Subversion Repositories HelenOS

Rev

Rev 2745 | Rev 3013 | 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,
  457.                         c - i))
  458.                         panic("Could not remove used "
  459.                             "space.\n");
  460.                 } else {
  461.                     /*
  462.                      * The interval of used space can be
  463.                      * completely removed.
  464.                      */
  465.                     if (!used_space_remove(area, b, c))
  466.                         panic("Could not remove used "
  467.                             "space.\n");
  468.                 }
  469.            
  470.                 for (; i < c; i++) {
  471.                     pte_t *pte;
  472.            
  473.                     page_table_lock(as, false);
  474.                     pte = page_mapping_find(as, b +
  475.                         i * PAGE_SIZE);
  476.                     ASSERT(pte && PTE_VALID(pte) &&
  477.                         PTE_PRESENT(pte));
  478.                     if (area->backend &&
  479.                         area->backend->frame_free) {
  480.                         area->backend->frame_free(area,
  481.                             b + i * PAGE_SIZE,
  482.                             PTE_GET_FRAME(pte));
  483.                     }
  484.                     page_mapping_remove(as, b +
  485.                         i * PAGE_SIZE);
  486.                     page_table_unlock(as, false);
  487.                 }
  488.             }
  489.         }
  490.  
  491.         /*
  492.          * Finish TLB shootdown sequence.
  493.          */
  494.  
  495.         tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,
  496.             area->pages - pages);
  497.         /*
  498.          * Invalidate software translation caches (e.g. TSB on sparc64).
  499.          */
  500.         as_invalidate_translation_cache(as, area->base +
  501.             pages * PAGE_SIZE, area->pages - pages);
  502.         tlb_shootdown_finalize();
  503.        
  504.     } else {
  505.         /*
  506.          * Growing the area.
  507.          * Check for overlaps with other address space areas.
  508.          */
  509.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE,
  510.             area)) {
  511.             mutex_unlock(&area->lock);
  512.             mutex_unlock(&as->lock);       
  513.             interrupts_restore(ipl);
  514.             return EADDRNOTAVAIL;
  515.         }
  516.     }
  517.  
  518.     area->pages = pages;
  519.    
  520.     mutex_unlock(&area->lock);
  521.     mutex_unlock(&as->lock);
  522.     interrupts_restore(ipl);
  523.  
  524.     return 0;
  525. }
  526.  
  527. /** Destroy address space area.
  528.  *
  529.  * @param as Address space.
  530.  * @param address Address withing the area to be deleted.
  531.  *
  532.  * @return Zero on success or a value from @ref errno.h on failure.
  533.  */
  534. int as_area_destroy(as_t *as, uintptr_t address)
  535. {
  536.     as_area_t *area;
  537.     uintptr_t base;
  538.     link_t *cur;
  539.     ipl_t ipl;
  540.  
  541.     ipl = interrupts_disable();
  542.     mutex_lock(&as->lock);
  543.  
  544.     area = find_area_and_lock(as, address);
  545.     if (!area) {
  546.         mutex_unlock(&as->lock);
  547.         interrupts_restore(ipl);
  548.         return ENOENT;
  549.     }
  550.  
  551.     base = area->base;
  552.  
  553.     /*
  554.      * Start TLB shootdown sequence.
  555.      */
  556.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  557.  
  558.     /*
  559.      * Visit only the pages mapped by used_space B+tree.
  560.      */
  561.     for (cur = area->used_space.leaf_head.next;
  562.         cur != &area->used_space.leaf_head; cur = cur->next) {
  563.         btree_node_t *node;
  564.         unsigned int i;
  565.        
  566.         node = list_get_instance(cur, btree_node_t, leaf_link);
  567.         for (i = 0; i < node->keys; i++) {
  568.             uintptr_t b = node->key[i];
  569.             count_t j;
  570.             pte_t *pte;
  571.            
  572.             for (j = 0; j < (count_t) node->value[i]; j++) {
  573.                 page_table_lock(as, false);
  574.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  575.                 ASSERT(pte && PTE_VALID(pte) &&
  576.                     PTE_PRESENT(pte));
  577.                 if (area->backend &&
  578.                     area->backend->frame_free) {
  579.                     area->backend->frame_free(area, b +
  580.                         j * PAGE_SIZE, PTE_GET_FRAME(pte));
  581.                 }
  582.                 page_mapping_remove(as, b + j * PAGE_SIZE);            
  583.                 page_table_unlock(as, false);
  584.             }
  585.         }
  586.     }
  587.  
  588.     /*
  589.      * Finish TLB shootdown sequence.
  590.      */
  591.  
  592.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  593.     /*
  594.      * Invalidate potential software translation caches (e.g. TSB on
  595.      * sparc64).
  596.      */
  597.     as_invalidate_translation_cache(as, area->base, area->pages);
  598.     tlb_shootdown_finalize();
  599.    
  600.     btree_destroy(&area->used_space);
  601.  
  602.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  603.    
  604.     if (area->sh_info)
  605.         sh_info_remove_reference(area->sh_info);
  606.        
  607.     mutex_unlock(&area->lock);
  608.  
  609.     /*
  610.      * Remove the empty area from address space.
  611.      */
  612.     btree_remove(&as->as_area_btree, base, NULL);
  613.    
  614.     free(area);
  615.    
  616.     mutex_unlock(&as->lock);
  617.     interrupts_restore(ipl);
  618.     return 0;
  619. }
  620.  
  621. /** Share address space area with another or the same address space.
  622.  *
  623.  * Address space area mapping is shared with a new address space area.
  624.  * If the source address space area has not been shared so far,
  625.  * a new sh_info is created. The new address space area simply gets the
  626.  * sh_info of the source area. The process of duplicating the
  627.  * mapping is done through the backend share function.
  628.  *
  629.  * @param src_as Pointer to source address space.
