Subversion Repositories HelenOS-historic

Rev

Rev 1587 | Rev 1705 | 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.     bool cond;
  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.     ipl = interrupts_disable();
  148.     spinlock_lock(&inactive_as_with_asid_lock);
  149.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  150.         if (as != AS && as->cpu_refcount == 0)
  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.      * The B+tee must be walked carefully because it is
  159.      * also being destroyed.
  160.      */
  161.     for (cond = true; cond; ) {
  162.         btree_node_t *node;
  163.  
  164.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  165.         node = list_get_instance(as->as_area_btree.leaf_head.next, btree_node_t, leaf_link);
  166.  
  167.         if ((cond = node->keys)) {
  168.             as_area_destroy(as, node->key[0]);
  169.         }
  170.     }
  171.  
  172.     btree_destroy(&as->as_area_btree);
  173.     page_table_destroy(as->page_table);
  174.  
  175.     interrupts_restore(ipl);
  176.    
  177.     free(as);
  178. }
  179.  
  180. /** Create address space area of common attributes.
  181.  *
  182.  * The created address space area is added to the target address space.
  183.  *
  184.  * @param as Target address space.
  185.  * @param flags Flags of the area memory.
  186.  * @param size Size of area.
  187.  * @param base Base address of area.
  188.  * @param attrs Attributes of the area.
  189.  * @param backend Address space area backend. NULL if no backend is used.
  190.  * @param backend_data NULL or a pointer to an array holding two void *.
  191.  *
  192.  * @return Address space area on success or NULL on failure.
  193.  */
  194. as_area_t *as_area_create(as_t *as, int flags, size_t size, __address base, int attrs,
  195.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  196. {
  197.     ipl_t ipl;
  198.     as_area_t *a;
  199.    
  200.     if (base % PAGE_SIZE)
  201.         return NULL;
  202.  
  203.     if (!size)
  204.         return NULL;
  205.  
  206.     /* Writeable executable areas are not supported. */
  207.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  208.         return NULL;
  209.    
  210.     ipl = interrupts_disable();
  211.     mutex_lock(&as->lock);
  212.    
  213.     if (!check_area_conflicts(as, base, size, NULL)) {
  214.         mutex_unlock(&as->lock);
  215.         interrupts_restore(ipl);
  216.         return NULL;
  217.     }
  218.    
  219.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  220.  
  221.     mutex_initialize(&a->lock);
  222.    
  223.     a->as = as;
  224.     a->flags = flags;
  225.     a->attributes = attrs;
  226.     a->pages = SIZE2FRAMES(size);
  227.     a->base = base;
  228.     a->sh_info = NULL;
  229.     a->backend = backend;
  230.     if (backend_data)
  231.         a->backend_data = *backend_data;
  232.     else
  233.         memsetb((__address) &a->backend_data, sizeof(a->backend_data), 0);
  234.  
  235.     btree_create(&a->used_space);
  236.    
  237.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  238.  
  239.     mutex_unlock(&as->lock);
  240.     interrupts_restore(ipl);
  241.  
  242.     return a;
  243. }
  244.  
  245. /** Find address space area and change it.
  246.  *
  247.  * @param as Address space.
  248.  * @param address Virtual address belonging to the area to be changed. Must be page-aligned.
  249.  * @param size New size of the virtual memory block starting at address.
  250.  * @param flags Flags influencing the remap operation. Currently unused.
  251.  *
  252.  * @return Zero on success or a value from @ref errno.h otherwise.
  253.  */
  254. int as_area_resize(as_t *as, __address address, size_t size, int flags)
  255. {
  256.     as_area_t *area;
  257.     ipl_t ipl;
  258.     size_t pages;
  259.    
  260.     ipl = interrupts_disable();
  261.     mutex_lock(&as->lock);
  262.    
  263.     /*
  264.      * Locate the area.
  265.      */
  266.     area = find_area_and_lock(as, address);
  267.     if (!area) {
  268.         mutex_unlock(&as->lock);
  269.         interrupts_restore(ipl);
  270.         return ENOENT;
  271.     }
  272.  
  273.     if (area->backend == &phys_backend) {
  274.         /*
  275.          * Remapping of address space areas associated
  276.          * with memory mapped devices is not supported.
