Subversion Repositories HelenOS

Rev

Rev 3129 | Rev 3156 | 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. #ifndef __OBJC__
  86. /**
  87.  * Each architecture decides what functions will be used to carry out
  88.  * address space operations such as creating or locking page tables.
  89.  */
  90. as_operations_t *as_operations = NULL;
  91.  
  92. /**
  93.  * Slab for as_t objects.
  94.  */
  95. static slab_cache_t *as_slab;
  96. #endif
  97.  
  98. /**
  99.  * This lock serializes access to the ASID subsystem.
  100.  * It protects:
  101.  * - inactive_as_with_asid_head list
  102.  * - as->asid for each as of the as_t type
  103.  * - asids_allocated counter
  104.  */
  105. SPINLOCK_INITIALIZE(asidlock);
  106.  
  107. /**
  108.  * This list contains address spaces that are not active on any
  109.  * processor and that have valid ASID.
  110.  */
  111. LIST_INITIALIZE(inactive_as_with_asid_head);
  112.  
  113. /** Kernel address space. */
  114. as_t *AS_KERNEL = NULL;
  115.  
  116. static int area_flags_to_page_flags(int aflags);
  117. static as_area_t *find_area_and_lock(as_t *as, uintptr_t va);
  118. static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  119.     as_area_t *avoid_area);
  120. static void sh_info_remove_reference(share_info_t *sh_info);
  121.  
  122. #ifndef __OBJC__
  123. static int as_constructor(void *obj, int flags)
  124. {
  125.     as_t *as = (as_t *) obj;
  126.     int rc;
  127.  
  128.     link_initialize(&as->inactive_as_with_asid_link);
  129.     mutex_initialize(&as->lock);   
  130.    
  131.     rc = as_constructor_arch(as, flags);
  132.    
  133.     return rc;
  134. }
  135.  
  136. static int as_destructor(void *obj)
  137. {
  138.     as_t *as = (as_t *) obj;
  139.  
  140.     return as_destructor_arch(as);
  141. }
  142. #endif
  143.  
  144. /** Initialize address space subsystem. */
  145. void as_init(void)
  146. {
  147.     as_arch_init();
  148.  
  149. #ifndef __OBJC__
  150.     as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,
  151.         as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);
  152. #endif
  153.    
  154.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  155.     if (!AS_KERNEL)
  156.         panic("can't create kernel address space\n");
  157.    
  158. }
  159.  
  160. /** Create address space.
  161.  *
  162.  * @param flags Flags that influence way in wich the address space is created.
  163.  */
  164. as_t *as_create(int flags)
  165. {
  166.     as_t *as;
  167.  
  168. #ifdef __OBJC__
  169.     as = [as_t new];
  170.     link_initialize(&as->inactive_as_with_asid_link);
  171.     mutex_initialize(&as->lock);   
  172.     (void) as_constructor_arch(as, flags);
  173. #else
  174.     as = (as_t *) slab_alloc(as_slab, 0);
  175. #endif
  176.     (void) as_create_arch(as, 0);
  177.    
  178.     btree_create(&as->as_area_btree);
  179.    
  180.     if (flags & FLAG_AS_KERNEL)
  181.         as->asid = ASID_KERNEL;
  182.     else
  183.         as->asid = ASID_INVALID;
  184.    
  185.     atomic_set(&as->refcount, 0);
  186.     as->cpu_refcount = 0;
  187. #ifdef AS_PAGE_TABLE
  188.     as->genarch.page_table = page_table_create(flags);
  189. #else
  190.     page_table_create(flags);
  191. #endif
  192.  
  193.     return as;
  194. }
  195.  
  196. /** Destroy adress space.
  197.  *
  198.  * When there are no tasks referencing this address space (i.e. its refcount is
  199.  * zero), the address space can be destroyed.
  200.  *
  201.  * We know that we don't hold any spinlock.
  202.  */
  203. void as_destroy(as_t *as)
  204. {
  205.     ipl_t ipl;
  206.     bool cond;
  207.     DEADLOCK_PROBE_INIT(p_asidlock);
  208.  
  209.     ASSERT(atomic_get(&as->refcount) == 0);
  210.    
  211.     /*
  212.      * Since there is no reference to this area,
  213.      * it is safe not to lock its mutex.
  214.      */
  215.  
  216.     /*
  217.      * We need to avoid deadlock between TLB shootdown and asidlock.
  218.      * We therefore try to take asid conditionally and if we don't succeed,
  219.      * we enable interrupts and try again. This is done while preemption is
  220.      * disabled to prevent nested context switches. We also depend on the
  221.      * fact that so far no spinlocks are held.
  222.      */
  223.     preemption_disable();
  224.     ipl = interrupts_read();
  225. retry:
  226.     interrupts_disable();
  227.     if (!spinlock_trylock(&asidlock)) {
  228.         interrupts_enable();
  229.         DEADLOCK_PROBE(p_asidlock, DEADLOCK_THRESHOLD);
  230.         goto retry;
  231.     }
  232.     preemption_enable();    /* Interrupts disabled, enable preemption */
  233.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  234.         if (as != AS && as->cpu_refcount == 0)
  235.             list_remove(&as->inactive_as_with_asid_link);
  236.         asid_put(as->asid);
  237.     }
  238.     spinlock_unlock(&asidlock);
  239.  
  240.     /*
  241.      * Destroy address space areas of the address space.
  242.      * The B+tree must be walked carefully because it is
  243.      * also being destroyed.
  244.      */
  245.     for (cond = true; cond; ) {
  246.         btree_node_t *node;
  247.  
  248.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  249.         node = list_get_instance(as->as_area_btree.leaf_head.next,
  250.             btree_node_t, leaf_link);
  251.  
  252.         if ((cond = node->keys)) {
  253.             as_area_destroy(as, node->key[0]);
  254.         }
  255.     }
  256.  
  257.     btree_destroy(&as->as_area_btree);
  258. #ifdef AS_PAGE_TABLE
  259.     page_table_destroy(as->genarch.page_table);
  260. #else
  261.     page_table_destroy(NULL);
  262. #endif
  263.  
  264.     interrupts_restore(ipl);
  265.  
  266. #ifdef __OBJC__
  267.     [as free];
  268. #else
  269.     slab_free(as_slab, as);
  270. #endif
  271. }
  272.  
  273. /** Create address space area of common attributes.
  274.  *
  275.  * The created address space area is added to the target address space.
  276.  *
  277.  * @param as Target address space.
  278.  * @param flags Flags of the area memory.
  279.  * @param size Size of area.
  280.  * @param base Base address of area.
  281.  * @param attrs Attributes of the area.
  282.  * @param backend Address space area backend. NULL if no backend is used.
  283.  * @param backend_data NULL or a pointer to an array holding two void *.
  284.  *
  285.  * @return Address space area on success or NULL on failure.
  286.  */
  287. as_area_t *
  288. as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
  289.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  290. {
  291.     ipl_t ipl;
  292.     as_area_t *a;
  293.    
  294.     if (base % PAGE_SIZE)
  295.         return NULL;
  296.  
  297.     if (!size)
  298.         return NULL;
  299.  
  300.     /* Writeable executable areas are not supported. */
  301.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  302.         return NULL;
  303.    
  304.     ipl = interrupts_disable();
  305.     mutex_lock(&as->lock);
  306.    
  307.     if (!check_area_conflicts(as, base, size, NULL)) {
  308.         mutex_unlock(&as->lock);
  309.         interrupts_restore(ipl);
  310.         return NULL;
  311.     }
  312.    
  313.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  314.  
  315.     mutex_initialize(&a->lock);
  316.    
  317.     a->as = as;
  318.     a->flags = flags;
  319.     a->attributes = attrs;
  320.     a->pages = SIZE2FRAMES(size);
  321.     a->base = base;
  322.     a->sh_info = NULL;
  323.     a->backend = backend;
  324.     if (backend_data)
  325.         a->backend_data = *backend_data;
  326.     else
  327.         memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data),
  328.             0);
  329.  
