Subversion Repositories HelenOS-historic

Rev

Rev 1483 | Rev 1587 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed

  1. /*
  2.  * Copyright (C) 2001-2006 Jakub Jermar
  3.  * All rights reserved.
  4.  *
  5.  * Redistribution and use in source and binary forms, with or without
  6.  * modification, are permitted provided that the following conditions
  7.  * are met:
  8.  *
  9.  * - Redistributions of source code must retain the above copyright
  10.  *   notice, this list of conditions and the following disclaimer.
  11.  * - Redistributions in binary form must reproduce the above copyright
  12.  *   notice, this list of conditions and the following disclaimer in the
  13.  *   documentation and/or other materials provided with the distribution.
  14.  * - The name of the author may not be used to endorse or promote products
  15.  *   derived from this software without specific prior written permission.
  16.  *
  17.  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  18.  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  19.  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  20.  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  21.  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  22.  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  23.  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  24.  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  25.  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  26.  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  27.  */
  28.  
  29. /**
  30.  * @file    as.c
  31.  * @brief   Address space related functions.
  32.  *
  33.  * This file contains address space manipulation functions.
  34.  * Roughly speaking, this is a higher-level client of
  35.  * Virtual Address Translation (VAT) subsystem.
  36.  *
  37.  * Functionality provided by this file allows one to
  38.  * create address space and create, resize and share
  39.  * address space areas.
  40.  *
  41.  * @see page.c
  42.  *
  43.  */
  44.  
  45. #include <mm/as.h>
  46. #include <arch/mm/as.h>
  47. #include <mm/page.h>
  48. #include <mm/frame.h>
  49. #include <mm/slab.h>
  50. #include <mm/tlb.h>
  51. #include <arch/mm/page.h>
  52. #include <genarch/mm/page_pt.h>
  53. #include <genarch/mm/page_ht.h>
  54. #include <mm/asid.h>
  55. #include <arch/mm/asid.h>
  56. #include <synch/spinlock.h>
  57. #include <synch/mutex.h>
  58. #include <adt/list.h>
  59. #include <adt/btree.h>
  60. #include <proc/task.h>
  61. #include <proc/thread.h>
  62. #include <arch/asm.h>
  63. #include <panic.h>
  64. #include <debug.h>
  65. #include <print.h>
  66. #include <memstr.h>
  67. #include <macros.h>
  68. #include <arch.h>
  69. #include <errno.h>
  70. #include <config.h>
  71. #include <align.h>
  72. #include <arch/types.h>
  73. #include <typedefs.h>
  74. #include <syscall/copy.h>
  75. #include <arch/interrupt.h>
  76.  
  77. as_operations_t *as_operations = NULL;
  78.  
  79. /** This lock protects inactive_as_with_asid_head list. It must be acquired before as_t mutex. */
  80. SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
  81.  
  82. /**
  83.  * This list contains address spaces that are not active on any
  84.  * processor and that have valid ASID.
  85.  */
  86. LIST_INITIALIZE(inactive_as_with_asid_head);
  87.  
  88. /** Kernel address space. */
  89. as_t *AS_KERNEL = NULL;
  90.  
  91. static int area_flags_to_page_flags(int aflags);
  92. static as_area_t *find_area_and_lock(as_t *as, __address va);
  93. static bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area);
  94. static void sh_info_remove_reference(share_info_t *sh_info);
  95.  
  96. /** Initialize address space subsystem. */
  97. void as_init(void)
  98. {
  99.     as_arch_init();
  100.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  101.     if (!AS_KERNEL)
  102.         panic("can't create kernel address space\n");
  103.    
  104. }
  105.  
  106. /** Create address space.
  107.  *
  108.  * @param flags Flags that influence way in wich the address space is created.
  109.  */
  110. as_t *as_create(int flags)
  111. {
  112.     as_t *as;
  113.  
  114.     as = (as_t *) malloc(sizeof(as_t), 0);
  115.     link_initialize(&as->inactive_as_with_asid_link);
  116.     mutex_initialize(&as->lock);
  117.     btree_create(&as->as_area_btree);
  118.    
  119.     if (flags & FLAG_AS_KERNEL)
  120.         as->asid = ASID_KERNEL;
  121.     else
  122.         as->asid = ASID_INVALID;
  123.    
  124.     as->refcount = 0;
  125.     as->cpu_refcount = 0;
  126.     as->page_table = page_table_create(flags);
  127.  
  128.     return as;
  129. }
  130.  
  131. /** Destroy adress space.
  132.  *
  133.  * When there are no tasks referencing this address space (i.e. its refcount is zero),
  134.  * the address space can be destroyed.
  135.  */
  136. void as_destroy(as_t *as)
  137. {
  138.     ipl_t ipl;
  139.     link_t *cur;
  140.  
  141.     ASSERT(as->refcount == 0);
  142.    
  143.     /*
  144.      * Since there is no reference to this area,
  145.      * it is safe not to lock its mutex.
  146.      */
  147.      
  148.     ipl = interrupts_disable();
  149.     spinlock_lock(&inactive_as_with_asid_lock);
  150.     if (as->asid != ASID_INVALID && as->asid != ASID_KERNEL) {
  151.         list_remove(&as->inactive_as_with_asid_link);
  152.         asid_put(as->asid);
  153.     }
  154.     spinlock_unlock(&inactive_as_with_asid_lock);
  155.  
  156.     /*
  157.      * Destroy address space areas of the address space.
  158.      */
  159.     for (cur = as->as_area_btree.leaf_head.next; cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  160.         btree_node_t *node;
  161.         int i;
  162.        
  163.         node = list_get_instance(cur, btree_node_t, leaf_link);
  164.         for (i = 0; i < node->keys; i++)
  165.             as_area_destroy(as, node->key[i]);
  166.     }
  167.  
  168.     btree_destroy(&as->as_area_btree);
  169.     page_table_destroy(as->page_table);
  170.  
  171.     interrupts_restore(ipl);
  172.    
  173.     free(as);
  174. }
  175.  
  176. /** Create address space area of common attributes.
  177.  *
  178.  * The created address space area is added to the target address space.
  179.  *
  180.  * @param as Target address space.