  277.          */
  278.         mutex_unlock(&area->lock);
  279.         mutex_unlock(&as->lock);
  280.         interrupts_restore(ipl);
  281.         return ENOTSUP;
  282.     }
  283.     if (area->sh_info) {
  284.         /*
  285.          * Remapping of shared address space areas
  286.          * is not supported.
  287.          */
  288.         mutex_unlock(&area->lock);
  289.         mutex_unlock(&as->lock);
  290.         interrupts_restore(ipl);
  291.         return ENOTSUP;
  292.     }
  293.  
  294.     pages = SIZE2FRAMES((address - area->base) + size);
  295.     if (!pages) {
  296.         /*
  297.          * Zero size address space areas are not allowed.
  298.          */
  299.         mutex_unlock(&area->lock);
  300.         mutex_unlock(&as->lock);
  301.         interrupts_restore(ipl);
  302.         return EPERM;
  303.     }
  304.    
  305.     if (pages < area->pages) {
  306.         bool cond;
  307.         __address start_free = area->base + pages*PAGE_SIZE;
  308.  
  309.         /*
  310.          * Shrinking the area.
  311.          * No need to check for overlaps.
  312.          */
  313.  
  314.         /*
  315.          * Start TLB shootdown sequence.
  316.          */
  317.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  318.  
  319.         /*
  320.          * Remove frames belonging to used space starting from
  321.          * the highest addresses downwards until an overlap with
  322.          * the resized address space area is found. Note that this
  323.          * is also the right way to remove part of the used_space
  324.          * B+tree leaf list.
  325.          */    
  326.         for (cond = true; cond;) {
  327.             btree_node_t *node;
  328.        
  329.             ASSERT(!list_empty(&area->used_space.leaf_head));
  330.             node = list_get_instance(area->used_space.leaf_head.prev, btree_node_t, leaf_link);
  331.             if ((cond = (bool) node->keys)) {
  332.                 __address b = node->key[node->keys - 1];
  333.                 count_t c = (count_t) node->value[node->keys - 1];
  334.                 int i = 0;
  335.            
  336.                 if (overlaps(b, c*PAGE_SIZE, area->base, pages*PAGE_SIZE)) {
  337.                    
  338.                     if (b + c*PAGE_SIZE <= start_free) {
  339.                         /*
  340.                          * The whole interval fits completely
  341.                          * in the resized address space area.
  342.                          */
  343.                         break;
  344.                     }
  345.        
  346.                     /*
  347.                      * Part of the interval corresponding to b and c
  348.                      * overlaps with the resized address space area.
  349.                      */
  350.        
  351.                     cond = false;   /* we are almost done */
  352.                     i = (start_free - b) >> PAGE_WIDTH;
  353.                     if (!used_space_remove(area, start_free, c - i))
  354.                         panic("Could not remove used space.");
  355.                 } else {
  356.                     /*
  357.                      * The interval of used space can be completely removed.
  358.                      */
  359.                     if (!used_space_remove(area, b, c))
  360.                         panic("Could not remove used space.\n");
  361.                 }
  362.            
  363.                 for (; i < c; i++) {
  364.                     pte_t *pte;
  365.            
  366.                     page_table_lock(as, false);
  367.                     pte = page_mapping_find(as, b + i*PAGE_SIZE);
  368.                     ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  369.                     if (area->backend && area->backend->frame_free) {
  370.                         area->backend->frame_free(area,
  371.                             b + i*PAGE_SIZE, PTE_GET_FRAME(pte));
  372.                     }
  373.                     page_mapping_remove(as, b + i*PAGE_SIZE);
  374.                     page_table_unlock(as, false);
  375.                 }
  376.             }
  377.         }
  378.  
  379.         /*
  380.          * Finish TLB shootdown sequence.
  381.          */
  382.         tlb_invalidate_pages(AS->asid, area->base + pages*PAGE_SIZE, area->pages - pages);
  383.         tlb_shootdown_finalize();
  384.     } else {
  385.         /*
  386.          * Growing the area.
  387.          * Check for overlaps with other address space areas.
  388.          */
  389.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE, area)) {
  390.             mutex_unlock(&area->lock);
  391.             mutex_unlock(&as->lock);       
  392.             interrupts_restore(ipl);
  393.             return EADDRNOTAVAIL;
  394.         }
  395.     }
  396.  
  397.     area->pages = pages;
  398.    
  399.     mutex_unlock(&area->lock);
  400.     mutex_unlock(&as->lock);
  401.     interrupts_restore(ipl);
  402.  