  630.  * @param src_base Base address of the source address space area.
  631.  * @param acc_size Expected size of the source area.
  632.  * @param dst_as Pointer to destination address space.
  633.  * @param dst_base Target base address.
  634.  * @param dst_flags_mask Destination address space area flags mask.
  635.  *
  636.  * @return Zero on success or ENOENT if there is no such task or if there is no
  637.  * such address space area, EPERM if there was a problem in accepting the area
  638.  * or ENOMEM if there was a problem in allocating destination address space
  639.  * area. ENOTSUP is returned if the address space area backend does not support
  640.  * sharing.
  641.  */
  642. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  643.     as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  644. {
  645.     ipl_t ipl;
  646.     int src_flags;
  647.     size_t src_size;
  648.     as_area_t *src_area, *dst_area;
  649.     share_info_t *sh_info;
  650.     mem_backend_t *src_backend;
  651.     mem_backend_data_t src_backend_data;
  652.    
  653.     ipl = interrupts_disable();
  654.     mutex_lock(&src_as->lock);
  655.     src_area = find_area_and_lock(src_as, src_base);
  656.     if (!src_area) {
  657.         /*
  658.          * Could not find the source address space area.
  659.          */
  660.         mutex_unlock(&src_as->lock);
  661.         interrupts_restore(ipl);
  662.         return ENOENT;
  663.     }
  664.  
  665.     if (!src_area->backend || !src_area->backend->share) {
  666.         /*
  667.          * There is no backend or the backend does not
  668.          * know how to share the area.
  669.          */
  670.         mutex_unlock(&src_area->lock);
  671.         mutex_unlock(&src_as->lock);
  672.         interrupts_restore(ipl);
  673.         return ENOTSUP;
  674.     }
  675.    
  676.     src_size = src_area->pages * PAGE_SIZE;
  677.     src_flags = src_area->flags;
  678.     src_backend = src_area->backend;
  679.     src_backend_data = src_area->backend_data;
  680.  
  681.     /* Share the cacheable flag from the original mapping */
  682.     if (src_flags & AS_AREA_CACHEABLE)
  683.         dst_flags_mask |= AS_AREA_CACHEABLE;
  684.  
  685.     if (src_size != acc_size ||
  686.         (src_flags & dst_flags_mask) != dst_flags_mask) {
  687.         mutex_unlock(&src_area->lock);
  688.         mutex_unlock(&src_as->lock);
  689.         interrupts_restore(ipl);
  690.         return EPERM;
  691.     }
  692.  
  693.     /*
  694.      * Now we are committed to sharing the area.
  695.      * First, prepare the area for sharing.
  696.      * Then it will be safe to unlock it.
  697.      */
  698.     sh_info = src_area->sh_info;
  699.     if (!sh_info) {
  700.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  701.         mutex_initialize(&sh_info->lock);
  702.         sh_info->refcount = 2;
  703.         btree_create(&sh_info->pagemap);
  704.         src_area->sh_info = sh_info;
  705.         /*
  706.          * Call the backend to setup sharing.
  707.          */
  708.         src_area->backend->share(src_area);
  709.     } else {
  710.         mutex_lock(&sh_info->lock);
  711.         sh_info->refcount++;
  712.         mutex_unlock(&sh_info->lock);
  713.     }
  714.  
  715.     mutex_unlock(&src_area->lock);
  716.     mutex_unlock(&src_as->lock);
  717.  
  718.     /*
  719.      * Create copy of the source address space area.
  720.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  721.      * attribute set which prevents race condition with
  722.      * preliminary as_page_fault() calls.
  723.      * The flags of the source area are masked against dst_flags_mask
  724.      * to support sharing in less privileged mode.
  725.      */
  726.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  727.         AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  728.     if (!dst_area) {
  729.         /*
  730.          * Destination address space area could not be created.
  731.          */
  732.         sh_info_remove_reference(sh_info);
  733.        
  734.         interrupts_restore(ipl);
  735.         return ENOMEM;
  736.     }
  737.  
  738.     /*
  739.      * Now the destination address space area has been
  740.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  741.      * attribute and set the sh_info.
  742.      */
  743.     mutex_lock(&dst_as->lock); 
  744.     mutex_lock(&dst_area->lock);
  745.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  746.     dst_area->sh_info = sh_info;
  747.     mutex_unlock(&dst_area->lock);
  748.     mutex_unlock(&dst_as->lock);   
  749.  
  750.     interrupts_restore(ipl);
  751.    
  752.     return 0;
  753. }
  754.  
  755. /** Check access mode for address space area.
  756.  *
  757.  * The address space area must be locked prior to this call.
  758.  *
  759.  * @param area Address space area.
  760.  * @param access Access mode.
  761.  *
  762.  * @return False if access violates area's permissions, true otherwise.
  763.  */
  764. bool as_area_check_access(as_area_t *area, pf_access_t access)
  765. {
  766.     int flagmap[] = {
  767.         [PF_ACCESS_READ] = AS_AREA_READ,
  768.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  769.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  770.     };
  771.  
  772.     if (!(area->flags & flagmap[access]))
  773.         return false;
  774.    
  775.     return true;
  776. }
  777.  
  778. /** Handle page fault within the current address space.
  779.  *
  780.  * This is the high-level page fault handler. It decides
  781.  * whether the page fault can be resolved by any backend
  782.  * and if so, it invokes the backend to resolve the page
  783.  * fault.
  784.  *
  785.  * Interrupts are assumed disabled.
  786.  *
  787.  * @param page Faulting page.
  788.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  789.  * @param istate Pointer to interrupted state.
  790.  *
  791.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  792.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  793.  */
  794. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  795. {
  796.     pte_t *pte;
  797.     as_area_t *area;
  798.    
  799.     if (!THREAD)
  800.         return AS_PF_FAULT;
  801.        
  802.     ASSERT(AS);
  803.  