  330.     btree_create(&a->used_space);
  331.    
  332.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  333.  
  334.     mutex_unlock(&as->lock);
  335.     interrupts_restore(ipl);
  336.  
  337.     return a;
  338. }
  339.  
  340. /** Find address space area and change it.
  341.  *
  342.  * @param as Address space.
  343.  * @param address Virtual address belonging to the area to be changed. Must be
  344.  *     page-aligned.
  345.  * @param size New size of the virtual memory block starting at address.
  346.  * @param flags Flags influencing the remap operation. Currently unused.
  347.  *
  348.  * @return Zero on success or a value from @ref errno.h otherwise.
  349.  */
  350. int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
  351. {
  352.     as_area_t *area;
  353.     ipl_t ipl;
  354.     size_t pages;
  355.    
  356.     ipl = interrupts_disable();
  357.     mutex_lock(&as->lock);
  358.    
  359.     /*
  360.      * Locate the area.
  361.      */
  362.     area = find_area_and_lock(as, address);
  363.     if (!area) {
  364.         mutex_unlock(&as->lock);
  365.         interrupts_restore(ipl);
  366.         return ENOENT;
  367.     }
  368.  
  369.     if (area->backend == &phys_backend) {
  370.         /*
  371.          * Remapping of address space areas associated
  372.          * with memory mapped devices is not supported.
  373.          */
  374.         mutex_unlock(&area->lock);
  375.         mutex_unlock(&as->lock);
  376.         interrupts_restore(ipl);
  377.         return ENOTSUP;
  378.     }
  379.     if (area->sh_info) {
  380.         /*
  381.          * Remapping of shared address space areas
  382.          * is not supported.
  383.          */
  384.         mutex_unlock(&area->lock);
  385.         mutex_unlock(&as->lock);
  386.         interrupts_restore(ipl);
  387.         return ENOTSUP;
  388.     }
  389.  
  390.     pages = SIZE2FRAMES((address - area->base) + size);
  391.     if (!pages) {
  392.         /*
  393.          * Zero size address space areas are not allowed.
  394.          */
  395.         mutex_unlock(&area->lock);
  396.         mutex_unlock(&as->lock);
  397.         interrupts_restore(ipl);
  398.         return EPERM;
  399.     }
  400.    
  401.     if (pages < area->pages) {
  402.         bool cond;
  403.         uintptr_t start_free = area->base + pages*PAGE_SIZE;
  404.  
  405.         /*
  406.          * Shrinking the area.
  407.          * No need to check for overlaps.
  408.          */
  409.  
  410.         /*
  411.          * Start TLB shootdown sequence.
  412.          */
  413.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base +
  414.             pages * PAGE_SIZE, area->pages - pages);
  415.  
  416.         /*
  417.          * Remove frames belonging to used space starting from
  418.          * the highest addresses downwards until an overlap with
  419.          * the resized address space area is found. Note that this
  420.          * is also the right way to remove part of the used_space
  421.          * B+tree leaf list.
  422.          */    
  423.         for (cond = true; cond;) {
  424.             btree_node_t *node;
  425.        
  426.             ASSERT(!list_empty(&area->used_space.leaf_head));
  427.             node =
  428.                 list_get_instance(area->used_space.leaf_head.prev,
  429.                 btree_node_t, leaf_link);
  430.             if ((cond = (bool) node->keys)) {
  431.                 uintptr_t b = node->key[node->keys - 1];
  432.                 count_t c =
  433.                     (count_t) node->value[node->keys - 1];
  434.                 unsigned int i = 0;
  435.            
  436.                 if (overlaps(b, c * PAGE_SIZE, area->base,
  437.                     pages * PAGE_SIZE)) {
  438.                    
  439.                     if (b + c * PAGE_SIZE <= start_free) {
  440.                         /*
  441.                          * The whole interval fits
  442.                          * completely in the resized
  443.                          * address space area.
  444.                          */
  445.                         break;
  446.                     }
  447.        
  448.                     /*
  449.                      * Part of the interval corresponding
  450.                      * to b and c overlaps with the resized
  451.                      * address space area.
  452.                      */
  453.        
  454.                     cond = false;   /* we are almost done */
  455.                     i = (start_free - b) >> PAGE_WIDTH;
  456.                     if (!used_space_remove(area, start_free, c - i))
  457.                         panic("Could not remove used space.\n");
  458.                 } else {
  459.                     /*
  460.                      * The interval of used space can be
  461.                      * completely removed.
  462.                      */
  463.                     if (!used_space_remove(area, b, c))
  464.                         panic("Could not remove used space.\n");
  465.                 }
  466.            
  467.                 for (; i < c; i++) {
  468.                     pte_t *pte;
  469.            
  470.                     page_table_lock(as, false);
  471.                     pte = page_mapping_find(as, b +
  472.                         i * PAGE_SIZE);
  473.                     ASSERT(pte && PTE_VALID(pte) &&
  474.                         PTE_PRESENT(pte));
  475.                     if (area->backend &&
  476.                         area->backend->frame_free) {
  477.                         area->backend->frame_free(area,
  478.                             b + i * PAGE_SIZE,
  479.                             PTE_GET_FRAME(pte));
  480.                     }
  481.                     page_mapping_remove(as, b +
  482.                         i * PAGE_SIZE);
  483.                     page_table_unlock(as, false);
  484.                 }
  485.             }
  486.         }
  487.  
  488.         /*
  489.          * Finish TLB shootdown sequence.
  490.          */
  491.  
  492.         tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,
  493.             area->pages - pages);
  494.         /*
  495.          * Invalidate software translation caches (e.g. TSB on sparc64).
  496.          */
  497.         as_invalidate_translation_cache(as, area->base +
  498.             pages * PAGE_SIZE, area->pages - pages);
  499.         tlb_shootdown_finalize();
  500.        
  501.     } else {
  502.         /*
  503.          * Growing the area.
  504.          * Check for overlaps with other address space areas.
  505.          */
  506.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE,
  507.             area)) {
  508.             mutex_unlock(&area->lock);
  509.             mutex_unlock(&as->lock);       
  510.             interrupts_restore(ipl);
  511.             return EADDRNOTAVAIL;
  512.         }
  513.     }
  514.  
  515.     area->pages = pages;
  516.    
  517.     mutex_unlock(&area->lock);
  518.     mutex_unlock(&as->lock);
  519.     interrupts_restore(ipl);
  520.  
  521.     return 0;
  522. }
  523.  
  524. /** Destroy address space area.
  525.  *
  526.  * @param as Address space.
  527.  * @param address Address withing the area to be deleted.
  528.  *
  529.  * @return Zero on success or a value from @ref errno.h on failure.
  530.  */
  531. int as_area_destroy(as_t *as, uintptr_t address)
  532. {
  533.     as_area_t *area;
  534.     uintptr_t base;
  535.     link_t *cur;
  536.     ipl_t ipl;
  537.  
  538.     ipl = interrupts_disable();
  539.     mutex_lock(&as->lock);
  540.  
  541.     area = find_area_and_lock(as, address);
  542.     if (!area) {
  543.         mutex_unlock(&as->lock);
  544.         interrupts_restore(ipl);
  545.         return ENOENT;
  546.     }
  547.  
  548.     base = area->base;
  549.  
  550.     /*
  551.      * Start TLB shootdown sequence.
  552.      */
  553.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  554.  
  555.     /*
  556.      * Visit only the pages mapped by used_space B+tree.
  557.      */
  558.     for (cur = area->used_space.leaf_head.next;
  559.         cur != &area->used_space.leaf_head; cur = cur->next) {
  560.         btree_node_t *node;
  561.         unsigned int i;
  562.        