  181.  * @param flags Flags of the area memory.
  182.  * @param size Size of area.
  183.  * @param base Base address of area.
  184.  * @param attrs Attributes of the area.
  185.  * @param backend Address space area backend. NULL if no backend is used.
  186.  * @param backend_data NULL or a pointer to an array holding two void *.
  187.  *
  188.  * @return Address space area on success or NULL on failure.
  189.  */
  190. as_area_t *as_area_create(as_t *as, int flags, size_t size, __address base, int attrs,
  191.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  192. {
  193.     ipl_t ipl;
  194.     as_area_t *a;
  195.    
  196.     if (base % PAGE_SIZE)
  197.         return NULL;
  198.  
  199.     if (!size)
  200.         return NULL;
  201.  
  202.     /* Writeable executable areas are not supported. */
  203.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  204.         return NULL;
  205.    
  206.     ipl = interrupts_disable();
  207.     mutex_lock(&as->lock);
  208.    
  209.     if (!check_area_conflicts(as, base, size, NULL)) {
  210.         mutex_unlock(&as->lock);
  211.         interrupts_restore(ipl);
  212.         return NULL;
  213.     }
  214.    
  215.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  216.  
  217.     mutex_initialize(&a->lock);
  218.    
  219.     a->as = as;
  220.     a->flags = flags;
  221.     a->attributes = attrs;
  222.     a->pages = SIZE2FRAMES(size);
  223.     a->base = base;
  224.     a->sh_info = NULL;
  225.     a->backend = backend;
  226.     if (backend_data)
  227.         a->backend_data = *backend_data;
  228.     else
  229.         memsetb((__address) &a->backend_data, sizeof(a->backend_data), 0);
  230.  
  231.     btree_create(&a->used_space);
  232.    
  233.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  234.  
  235.     mutex_unlock(&as->lock);
  236.     interrupts_restore(ipl);
  237.  
  238.     return a;
  239. }
  240.  
  241. /** Find address space area and change it.
  242.  *
  243.  * @param as Address space.
  244.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  245.  * @param size New size of the virtual memory block starting at address.
  246.  * @param flags Flags influencing the remap operation. Currently unused.
  247.  *
  248.  * @return Zero on success or a value from @ref errno.h otherwise.
  249.  */
  250. int as_area_resize(as_t *as, __address address, size_t size, int flags)
  251. {
  252.     as_area_t *area;
  253.     ipl_t ipl;
  254.     size_t pages;
  255.    
  256.     ipl = interrupts_disable();
  257.     mutex_lock(&as->lock);
  258.    
  259.     /*
  260.      * Locate the area.
  261.      */
  262.     area = find_area_and_lock(as, address);
  263.     if (!area) {
  264.         mutex_unlock(&as->lock);
  265.         interrupts_restore(ipl);
  266.         return ENOENT;
  267.     }
  268.  
  269.     if (area->backend == &phys_backend) {
  270.         /*
  271.          * Remapping of address space areas associated
  272.          * with memory mapped devices is not supported.
  273.          */
  274.         mutex_unlock(&area->lock);
  275.         mutex_unlock(&as->lock);
  276.         interrupts_restore(ipl);
  277.         return ENOTSUP;
  278.     }
  279.     if (area->sh_info) {
  280.         /*
  281.          * Remapping of shared address space areas
  282.          * is not supported.
  283.          */
  284.         mutex_unlock(&area->lock);
  285.         mutex_unlock(&as->lock);
  286.         interrupts_restore(ipl);
  287.         return ENOTSUP;
  288.     }
  289.  
  290.     pages = SIZE2FRAMES((address - area->base) + size);
  291.     if (!pages) {
  292.         /*
  293.          * Zero size address space areas are not allowed.
  294.          */
  295.         mutex_unlock(&area->lock);
  296.         mutex_unlock(&as->lock);
  297.         interrupts_restore(ipl);
  298.         return EPERM;
  299.     }
  300.    
  301.     if (pages < area->pages) {
  302.         bool cond;
  303.         __address start_free = area->base + pages*PAGE_SIZE;
  304.  
  305.         /*
  306.          * Shrinking the area.
  307.          * No need to check for overlaps.
  308.          */
  309.  
  310.         /*
  311.          * Start TLB shootdown sequence.
  312.          */
  313.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  314.  
  315.         /*
  316.          * Remove frames belonging to used space starting from
  317.          * the highest addresses downwards until an overlap with
  318.          * the resized address space area is found. Note that this
  319.          * is also the right way to remove part of the used_space
  320.          * B+tree leaf list.
  321.          */    
  322.         for (cond = true; cond;) {
  323.             btree_node_t *node;
  324.        
  325.             ASSERT(!list_empty(&area->used_space.leaf_head));
  326.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  327.             if ((cond = (bool) node->keys)) {
  328.                 __address b = node->key[node->keys - 1];
  329.                 count_t c = (count_t) node->value[node->keys - 1];
  330.                 int i = 0;
  331.            
  332.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  333.                    
  334.                     if (b + c*PAGE_SIZE <= start_free) {
  335.                         /*
  336.                          * The whole interval fits completely
  337.                          * in the resized address space area.
  338.                          */
  339.                         break;
  340.                     }
  341.        
  342.                     /*
  343.                      * Part of the interval corresponding to b and c
  344.                      * overlaps with the resized address space area.
  345.                      */
  346.        
  347.                     cond = false;   /* we are almost done */
  348.                     i = (start_free - b) >> PAGE_WIDTH;
  349.                     if (!used_space_remove(area, start_free, c - i))
  350.                         panic("Could not remove used space.");
  351.                 } else {
  352.                     /*
  353.                      * The interval of used space can be completely removed.
  354.                      */
  355.                     if (!used_space_remove(area, b, c))
  356.                         panic("Could not remove used space.\n");
  357.                 }
  358.            
  359.                 for (; i < c; i++) {
  360.                     pte_t *pte;
  361.            
  362.                     page_table_lock(as, false);
  363.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  364.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  365.                     if (area->backend && area->backend->frame_free) {
  366.                         area->backend->frame_free(area,
  367.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  368.                     }
  369.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  370.                     page_table_unlock(as, false);
  371.                 }
  372.             }
  373.         }
  374.  