  804.     mutex_lock(&AS->lock);
  805.     area = find_area_and_lock(AS, page);   
  806.     if (!area) {
  807.         /*
  808.          * No area contained mapping for 'page'.
  809.          * Signal page fault to low-level handler.
  810.          */
  811.         mutex_unlock(&AS->lock);
  812.         goto page_fault;
  813.     }
  814.  
  815.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  816.         /*
  817.          * The address space area is not fully initialized.
  818.          * Avoid possible race by returning error.
  819.          */
  820.         mutex_unlock(&area->lock);
  821.         mutex_unlock(&AS->lock);
  822.         goto page_fault;       
  823.     }
  824.  
  825.     if (!area->backend || !area->backend->page_fault) {
  826.         /*
  827.          * The address space area is not backed by any backend
  828.          * or the backend cannot handle page faults.
  829.          */
  830.         mutex_unlock(&area->lock);
  831.         mutex_unlock(&AS->lock);
  832.         goto page_fault;       
  833.     }
  834.  
  835.     page_table_lock(AS, false);
  836.    
  837.     /*
  838.      * To avoid race condition between two page faults
  839.      * on the same address, we need to make sure
  840.      * the mapping has not been already inserted.
  841.      */
  842.     if ((pte = page_mapping_find(AS, page))) {
  843.         if (PTE_PRESENT(pte)) {
  844.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  845.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  846.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  847.                 page_table_unlock(AS, false);
  848.                 mutex_unlock(&area->lock);
  849.                 mutex_unlock(&AS->lock);
  850.                 return AS_PF_OK;
  851.             }
  852.         }
  853.     }
  854.    
  855.     /*
  856.      * Resort to the backend page fault handler.
  857.      */
  858.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  859.         page_table_unlock(AS, false);
  860.         mutex_unlock(&area->lock);
  861.         mutex_unlock(&AS->lock);
  862.         goto page_fault;
  863.     }
  864.    
  865.     page_table_unlock(AS, false);
  866.     mutex_unlock(&area->lock);
  867.     mutex_unlock(&AS->lock);
  868.     return AS_PF_OK;
  869.  
  870. page_fault:
  871.     if (THREAD->in_copy_from_uspace) {
  872.         THREAD->in_copy_from_uspace = false;
  873.         istate_set_retaddr(istate,
  874.             (uintptr_t) &memcpy_from_uspace_failover_address);
  875.     } else if (THREAD->in_copy_to_uspace) {
  876.         THREAD->in_copy_to_uspace = false;
  877.         istate_set_retaddr(istate,
  878.             (uintptr_t) &memcpy_to_uspace_failover_address);
  879.     } else {
  880.         return AS_PF_FAULT;
  881.     }
  882.  
  883.     return AS_PF_DEFER;
  884. }
  885.  
  886. /** Switch address spaces.
  887.  *
  888.  * Note that this function cannot sleep as it is essentially a part of
  889.  * scheduling. Sleeping here would lead to deadlock on wakeup. Another
  890.  * thing which is forbidden in this context is locking the address space.
  891.  *
  892.  * When this function is enetered, no spinlocks may be held.
  893.  *
  894.  * @param old Old address space or NULL.
  895.  * @param new New address space.
  896.  */
  897. void as_switch(as_t *old_as, as_t *new_as)
  898. {
  899.     DEADLOCK_PROBE_INIT(p_asidlock);
  900.     preemption_disable();
  901. retry:
  902.     (void) interrupts_disable();
  903.     if (!spinlock_trylock(&asidlock)) {
  904.         /*
  905.          * Avoid deadlock with TLB shootdown.
  906.          * We can enable interrupts here because
  907.          * preemption is disabled. We should not be
  908.          * holding any other lock.
  909.          */
  910.         (void) interrupts_enable();
  911.         DEADLOCK_PROBE(p_asidlock, DEADLOCK_THRESHOLD);
  912.         goto retry;
  913.     }
  914.     preemption_enable();
  915.  
  916.     /*
  917.      * First, take care of the old address space.
  918.      */
  919.     if (old_as) {
  920.         ASSERT(old_as->cpu_refcount);
  921.         if((--old_as->cpu_refcount == 0) && (old_as != AS_KERNEL)) {
  922.             /*
  923.              * The old address space is no longer active on
  924.              * any processor. It can be appended to the
  925.              * list of inactive address spaces with assigned
  926.              * ASID.
  927.              */
  928.             ASSERT(old_as->asid != ASID_INVALID);
  929.             list_append(&old_as->inactive_as_with_asid_link,
  930.                 &inactive_as_with_asid_head);
  931.         }
  932.  
  933.         /*
  934.          * Perform architecture-specific tasks when the address space
  935.          * is being removed from the CPU.
  936.          */
  937.         as_deinstall_arch(old_as);
  938.     }
  939.  
  940.     /*
  941.      * Second, prepare the new address space.
  942.      */
  943.     if ((new_as->cpu_refcount++ == 0) && (new_as != AS_KERNEL)) {
  944.         if (new_as->asid != ASID_INVALID)
  945.             list_remove(&new_as->inactive_as_with_asid_link);
  946.         else
  947.             new_as->asid = asid_get();
  948.     }
  949. #ifdef AS_PAGE_TABLE
  950.     SET_PTL0_ADDRESS(new_as->genarch.page_table);
  951. #endif
  952.    
  953.     /*
  954.      * Perform architecture-specific steps.
  955.      * (e.g. write ASID to hardware register etc.)
  956.      */
  957.     as_install_arch(new_as);
  958.  
  959.     spinlock_unlock(&asidlock);
  960.    
  961.     AS = new_as;
  962. }
  963.  
  964. /** Convert address space area flags to page flags.
  965.  *
  966.  * @param aflags Flags of some address space area.
  967.  *
  968.  * @return Flags to be passed to page_mapping_insert().