  563.         node = list_get_instance(cur, btree_node_t, leaf_link);
  564.         for (i = 0; i < node->keys; i++) {
  565.             uintptr_t b = node->key[i];
  566.             count_t j;
  567.             pte_t *pte;
  568.            
  569.             for (j = 0; j < (count_t) node->value[i]; j++) {
  570.                 page_table_lock(as, false);
  571.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  572.                 ASSERT(pte && PTE_VALID(pte) &&
  573.                     PTE_PRESENT(pte));
  574.                 if (area->backend &&
  575.                     area->backend->frame_free) {
  576.                     area->backend->frame_free(area, b +
  577.                         j * PAGE_SIZE, PTE_GET_FRAME(pte));
  578.                 }
  579.                 page_mapping_remove(as, b + j * PAGE_SIZE);            
  580.                 page_table_unlock(as, false);
  581.             }
  582.         }
  583.     }
  584.  
  585.     /*
  586.      * Finish TLB shootdown sequence.
  587.      */
  588.  
  589.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  590.     /*
  591.      * Invalidate potential software translation caches (e.g. TSB on
  592.      * sparc64).
  593.      */
  594.     as_invalidate_translation_cache(as, area->base, area->pages);
  595.     tlb_shootdown_finalize();
  596.    
  597.     btree_destroy(&area->used_space);
  598.  
  599.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  600.    
  601.     if (area->sh_info)
  602.         sh_info_remove_reference(area->sh_info);
  603.        
  604.     mutex_unlock(&area->lock);
  605.  
  606.     /*
  607.      * Remove the empty area from address space.
  608.      */
  609.     btree_remove(&as->as_area_btree, base, NULL);
  610.    
  611.     free(area);
  612.    
  613.     mutex_unlock(&as->lock);
  614.     interrupts_restore(ipl);
  615.     return 0;
  616. }
  617.  
  618. /** Share address space area with another or the same address space.
  619.  *
  620.  * Address space area mapping is shared with a new address space area.
  621.  * If the source address space area has not been shared so far,
  622.  * a new sh_info is created. The new address space area simply gets the
  623.  * sh_info of the source area. The process of duplicating the
  624.  * mapping is done through the backend share function.
  625.  *
  626.  * @param src_as Pointer to source address space.
  627.  * @param src_base Base address of the source address space area.
  628.  * @param acc_size Expected size of the source area.
  629.  * @param dst_as Pointer to destination address space.
  630.  * @param dst_base Target base address.
  631.  * @param dst_flags_mask Destination address space area flags mask.
  632.  *
  633.  * @return Zero on success or ENOENT if there is no such task or if there is no
  634.  * such address space area, EPERM if there was a problem in accepting the area
  635.  * or ENOMEM if there was a problem in allocating destination address space
  636.  * area. ENOTSUP is returned if the address space area backend does not support
  637.  * sharing.
  638.  */
  639. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  640.     as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  641. {
  642.     ipl_t ipl;
  643.     int src_flags;
  644.     size_t src_size;
  645.     as_area_t *src_area, *dst_area;
  646.     share_info_t *sh_info;
  647.     mem_backend_t *src_backend;
  648.     mem_backend_data_t src_backend_data;
  649.    
  650.     ipl = interrupts_disable();
  651.     mutex_lock(&src_as->lock);
  652.     src_area = find_area_and_lock(src_as, src_base);
  653.     if (!src_area) {
  654.         /*
  655.          * Could not find the source address space area.
  656.          */
  657.         mutex_unlock(&src_as->lock);
  658.         interrupts_restore(ipl);
  659.         return ENOENT;
  660.     }
  661.  
  662.     if (!src_area->backend || !src_area->backend->share) {
  663.         /*
  664.          * There is no backend or the backend does not
  665.          * know how to share the area.
  666.          */
  667.         mutex_unlock(&src_area->lock);
  668.         mutex_unlock(&src_as->lock);
  669.         interrupts_restore(ipl);
  670.         return ENOTSUP;
  671.     }
  672.    
  673.     src_size = src_area->pages * PAGE_SIZE;
  674.     src_flags = src_area->flags;
  675.     src_backend = src_area->backend;
  676.     src_backend_data = src_area->backend_data;
  677.  
  678.     /* Share the cacheable flag from the original mapping */
  679.     if (src_flags & AS_AREA_CACHEABLE)
  680.         dst_flags_mask |= AS_AREA_CACHEABLE;
  681.  
  682.     if (src_size != acc_size ||
  683.         (src_flags & dst_flags_mask) != dst_flags_mask) {
  684.         mutex_unlock(&src_area->lock);
  685.         mutex_unlock(&src_as->lock);
  686.         interrupts_restore(ipl);
  687.         return EPERM;
  688.     }
  689.  
  690.     /*
  691.      * Now we are committed to sharing the area.
  692.      * First, prepare the area for sharing.
  693.      * Then it will be safe to unlock it.
  694.      */
  695.     sh_info = src_area->sh_info;
  696.     if (!sh_info) {
  697.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  698.         mutex_initialize(&sh_info->lock);
  699.         sh_info->refcount = 2;
  700.         btree_create(&sh_info->pagemap);
  701.         src_area->sh_info = sh_info;
  702.         /*
  703.          * Call the backend to setup sharing.
  704.          */
  705.         src_area->backend->share(src_area);
  706.     } else {
  707.         mutex_lock(&sh_info->lock);
  708.         sh_info->refcount++;
  709.         mutex_unlock(&sh_info->lock);
  710.     }
  711.  
  712.     mutex_unlock(&src_area->lock);
  713.     mutex_unlock(&src_as->lock);
  714.  
  715.     /*
  716.      * Create copy of the source address space area.
  717.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  718.      * attribute set which prevents race condition with
  719.      * preliminary as_page_fault() calls.
  720.      * The flags of the source area are masked against dst_flags_mask
  721.      * to support sharing in less privileged mode.
  722.      */
  723.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  724.         AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  725.     if (!dst_area) {
  726.         /*
  727.          * Destination address space area could not be created.
  728.          */
  729.         sh_info_remove_reference(sh_info);
  730.        
  731.         interrupts_restore(ipl);
  732.         return ENOMEM;
  733.     }
  734.  
  735.     /*
  736.      * Now the destination address space area has been
  737.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  738.      * attribute and set the sh_info.
  739.      */
  740.     mutex_lock(&dst_as->lock); 
  741.     mutex_lock(&dst_area->lock);
  742.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  743.     dst_area->sh_info = sh_info;
  744.     mutex_unlock(&dst_area->lock);
  745.     mutex_unlock(&dst_as->lock);   
  746.  
  747.     interrupts_restore(ipl);
  748.    
  749.     return 0;
  750. }
  751.  
  752. /** Check access mode for address space area.
  753.  *
  754.  * The address space area must be locked prior to this call.
  755.  *
  756.  * @param area Address space area.
  757.  * @param access Access mode.
  758.  *
  759.  * @return False if access violates area's permissions, true otherwise.
  760.  */
  761. bool as_area_check_access(as_area_t *area, pf_access_t access)
  762. {
  763.     int flagmap[] = {
  764.         [PF_ACCESS_READ] = AS_AREA_READ,
  765.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  766.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  767.     };
  768.  
  769.     if (!(area->flags & flagmap[access]))
  770.         return false;
  771.    
  772.     return true;
  773. }
  774.  
  775. /** Change adress area flags.
  776.  *
  777.  * The idea is to have the same data, but with a different access mode.
  778.  * This is needed e.g. for writing code into memory and then executing it.
  779.  * In order for this to work properly, this may copy the data
  780.  * into private anonymous memory (unless it's already there).
  781.  *
  782.  * @param as Address space.
  783.  * @param flags Flags of the area memory.
  784.  * @param address Address withing the area to be changed.
  785.  *
  786.  * @return Zero on success or a value from @ref errno.h on failure.