  375.         /*
  376.          * Finish TLB shootdown sequence.
  377.          */
  378.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  379.         tlb_shootdown_finalize();
  380.     } else {
  381.         /*
  382.          * Growing the area.
  383.          * Check for overlaps with other address space areas.
  384.          */
  385.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  386.             mutex_unlock(&area->lock);
  387.             mutex_unlock(&as->lock);       
  388.             interrupts_restore(ipl);
  389.             return EADDRNOTAVAIL;
  390.         }
  391.     }
  392.  
  393.     area->pages = pages;
  394.    
  395.     mutex_unlock(&area->lock);
  396.     mutex_unlock(&as->lock);
  397.     interrupts_restore(ipl);
  398.  
  399.     return 0;
  400. }
  401.  
  402. /** Destroy address space area.
  403.  *
  404.  * @param as Address space.
  405.  * @param address Address withing the area to be deleted.
  406.  *
  407.  * @return Zero on success or a value from @ref errno.h on failure.
  408.  */
  409. int as_area_destroy(as_t *as, __address address)
  410. {
  411.     as_area_t *area;
  412.     __address base;
  413.     link_t *cur;
  414.     ipl_t ipl;
  415.  
  416.     ipl = interrupts_disable();
  417.     mutex_lock(&as->lock);
  418.  
  419.     area = find_area_and_lock(as, address);
  420.     if (!area) {
  421.         mutex_unlock(&as->lock);
  422.         interrupts_restore(ipl);
  423.         return ENOENT;
  424.     }
  425.  
  426.     base = area->base;
  427.  
  428.     /*
  429.      * Start TLB shootdown sequence.
  430.      */
  431.     tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
  432.  
  433.     /*
  434.      * Visit only the pages mapped by used_space B+tree.
  435.      */
  436.     for (cur = area->used_space.leaf_head.next; cur != &area->used_space.leaf_head; cur = cur->next) {
  437.         btree_node_t *node;
  438.         int i;
  439.        
  440.         node = list_get_instance(cur, btree_node_t, leaf_link);
  441.         for (i = 0; i < node->keys; i++) {
  442.             __address b = node->key[i];
  443.             count_t j;
  444.             pte_t *pte;
  445.            
  446.             for (j = 0; j < (count_t) node->value[i]; j++) {
  447.                 page_table_lock(as, false);
  448.                 pte = page_mapping_find(as, b + j*PAGE_SIZE);
  449.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  450.                 if (area->backend && area->backend->frame_free) {
  451.                     area->backend->frame_free(area,
  452.                         b + j*PAGE_SIZE, PTE_GET_FRAME(pte));
  453.                 }
  454.                 page_mapping_remove(as, b + j*PAGE_SIZE);
  455.                 page_table_unlock(as, false);
  456.             }
  457.         }
  458.     }
  459.  
  460.     /*
  461.      * Finish TLB shootdown sequence.
  462.      */
  463.     tlb_invalidate_pages(AS->asid, area->base, area->pages);
  464.     tlb_shootdown_finalize();
  465.    
  466.     btree_destroy(&area->used_space);
  467.  
  468.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  469.    
  470.     if (area->sh_info)
  471.         sh_info_remove_reference(area->sh_info);
  472.        
  473.     mutex_unlock(&area->lock);
  474.  
  475.     /*
  476.      * Remove the empty area from address space.
  477.      */
  478.     btree_remove(&AS->as_area_btree, base, NULL);
  479.    
  480.     free(area);
  481.    
  482.     mutex_unlock(&AS->lock);
  483.     interrupts_restore(ipl);
  484.     return 0;
  485. }
  486.  
  487. /** Share address space area with another or the same address space.
  488.  *
  489.  * Address space area mapping is shared with a new address space area.
  490.  * If the source address space area has not been shared so far,
  491.  * a new sh_info is created. The new address space area simply gets the
  492.  * sh_info of the source area. The process of duplicating the
  493.  * mapping is done through the backend share function.
  494.  *
  495.  * @param src_as Pointer to source address space.
  496.  * @param src_base Base address of the source address space area.
  497.  * @param acc_size Expected size of the source area.
  498.  * @param dst_as Pointer to destination address space.
  499.  * @param dst_base Target base address.
  500.  * @param dst_flags_mask Destination address space area flags mask.
  501.  *
  502.  * @return Zero on success or ENOENT if there is no such task or
  503.  *     if there is no such address space area,
  504.  *     EPERM if there was a problem in accepting the area or
  505.  *     ENOMEM if there was a problem in allocating destination
  506.  *     address space area. ENOTSUP is returned if an attempt
  507.  *     to share non-anonymous address space area is detected.
  508.  */
  509. int as_area_share(as_t *src_as, __address src_base, size_t acc_size,
  510.           as_t *dst_as, __address dst_base, int dst_flags_mask)
  511. {
  512.     ipl_t ipl;
  513.     int src_flags;
  514.     size_t src_size;
  515.     as_area_t *src_area, *dst_area;
  516.     share_info_t *sh_info;
  517.     mem_backend_t *src_backend;
  518.     mem_backend_data_t src_backend_data;
  519.    
  520.     ipl = interrupts_disable();
  521.     mutex_lock(&src_as->lock);
  522.     src_area = find_area_and_lock(src_as, src_base);
  523.     if (!src_area) {
  524.         /*
  525.          * Could not find the source address space area.
  526.          */
  527.         mutex_unlock(&src_as->lock);
  528.         interrupts_restore(ipl);
  529.         return ENOENT;
  530.     }
  531.    
  532.     if (!src_area->backend || !src_area->backend->share) {
  533.         /*
  534.          * There is now backend or the backend does not
  535.          * know how to share the area.
  536.          */
  537.         mutex_unlock(&src_area->lock);
  538.         mutex_unlock(&src_as->lock);
  539.         interrupts_restore(ipl);
  540.         return ENOTSUP;
  541.     }
  542.    
  543.     src_size = src_area->pages * PAGE_SIZE;
  544.     src_flags = src_area->flags;
  545.     src_backend = src_area->backend;
  546.     src_backend_data = src_area->backend_data;
  547.    