  969.  */
  970. int area_flags_to_page_flags(int aflags)
  971. {
  972.     int flags;
  973.  
  974.     flags = PAGE_USER | PAGE_PRESENT;
  975.    
  976.     if (aflags & AS_AREA_READ)
  977.         flags |= PAGE_READ;
  978.        
  979.     if (aflags & AS_AREA_WRITE)
  980.         flags |= PAGE_WRITE;
  981.    
  982.     if (aflags & AS_AREA_EXEC)
  983.         flags |= PAGE_EXEC;
  984.    
  985.     if (aflags & AS_AREA_CACHEABLE)
  986.         flags |= PAGE_CACHEABLE;
  987.        
  988.     return flags;
  989. }
  990.  
  991. /** Compute flags for virtual address translation subsytem.
  992.  *
  993.  * The address space area must be locked.
  994.  * Interrupts must be disabled.
  995.  *
  996.  * @param a Address space area.
  997.  *
  998.  * @return Flags to be used in page_mapping_insert().
  999.  */
  1000. int as_area_get_flags(as_area_t *a)
  1001. {
  1002.     return area_flags_to_page_flags(a->flags);
  1003. }
  1004.  
  1005. /** Create page table.
  1006.  *
  1007.  * Depending on architecture, create either address space
  1008.  * private or global page table.
  1009.  *
  1010.  * @param flags Flags saying whether the page table is for kernel address space.
  1011.  *
  1012.  * @return First entry of the page table.
  1013.  */
  1014. pte_t *page_table_create(int flags)
  1015. {
  1016. #ifdef __OBJC__
  1017.     return [as_t page_table_create: flags];
  1018. #else
  1019.     ASSERT(as_operations);
  1020.     ASSERT(as_operations->page_table_create);
  1021.    
  1022.     return as_operations->page_table_create(flags);
  1023. #endif
  1024. }
  1025.  
  1026. /** Destroy page table.
  1027.  *
  1028.  * Destroy page table in architecture specific way.
  1029.  *
  1030.  * @param page_table Physical address of PTL0.
  1031.  */
  1032. void page_table_destroy(pte_t *page_table)
  1033. {
  1034. #ifdef __OBJC__
  1035.     return [as_t page_table_destroy: page_table];
  1036. #else
  1037.     ASSERT(as_operations);
  1038.     ASSERT(as_operations->page_table_destroy);
  1039.    
  1040.     as_operations->page_table_destroy(page_table);
  1041. #endif
  1042. }
  1043.  
  1044. /** Lock page table.
  1045.  *
  1046.  * This function should be called before any page_mapping_insert(),
  1047.  * page_mapping_remove() and page_mapping_find().
  1048.  *
  1049.  * Locking order is such that address space areas must be locked
  1050.  * prior to this call. Address space can be locked prior to this
  1051.  * call in which case the lock argument is false.
  1052.  *
  1053.  * @param as Address space.
  1054.  * @param lock If false, do not attempt to lock as->lock.
  1055.  */
  1056. void page_table_lock(as_t *as, bool lock)
  1057. {
  1058. #ifdef __OBJC__
  1059.     [as page_table_lock: lock];
  1060. #else
  1061.     ASSERT(as_operations);
  1062.     ASSERT(as_operations->page_table_lock);
  1063.    
  1064.     as_operations->page_table_lock(as, lock);
  1065. #endif
  1066. }
  1067.  
  1068. /** Unlock page table.
  1069.  *
  1070.  * @param as Address space.
  1071.  * @param unlock If false, do not attempt to unlock as->lock.
  1072.  */
  1073. void page_table_unlock(as_t *as, bool unlock)
  1074. {
  1075. #ifdef __OBJC__
  1076.     [as page_table_unlock: unlock];
  1077. #else
  1078.     ASSERT(as_operations);
  1079.     ASSERT(as_operations->page_table_unlock);
  1080.    
  1081.     as_operations->page_table_unlock(as, unlock);
  1082. #endif
  1083. }
  1084.  
  1085.  
  1086. /** Find address space area and lock it.
  1087.  *
  1088.  * The address space must be locked and interrupts must be disabled.
  1089.  *
  1090.  * @param as Address space.
  1091.  * @param va Virtual address.
  1092.  *
  1093.  * @return Locked address space area containing va on success or NULL on
  1094.  *     failure.
  1095.  */
  1096. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  1097. {
  1098.     as_area_t *a;
  1099.     btree_node_t *leaf, *lnode;
  1100.     unsigned int i;
  1101.    
  1102.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  1103.     if (a) {
  1104.         /* va is the base address of an address space area */
  1105.         mutex_lock(&a->lock);
  1106.         return a;
  1107.     }
  1108.    
  1109.     /*
  1110.      * Search the leaf node and the righmost record of its left neighbour
  1111.      * to find out whether this is a miss or va belongs to an address
  1112.      * space area found there.
  1113.      */
  1114.    
  1115.     /* First, search the leaf node itself. */
  1116.     for (i = 0; i < leaf->keys; i++) {
  1117.         a = (as_area_t *) leaf->value[i];
  1118.         mutex_lock(&a->lock);
  1119.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  1120.             return a;
  1121.         }
  1122.         mutex_unlock(&a->lock);
  1123.     }
  1124.  
  1125.     /*
  1126.      * Second, locate the left neighbour and test its last record.
  1127.      * Because of its position in the B+tree, it must have base < va.
  1128.      */
  1129.     lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf);
  1130.     if (lnode) {
  1131.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  1132.         mutex_lock(&a->lock);
  1133.         if (va < a->base + a->pages * PAGE_SIZE) {
  1134.             return a;
  1135.         }
  1136.         mutex_unlock(&a->lock);
  1137.     }
  1138.  
  1139.     return NULL;
  1140. }
  1141.  
  1142. /** Check area conflicts with other areas.