  403.     return 0;
  404. }
  405.  
  406. /** Destroy address space area.
  407.  *
  408.  * @param as Address space.
  409.  * @param address Address withing the area to be deleted.
  410.  *
  411.  * @return Zero on success or a value from @ref errno.h on failure.
  412.  */
  413. int as_area_destroy(as_t *as, __address address)
  414. {
  415.     as_area_t *area;
  416.     __address base;
  417.     link_t *cur;
  418.     ipl_t ipl;
  419.  
  420.     ipl = interrupts_disable();
  421.     mutex_lock(&as->lock);
  422.  
  423.     area = find_area_and_lock(as, address);
  424.     if (!area) {
  425.         mutex_unlock(&as->lock);
  426.         interrupts_restore(ipl);
  427.         return ENOENT;
  428.     }
  429.  
  430.     base = area->base;
  431.  
  432.     /*
  433.      * Start TLB shootdown sequence.
  434.      */
  435.     tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base, area->pages);
  436.  
  437.     /*
  438.      * Visit only the pages mapped by used_space B+tree.
  439.      */
  440.     for (cur = area->used_space.leaf_head.next; cur != &area->used_space.leaf_head; cur = cur->next) {
  441.         btree_node_t *node;
  442.         int i;
  443.        
  444.         node = list_get_instance(cur, btree_node_t, leaf_link);
  445.         for (i = 0; i < node->keys; i++) {
  446.             __address b = node->key[i];
  447.             count_t j;
  448.             pte_t *pte;
  449.            
  450.             for (j = 0; j < (count_t) node->value[i]; j++) {
  451.                 page_table_lock(as, false);
  452.                 pte = page_mapping_find(as, b + j*PAGE_SIZE);
  453.                 ASSERT(pte && PTE_VALID(pte) && PTE_PRESENT(pte));
  454.                 if (area->backend && area->backend->frame_free) {
  455.                     area->backend->frame_free(area,
  456.                         b + j*PAGE_SIZE, PTE_GET_FRAME(pte));
  457.                 }
  458.                 page_mapping_remove(as, b + j*PAGE_SIZE);
  459.                 page_table_unlock(as, false);
  460.             }
  461.         }
  462.     }
  463.  
  464.     /*
  465.      * Finish TLB shootdown sequence.
  466.      */
  467.     tlb_invalidate_pages(AS->asid, area->base, area->pages);
  468.     tlb_shootdown_finalize();
  469.    
  470.     btree_destroy(&area->used_space);
  471.  
  472.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  473.    
  474.     if (area->sh_info)
  475.         sh_info_remove_reference(area->sh_info);
  476.        
  477.     mutex_unlock(&area->lock);
  478.  
  479.     /*
  480.      * Remove the empty area from address space.
  481.      */
  482.     btree_remove(&AS->as_area_btree, base, NULL);
  483.    
  484.     free(area);
  485.    
  486.     mutex_unlock(&AS->lock);
  487.     interrupts_restore(ipl);
  488.     return 0;
  489. }
  490.  
  491. /** Share address space area with another or the same address space.
  492.  *
  493.  * Address space area mapping is shared with a new address space area.
  494.  * If the source address space area has not been shared so far,
  495.  * a new sh_info is created. The new address space area simply gets the
  496.  * sh_info of the source area. The process of duplicating the
  497.  * mapping is done through the backend share function.
  498.  *
  499.  * @param src_as Pointer to source address space.
  500.  * @param src_base Base address of the source address space area.
  501.  * @param acc_size Expected size of the source area.
  502.  * @param dst_as Pointer to destination address space.
  503.  * @param dst_base Target base address.
  504.  * @param dst_flags_mask Destination address space area flags mask.
  505.  *
  506.  * @return Zero on success or ENOENT if there is no such task or
  507.  *     if there is no such address space area,
  508.  *     EPERM if there was a problem in accepting the area or
  509.  *     ENOMEM if there was a problem in allocating destination
  510.  *     address space area. ENOTSUP is returned if an attempt
  511.  *     to share non-anonymous address space area is detected.
  512.  */
  513. int as_area_share(as_t *src_as, __address src_base, size_t acc_size,
  514.           as_t *dst_as, __address dst_base, int dst_flags_mask)
  515. {
  516.     ipl_t ipl;
  517.     int src_flags;
  518.     size_t src_size;
  519.     as_area_t *src_area, *dst_area;
  520.     share_info_t *sh_info;
  521.     mem_backend_t *src_backend;
  522.     mem_backend_data_t src_backend_data;
  523.    