  548.     if (src_size != acc_size || (src_flags & dst_flags_mask) != dst_flags_mask) {
  549.         mutex_unlock(&src_area->lock);
  550.         mutex_unlock(&src_as->lock);
  551.         interrupts_restore(ipl);
  552.         return EPERM;
  553.     }
  554.  
  555.     /*
  556.      * Now we are committed to sharing the area.
  557.      * First prepare the area for sharing.
  558.      * Then it will be safe to unlock it.
  559.      */
  560.     sh_info = src_area->sh_info;
  561.     if (!sh_info) {
  562.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  563.         mutex_initialize(&sh_info->lock);
  564.         sh_info->refcount = 2;
  565.         btree_create(&sh_info->pagemap);
  566.         src_area->sh_info = sh_info;
  567.     } else {
  568.         mutex_lock(&sh_info->lock);
  569.         sh_info->refcount++;
  570.         mutex_unlock(&sh_info->lock);
  571.     }
  572.  
  573.     src_area->backend->share(src_area);
  574.  
  575.     mutex_unlock(&src_area->lock);
  576.     mutex_unlock(&src_as->lock);
  577.  
  578.     /*
  579.      * Create copy of the source address space area.
  580.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  581.      * attribute set which prevents race condition with
  582.      * preliminary as_page_fault() calls.
  583.      * The flags of the source area are masked against dst_flags_mask
  584.      * to support sharing in less privileged mode.
  585.      */
  586.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  587.                   AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  588.     if (!dst_area) {
  589.         /*
  590.          * Destination address space area could not be created.
  591.          */
  592.         sh_info_remove_reference(sh_info);
  593.        
  594.         interrupts_restore(ipl);
  595.         return ENOMEM;
  596.     }
  597.    
  598.     /*
  599.      * Now the destination address space area has been
  600.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  601.      * attribute and set the sh_info.
  602.      */
  603.     mutex_lock(&dst_area->lock);
  604.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  605.     dst_area->sh_info = sh_info;
  606.     mutex_unlock(&dst_area->lock);
  607.    
  608.     interrupts_restore(ipl);
  609.    
  610.     return 0;
  611. }
  612.  
  613. /** Check access mode for address space area.
  614.  *
  615.  * The address space area must be locked prior to this call.
  616.  *
  617.  * @param area Address space area.
  618.  * @param access Access mode.
  619.  *
  620.  * @return False if access violates area's permissions, true otherwise.
  621.  */
  622. bool as_area_check_access(as_area_t *area, pf_access_t access)
  623. {
  624.     int flagmap[] = {
  625.         [PF_ACCESS_READ] = AS_AREA_READ,
  626.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  627.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  628.     };
  629.  
  630.     if (!(area->flags & flagmap[access]))
  631.         return false;
  632.    
  633.     return true;
  634. }
  635.  
  636. /** Handle page fault within the current address space.
  637.  *
  638.  * This is the high-level page fault handler. It decides
  639.  * whether the page fault can be resolved by any backend
  640.  * and if so, it invokes the backend to resolve the page
  641.  * fault.
  642.  *
  643.  * Interrupts are assumed disabled.
  644.  *
  645.  * @param page Faulting page.
  646.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  647.  * @param istate Pointer to interrupted state.
  648.  *
  649.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  650.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  651.  */
  652. int as_page_fault(__address page, pf_access_t access, istate_t *istate)
  653. {
  654.     pte_t *pte;
  655.     as_area_t *area;
  656.    
  657.     if (!THREAD)
  658.         return AS_PF_FAULT;
  659.        
  660.     ASSERT(AS);
  661.  
  662.     mutex_lock(&AS->lock);
  663.     area = find_area_and_lock(AS, page);   
  664.     if (!area) {
  665.         /*
  666.          * No area contained mapping for 'page'.
  667.          * Signal page fault to low-level handler.
  668.          */
  669.         mutex_unlock(&AS->lock);
  670.         goto page_fault;
  671.     }
  672.  
  673.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  674.         /*
  675.          * The address space area is not fully initialized.
  676.          * Avoid possible race by returning error.
  677.          */
  678.         mutex_unlock(&area->lock);
  679.         mutex_unlock(&AS->lock);
  680.         goto page_fault;       
  681.     }
  682.  
  683.     if (!area->backend || !area->backend->page_fault) {
  684.         /*
  685.          * The address space area is not backed by any backend
  686.          * or the backend cannot handle page faults.
  687.          */
  688.         mutex_unlock(&area->lock);
  689.         mutex_unlock(&AS->lock);
  690.         goto page_fault;       
  691.     }
  692.  
  693.     page_table_lock(AS, false);
  694.    
  695.     /*
  696.      * To avoid race condition between two page faults
  697.      * on the same address, we need to make sure
  698.      * the mapping has not been already inserted.
  699.      */
  700.     if ((pte = page_mapping_find(AS, page))) {
  701.         if (PTE_PRESENT(pte)) {
  702.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  703.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  704.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  705.                 page_table_unlock(AS, false);
  706.                 mutex_unlock(&area->lock);
  707.                 mutex_unlock(&AS->lock);
  708.                 return AS_PF_OK;
  709.             }
  710.         }
  711.     }
  712.    
  713.     /*
  714.      * Resort to the backend page fault handler.
  715.      */
  716.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  717.         page_table_unlock(AS, false);
  718.         mutex_unlock(&area->lock);
  719.         mutex_unlock(&AS->lock);
  720.         goto page_fault;
  721.     }
  722.    
  723.     page_table_unlock(AS, false);
  724.     mutex_unlock(&area->lock);
  725.     mutex_unlock(&AS->lock);
  726.     return AS_PF_OK;
  727.  
  728. page_fault:
  729.     if (THREAD->in_copy_from_uspace) {
  730.         THREAD->in_copy_from_uspace = false;
  731.         istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
  732.     } else if (THREAD->in_copy_to_uspace) {
  733.         THREAD->in_copy_to_uspace = false;
  734.         istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
  735.     } else {
  736.         return AS_PF_FAULT;
  737.     }
  738.  
  739.     return AS_PF_DEFER;
  740. }
  741.  
  742. /** Switch address spaces.
  743.  *
  744.  * Note that this function cannot sleep as it is essentially a part of
  745.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  746.  *
  747.  * @param old Old address space or NULL.