  1143.  *
  1144.  * The address space must be locked and interrupts must be disabled.
  1145.  *
  1146.  * @param as Address space.
  1147.  * @param va Starting virtual address of the area being tested.
  1148.  * @param size Size of the area being tested.
  1149.  * @param avoid_area Do not touch this area.
  1150.  *
  1151.  * @return True if there is no conflict, false otherwise.
  1152.  */
  1153. bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  1154.               as_area_t *avoid_area)
  1155. {
  1156.     as_area_t *a;
  1157.     btree_node_t *leaf, *node;
  1158.     unsigned int i;
  1159.    
  1160.     /*
  1161.      * We don't want any area to have conflicts with NULL page.
  1162.      */
  1163.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1164.         return false;
  1165.    
  1166.     /*
  1167.      * The leaf node is found in O(log n), where n is proportional to
  1168.      * the number of address space areas belonging to as.
  1169.      * The check for conflicts is then attempted on the rightmost
  1170.      * record in the left neighbour, the leftmost record in the right
  1171.      * neighbour and all records in the leaf node itself.
  1172.      */
  1173.    
  1174.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1175.         if (a != avoid_area)
  1176.             return false;
  1177.     }
  1178.    
  1179.     /* First, check the two border cases. */
  1180.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1181.         a = (as_area_t *) node->value[node->keys - 1];
  1182.         mutex_lock(&a->lock);
  1183.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1184.             mutex_unlock(&a->lock);
  1185.             return false;
  1186.         }
  1187.         mutex_unlock(&a->lock);
  1188.     }
  1189.     node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf);
  1190.     if (node) {
  1191.         a = (as_area_t *) node->value[0];
  1192.         mutex_lock(&a->lock);
  1193.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1194.             mutex_unlock(&a->lock);
  1195.             return false;
  1196.         }
  1197.         mutex_unlock(&a->lock);
  1198.     }
  1199.    
  1200.     /* Second, check the leaf node. */
  1201.     for (i = 0; i < leaf->keys; i++) {
  1202.         a = (as_area_t *) leaf->value[i];
  1203.    
  1204.         if (a == avoid_area)
  1205.             continue;
  1206.    
  1207.         mutex_lock(&a->lock);
  1208.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1209.             mutex_unlock(&a->lock);
  1210.             return false;
  1211.         }
  1212.         mutex_unlock(&a->lock);
  1213.     }
  1214.  
  1215.     /*
  1216.      * So far, the area does not conflict with other areas.
  1217.      * Check if it doesn't conflict with kernel address space.
  1218.      */  
  1219.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1220.         return !overlaps(va, size,
  1221.             KERNEL_ADDRESS_SPACE_START,
  1222.             KERNEL_ADDRESS_SPACE_END - KERNEL_ADDRESS_SPACE_START);
  1223.     }
  1224.  
  1225.     return true;
  1226. }
  1227.  
  1228. /** Return size of the address space area with given base.
  1229.  *
  1230.  * @param base      Arbitrary address insede the address space area.
  1231.  *
  1232.  * @return      Size of the address space area in bytes or zero if it
  1233.  *          does not exist.
  1234.  */
  1235. size_t as_area_get_size(uintptr_t base)
  1236. {
  1237.     ipl_t ipl;
  1238.     as_area_t *src_area;
  1239.     size_t size;
  1240.  
  1241.     ipl = interrupts_disable();
  1242.     src_area = find_area_and_lock(AS, base);
  1243.     if (src_area){
  1244.         size = src_area->pages * PAGE_SIZE;
  1245.         mutex_unlock(&src_area->lock);
  1246.     } else {
  1247.         size = 0;
  1248.     }
  1249.     interrupts_restore(ipl);
  1250.     return size;
  1251. }
  1252.  
  1253. /** Mark portion of address space area as used.
  1254.  *
  1255.  * The address space area must be already locked.
  1256.  *
  1257.  * @param a Address space area.
  1258.  * @param page First page to be marked.
  1259.  * @param count Number of page to be marked.
  1260.  *
  1261.  * @return 0 on failure and 1 on success.
  1262.  */
  1263. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1264. {
  1265.     btree_node_t *leaf, *node;
  1266.     count_t pages;
  1267.     unsigned int i;
  1268.  
  1269.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1270.     ASSERT(count);
  1271.  
  1272.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1273.     if (pages) {
  1274.         /*
  1275.          * We hit the beginning of some used space.
  1276.          */
  1277.         return 0;
  1278.     }
  1279.  
  1280.     if (!leaf->keys) {
  1281.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1282.         return 1;
  1283.     }
  1284.  
  1285.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1286.     if (node) {
  1287.         uintptr_t left_pg = node->key[node->keys - 1];
  1288.         uintptr_t right_pg = leaf->key[0];
  1289.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1290.         count_t right_cnt = (count_t) leaf->value[0];
  1291.        
  1292.         /*
  1293.          * Examine the possibility that the interval fits
  1294.          * somewhere between the rightmost interval of
  1295.          * the left neigbour and the first interval of the leaf.
  1296.          */
  1297.          
  1298.         if (page >= right_pg) {
  1299.             /* Do nothing. */
  1300.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1301.             left_cnt * PAGE_SIZE)) {
  1302.             /* The interval intersects with the left interval. */
  1303.             return 0;
  1304.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1305.             right_cnt * PAGE_SIZE)) {
  1306.             /* The interval intersects with the right interval. */
  1307.             return 0;          
  1308.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1309.             (page + count * PAGE_SIZE == right_pg)) {
  1310.             /*
  1311.              * The interval can be added by merging the two already
  1312.              * present intervals.
  1313.              */
  1314.             node->value[node->keys - 1] += count + right_cnt;
  1315.             btree_remove(&a->used_space, right_pg, leaf);
  1316.             return 1;
  1317.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1318.             /*
  1319.              * The interval can be added by simply growing the left
  1320.              * interval.