  787.  */
  788. int as_area_change_flags(as_t *as, int flags, uintptr_t address)
  789. {
  790.     as_area_t *area;
  791.     uintptr_t base;
  792.     link_t *cur;
  793.     ipl_t ipl;
  794.     int page_flags;
  795.     uintptr_t old_frame;
  796.  
  797.     /* Flags for the new memory mapping */
  798.     page_flags = area_flags_to_page_flags(flags);
  799.  
  800.     ipl = interrupts_disable();
  801.     mutex_lock(&as->lock);
  802.  
  803.     area = find_area_and_lock(as, address);
  804.     if (!area) {
  805.         mutex_unlock(&as->lock);
  806.         interrupts_restore(ipl);
  807.         return ENOENT;
  808.     }
  809.  
  810.     if (area->sh_info || area->backend != &anon_backend) {
  811.         /* Copying shared areas not supported yet */
  812.         /* Copying non-anonymous memory not supported yet */
  813.         mutex_unlock(&area->lock);
  814.         mutex_unlock(&as->lock);
  815.         interrupts_restore(ipl);
  816.         return ENOTSUP;
  817.     }
  818.  
  819.     base = area->base;
  820.  
  821.     /*
  822.      * Start TLB shootdown sequence.
  823.      */
  824.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  825.  
  826.     /*
  827.      * Visit only the pages mapped by used_space B+tree.
  828.      */
  829.     for (cur = area->used_space.leaf_head.next;
  830.         cur != &area->used_space.leaf_head; cur = cur->next) {
  831.         btree_node_t *node;
  832.         unsigned int i;
  833.        
  834.         node = list_get_instance(cur, btree_node_t, leaf_link);
  835.         for (i = 0; i < node->keys; i++) {
  836.             uintptr_t b = node->key[i];
  837.             count_t j;
  838.             pte_t *pte;
  839.            
  840.             for (j = 0; j < (count_t) node->value[i]; j++) {
  841.                 page_table_lock(as, false);
  842.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  843.                 ASSERT(pte && PTE_VALID(pte) &&
  844.                     PTE_PRESENT(pte));
  845.                 old_frame = PTE_GET_FRAME(pte);
  846.  
  847.                 /* Remove old mapping and insert the new one */
  848.                 page_mapping_remove(as, b + j * PAGE_SIZE);
  849.                 page_mapping_insert(as, b + j * PAGE_SIZE,
  850.                     old_frame, page_flags);
  851.  
  852.                 page_table_unlock(as, false);
  853.             }
  854.         }
  855.     }
  856.  
  857.     /*
  858.      * Finish TLB shootdown sequence.
  859.      */
  860.  
  861.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  862.     /*
  863.      * Invalidate potential software translation caches (e.g. TSB on
  864.      * sparc64).
  865.      */
  866.     as_invalidate_translation_cache(as, area->base, area->pages);
  867.     tlb_shootdown_finalize();
  868.    
  869.     mutex_unlock(&area->lock);
  870.     mutex_unlock(&as->lock);
  871.     interrupts_restore(ipl);
  872.  
  873.     return 0;
  874. }
  875.  
  876.  
  877. /** Handle page fault within the current address space.
  878.  *
  879.  * This is the high-level page fault handler. It decides
  880.  * whether the page fault can be resolved by any backend
  881.  * and if so, it invokes the backend to resolve the page
  882.  * fault.
  883.  *
  884.  * Interrupts are assumed disabled.
  885.  *
  886.  * @param page Faulting page.
  887.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  888.  * @param istate Pointer to interrupted state.
  889.  *
  890.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  891.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  892.  */
  893. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  894. {
  895.     pte_t *pte;
  896.     as_area_t *area;
  897.    
  898.     if (!THREAD)
  899.         return AS_PF_FAULT;
  900.        
  901.     ASSERT(AS);
  902.  
  903.     mutex_lock(&AS->lock);
  904.     area = find_area_and_lock(AS, page);   
  905.     if (!area) {
  906.         /*
  907.          * No area contained mapping for 'page'.
  908.          * Signal page fault to low-level handler.
  909.          */
  910.         mutex_unlock(&AS->lock);
  911.         goto page_fault;
  912.     }
  913.  
  914.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  915.         /*
  916.          * The address space area is not fully initialized.
  917.          * Avoid possible race by returning error.
  918.          */
  919.         mutex_unlock(&area->lock);
  920.         mutex_unlock(&AS->lock);
  921.         goto page_fault;       
  922.     }
  923.  
  924.     if (!area->backend || !area->backend->page_fault) {
  925.         /*
  926.          * The address space area is not backed by any backend
  927.          * or the backend cannot handle page faults.
  928.          */
  929.         mutex_unlock(&area->lock);
  930.         mutex_unlock(&AS->lock);
  931.         goto page_fault;       
  932.     }
  933.  
  934.     page_table_lock(AS, false);
  935.    
  936.     /*
  937.      * To avoid race condition between two page faults
  938.      * on the same address, we need to make sure
  939.      * the mapping has not been already inserted.
  940.      */
  941.     if ((pte = page_mapping_find(AS, page))) {
  942.         if (PTE_PRESENT(pte)) {
  943.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  944.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  945.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  946.                 page_table_unlock(AS, false);
  947.                 mutex_unlock(&area->lock);
  948.                 mutex_unlock(&AS->lock);
  949.                 return AS_PF_OK;
  950.             }
  951.         }
  952.     }
  953.    
  954.     /*
  955.      * Resort to the backend page fault handler.
  956.      */
  957.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  958.         page_table_unlock(AS, false);
  959.         mutex_unlock(&area->lock);
  960.         mutex_unlock(&AS->lock);
  961.         goto page_fault;
  962.     }
  963.    
  964.     page_table_unlock(AS, false);
  965.     mutex_unlock(&area->lock);
  966.     mutex_unlock(&AS->lock);
  967.     return AS_PF_OK;
  968.  
  969. page_fault:
  970.     if (THREAD->in_copy_from_uspace) {
  971.         THREAD->in_copy_from_uspace = false;
  972.         istate_set_retaddr(istate,
  973.             (uintptr_t) &memcpy_from_uspace_failover_address);
  974.     } else if (THREAD->in_copy_to_uspace) {
  975.         THREAD->in_copy_to_uspace = false;
  976.         istate_set_retaddr(istate,
  977.             (uintptr_t) &memcpy_to_uspace_failover_address);
  978.     } else {
  979.         return AS_PF_FAULT;
  980.     }
  981.  
  982.     return AS_PF_DEFER;
  983. }
  984.  
  985. /** Switch address spaces.
  986.  *
  987.  * Note that this function cannot sleep as it is essentially a part of
  988.  * scheduling. Sleeping here would lead to deadlock on wakeup. Another
  989.  * thing which is forbidden in this context is locking the address space.
  990.  *
  991.  * When this function is enetered, no spinlocks may be held.
  992.  *
  993.  * @param old Old address space or NULL.
  994.  * @param new New address space.
  995.  */
  996. void as_switch(as_t *old_as, as_t *new_as)
  997. {
  998.     DEADLOCK_PROBE_INIT(p_asidlock);
  999.     preemption_disable();
  1000. retry:
  1001.     (void) interrupts_disable();
  1002.     if (!spinlock_trylock(&asidlock)) {
  1003.         /*
  1004.          * Avoid deadlock with TLB shootdown.
  1005.          * We can enable interrupts here because
  1006.          * preemption is disabled. We should not be
  1007.          * holding any other lock.
  1008.          */
  1009.         (void) interrupts_enable();
  1010.         DEADLOCK_PROBE(p_asidlock, DEADLOCK_THRESHOLD);
  1011.         goto retry;
  1012.     }
  1013.     preemption_enable();
  1014.  
  1015.     /*
  1016.      * First, take care of the old address space.