  748.  * @param new New address space.
  749.  */
  750. void as_switch(as_t *old, as_t *new)
  751. {
  752.     ipl_t ipl;
  753.     bool needs_asid = false;
  754.    
  755.     ipl = interrupts_disable();
  756.     spinlock_lock(&inactive_as_with_asid_lock);
  757.  
  758.     /*
  759.      * First, take care of the old address space.
  760.      */
  761.     if (old) {
  762.         mutex_lock_active(&old->lock);
  763.         ASSERT(old->cpu_refcount);
  764.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  765.             /*
  766.              * The old address space is no longer active on
  767.              * any processor. It can be appended to the
  768.              * list of inactive address spaces with assigned
  769.              * ASID.
  770.              */
  771.              ASSERT(old->asid != ASID_INVALID);
  772.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  773.         }
  774.         mutex_unlock(&old->lock);
  775.     }
  776.  
  777.     /*
  778.      * Second, prepare the new address space.
  779.      */
  780.     mutex_lock_active(&new->lock);
  781.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  782.         if (new->asid != ASID_INVALID)
  783.             list_remove(&new->inactive_as_with_asid_link);
  784.         else
  785.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  786.     }
  787.     SET_PTL0_ADDRESS(new->page_table);
  788.     mutex_unlock(&new->lock);
  789.  
  790.     if (needs_asid) {
  791.         /*
  792.          * Allocation of new ASID was deferred
  793.          * until now in order to avoid deadlock.
  794.          */
  795.         asid_t asid;
  796.        
  797.         asid = asid_get();
  798.         mutex_lock_active(&new->lock);
  799.         new->asid = asid;
  800.         mutex_unlock(&new->lock);
  801.     }
  802.     spinlock_unlock(&inactive_as_with_asid_lock);
  803.     interrupts_restore(ipl);
  804.    
  805.     /*
  806.      * Perform architecture-specific steps.
  807.      * (e.g. write ASID to hardware register etc.)
  808.      */
  809.     as_install_arch(new);
  810.    
  811.     AS = new;
  812. }
  813.  
  814. /** Convert address space area flags to page flags.
  815.  *
  816.  * @param aflags Flags of some address space area.
  817.  *
  818.  * @return Flags to be passed to page_mapping_insert().
  819.  */
  820. int area_flags_to_page_flags(int aflags)
  821. {
  822.     int flags;
  823.  
  824.     flags = PAGE_USER | PAGE_PRESENT;
  825.    
  826.     if (aflags & AS_AREA_READ)
  827.         flags |= PAGE_READ;
  828.        
  829.     if (aflags & AS_AREA_WRITE)
  830.         flags |= PAGE_WRITE;
  831.    
  832.     if (aflags & AS_AREA_EXEC)
  833.         flags |= PAGE_EXEC;
  834.    
  835.     if (aflags & AS_AREA_CACHEABLE)
  836.         flags |= PAGE_CACHEABLE;
  837.        
  838.     return flags;
  839. }
  840.  
  841. /** Compute flags for virtual address translation subsytem.
  842.  *
  843.  * The address space area must be locked.
  844.  * Interrupts must be disabled.
  845.  *
  846.  * @param a Address space area.
  847.  *
  848.  * @return Flags to be used in page_mapping_insert().
  849.  */
  850. int as_area_get_flags(as_area_t *a)
  851. {
  852.     return area_flags_to_page_flags(a->flags);
  853. }
  854.  
  855. /** Create page table.
  856.  *
  857.  * Depending on architecture, create either address space
  858.  * private or global page table.
  859.  *
  860.  * @param flags Flags saying whether the page table is for kernel address space.
  861.  *
  862.  * @return First entry of the page table.
  863.  */
  864. pte_t *page_table_create(int flags)
  865. {
  866.         ASSERT(as_operations);
  867.         ASSERT(as_operations->page_table_create);
  868.  
  869.         return as_operations->page_table_create(flags);
  870. }
  871.  
  872. /** Destroy page table.
  873.  *
  874.  * Destroy page table in architecture specific way.
  875.  *
  876.  * @param page_table Physical address of PTL0.
  877.  */
  878. void page_table_destroy(pte_t *page_table)
  879. {
  880.         ASSERT(as_operations);
  881.         ASSERT(as_operations->page_table_destroy);
  882.  
  883.         as_operations->page_table_destroy(page_table);
  884. }
  885.  
  886. /** Lock page table.
  887.  *
  888.  * This function should be called before any page_mapping_insert(),
  889.  * page_mapping_remove() and page_mapping_find().
  890.  *
  891.  * Locking order is such that address space areas must be locked
  892.  * prior to this call. Address space can be locked prior to this
  893.  * call in which case the lock argument is false.
  894.  *
  895.  * @param as Address space.
  896.  * @param lock If false, do not attempt to lock as->lock.
  897.  */
  898. void page_table_lock(as_t *as, bool lock)
  899. {
  900.     ASSERT(as_operations);
  901.     ASSERT(as_operations->page_table_lock);
  902.  
  903.     as_operations->page_table_lock(as, lock);
  904. }
  905.  
  906. /** Unlock page table.
  907.  *
  908.  * @param as Address space.
  909.  * @param unlock If false, do not attempt to unlock as->lock.
  910.  */
  911. void page_table_unlock(as_t *as, bool unlock)
  912. {
  913.     ASSERT(as_operations);
  914.     ASSERT(as_operations->page_table_unlock);
  915.  
  916.     as_operations->page_table_unlock(as, unlock);
  917. }
  918.  
  919.  
  920. /** Find address space area and lock it.
  921.  *
  922.  * The address space must be locked and interrupts must be disabled.
  923.  *
  924.  * @param as Address space.
  925.  * @param va Virtual address.
  926.  *
  927.  * @return Locked address space area containing va on success or NULL on failure.
  928.  */
  929. as_area_t *find_area_and_lock(as_t *as, __address va)
  930. {
  931.     as_area_t *a;
  932.     btree_node_t *leaf, *lnode;
  933.     int i;
  934.    
  935.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  936.     if (a) {
  937.         /* va is the base address of an address space area */
  938.         mutex_lock(&a->lock);
  939.         return a;
  940.     }
  941.    