  1321.              */
  1322.             node->value[node->keys - 1] += count;
  1323.             return 1;
  1324.         } else if (page + count * PAGE_SIZE == right_pg) {
  1325.             /*
  1326.              * The interval can be addded by simply moving base of
  1327.              * the right interval down and increasing its size
  1328.              * accordingly.
  1329.              */
  1330.             leaf->value[0] += count;
  1331.             leaf->key[0] = page;
  1332.             return 1;
  1333.         } else {
  1334.             /*
  1335.              * The interval is between both neigbouring intervals,
  1336.              * but cannot be merged with any of them.
  1337.              */
  1338.             btree_insert(&a->used_space, page, (void *) count,
  1339.                 leaf);
  1340.             return 1;
  1341.         }
  1342.     } else if (page < leaf->key[0]) {
  1343.         uintptr_t right_pg = leaf->key[0];
  1344.         count_t right_cnt = (count_t) leaf->value[0];
  1345.    
  1346.         /*
  1347.          * Investigate the border case in which the left neighbour does
  1348.          * not exist but the interval fits from the left.
  1349.          */
  1350.          
  1351.         if (overlaps(page, count * PAGE_SIZE, right_pg,
  1352.             right_cnt * PAGE_SIZE)) {
  1353.             /* The interval intersects with the right interval. */
  1354.             return 0;
  1355.         } else if (page + count * PAGE_SIZE == right_pg) {
  1356.             /*
  1357.              * The interval can be added by moving the base of the
  1358.              * right interval down and increasing its size
  1359.              * accordingly.
  1360.              */
  1361.             leaf->key[0] = page;
  1362.             leaf->value[0] += count;
  1363.             return 1;
  1364.         } else {
  1365.             /*
  1366.              * The interval doesn't adjoin with the right interval.
  1367.              * It must be added individually.
  1368.              */
  1369.             btree_insert(&a->used_space, page, (void *) count,
  1370.                 leaf);
  1371.             return 1;
  1372.         }
  1373.     }
  1374.  
  1375.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1376.     if (node) {
  1377.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1378.         uintptr_t right_pg = node->key[0];
  1379.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1380.         count_t right_cnt = (count_t) node->value[0];
  1381.        
  1382.         /*
  1383.          * Examine the possibility that the interval fits
  1384.          * somewhere between the leftmost interval of
  1385.          * the right neigbour and the last interval of the leaf.
  1386.          */
  1387.  
  1388.         if (page < left_pg) {
  1389.             /* Do nothing. */
  1390.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1391.             left_cnt * PAGE_SIZE)) {
  1392.             /* The interval intersects with the left interval. */
  1393.             return 0;
  1394.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1395.             right_cnt * PAGE_SIZE)) {
  1396.             /* The interval intersects with the right interval. */
  1397.             return 0;          
  1398.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1399.             (page + count * PAGE_SIZE == right_pg)) {
  1400.             /*
  1401.              * The interval can be added by merging the two already
  1402.              * present intervals.
  1403.              * */
  1404.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1405.             btree_remove(&a->used_space, right_pg, node);
  1406.             return 1;
  1407.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1408.             /*
  1409.              * The interval can be added by simply growing the left
  1410.              * interval.
  1411.              * */
  1412.             leaf->value[leaf->keys - 1] +=  count;
  1413.             return 1;
  1414.         } else if (page + count * PAGE_SIZE == right_pg) {
  1415.             /*
  1416.              * The interval can be addded by simply moving base of
  1417.              * the right interval down and increasing its size
  1418.              * accordingly.
  1419.              */
  1420.             node->value[0] += count;
  1421.             node->key[0] = page;
  1422.             return 1;
  1423.         } else {
  1424.             /*
  1425.              * The interval is between both neigbouring intervals,
  1426.              * but cannot be merged with any of them.
  1427.              */
  1428.             btree_insert(&a->used_space, page, (void *) count,
  1429.                 leaf);
  1430.             return 1;
  1431.         }
  1432.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1433.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1434.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1435.    
  1436.         /*
  1437.          * Investigate the border case in which the right neighbour
  1438.          * does not exist but the interval fits from the right.
  1439.          */
  1440.          
  1441.         if (overlaps(page, count * PAGE_SIZE, left_pg,
  1442.             left_cnt * PAGE_SIZE)) {
  1443.             /* The interval intersects with the left interval. */
  1444.             return 0;
  1445.         } else if (left_pg + left_cnt * PAGE_SIZE == page) {
  1446.             /*
  1447.              * The interval can be added by growing the left
  1448.              * interval.
  1449.              */
  1450.             leaf->value[leaf->keys - 1] += count;
  1451.             return 1;
  1452.         } else {
  1453.             /*
  1454.              * The interval doesn't adjoin with the left interval.
  1455.              * It must be added individually.
  1456.              */
  1457.             btree_insert(&a->used_space, page, (void *) count,
  1458.                 leaf);
  1459.             return 1;
  1460.         }
  1461.     }
  1462.    
  1463.     /*
  1464.      * Note that if the algorithm made it thus far, the interval can fit
  1465.      * only between two other intervals of the leaf. The two border cases
  1466.      * were already resolved.
  1467.      */
  1468.     for (i = 1; i < leaf->keys; i++) {
  1469.         if (page < leaf->key[i]) {
  1470.             uintptr_t left_pg = leaf->key[i - 1];
  1471.             uintptr_t right_pg = leaf->key[i];
  1472.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1473.             count_t right_cnt = (count_t) leaf->value[i];
  1474.  
  1475.             /*
  1476.              * The interval fits between left_pg and right_pg.
  1477.              */
  1478.  
  1479.             if (overlaps(page, count * PAGE_SIZE, left_pg,
  1480.                 left_cnt * PAGE_SIZE)) {
  1481.                 /*
  1482.                  * The interval intersects with the left
  1483.                  * interval.