  1017.      */
  1018.     if (old_as) {
  1019.         ASSERT(old_as->cpu_refcount);
  1020.         if((--old_as->cpu_refcount == 0) && (old_as != AS_KERNEL)) {
  1021.             /*
  1022.              * The old address space is no longer active on
  1023.              * any processor. It can be appended to the
  1024.              * list of inactive address spaces with assigned
  1025.              * ASID.
  1026.              */
  1027.             ASSERT(old_as->asid != ASID_INVALID);
  1028.             list_append(&old_as->inactive_as_with_asid_link,
  1029.                 &inactive_as_with_asid_head);
  1030.         }
  1031.  
  1032.         /*
  1033.          * Perform architecture-specific tasks when the address space
  1034.          * is being removed from the CPU.
  1035.          */
  1036.         as_deinstall_arch(old_as);
  1037.     }
  1038.  
  1039.     /*
  1040.      * Second, prepare the new address space.
  1041.      */
  1042.     if ((new_as->cpu_refcount++ == 0) && (new_as != AS_KERNEL)) {
  1043.         if (new_as->asid != ASID_INVALID)
  1044.             list_remove(&new_as->inactive_as_with_asid_link);
  1045.         else
  1046.             new_as->asid = asid_get();
  1047.     }
  1048. #ifdef AS_PAGE_TABLE
  1049.     SET_PTL0_ADDRESS(new_as->genarch.page_table);
  1050. #endif
  1051.    
  1052.     /*
  1053.      * Perform architecture-specific steps.
  1054.      * (e.g. write ASID to hardware register etc.)
  1055.      */
  1056.     as_install_arch(new_as);
  1057.  
  1058.     spinlock_unlock(&asidlock);
  1059.    
  1060.     AS = new_as;
  1061. }
  1062.  
  1063. /** Convert address space area flags to page flags.
  1064.  *
  1065.  * @param aflags Flags of some address space area.
  1066.  *
  1067.  * @return Flags to be passed to page_mapping_insert().
  1068.  */
  1069. int area_flags_to_page_flags(int aflags)
  1070. {
  1071.     int flags;
  1072.  
  1073.     flags = PAGE_USER | PAGE_PRESENT;
  1074.    
  1075.     if (aflags & AS_AREA_READ)
  1076.         flags |= PAGE_READ;
  1077.        
  1078.     if (aflags & AS_AREA_WRITE)
  1079.         flags |= PAGE_WRITE;
  1080.    
  1081.     if (aflags & AS_AREA_EXEC)
  1082.         flags |= PAGE_EXEC;
  1083.    
  1084.     if (aflags & AS_AREA_CACHEABLE)
  1085.         flags |= PAGE_CACHEABLE;
  1086.        
  1087.     return flags;
  1088. }
  1089.  
  1090. /** Compute flags for virtual address translation subsytem.
  1091.  *
  1092.  * The address space area must be locked.
  1093.  * Interrupts must be disabled.
  1094.  *
  1095.  * @param a Address space area.
  1096.  *
  1097.  * @return Flags to be used in page_mapping_insert().
  1098.  */
  1099. int as_area_get_flags(as_area_t *a)
  1100. {
  1101.     return area_flags_to_page_flags(a->flags);
  1102. }
  1103.  
  1104. /** Create page table.
  1105.  *
  1106.  * Depending on architecture, create either address space
  1107.  * private or global page table.
  1108.  *
  1109.  * @param flags Flags saying whether the page table is for kernel address space.
  1110.  *
  1111.  * @return First entry of the page table.
  1112.  */
  1113. pte_t *page_table_create(int flags)
  1114. {
  1115. #ifdef __OBJC__
  1116.     return [as_t page_table_create: flags];
  1117. #else
  1118.     ASSERT(as_operations);
  1119.     ASSERT(as_operations->page_table_create);
  1120.    
  1121.     return as_operations->page_table_create(flags);
  1122. #endif
  1123. }
  1124.  
  1125. /** Destroy page table.
  1126.  *
  1127.  * Destroy page table in architecture specific way.
  1128.  *
  1129.  * @param page_table Physical address of PTL0.
  1130.  */
  1131. void page_table_destroy(pte_t *page_table)
  1132. {
  1133. #ifdef __OBJC__
  1134.     return [as_t page_table_destroy: page_table];
  1135. #else
  1136.     ASSERT(as_operations);
  1137.     ASSERT(as_operations->page_table_destroy);
  1138.    
  1139.     as_operations->page_table_destroy(page_table);
  1140. #endif
  1141. }
  1142.  
  1143. /** Lock page table.
  1144.  *
  1145.  * This function should be called before any page_mapping_insert(),
  1146.  * page_mapping_remove() and page_mapping_find().
  1147.  *
  1148.  * Locking order is such that address space areas must be locked
  1149.  * prior to this call. Address space can be locked prior to this
  1150.  * call in which case the lock argument is false.
  1151.  *
  1152.  * @param as Address space.
  1153.  * @param lock If false, do not attempt to lock as->lock.
  1154.  */
  1155. void page_table_lock(as_t *as, bool lock)
  1156. {
  1157. #ifdef __OBJC__
  1158.     [as page_table_lock: lock];
  1159. #else
  1160.     ASSERT(as_operations);
  1161.     ASSERT(as_operations->page_table_lock);
  1162.    
  1163.     as_operations->page_table_lock(as, lock);
  1164. #endif
  1165. }
  1166.  
  1167. /** Unlock page table.
  1168.  *
  1169.  * @param as Address space.
  1170.  * @param unlock If false, do not attempt to unlock as->lock.
  1171.  */
  1172. void page_table_unlock(as_t *as, bool unlock)
  1173. {
  1174. #ifdef __OBJC__
  1175.     [as page_table_unlock: unlock];
  1176. #else
  1177.     ASSERT(as_operations);
  1178.     ASSERT(as_operations->page_table_unlock);
  1179.    
  1180.     as_operations->page_table_unlock(as, unlock);
  1181. #endif
  1182. }
  1183.  
  1184.  
  1185. /** Find address space area and lock it.
  1186.  *
  1187.  * The address space must be locked and interrupts must be disabled.
  1188.  *
  1189.  * @param as Address space.
  1190.  * @param va Virtual address.
  1191.  *
  1192.  * @return Locked address space area containing va on success or NULL on
  1193.  *     failure.
  1194.  */
  1195. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  1196. {
  1197.     as_area_t *a;
  1198.     btree_node_t *leaf, *lnode;
  1199.     unsigned int i;
  1200.    
  1201.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  1202.     if (a) {
  1203.         /* va is the base address of an address space area */
  1204.         mutex_lock(&a->lock);
  1205.         return a;
  1206.     }
  1207.    
  1208.     /*
  1209.      * Search the leaf node and the righmost record of its left neighbour
  1210.      * to find out whether this is a miss or va belongs to an address
  1211.      * space area found there.
  1212.      */
  1213.    
  1214.     /* First, search the leaf node itself. */
  1215.     for (i = 0; i < leaf->keys; i++) {
  1216.         a = (as_area_t *) leaf->value[i];
  1217.         mutex_lock(&a->lock);
  1218.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  1219.             return a;
  1220.         }
  1221.         mutex_unlock(&a->lock);
  1222.     }
  1223.  
  1224.     /*
  1225.      * Second, locate the left neighbour and test its last record.
  1226.      * Because of its position in the B+tree, it must have base < va.
  1227.      */
  1228.     lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf);
  1229.     if (lnode) {
  1230.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  1231.         mutex_lock(&a->lock);
  1232.         if (va < a->base + a->pages * PAGE_SIZE) {
  1233.             return a;
  1234.         }
  1235.         mutex_unlock(&a->lock);
  1236.     }
  1237.  
  1238.     return NULL;
  1239. }
  1240.  
  1241. /** Check area conflicts with other areas.