  942.     /*
  943.      * Search the leaf node and the righmost record of its left neighbour
  944.      * to find out whether this is a miss or va belongs to an address
  945.      * space area found there.
  946.      */
  947.    
  948.     /* First, search the leaf node itself. */
  949.     for (i = 0; i < leaf->keys; i++) {
  950.         a = (as_area_t *) leaf->value[i];
  951.         mutex_lock(&a->lock);
  952.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  953.             return a;
  954.         }
  955.         mutex_unlock(&a->lock);
  956.     }
  957.  
  958.     /*
  959.      * Second, locate the left neighbour and test its last record.
  960.      * Because of its position in the B+tree, it must have base < va.
  961.      */
  962.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  963.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  964.         mutex_lock(&a->lock);
  965.         if (va < a->base + a->pages * PAGE_SIZE) {
  966.             return a;
  967.         }
  968.         mutex_unlock(&a->lock);
  969.     }
  970.  
  971.     return NULL;
  972. }
  973.  
  974. /** Check area conflicts with other areas.
  975.  *
  976.  * The address space must be locked and interrupts must be disabled.
  977.  *
  978.  * @param as Address space.
  979.  * @param va Starting virtual address of the area being tested.
  980.  * @param size Size of the area being tested.
  981.  * @param avoid_area Do not touch this area.
  982.  *
  983.  * @return True if there is no conflict, false otherwise.
  984.  */
  985. bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area)
  986. {
  987.     as_area_t *a;
  988.     btree_node_t *leaf, *node;
  989.     int i;
  990.    
  991.     /*
  992.      * We don't want any area to have conflicts with NULL page.
  993.      */
  994.     if (overlaps(va, size, NULL, PAGE_SIZE))
  995.         return false;
  996.    
  997.     /*
  998.      * The leaf node is found in O(log n), where n is proportional to
  999.      * the number of address space areas belonging to as.
  1000.      * The check for conflicts is then attempted on the rightmost
  1001.      * record in the left neighbour, the leftmost record in the right
  1002.      * neighbour and all records in the leaf node itself.
  1003.      */
  1004.    
  1005.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1006.         if (a != avoid_area)
  1007.             return false;
  1008.     }
  1009.    
  1010.     /* First, check the two border cases. */
  1011.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1012.         a = (as_area_t *) node->value[node->keys - 1];
  1013.         mutex_lock(&a->lock);
  1014.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1015.             mutex_unlock(&a->lock);
  1016.             return false;
  1017.         }
  1018.         mutex_unlock(&a->lock);
  1019.     }
  1020.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1021.         a = (as_area_t *) node->value[0];
  1022.         mutex_lock(&a->lock);
  1023.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1024.             mutex_unlock(&a->lock);
  1025.             return false;
  1026.         }
  1027.         mutex_unlock(&a->lock);
  1028.     }
  1029.    
  1030.     /* Second, check the leaf node. */
  1031.     for (i = 0; i < leaf->keys; i++) {
  1032.         a = (as_area_t *) leaf->value[i];
  1033.    
  1034.         if (a == avoid_area)
  1035.             continue;
  1036.    
  1037.         mutex_lock(&a->lock);
  1038.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1039.             mutex_unlock(&a->lock);
  1040.             return false;
  1041.         }
  1042.         mutex_unlock(&a->lock);
  1043.     }
  1044.  
  1045.     /*
  1046.      * So far, the area does not conflict with other areas.
  1047.      * Check if it doesn't conflict with kernel address space.
  1048.      */  
  1049.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1050.         return !overlaps(va, size,
  1051.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1052.     }
  1053.  
  1054.     return true;
  1055. }
  1056.  
  1057. /** Return size of the address space area with given base.  */
  1058. size_t as_get_size(__address base)
  1059. {
  1060.     ipl_t ipl;
  1061.     as_area_t *src_area;
  1062.     size_t size;
  1063.  
  1064.     ipl = interrupts_disable();
  1065.     src_area = find_area_and_lock(AS, base);
  1066.     if (src_area){
  1067.         size = src_area->pages * PAGE_SIZE;
  1068.         mutex_unlock(&src_area->lock);
  1069.     } else {
  1070.         size = 0;
  1071.     }
  1072.     interrupts_restore(ipl);
  1073.     return size;
  1074. }
  1075.  
  1076. /** Mark portion of address space area as used.
  1077.  *
  1078.  * The address space area must be already locked.
  1079.  *
  1080.  * @param a Address space area.
  1081.  * @param page First page to be marked.
  1082.  * @param count Number of page to be marked.
  1083.  *
  1084.  * @return 0 on failure and 1 on success.
  1085.  */
  1086. int used_space_insert(as_area_t *a, __address page, count_t count)
  1087. {
  1088.     btree_node_t *leaf, *node;
  1089.     count_t pages;
  1090.     int i;
  1091.  
  1092.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1093.     ASSERT(count);
  1094.  
  1095.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1096.     if (pages) {
  1097.         /*
  1098.          * We hit the beginning of some used space.
  1099.          */
  1100.         return 0;
  1101.     }
  1102.  
  1103.     if (!leaf->keys) {
  1104.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1105.         return 1;
  1106.     }
  1107.  
  1108.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1109.     if (node) {
  1110.         __address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1111.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1112.        
  1113.         /*
  1114.          * Examine the possibility that the interval fits
  1115.          * somewhere between the rightmost interval of
  1116.          * the left neigbour and the first interval of the leaf.
  1117.          */
  1118.          
  1119.         if (page >= right_pg) {
  1120.             /* Do nothing. */
  1121.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1122.             /* The interval intersects with the left interval. */
  1123.             return 0;
  1124.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1125.             /* The interval intersects with the right interval. */
  1126.             return 0;          
  1127.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1128.             /* The interval can be added by merging the two already present intervals. */
  1129.             node->value[node->keys - 1] += count + right_cnt;
  1130.             btree_remove(&a->used_space, right_pg, leaf);
  1131.             return 1;
  1132.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1133.             /* The interval can be added by simply growing the left interval. */
  1134.             node->value[node->keys - 1] += count;
  1135.             return 1;
  1136.         } else if (page + count*PAGE_SIZE == right_pg) {
  1137.             /*
  1138.              * The interval can be addded by simply moving base of the right
  1139.              * interval down and increasing its size accordingly.