  1484.                  */
  1485.                 return 0;
  1486.             } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1487.                 right_cnt * PAGE_SIZE)) {
  1488.                 /*
  1489.                  * The interval intersects with the right
  1490.                  * interval.
  1491.                  */
  1492.                 return 0;          
  1493.             } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1494.                 (page + count * PAGE_SIZE == right_pg)) {
  1495.                 /*
  1496.                  * The interval can be added by merging the two
  1497.                  * already present intervals.
  1498.                  */
  1499.                 leaf->value[i - 1] += count + right_cnt;
  1500.                 btree_remove(&a->used_space, right_pg, leaf);
  1501.                 return 1;
  1502.             } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1503.                 /*
  1504.                  * The interval can be added by simply growing
  1505.                  * the left interval.
  1506.                  */
  1507.                 leaf->value[i - 1] += count;
  1508.                 return 1;
  1509.             } else if (page + count * PAGE_SIZE == right_pg) {
  1510.                 /*
  1511.                      * The interval can be addded by simply moving
  1512.                  * base of the right interval down and
  1513.                  * increasing its size accordingly.
  1514.                  */
  1515.                 leaf->value[i] += count;
  1516.                 leaf->key[i] = page;
  1517.                 return 1;
  1518.             } else {
  1519.                 /*
  1520.                  * The interval is between both neigbouring
  1521.                  * intervals, but cannot be merged with any of
  1522.                  * them.
  1523.                  */
  1524.                 btree_insert(&a->used_space, page,
  1525.                     (void *) count, leaf);
  1526.                 return 1;
  1527.             }
  1528.         }
  1529.     }
  1530.  
  1531.     panic("Inconsistency detected while adding %d pages of used space at "
  1532.         "%p.\n", count, page);
  1533. }
  1534.  
  1535. /** Mark portion of address space area as unused.
  1536.  *
  1537.  * The address space area must be already locked.
  1538.  *
  1539.  * @param a Address space area.
  1540.  * @param page First page to be marked.
  1541.  * @param count Number of page to be marked.
  1542.  *
  1543.  * @return 0 on failure and 1 on success.
  1544.  */
  1545. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1546. {
  1547.     btree_node_t *leaf, *node;
  1548.     count_t pages;
  1549.     unsigned int i;
  1550.  
  1551.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1552.     ASSERT(count);
  1553.  
  1554.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1555.     if (pages) {
  1556.         /*
  1557.          * We are lucky, page is the beginning of some interval.
  1558.          */
  1559.         if (count > pages) {
  1560.             return 0;
  1561.         } else if (count == pages) {
  1562.             btree_remove(&a->used_space, page, leaf);
  1563.             return 1;
  1564.         } else {
  1565.             /*
  1566.              * Find the respective interval.
  1567.              * Decrease its size and relocate its start address.
  1568.              */
  1569.             for (i = 0; i < leaf->keys; i++) {
  1570.                 if (leaf->key[i] == page) {
  1571.                     leaf->key[i] += count * PAGE_SIZE;
  1572.                     leaf->value[i] -= count;
  1573.                     return 1;
  1574.                 }
  1575.             }
  1576.             goto error;
  1577.         }
  1578.     }
  1579.  
  1580.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1581.     if (node && page < leaf->key[0]) {
  1582.         uintptr_t left_pg = node->key[node->keys - 1];
  1583.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1584.  
  1585.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1586.             count * PAGE_SIZE)) {
  1587.             if (page + count * PAGE_SIZE ==
  1588.                 left_pg + left_cnt * PAGE_SIZE) {
  1589.                 /*
  1590.                  * The interval is contained in the rightmost
  1591.                  * interval of the left neighbour and can be
  1592.                  * removed by updating the size of the bigger
  1593.                  * interval.
  1594.                  */
  1595.                 node->value[node->keys - 1] -= count;
  1596.                 return 1;
  1597.             } else if (page + count * PAGE_SIZE <
  1598.                 left_pg + left_cnt*PAGE_SIZE) {
  1599.                 count_t new_cnt;
  1600.                
  1601.                 /*
  1602.                  * The interval is contained in the rightmost
  1603.                  * interval of the left neighbour but its
  1604.                  * removal requires both updating the size of
  1605.                  * the original interval and also inserting a
  1606.                  * new interval.
  1607.                  */
  1608.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1609.                     (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1610.                 node->value[node->keys - 1] -= count + new_cnt;
  1611.                 btree_insert(&a->used_space, page +
  1612.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1613.                 return 1;
  1614.             }
  1615.         }
  1616.         return 0;
  1617.     } else if (page < leaf->key[0]) {
  1618.         return 0;
  1619.     }
  1620.    
  1621.     if (page > leaf->key[leaf->keys - 1]) {
  1622.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1623.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1624.  
  1625.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1626.             count * PAGE_SIZE)) {
  1627.             if (page + count * PAGE_SIZE ==
  1628.                 left_pg + left_cnt * PAGE_SIZE) {
  1629.                 /*
  1630.                  * The interval is contained in the rightmost
  1631.                  * interval of the leaf and can be removed by
  1632.                  * updating the size of the bigger interval.
  1633.                  */
  1634.                 leaf->value[leaf->keys - 1] -= count;
  1635.                 return 1;
  1636.             } else if (page + count * PAGE_SIZE < left_pg +
  1637.                 left_cnt * PAGE_SIZE) {
  1638.                 count_t new_cnt;
  1639.                
  1640.                 /*
  1641.                  * The interval is contained in the rightmost
  1642.                  * interval of the leaf but its removal
  1643.                  * requires both updating the size of the
  1644.                  * original interval and also inserting a new
  1645.                  * interval.