  1242.  *
  1243.  * The address space must be locked and interrupts must be disabled.
  1244.  *
  1245.  * @param as Address space.
  1246.  * @param va Starting virtual address of the area being tested.
  1247.  * @param size Size of the area being tested.
  1248.  * @param avoid_area Do not touch this area.
  1249.  *
  1250.  * @return True if there is no conflict, false otherwise.
  1251.  */
  1252. bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  1253.               as_area_t *avoid_area)
  1254. {
  1255.     as_area_t *a;
  1256.     btree_node_t *leaf, *node;
  1257.     unsigned int i;
  1258.    
  1259.     /*
  1260.      * We don't want any area to have conflicts with NULL page.
  1261.      */
  1262.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1263.         return false;
  1264.    
  1265.     /*
  1266.      * The leaf node is found in O(log n), where n is proportional to
  1267.      * the number of address space areas belonging to as.
  1268.      * The check for conflicts is then attempted on the rightmost
  1269.      * record in the left neighbour, the leftmost record in the right
  1270.      * neighbour and all records in the leaf node itself.
  1271.      */
  1272.    
  1273.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1274.         if (a != avoid_area)
  1275.             return false;
  1276.     }
  1277.    
  1278.     /* First, check the two border cases. */
  1279.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1280.         a = (as_area_t *) node->value[node->keys - 1];
  1281.         mutex_lock(&a->lock);
  1282.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1283.             mutex_unlock(&a->lock);
  1284.             return false;
  1285.         }
  1286.         mutex_unlock(&a->lock);
  1287.     }
  1288.     node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf);
  1289.     if (node) {
  1290.         a = (as_area_t *) node->value[0];
  1291.         mutex_lock(&a->lock);
  1292.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1293.             mutex_unlock(&a->lock);
  1294.             return false;
  1295.         }
  1296.         mutex_unlock(&a->lock);
  1297.     }
  1298.    
  1299.     /* Second, check the leaf node. */
  1300.     for (i = 0; i < leaf->keys; i++) {
  1301.         a = (as_area_t *) leaf->value[i];
  1302.    
  1303.         if (a == avoid_area)
  1304.             continue;
  1305.    
  1306.         mutex_lock(&a->lock);
  1307.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1308.             mutex_unlock(&a->lock);
  1309.             return false;
  1310.         }
  1311.         mutex_unlock(&a->lock);
  1312.     }
  1313.  
  1314.     /*
  1315.      * So far, the area does not conflict with other areas.
  1316.      * Check if it doesn't conflict with kernel address space.
  1317.      */  
  1318.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1319.         return !overlaps(va, size,
  1320.             KERNEL_ADDRESS_SPACE_START,
  1321.             KERNEL_ADDRESS_SPACE_END - KERNEL_ADDRESS_SPACE_START);
  1322.     }
  1323.  
  1324.     return true;
  1325. }
  1326.  
  1327. /** Return size of the address space area with given base.
  1328.  *
  1329.  * @param base      Arbitrary address insede the address space area.
  1330.  *
  1331.  * @return      Size of the address space area in bytes or zero if it
  1332.  *          does not exist.
  1333.  */
  1334. size_t as_area_get_size(uintptr_t base)
  1335. {
  1336.     ipl_t ipl;
  1337.     as_area_t *src_area;
  1338.     size_t size;
  1339.  
  1340.     ipl = interrupts_disable();
  1341.     src_area = find_area_and_lock(AS, base);
  1342.     if (src_area){
  1343.         size = src_area->pages * PAGE_SIZE;
  1344.         mutex_unlock(&src_area->lock);
  1345.     } else {
  1346.         size = 0;
  1347.     }
  1348.     interrupts_restore(ipl);
  1349.     return size;
  1350. }
  1351.  
  1352. /** Mark portion of address space area as used.
  1353.  *
  1354.  * The address space area must be already locked.
  1355.  *
  1356.  * @param a Address space area.
  1357.  * @param page First page to be marked.
  1358.  * @param count Number of page to be marked.
  1359.  *
  1360.  * @return 0 on failure and 1 on success.
  1361.  */
  1362. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1363. {
  1364.     btree_node_t *leaf, *node;
  1365.     count_t pages;
  1366.     unsigned int i;
  1367.  
  1368.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1369.     ASSERT(count);
  1370.  
  1371.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1372.     if (pages) {
  1373.         /*
  1374.          * We hit the beginning of some used space.
  1375.          */
  1376.         return 0;
  1377.     }
  1378.  
  1379.     if (!leaf->keys) {
  1380.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1381.         return 1;
  1382.     }
  1383.  
  1384.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1385.     if (node) {
  1386.         uintptr_t left_pg = node->key[node->keys - 1];
  1387.         uintptr_t right_pg = leaf->key[0];
  1388.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1389.         count_t right_cnt = (count_t) leaf->value[0];
  1390.        
  1391.         /*
  1392.          * Examine the possibility that the interval fits
  1393.          * somewhere between the rightmost interval of
  1394.          * the left neigbour and the first interval of the leaf.
  1395.          */
  1396.          
  1397.         if (page >= right_pg) {
  1398.             /* Do nothing. */
  1399.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1400.             left_cnt * PAGE_SIZE)) {
  1401.             /* The interval intersects with the left interval. */
  1402.             return 0;
  1403.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1404.             right_cnt * PAGE_SIZE)) {
  1405.             /* The interval intersects with the right interval. */
  1406.             return 0;          
  1407.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1408.             (page + count * PAGE_SIZE == right_pg)) {
  1409.             /*
  1410.              * The interval can be added by merging the two already
  1411.              * present intervals.
  1412.              */
  1413.             node->value[node->keys - 1] += count + right_cnt;
  1414.             btree_remove(&a->used_space, right_pg, leaf);
  1415.             return 1;
  1416.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1417.             /*
  1418.              * The interval can be added by simply growing the left
  1419.              * interval.
  1420.              */
  1421.             node->value[node->keys - 1] += count;
  1422.             return 1;
  1423.         } else if (page + count * PAGE_SIZE == right_pg) {
  1424.             /*
  1425.              * The interval can be addded by simply moving base of
  1426.              * the right interval down and increasing its size
  1427.              * accordingly.
  1428.              */
  1429.             leaf->value[0] += count;
  1430.             leaf->key[0] = page;
  1431.             return 1;
  1432.         } else {
  1433.             /*
  1434.              * The interval is between both neigbouring intervals,
  1435.              * but cannot be merged with any of them.
  1436.              */
  1437.             btree_insert(&a->used_space, page, (void *) count,
  1438.                 leaf);
  1439.             return 1;
  1440.         }
  1441.     } else if (page < leaf->key[0]) {
  1442.         uintptr_t right_pg = leaf->key[0];
  1443.         count_t right_cnt = (count_t) leaf->value[0];
  1444.    
  1445.         /*
  1446.          * Investigate the border case in which the left neighbour does
  1447.          * not exist but the interval fits from the left.
  1448.          */
  1449.          
  1450.         if (overlaps(page, count * PAGE_SIZE, right_pg,
  1451.             right_cnt * PAGE_SIZE)) {
  1452.             /* The interval intersects with the right interval. */
  1453.             return 0;
  1454.         } else if (page + count * PAGE_SIZE == right_pg) {
  1455.             /*
  1456.              * The interval can be added by moving the base of the
  1457.              * right interval down and increasing its size
  1458.              * accordingly.
  1459.              */
  1460.             leaf->key[0] = page;
  1461.             leaf->value[0] += count;
  1462.             return 1;
  1463.         } else {
  1464.             /*
  1465.              * The interval doesn't adjoin with the right interval.
  1466.              * It must be added individually.
  1467.              */
  1468.             btree_insert(&a->used_space, page, (void *) count,
  1469.                 leaf);
  1470.             return 1;
  1471.         }
  1472.     }
  1473.  