  524.     ipl = interrupts_disable();
  525.     mutex_lock(&src_as->lock);
  526.     src_area = find_area_and_lock(src_as, src_base);
  527.     if (!src_area) {
  528.         /*
  529.          * Could not find the source address space area.
  530.          */
  531.         mutex_unlock(&src_as->lock);
  532.         interrupts_restore(ipl);
  533.         return ENOENT;
  534.     }
  535.    
  536.     if (!src_area->backend || !src_area->backend->share) {
  537.         /*
  538.          * There is now backend or the backend does not
  539.          * know how to share the area.
  540.          */
  541.         mutex_unlock(&src_area->lock);
  542.         mutex_unlock(&src_as->lock);
  543.         interrupts_restore(ipl);
  544.         return ENOTSUP;
  545.     }
  546.    
  547.     src_size = src_area->pages * PAGE_SIZE;
  548.     src_flags = src_area->flags;
  549.     src_backend = src_area->backend;
  550.     src_backend_data = src_area->backend_data;
  551.  
  552.     /* Share the cacheable flag from the original mapping */
  553.     if (src_flags & AS_AREA_CACHEABLE)
  554.         dst_flags_mask |= AS_AREA_CACHEABLE;
  555.  
  556.     if (src_size != acc_size || (src_flags & dst_flags_mask) != dst_flags_mask) {
  557.         mutex_unlock(&src_area->lock);
  558.         mutex_unlock(&src_as->lock);
  559.         interrupts_restore(ipl);
  560.         return EPERM;
  561.     }
  562.  
  563.     /*
  564.      * Now we are committed to sharing the area.
  565.      * First prepare the area for sharing.
  566.      * Then it will be safe to unlock it.
  567.      */
  568.     sh_info = src_area->sh_info;
  569.     if (!sh_info) {
  570.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  571.         mutex_initialize(&sh_info->lock);
  572.         sh_info->refcount = 2;
  573.         btree_create(&sh_info->pagemap);
  574.         src_area->sh_info = sh_info;
  575.     } else {
  576.         mutex_lock(&sh_info->lock);
  577.         sh_info->refcount++;
  578.         mutex_unlock(&sh_info->lock);
  579.     }
  580.  
  581.     src_area->backend->share(src_area);
  582.  
  583.     mutex_unlock(&src_area->lock);
  584.     mutex_unlock(&src_as->lock);
  585.  
  586.     /*
  587.      * Create copy of the source address space area.
  588.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  589.      * attribute set which prevents race condition with
  590.      * preliminary as_page_fault() calls.
  591.      * The flags of the source area are masked against dst_flags_mask
  592.      * to support sharing in less privileged mode.
  593.      */
  594.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  595.                   AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  596.     if (!dst_area) {
  597.         /*
  598.          * Destination address space area could not be created.
  599.          */
  600.         sh_info_remove_reference(sh_info);
  601.        
  602.         interrupts_restore(ipl);
  603.         return ENOMEM;
  604.     }
  605.    
  606.     /*
  607.      * Now the destination address space area has been
  608.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  609.      * attribute and set the sh_info.
  610.      */
  611.     mutex_lock(&dst_area->lock);
  612.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  613.     dst_area->sh_info = sh_info;
  614.     mutex_unlock(&dst_area->lock);
  615.    
  616.     interrupts_restore(ipl);
  617.    
  618.     return 0;
  619. }
  620.  
  621. /** Check access mode for address space area.
  622.  *
  623.  * The address space area must be locked prior to this call.
  624.  *
  625.  * @param area Address space area.
  626.  * @param access Access mode.
  627.  *
  628.  * @return False if access violates area's permissions, true otherwise.
  629.  */
  630. bool as_area_check_access(as_area_t *area, pf_access_t access)
  631. {
  632.     int flagmap[] = {
  633.         [PF_ACCESS_READ] = AS_AREA_READ,
  634.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  635.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  636.     };
  637.  
  638.     if (!(area->flags & flagmap[access]))
  639.         return false;
  640.    
  641.     return true;
  642. }
  643.  
  644. /** Handle page fault within the current address space.
