
  1865.         for (cur = sh_info->pagemap.leaf_head.next;
  1866.             cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1867.             btree_node_t *node;
  1868.             unsigned int i;
  1869.            
  1870.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1871.             for (i = 0; i < node->keys; i++)
  1872.                 frame_free((uintptr_t) node->value[i]);
  1873.         }
  1874.        
  1875.     }
  1876.     mutex_unlock(&sh_info->lock);
  1877.    
  1878.     if (dealloc) {
  1879.         btree_destroy(&sh_info->pagemap);
  1880.         free(sh_info);
  1881.     }
  1882. }
  1883.  
  1884. /*
  1885.  * Address space related syscalls.
  1886.  */
  1887.  
  1888. /** Wrapper for as_area_create(). */
  1889. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1890. {
  1891.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address,
  1892.         AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1893.         return (unative_t) address;
  1894.     else
  1895.         return (unative_t) -1;
  1896. }
  1897.  
  1898. /** Wrapper for as_area_resize(). */
  1899. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1900. {
  1901.     return (unative_t) as_area_resize(AS, address, size, 0);
  1902. }
  1903.  
  1904. /** Wrapper for as_area_change_flags(). */
  1905. unative_t sys_as_area_change_flags(uintptr_t address, int flags)
  1906. {
  1907.     return (unative_t) as_area_change_flags(AS, flags, address);
  1908. }
  1909.  
  1910. /** Wrapper for as_area_destroy(). */
  1911. unative_t sys_as_area_destroy(uintptr_t address)
  1912. {
  1913.     return (unative_t) as_area_destroy(AS, address);
  1914. }
  1915.  
  1916. /** Print out information about address space.
  1917.  *
  1918.  * @param as        Address space.
  1919.  */
  1920. void as_print(as_t *as)
  1921. {
  1922.     ipl_t ipl;
  1923.    
  1924.     ipl = interrupts_disable();
  1925.     mutex_lock(&as->lock);
  1926.    
  1927.     /* print out info about address space areas */
  1928.     link_t *cur;
  1929.     for (cur = as->as_area_btree.leaf_head.next;
  1930.         cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  1931.         btree_node_t *node;
  1932.        
  1933.         node = list_get_instance(cur, btree_node_t, leaf_link);
  1934.        
  1935.         unsigned int i;
  1936.         for (i = 0; i < node->keys; i++) {
  1937.             as_area_t *area = node->value[i];
  1938.        
  1939.             mutex_lock(&area->lock);
  1940.             printf("as_area: %p, base=%p, pages=%" PRIc
  1941.                 " (%p - %p)\n", area, area->base, area->pages,
  1942.                 area->base, area->base + FRAMES2SIZE(area->pages));
  1943.             mutex_unlock(&area->lock);
  1944.         }
  1945.     }
  1946.    
  1947.     mutex_unlock(&as->lock);
  1948.     interrupts_restore(ipl);
  1949. }
  1950.  
  1951. /** @}
  1952.  */
  1953.