  1474.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1475.     if (node) {
  1476.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1477.         uintptr_t right_pg = node->key[0];
  1478.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1479.         count_t right_cnt = (count_t) node->value[0];
  1480.        
  1481.         /*
  1482.          * Examine the possibility that the interval fits
  1483.          * somewhere between the leftmost interval of
  1484.          * the right neigbour and the last interval of the leaf.
  1485.          */
  1486.  
  1487.         if (page < left_pg) {
  1488.             /* Do nothing. */
  1489.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1490.             left_cnt * PAGE_SIZE)) {
  1491.             /* The interval intersects with the left interval. */
  1492.             return 0;
  1493.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1494.             right_cnt * PAGE_SIZE)) {
  1495.             /* The interval intersects with the right interval. */
  1496.             return 0;          
  1497.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1498.             (page + count * PAGE_SIZE == right_pg)) {
  1499.             /*
  1500.              * The interval can be added by merging the two already
  1501.              * present intervals.
  1502.              * */
  1503.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1504.             btree_remove(&a->used_space, right_pg, node);
  1505.             return 1;
  1506.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1507.             /*
  1508.              * The interval can be added by simply growing the left
  1509.              * interval.
  1510.              * */
  1511.             leaf->value[leaf->keys - 1] +=  count;
  1512.             return 1;
  1513.         } else if (page + count * PAGE_SIZE == right_pg) {
  1514.             /*
  1515.              * The interval can be addded by simply moving base of
  1516.              * the right interval down and increasing its size
  1517.              * accordingly.
  1518.              */
  1519.             node->value[0] += count;
  1520.             node->key[0] = page;
  1521.             return 1;
  1522.         } else {
  1523.             /*
  1524.              * The interval is between both neigbouring intervals,
  1525.              * but cannot be merged with any of them.
  1526.              */
  1527.             btree_insert(&a->used_space, page, (void *) count,
  1528.                 leaf);
  1529.             return 1;
  1530.         }
  1531.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1532.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1533.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1534.    
  1535.         /*
  1536.          * Investigate the border case in which the right neighbour
  1537.          * does not exist but the interval fits from the right.
  1538.          */
  1539.          
  1540.         if (overlaps(page, count * PAGE_SIZE, left_pg,
  1541.             left_cnt * PAGE_SIZE)) {
  1542.             /* The interval intersects with the left interval. */
  1543.             return 0;
  1544.         } else if (left_pg + left_cnt * PAGE_SIZE == page) {
  1545.             /*
  1546.              * The interval can be added by growing the left
  1547.              * interval.
  1548.              */
  1549.             leaf->value[leaf->keys - 1] += count;
  1550.             return 1;
  1551.         } else {
  1552.             /*
  1553.              * The interval doesn't adjoin with the left interval.
  1554.              * It must be added individually.
  1555.              */
  1556.             btree_insert(&a->used_space, page, (void *) count,
  1557.                 leaf);
  1558.             return 1;
  1559.         }
  1560.     }
  1561.    
  1562.     /*
  1563.      * Note that if the algorithm made it thus far, the interval can fit
  1564.      * only between two other intervals of the leaf. The two border cases
  1565.      * were already resolved.
  1566.      */
  1567.     for (i = 1; i < leaf->keys; i++) {
  1568.         if (page < leaf->key[i]) {
  1569.             uintptr_t left_pg = leaf->key[i - 1];
  1570.             uintptr_t right_pg = leaf->key[i];
  1571.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1572.             count_t right_cnt = (count_t) leaf->value[i];
  1573.  
  1574.             /*
  1575.              * The interval fits between left_pg and right_pg.
  1576.              */
  1577.  
  1578.             if (overlaps(page, count * PAGE_SIZE, left_pg,
  1579.                 left_cnt * PAGE_SIZE)) {
  1580.                 /*
  1581.                  * The interval intersects with the left
  1582.                  * interval.
  1583.                  */
  1584.                 return 0;
  1585.             } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1586.                 right_cnt * PAGE_SIZE)) {
  1587.                 /*
  1588.                  * The interval intersects with the right
  1589.                  * interval.
  1590.                  */
  1591.                 return 0;          
  1592.             } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1593.                 (page + count * PAGE_SIZE == right_pg)) {
  1594.                 /*
  1595.                  * The interval can be added by merging the two
  1596.                  * already present intervals.
  1597.                  */
  1598.                 leaf->value[i - 1] += count + right_cnt;
  1599.                 btree_remove(&a->used_space, right_pg, leaf);
  1600.                 return 1;
  1601.             } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1602.                 /*
  1603.                  * The interval can be added by simply growing
  1604.                  * the left interval.
  1605.                  */
  1606.                 leaf->value[i - 1] += count;
  1607.                 return 1;
  1608.             } else if (page + count * PAGE_SIZE == right_pg) {
  1609.                 /*
  1610.                      * The interval can be addded by simply moving
  1611.                  * base of the right interval down and
  1612.                  * increasing its size accordingly.
  1613.                  */
  1614.                 leaf->value[i] += count;
  1615.                 leaf->key[i] = page;
  1616.                 return 1;
  1617.             } else {
  1618.                 /*
  1619.                  * The interval is between both neigbouring
  1620.                  * intervals, but cannot be merged with any of
  1621.                  * them.
  1622.                  */
  1623.                 btree_insert(&a->used_space, page,
  1624.                     (void *) count, leaf);
  1625.                 return 1;
  1626.             }
  1627.         }
  1628.     }
  1629.  
  1630.     panic("Inconsistency detected while adding %" PRIc " pages of used space at "
  1631.         "%p.\n", count, page);
  1632. }
  1633.  
  1634. /** Mark portion of address space area as unused.
  1635.  *
  1636.  * The address space area must be already locked.
  1637.  *
  1638.  * @param a Address space area.
  1639.  * @param page First page to be marked.
  1640.  * @param count Number of page to be marked.
  1641.  *
  1642.  * @return 0 on failure and 1 on success.
  1643.  */
  1644. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1645. {
  1646.     btree_node_t *leaf, *node;
  1647.     count_t pages;
  1648.     unsigned int i;
  1649.  
  1650.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1651.     ASSERT(count);
  1652.  
  1653.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1654.     if (pages) {
  1655.         /*
  1656.          * We are lucky, page is the beginning of some interval.
  1657.          */
  1658.         if (count > pages) {
  1659.             return 0;
  1660.         } else if (count == pages) {
  1661.             btree_remove(&a->used_space, page, leaf);
  1662.             return 1;
  1663.         } else {
  1664.             /*
  1665.              * Find the respective interval.
  1666.              * Decrease its size and relocate its start address.
  1667.              */
  1668.             for (i = 0; i < leaf->keys; i++) {
  1669.                 if (leaf->key[i] == page) {
  1670.                     leaf->key[i] += count * PAGE_SIZE;
  1671.                     leaf->value[i] -= count;
  1672.                     return 1;
  1673.                 }
  1674.             }
  1675.             goto error;
  1676.         }
  1677.     }
  1678.  
  1679.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1680.     if (node && page < leaf->key[0]) {
  1681.         uintptr_t left_pg = node->key[node->keys - 1];
  1682.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1683.  
  1684.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1685.             count * PAGE_SIZE)) {
  1686.             if (page + count * PAGE_SIZE ==
  1687.                 left_pg + left_cnt * PAGE_SIZE) {
  1688.                 /*
  1689.                  * The interval is contained in the rightmost
  1690.                  * interval of the left neighbour and can be
  1691.                  * removed by updating the size of the bigger
  1692.                  * interval.
  1693.                  */
  1694.                 node->value[node->keys - 1] -= count;
  1695.                 return 1;
  1696.             } else if (page + count * PAGE_SIZE <
  1697.                 left_pg + left_cnt*PAGE_SIZE) {
  1698.                 count_t new_cnt;
  1699.                