  1140.              */
  1141.             leaf->value[0] += count;
  1142.             leaf->key[0] = page;
  1143.             return 1;
  1144.         } else {
  1145.             /*
  1146.              * The interval is between both neigbouring intervals,
  1147.              * but cannot be merged with any of them.
  1148.              */
  1149.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1150.             return 1;
  1151.         }
  1152.     } else if (page < leaf->key[0]) {
  1153.         __address right_pg = leaf->key[0];
  1154.         count_t right_cnt = (count_t) leaf->value[0];
  1155.    
  1156.         /*
  1157.          * Investigate the border case in which the left neighbour does not
  1158.          * exist but the interval fits from the left.
  1159.          */
  1160.          
  1161.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1162.             /* The interval intersects with the right interval. */
  1163.             return 0;
  1164.         } else if (page + count*PAGE_SIZE == right_pg) {
  1165.             /*
  1166.              * The interval can be added by moving the base of the right interval down
  1167.              * and increasing its size accordingly.
  1168.              */
  1169.             leaf->key[0] = page;
  1170.             leaf->value[0] += count;
  1171.             return 1;
  1172.         } else {
  1173.             /*
  1174.              * The interval doesn't adjoin with the right interval.
  1175.              * It must be added individually.
  1176.              */
  1177.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1178.             return 1;
  1179.         }
  1180.     }
  1181.  
  1182.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1183.     if (node) {
  1184.         __address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1185.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1186.        
  1187.         /*
  1188.          * Examine the possibility that the interval fits
  1189.          * somewhere between the leftmost interval of
  1190.          * the right neigbour and the last interval of the leaf.
  1191.          */
  1192.  
  1193.         if (page < left_pg) {
  1194.             /* Do nothing. */
  1195.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1196.             /* The interval intersects with the left interval. */
  1197.             return 0;
  1198.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1199.             /* The interval intersects with the right interval. */
  1200.             return 0;          
  1201.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1202.             /* The interval can be added by merging the two already present intervals. */
  1203.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1204.             btree_remove(&a->used_space, right_pg, node);
  1205.             return 1;
  1206.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1207.             /* The interval can be added by simply growing the left interval. */
  1208.             leaf->value[leaf->keys - 1] +=  count;
  1209.             return 1;
  1210.         } else if (page + count*PAGE_SIZE == right_pg) {
  1211.             /*
  1212.              * The interval can be addded by simply moving base of the right
  1213.              * interval down and increasing its size accordingly.
  1214.              */
  1215.             node->value[0] += count;
  1216.             node->key[0] = page;
  1217.             return 1;
  1218.         } else {
  1219.             /*
  1220.              * The interval is between both neigbouring intervals,
  1221.              * but cannot be merged with any of them.
  1222.              */
  1223.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1224.             return 1;
  1225.         }
  1226.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1227.         __address left_pg = leaf->key[leaf->keys - 1];
  1228.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1229.    
  1230.         /*
  1231.          * Investigate the border case in which the right neighbour does not
  1232.          * exist but the interval fits from the right.
  1233.          */
  1234.          
  1235.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1236.             /* The interval intersects with the left interval. */
  1237.             return 0;
  1238.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1239.             /* The interval can be added by growing the left interval. */
  1240.             leaf->value[leaf->keys - 1] += count;
  1241.             return 1;
  1242.         } else {
  1243.             /*
  1244.              * The interval doesn't adjoin with the left interval.
  1245.              * It must be added individually.
  1246.              */
  1247.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1248.             return 1;
  1249.         }
  1250.     }
  1251.    
  1252.     /*
  1253.      * Note that if the algorithm made it thus far, the interval can fit only
  1254.      * between two other intervals of the leaf. The two border cases were already
  1255.      * resolved.
  1256.      */
  1257.     for (i = 1; i < leaf->keys; i++) {
  1258.         if (page < leaf->key[i]) {
  1259.             __address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1260.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1261.  
  1262.             /*
  1263.              * The interval fits between left_pg and right_pg.
  1264.              */
  1265.  
  1266.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1267.                 /* The interval intersects with the left interval. */
  1268.                 return 0;
  1269.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1270.                 /* The interval intersects with the right interval. */
  1271.                 return 0;          
  1272.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1273.                 /* The interval can be added by merging the two already present intervals. */
  1274.                 leaf->value[i - 1] += count + right_cnt;
  1275.                 btree_remove(&a->used_space, right_pg, leaf);
  1276.                 return 1;
  1277.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1278.                 /* The interval can be added by simply growing the left interval. */
  1279.                 leaf->value[i - 1] += count;
  1280.                 return 1;
  1281.             } else if (page + count*PAGE_SIZE == right_pg) {
  1282.                 /*
  1283.                      * The interval can be addded by simply moving base of the right
  1284.                  * interval down and increasing its size accordingly.
  1285.                  */
  1286.                 leaf->value[i] += count;
  1287.                 leaf->key[i] = page;
  1288.                 return 1;
  1289.             } else {
  1290.                 /*
  1291.                  * The interval is between both neigbouring intervals,
  1292.                  * but cannot be merged with any of them.
  1293.                  */
  1294.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1295.                 return 1;
  1296.             }
  1297.         }
  1298.     }
  1299.  
  1300.     panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
  1301. }
  1302.  
  1303. /** Mark portion of address space area as unused.
  1304.  *
  1305.  * The address space area must be already locked.
  1306.  *
  1307.  * @param a Address space area.
  1308.  * @param page First page to be marked.
  1309.  * @param count Number of page to be marked.
  1310.  *
  1311.  * @return 0 on failure and 1 on success.
  1312.  */
  1313. int used_space_remove(as_area_t *a, __address page, count_t count)
  1314. {
  1315.     btree_node_t *leaf, *node;
  1316.     count_t pages;
  1317.     int i;
  1318.  
  1319.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1320.     ASSERT(count);
  1321.  
  1322.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1323.     if (pages) {
  1324.         /*
  1325.          * We are lucky, page is the beginning of some interval.