  645.  *
  646.  * This is the high-level page fault handler. It decides
  647.  * whether the page fault can be resolved by any backend
  648.  * and if so, it invokes the backend to resolve the page
  649.  * fault.
  650.  *
  651.  * Interrupts are assumed disabled.
  652.  *
  653.  * @param page Faulting page.
  654.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  655.  * @param istate Pointer to interrupted state.
  656.  *
  657.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  658.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  659.  */
  660. int as_page_fault(__address page, pf_access_t access, istate_t *istate)
  661. {
  662.     pte_t *pte;
  663.     as_area_t *area;
  664.    
  665.     if (!THREAD)
  666.         return AS_PF_FAULT;
  667.        
  668.     ASSERT(AS);
  669.  
  670.     mutex_lock(&AS->lock);
  671.     area = find_area_and_lock(AS, page);   
  672.     if (!area) {
  673.         /*
  674.          * No area contained mapping for 'page'.
  675.          * Signal page fault to low-level handler.
  676.          */
  677.         mutex_unlock(&AS->lock);
  678.         goto page_fault;
  679.     }
  680.  
  681.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  682.         /*
  683.          * The address space area is not fully initialized.
  684.          * Avoid possible race by returning error.
  685.          */
  686.         mutex_unlock(&area->lock);
  687.         mutex_unlock(&AS->lock);
  688.         goto page_fault;       
  689.     }
  690.  
  691.     if (!area->backend || !area->backend->page_fault) {
  692.         /*
  693.          * The address space area is not backed by any backend
  694.          * or the backend cannot handle page faults.
  695.          */
  696.         mutex_unlock(&area->lock);
  697.         mutex_unlock(&AS->lock);
  698.         goto page_fault;       
  699.     }
  700.  
  701.     page_table_lock(AS, false);
  702.    
  703.     /*
  704.      * To avoid race condition between two page faults
  705.      * on the same address, we need to make sure
  706.      * the mapping has not been already inserted.
  707.      */
  708.     if ((pte = page_mapping_find(AS, page))) {
  709.         if (PTE_PRESENT(pte)) {
  710.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  711.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  712.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  713.                 page_table_unlock(AS, false);
  714.                 mutex_unlock(&area->lock);
  715.                 mutex_unlock(&AS->lock);
  716.                 return AS_PF_OK;
  717.             }
  718.         }
  719.     }
  720.    
  721.     /*
  722.      * Resort to the backend page fault handler.
  723.      */
  724.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  725.         page_table_unlock(AS, false);
  726.         mutex_unlock(&area->lock);
  727.         mutex_unlock(&AS->lock);
  728.         goto page_fault;
  729.     }
  730.    
  731.     page_table_unlock(AS, false);
  732.     mutex_unlock(&area->lock);
  733.     mutex_unlock(&AS->lock);
  734.     return AS_PF_OK;
  735.  
  736. page_fault:
  737.     if (THREAD->in_copy_from_uspace) {
  738.         THREAD->in_copy_from_uspace = false;
  739.         istate_set_retaddr(istate, (__address) &memcpy_from_uspace_failover_address);
  740.     } else if (THREAD->in_copy_to_uspace) {
  741.         THREAD->in_copy_to_uspace = false;
  742.         istate_set_retaddr(istate, (__address) &memcpy_to_uspace_failover_address);
  743.     } else {
  744.         return AS_PF_FAULT;
  745.     }
  746.  
  747.     return AS_PF_DEFER;
  748. }
  749.  
  750. /** Switch address spaces.
  751.  *
  752.  * Note that this function cannot sleep as it is essentially a part of
  753.  * scheduling. Sleeping here would lead to deadlock on wakeup.
  754.  *
  755.  * @param old Old address space or NULL.
  756.  * @param new New address space.
  757.  */
  758. void as_switch(as_t *old, as_t *new)
  759. {
  760.     ipl_t ipl;
  761.     bool needs_asid = false;
  762.    
  763.     ipl = interrupts_disable();
  764.     spinlock_lock(&inactive_as_with_asid_lock);
  765.  
  766.     /*
  767.      * First, take care of the old address space.
  768.      */
  769.     if (old) {
  770.         mutex_lock_active(&old->lock);
  771.         ASSERT(old->cpu_refcount);
  772.         if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
  773.             /*
  774.              * The old address space is no longer active on
  775.              * any processor. It can be appended to the
  776.              * list of inactive address spaces with assigned
  777.              * ASID.