  1700.                 /*
  1701.                  * The interval is contained in the rightmost
  1702.                  * interval of the left neighbour but its
  1703.                  * removal requires both updating the size of
  1704.                  * the original interval and also inserting a
  1705.                  * new interval.
  1706.                  */
  1707.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1708.                     (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1709.                 node->value[node->keys - 1] -= count + new_cnt;
  1710.                 btree_insert(&a->used_space, page +
  1711.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1712.                 return 1;
  1713.             }
  1714.         }
  1715.         return 0;
  1716.     } else if (page < leaf->key[0]) {
  1717.         return 0;
  1718.     }
  1719.    
  1720.     if (page > leaf->key[leaf->keys - 1]) {
  1721.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1722.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1723.  
  1724.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1725.             count * PAGE_SIZE)) {
  1726.             if (page + count * PAGE_SIZE ==
  1727.                 left_pg + left_cnt * PAGE_SIZE) {
  1728.                 /*
  1729.                  * The interval is contained in the rightmost
  1730.                  * interval of the leaf and can be removed by
  1731.                  * updating the size of the bigger interval.
  1732.                  */
  1733.                 leaf->value[leaf->keys - 1] -= count;
  1734.                 return 1;
  1735.             } else if (page + count * PAGE_SIZE < left_pg +
  1736.                 left_cnt * PAGE_SIZE) {
  1737.                 count_t new_cnt;
  1738.                
  1739.                 /*
  1740.                  * The interval is contained in the rightmost
  1741.                  * interval of the leaf but its removal
  1742.                  * requires both updating the size of the
  1743.                  * original interval and also inserting a new
  1744.                  * interval.
  1745.                  */
  1746.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1747.                     (page + count * PAGE_SIZE)) >> PAGE_WIDTH;
  1748.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1749.                 btree_insert(&a->used_space, page +
  1750.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1751.                 return 1;
  1752.             }
  1753.         }
  1754.         return 0;
  1755.     }  
  1756.    
  1757.     /*
  1758.      * The border cases have been already resolved.
  1759.      * Now the interval can be only between intervals of the leaf.
  1760.      */
  1761.     for (i = 1; i < leaf->keys - 1; i++) {
  1762.         if (page < leaf->key[i]) {
  1763.             uintptr_t left_pg = leaf->key[i - 1];
  1764.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1765.  
  1766.             /*
  1767.              * Now the interval is between intervals corresponding
  1768.              * to (i - 1) and i.
  1769.              */
  1770.             if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1771.                 count * PAGE_SIZE)) {
  1772.                 if (page + count * PAGE_SIZE ==
  1773.                     left_pg + left_cnt*PAGE_SIZE) {
  1774.                     /*
  1775.                      * The interval is contained in the
  1776.                      * interval (i - 1) of the leaf and can
  1777.                      * be removed by updating the size of
  1778.                      * the bigger interval.
  1779.                      */
  1780.                     leaf->value[i - 1] -= count;
  1781.                     return 1;
  1782.                 } else if (page + count * PAGE_SIZE <
  1783.                     left_pg + left_cnt * PAGE_SIZE) {
  1784.                     count_t new_cnt;
  1785.                
  1786.                     /*
  1787.                      * The interval is contained in the
  1788.                      * interval (i - 1) of the leaf but its
  1789.                      * removal requires both updating the
  1790.                      * size of the original interval and
  1791.                      * also inserting a new interval.
  1792.                      */
  1793.                     new_cnt = ((left_pg +
  1794.                         left_cnt * PAGE_SIZE) -
  1795.                         (page + count * PAGE_SIZE)) >>
  1796.                         PAGE_WIDTH;
  1797.                     leaf->value[i - 1] -= count + new_cnt;
  1798.                     btree_insert(&a->used_space, page +
  1799.                         count * PAGE_SIZE, (void *) new_cnt,
  1800.                         leaf);
  1801.                     return 1;
  1802.                 }
  1803.             }
  1804.             return 0;
  1805.         }
  1806.     }
  1807.  
  1808. error:
  1809.     panic("Inconsistency detected while removing %" PRIc " pages of used space "
  1810.         "from %p.\n", count, page);
  1811. }
  1812.  
  1813. /** Remove reference to address space area share info.
  1814.  *
  1815.  * If the reference count drops to 0, the sh_info is deallocated.
  1816.  *
  1817.  * @param sh_info Pointer to address space area share info.
  1818.  */
  1819. void sh_info_remove_reference(share_info_t *sh_info)
  1820. {
  1821.     bool dealloc = false;
  1822.  
  1823.     mutex_lock(&sh_info->lock);
  1824.     ASSERT(sh_info->refcount);
  1825.     if (--sh_info->refcount == 0) {
  1826.         dealloc = true;
  1827.         link_t *cur;
  1828.        
  1829.         /*
  1830.          * Now walk carefully the pagemap B+tree and free/remove
  1831.          * reference from all frames found there.
  1832.          */
  1833.         for (cur = sh_info->pagemap.leaf_head.next;
  1834.             cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1835.             btree_node_t *node;
  1836.             unsigned int i;
  1837.            
  1838.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1839.             for (i = 0; i < node->keys; i++)
  1840.                 frame_free((uintptr_t) node->value[i]);
  1841.         }
  1842.        
  1843.     }
  1844.     mutex_unlock(&sh_info->lock);
  1845.    
  1846.     if (dealloc) {
  1847.         btree_destroy(&sh_info->pagemap);
  1848.         free(sh_info);
  1849.     }
  1850. }
  1851.  
  1852. /*
  1853.  * Address space related syscalls.
  1854.  */
  1855.  
  1856. /** Wrapper for as_area_create(). */
  1857. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1858. {
  1859.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address,
  1860.         AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1861.         return (unative_t) address;
  1862.     else
  1863.         return (unative_t) -1;
  1864. }
  1865.  
  1866. /** Wrapper for as_area_resize(). */
  1867. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1868. {
  1869.     return (unative_t) as_area_resize(AS, address, size, 0);
  1870. }
  1871.  
  1872. /** Wrapper for as_area_change_flags(). */
  1873. unative_t sys_as_area_change_flags(uintptr_t address, int flags)
  1874. {
  1875.     return (unative_t) as_area_change_flags(AS, flags, address);
  1876. }
  1877.  
  1878. /** Wrapper for as_area_destroy(). */
  1879. unative_t sys_as_area_destroy(uintptr_t address)
  1880. {
  1881.     return (unative_t) as_area_destroy(AS, address);
  1882. }
  1883.  
  1884. /** Print out information about address space.
  1885.  *
  1886.  * @param as Address space.
  1887.  */
  1888. void as_print(as_t *as)
  1889. {
  1890.     ipl_t ipl;
  1891.    
  1892.     ipl = interrupts_disable();
  1893.     mutex_lock(&as->lock);
  1894.    
  1895.     /* print out info about address space areas */
  1896.     link_t *cur;
  1897.     for (cur = as->as_area_btree.leaf_head.next;
  1898.         cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  1899.         btree_node_t *node;
  1900.        
  1901.         node = list_get_instance(cur, btree_node_t, leaf_link);
  1902.        
  1903.         unsigned int i;
  1904.         for (i = 0; i < node->keys; i++) {
  1905.             as_area_t *area = node->value[i];
  1906.        
  1907.             mutex_lock(&area->lock);
  1908.             printf("as_area: %p, base=%p, pages=%" PRIc " (%p - %p)\n",
  1909.                 area, area->base, area->pages, area->base,
  1910.                 area->base + FRAMES2SIZE(area->pages));
  1911.             mutex_unlock(&area->lock);
  1912.         }
  1913.     }
  1914.    
  1915.     mutex_unlock(&as->lock);
  1916.     interrupts_restore(ipl);
  1917. }
  1918.  
  1919. /** @}
  1920.  */
  1921.