  1326.          */
  1327.         if (count > pages) {
  1328.             return 0;
  1329.         } else if (count == pages) {
  1330.             btree_remove(&a->used_space, page, leaf);
  1331.             return 1;
  1332.         } else {
  1333.             /*
  1334.              * Find the respective interval.
  1335.              * Decrease its size and relocate its start address.
  1336.              */
  1337.             for (i = 0; i < leaf->keys; i++) {
  1338.                 if (leaf->key[i] == page) {
  1339.                     leaf->key[i] += count*PAGE_SIZE;
  1340.                     leaf->value[i] -= count;
  1341.                     return 1;
  1342.                 }
  1343.             }
  1344.             goto error;
  1345.         }
  1346.     }
  1347.  
  1348.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1349.     if (node && page < leaf->key[0]) {
  1350.         __address left_pg = node->key[node->keys - 1];
  1351.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1352.  
  1353.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1354.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1355.                 /*
  1356.                  * The interval is contained in the rightmost interval
  1357.                  * of the left neighbour and can be removed by
  1358.                  * updating the size of the bigger interval.
  1359.                  */
  1360.                 node->value[node->keys - 1] -= count;
  1361.                 return 1;
  1362.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1363.                 count_t new_cnt;
  1364.                
  1365.                 /*
  1366.                  * The interval is contained in the rightmost interval
  1367.                  * of the left neighbour but its removal requires
  1368.                  * both updating the size of the original interval and
  1369.                  * also inserting a new interval.
  1370.                  */
  1371.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1372.                 node->value[node->keys - 1] -= count + new_cnt;
  1373.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1374.                 return 1;
  1375.             }
  1376.         }
  1377.         return 0;
  1378.     } else if (page < leaf->key[0]) {
  1379.         return 0;
  1380.     }
  1381.    
  1382.     if (page > leaf->key[leaf->keys - 1]) {
  1383.         __address left_pg = leaf->key[leaf->keys - 1];
  1384.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1385.  
  1386.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1387.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1388.                 /*
  1389.                  * The interval is contained in the rightmost interval
  1390.                  * of the leaf and can be removed by updating the size
  1391.                  * of the bigger interval.
  1392.                  */
  1393.                 leaf->value[leaf->keys - 1] -= count;
  1394.                 return 1;
  1395.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1396.                 count_t new_cnt;
  1397.                
  1398.                 /*
  1399.                  * The interval is contained in the rightmost interval
  1400.                  * of the leaf but its removal requires both updating
  1401.                  * the size of the original interval and
  1402.                  * also inserting a new interval.
  1403.                  */
  1404.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1405.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1406.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1407.                 return 1;
  1408.             }
  1409.         }
  1410.         return 0;
  1411.     }  
  1412.    
  1413.     /*
  1414.      * The border cases have been already resolved.
  1415.      * Now the interval can be only between intervals of the leaf.
  1416.      */
  1417.     for (i = 1; i < leaf->keys - 1; i++) {
  1418.         if (page < leaf->key[i]) {
  1419.             __address left_pg = leaf->key[i - 1];
  1420.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1421.  
  1422.             /*
  1423.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1424.              */
  1425.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1426.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1427.                     /*
  1428.                     * The interval is contained in the interval (i - 1)
  1429.                      * of the leaf and can be removed by updating the size
  1430.                      * of the bigger interval.
  1431.                      */
  1432.                     leaf->value[i - 1] -= count;
  1433.                     return 1;
  1434.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1435.                     count_t new_cnt;
  1436.                
  1437.                     /*
  1438.                      * The interval is contained in the interval (i - 1)
  1439.                      * of the leaf but its removal requires both updating
  1440.                      * the size of the original interval and
  1441.                      * also inserting a new interval.
  1442.                      */
  1443.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1444.                     leaf->value[i - 1] -= count + new_cnt;
  1445.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1446.                     return 1;
  1447.                 }
  1448.             }
  1449.             return 0;
  1450.         }
  1451.     }
  1452.  
  1453. error:
  1454.     panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
  1455. }
  1456.  
  1457. /** Remove reference to address space area share info.
  1458.  *
  1459.  * If the reference count drops to 0, the sh_info is deallocated.
  1460.  *
  1461.  * @param sh_info Pointer to address space area share info.
  1462.  */
  1463. void sh_info_remove_reference(share_info_t *sh_info)
  1464. {
  1465.     bool dealloc = false;
  1466.  
  1467.     mutex_lock(&sh_info->lock);
  1468.     ASSERT(sh_info->refcount);
  1469.     if (--sh_info->refcount == 0) {
  1470.         dealloc = true;
  1471.         link_t *cur;
  1472.        
  1473.         /*
  1474.          * Now walk carefully the pagemap B+tree and free/remove
  1475.          * reference from all frames found there.
  1476.          */
  1477.         for (cur = sh_info->pagemap.leaf_head.next; cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1478.             btree_node_t *node;
  1479.             int i;
  1480.            
  1481.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1482.             for (i = 0; i < node->keys; i++)
  1483.                 frame_free(ADDR2PFN((__address) node->value[i]));
  1484.         }
  1485.        
  1486.     }
  1487.     mutex_unlock(&sh_info->lock);
  1488.    
  1489.     if (dealloc) {
  1490.         btree_destroy(&sh_info->pagemap);
  1491.         free(sh_info);
  1492.     }
  1493. }
  1494.  
  1495. /*
  1496.  * Address space related syscalls.
  1497.  */
  1498.  
  1499. /** Wrapper for as_area_create(). */
  1500. __native sys_as_area_create(__address address, size_t size, int flags)
  1501. {
  1502.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1503.         return (__native) address;
  1504.     else
  1505.         return (__native) -1;
  1506. }
  1507.  
  1508. /** Wrapper for as_area_resize. */
  1509. __native sys_as_area_resize(__address address, size_t size, int flags)
  1510. {
  1511.     return (__native) as_area_resize(AS, address, size, 0);
  1512. }
  1513.  
  1514. /** Wrapper for as_area_destroy. */
  1515. __native sys_as_area_destroy(__address address)
  1516. {
  1517.     return (__native) as_area_destroy(AS, address);
  1518. }
  1519.