  778.              */
  779.              ASSERT(old->asid != ASID_INVALID);
  780.              list_append(&old->inactive_as_with_asid_link, &inactive_as_with_asid_head);
  781.         }
  782.         mutex_unlock(&old->lock);
  783.     }
  784.  
  785.     /*
  786.      * Second, prepare the new address space.
  787.      */
  788.     mutex_lock_active(&new->lock);
  789.     if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
  790.         if (new->asid != ASID_INVALID)
  791.             list_remove(&new->inactive_as_with_asid_link);
  792.         else
  793.             needs_asid = true;  /* defer call to asid_get() until new->lock is released */
  794.     }
  795.     SET_PTL0_ADDRESS(new->page_table);
  796.     mutex_unlock(&new->lock);
  797.  
  798.     if (needs_asid) {
  799.         /*
  800.          * Allocation of new ASID was deferred
  801.          * until now in order to avoid deadlock.
  802.          */
  803.         asid_t asid;
  804.        
  805.         asid = asid_get();
  806.         mutex_lock_active(&new->lock);
  807.         new->asid = asid;
  808.         mutex_unlock(&new->lock);
  809.     }
  810.     spinlock_unlock(&inactive_as_with_asid_lock);
  811.     interrupts_restore(ipl);
  812.    
  813.     /*
  814.      * Perform architecture-specific steps.
  815.      * (e.g. write ASID to hardware register etc.)
  816.      */
  817.     as_install_arch(new);
  818.    
  819.     AS = new;
  820. }
  821.  
  822. /** Convert address space area flags to page flags.
  823.  *
  824.  * @param aflags Flags of some address space area.
  825.  *
  826.  * @return Flags to be passed to page_mapping_insert().
  827.  */
  828. int area_flags_to_page_flags(int aflags)
  829. {
  830.     int flags;
  831.  
  832.     flags = PAGE_USER | PAGE_PRESENT;
  833.    
  834.     if (aflags & AS_AREA_READ)
  835.         flags |= PAGE_READ;
  836.        
  837.     if (aflags & AS_AREA_WRITE)
  838.         flags |= PAGE_WRITE;
  839.    
  840.     if (aflags & AS_AREA_EXEC)
  841.         flags |= PAGE_EXEC;
  842.    
  843.     if (aflags & AS_AREA_CACHEABLE)
  844.         flags |= PAGE_CACHEABLE;
  845.        
  846.     return flags;
  847. }
  848.  
  849. /** Compute flags for virtual address translation subsytem.
  850.  *
  851.  * The address space area must be locked.
  852.  * Interrupts must be disabled.
  853.  *
  854.  * @param a Address space area.
  855.  *
  856.  * @return Flags to be used in page_mapping_insert().
  857.  */
  858. int as_area_get_flags(as_area_t *a)
  859. {
  860.     return area_flags_to_page_flags(a->flags);
  861. }
  862.  
  863. /** Create page table.
  864.  *
  865.  * Depending on architecture, create either address space
  866.  * private or global page table.
  867.  *
  868.  * @param flags Flags saying whether the page table is for kernel address space.
  869.  *
  870.  * @return First entry of the page table.
  871.  */
  872. pte_t *page_table_create(int flags)
  873. {
  874.         ASSERT(as_operations);
  875.         ASSERT(as_operations->page_table_create);
  876.  
  877.         return as_operations->page_table_create(flags);
  878. }
  879.  
  880. /** Destroy page table.
  881.  *
  882.  * Destroy page table in architecture specific way.
  883.  *
  884.  * @param page_table Physical address of PTL0.
  885.  */
  886. void page_table_destroy(pte_t *page_table)
  887. {
  888.         ASSERT(as_operations);
  889.         ASSERT(as_operations->page_table_destroy);
  890.  
  891.         as_operations->page_table_destroy(page_table);
  892. }
  893.  
  894. /** Lock page table.
  895.  *
  896.  * This function should be called before any page_mapping_insert(),
  897.  * page_mapping_remove() and page_mapping_find().
  898.  *
  899.  * Locking order is such that address space areas must be locked
  900.  * prior to this call. Address space can be locked prior to this
  901.  * call in which case the lock argument is false.
  902.  *
  903.  * @param as Address space.
  904.  * @param lock If false, do not attempt to lock as->lock.
  905.  */
  906. void page_table_lock(as_t *as, bool lock)
  907. {
  908.     ASSERT(as_operations);
  909.     ASSERT(as_operations->page_table_lock);
  910.  