  1646.                  */
  1647.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1648.                     (page + count * PAGE_SIZE)) >> PAGE_WIDTH;
  1649.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1650.                 btree_insert(&a->used_space, page +
  1651.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1652.                 return 1;
  1653.             }
  1654.         }
  1655.         return 0;
  1656.     }  
  1657.    
  1658.     /*
  1659.      * The border cases have been already resolved.
  1660.      * Now the interval can be only between intervals of the leaf.
  1661.      */
  1662.     for (i = 1; i < leaf->keys - 1; i++) {
  1663.         if (page < leaf->key[i]) {
  1664.             uintptr_t left_pg = leaf->key[i - 1];
  1665.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1666.  
  1667.             /*
  1668.              * Now the interval is between intervals corresponding
  1669.              * to (i - 1) and i.
  1670.              */
  1671.             if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1672.                 count * PAGE_SIZE)) {
  1673.                 if (page + count * PAGE_SIZE ==
  1674.                     left_pg + left_cnt*PAGE_SIZE) {
  1675.                     /*
  1676.                      * The interval is contained in the
  1677.                      * interval (i - 1) of the leaf and can
  1678.                      * be removed by updating the size of
  1679.                      * the bigger interval.
  1680.                      */
  1681.                     leaf->value[i - 1] -= count;
  1682.                     return 1;
  1683.                 } else if (page + count * PAGE_SIZE <
  1684.                     left_pg + left_cnt * PAGE_SIZE) {
  1685.                     count_t new_cnt;
  1686.                
  1687.                     /*
  1688.                      * The interval is contained in the
  1689.                      * interval (i - 1) of the leaf but its
  1690.                      * removal requires both updating the
  1691.                      * size of the original interval and
  1692.                      * also inserting a new interval.
  1693.                      */
  1694.                     new_cnt = ((left_pg +
  1695.                         left_cnt * PAGE_SIZE) -
  1696.                         (page + count * PAGE_SIZE)) >>
  1697.                         PAGE_WIDTH;
  1698.                     leaf->value[i - 1] -= count + new_cnt;
  1699.                     btree_insert(&a->used_space, page +
  1700.                         count * PAGE_SIZE, (void *) new_cnt,
  1701.                         leaf);
  1702.                     return 1;
  1703.                 }
  1704.             }
  1705.             return 0;
  1706.         }
  1707.     }
  1708.  
  1709. error:
  1710.     panic("Inconsistency detected while removing %d pages of used space "
  1711.         "from %p.\n", count, page);
  1712. }
  1713.  
  1714. /** Remove reference to address space area share info.
  1715.  *
  1716.  * If the reference count drops to 0, the sh_info is deallocated.
  1717.  *
  1718.  * @param sh_info Pointer to address space area share info.
  1719.  */
  1720. void sh_info_remove_reference(share_info_t *sh_info)
  1721. {
  1722.     bool dealloc = false;
  1723.  
  1724.     mutex_lock(&sh_info->lock);
  1725.     ASSERT(sh_info->refcount);
  1726.     if (--sh_info->refcount == 0) {
  1727.         dealloc = true;
  1728.         link_t *cur;
  1729.        
  1730.         /*
  1731.          * Now walk carefully the pagemap B+tree and free/remove
  1732.          * reference from all frames found there.
  1733.          */
  1734.         for (cur = sh_info->pagemap.leaf_head.next;
  1735.             cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1736.             btree_node_t *node;
  1737.             unsigned int i;
  1738.            
  1739.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1740.             for (i = 0; i < node->keys; i++)
  1741.                 frame_free((uintptr_t) node->value[i]);
  1742.         }
  1743.        
  1744.     }
  1745.     mutex_unlock(&sh_info->lock);
  1746.    
  1747.     if (dealloc) {
  1748.         btree_destroy(&sh_info->pagemap);
  1749.         free(sh_info);
  1750.     }
  1751. }
  1752.  
  1753. /*
  1754.  * Address space related syscalls.
  1755.  */
  1756.  
  1757. /** Wrapper for as_area_create(). */
  1758. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1759. {
  1760.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address,
  1761.         AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1762.         return (unative_t) address;
  1763.     else
  1764.         return (unative_t) -1;
  1765. }
  1766.  
  1767. /** Wrapper for as_area_resize(). */
  1768. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1769. {
  1770.     return (unative_t) as_area_resize(AS, address, size, 0);
  1771. }
  1772.  
  1773. /** Wrapper for as_area_destroy(). */
  1774. unative_t sys_as_area_destroy(uintptr_t address)
  1775. {
  1776.     return (unative_t) as_area_destroy(AS, address);
  1777. }
  1778.  
  1779. /** Print out information about address space.
  1780.  *
  1781.  * @param as Address space.
  1782.  */
  1783. void as_print(as_t *as)
  1784. {
  1785.     ipl_t ipl;
  1786.    
  1787.     ipl = interrupts_disable();
  1788.     mutex_lock(&as->lock);
  1789.    
  1790.     /* print out info about address space areas */
  1791.     link_t *cur;
  1792.     for (cur = as->as_area_btree.leaf_head.next;
  1793.         cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  1794.         btree_node_t *node;
  1795.        
  1796.         node = list_get_instance(cur, btree_node_t, leaf_link);
  1797.        
  1798.         unsigned int i;
  1799.         for (i = 0; i < node->keys; i++) {
  1800.             as_area_t *area = node->value[i];
  1801.        
  1802.             mutex_lock(&area->lock);
  1803.             printf("as_area: %p, base=%p, pages=%d (%p - %p)\n",
  1804.                 area, area->base, area->pages, area->base,
  1805.                 area->base + area->pages*PAGE_SIZE);
  1806.             mutex_unlock(&area->lock);
  1807.         }
  1808.     }
  1809.    
  1810.     mutex_unlock(&as->lock);
  1811.     interrupts_restore(ipl);
  1812. }
  1813.  
  1814. /** @}
  1815.  */
  1816.