  911.     as_operations->page_table_lock(as, lock);
  912. }
  913.  
  914. /** Unlock page table.
  915.  *
  916.  * @param as Address space.
  917.  * @param unlock If false, do not attempt to unlock as->lock.
  918.  */
  919. void page_table_unlock(as_t *as, bool unlock)
  920. {
  921.     ASSERT(as_operations);
  922.     ASSERT(as_operations->page_table_unlock);
  923.  
  924.     as_operations->page_table_unlock(as, unlock);
  925. }
  926.  
  927.  
  928. /** Find address space area and lock it.
  929.  *
  930.  * The address space must be locked and interrupts must be disabled.
  931.  *
  932.  * @param as Address space.
  933.  * @param va Virtual address.
  934.  *
  935.  * @return Locked address space area containing va on success or NULL on failure.
  936.  */
  937. as_area_t *find_area_and_lock(as_t *as, __address va)
  938. {
  939.     as_area_t *a;
  940.     btree_node_t *leaf, *lnode;
  941.     int i;
  942.    
  943.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  944.     if (a) {
  945.         /* va is the base address of an address space area */
  946.         mutex_lock(&a->lock);
  947.         return a;
  948.     }
  949.    
  950.     /*
  951.      * Search the leaf node and the righmost record of its left neighbour
  952.      * to find out whether this is a miss or va belongs to an address
  953.      * space area found there.
  954.      */
  955.    
  956.     /* First, search the leaf node itself. */
  957.     for (i = 0; i < leaf->keys; i++) {
  958.         a = (as_area_t *) leaf->value[i];
  959.         mutex_lock(&a->lock);
  960.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  961.             return a;
  962.         }
  963.         mutex_unlock(&a->lock);
  964.     }
  965.  
  966.     /*
  967.      * Second, locate the left neighbour and test its last record.
  968.      * Because of its position in the B+tree, it must have base < va.
  969.      */
  970.     if ((lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  971.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  972.         mutex_lock(&a->lock);
  973.         if (va < a->base + a->pages * PAGE_SIZE) {
  974.             return a;
  975.         }
  976.         mutex_unlock(&a->lock);
  977.     }
  978.  
  979.     return NULL;
  980. }
  981.  
  982. /** Check area conflicts with other areas.
  983.  *
  984.  * The address space must be locked and interrupts must be disabled.
  985.  *
  986.  * @param as Address space.
  987.  * @param va Starting virtual address of the area being tested.
  988.  * @param size Size of the area being tested.
  989.  * @param avoid_area Do not touch this area.
  990.  *
  991.  * @return True if there is no conflict, false otherwise.
  992.  */
  993. bool check_area_conflicts(as_t *as, __address va, size_t size, as_area_t *avoid_area)
  994. {
  995.     as_area_t *a;
  996.     btree_node_t *leaf, *node;
  997.     int i;
  998.    
  999.     /*
  1000.      * We don't want any area to have conflicts with NULL page.
  1001.      */
  1002.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1003.         return false;
  1004.    
  1005.     /*
  1006.      * The leaf node is found in O(log n), where n is proportional to
  1007.      * the number of address space areas belonging to as.
  1008.      * The check for conflicts is then attempted on the rightmost
  1009.      * record in the left neighbour, the leftmost record in the right
  1010.      * neighbour and all records in the leaf node itself.
  1011.      */
  1012.    
  1013.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1014.         if (a != avoid_area)
  1015.             return false;
  1016.     }
  1017.    
  1018.     /* First, check the two border cases. */
  1019.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1020.         a = (as_area_t *) node->value[node->keys - 1];
  1021.         mutex_lock(&a->lock);
  1022.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1023.             mutex_unlock(&a->lock);
  1024.             return false;
  1025.         }
  1026.         mutex_unlock(&a->lock);
  1027.     }
  1028.     if ((node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf))) {
  1029.         a = (as_area_t *) node->value[0];
  1030.         mutex_lock(&a->lock);
  1031.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1032.             mutex_unlock(&a->lock);
  1033.             return false;
  1034.         }
  1035.         mutex_unlock(&a->lock);
  1036.     }
  1037.    
  1038.     /* Second, check the leaf node. */
  1039.     for (i = 0; i < leaf->keys; i++) {
  1040.         a = (as_area_t *) leaf->value[i];
  1041.    
  1042.         if (a == avoid_area)
  1043.             continue;
  1044.    
  1045.         mutex_lock(&a->lock);
  1046.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1047.             mutex_unlock(&a->lock);
  1048.             return false;
  1049.         }
  1050.         mutex_unlock(&a->lock);
  1051.     }
  1052.  
  1053.     /*
  1054.      * So far, the area does not conflict with other areas.
  1055.      * Check if it doesn't conflict with kernel address space.
  1056.      */  
  1057.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1058.         return !overlaps(va, size,
  1059.             KERNEL_ADDRESS_SPACE_START, KERNEL_ADDRESS_SPACE_END-KERNEL_ADDRESS_SPACE_START);
  1060.     }
  1061.  
  1062.     return true;
  1063. }
  1064.  
  1065. /** Return size of the address space area with given base.  */
  1066. size_t as_get_size(__address base)
  1067. {
  1068.     ipl_t ipl;
  1069.     as_area_t *src_area;
  1070.     size_t size;
  1071.  
  1072.     ipl = interrupts_disable();
  1073.     src_area = find_area_and_lock(AS, base);
  1074.     if (src_area){
  1075.         size = src_area->pages * PAGE_SIZE;
  1076.         mutex_unlock(&src_area->lock);
  1077.     } else {
  1078.         size = 0;
  1079.     }
  1080.     interrupts_restore(ipl);
  1081.     return size;
  1082. }
  1083.  
  1084. /** Mark portion of address space area as used.
  1085.  *
  1086.  * The address space area must be already locked.
  1087.  *
  1088.  * @param a Address space area.
  1089.  * @param page First page to be marked.
  1090.  * @param count Number of page to be marked.
  1091.  *
  1092.  * @return 0 on failure and 1 on success.
  1093.  */
  1094. int used_space_insert(as_area_t *a, __address page, count_t count)
  1095. {
  1096.     btree_node_t *leaf, *node;
  1097.     count_t pages;
  1098.     int i;
  1099.  
  1100.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1101.     ASSERT(count);
  1102.  
  1103.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1104.     if (pages) {
  1105.         /*
  1106.          * We hit the beginning of some used space.
  1107.          */
  1108.         return 0;
  1109.     }
  1110.  
  1111.     if (!leaf->keys) {
  1112.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1113.         return 1;
  1114.     }
  1115.  
  1116.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1117.     if (node) {
  1118.         __address left_pg = node->key[node->keys - 1], right_pg = leaf->key[0];
  1119.         count_t left_cnt = (count_t) node->value[node->keys - 1], right_cnt = (count_t) leaf->value[0];
  1120.        
  1121.         /*
  1122.          * Examine the possibility that the interval fits
  1123.          * somewhere between the rightmost interval of
  1124.          * the left neigbour and the first interval of the leaf.
  1125.          */
  1126.          
  1127.         if (page >= right_pg) {
  1128.             /* Do nothing. */
  1129.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1130.             /* The interval intersects with the left interval. */
  1131.             return 0;
  1132.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1133.             /* The interval intersects with the right interval. */
  1134.             return 0;          
  1135.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1136.             /* The interval can be added by merging the two already present intervals. */
  1137.             node->value[node->keys - 1] += count + right_cnt;
  1138.             btree_remove(&a->used_space, right_pg, leaf);
  1139.             return 1;
  1140.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1141.             /* The interval can be added by simply growing the left interval. */
  1142.             node->value[node->keys - 1] += count;
  1143.             return 1;
  1144.         } else if (page + count*PAGE_SIZE == right_pg) {
  1145.             /*
  1146.              * The interval can be addded by simply moving base of the right
  1147.              * interval down and increasing its size accordingly.
  1148.              */
  1149.             leaf->value[0] += count;
  1150.             leaf->key[0] = page;
  1151.             return 1;
  1152.         } else {
  1153.             /*
  1154.              * The interval is between both neigbouring intervals,
  1155.              * but cannot be merged with any of them.
  1156.              */
  1157.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1158.             return 1;
  1159.         }
  1160.     } else if (page < leaf->key[0]) {
  1161.         __address right_pg = leaf->key[0];
  1162.         count_t right_cnt = (count_t) leaf->value[0];
  1163.    
  1164.         /*
  1165.          * Investigate the border case in which the left neighbour does not
  1166.          * exist but the interval fits from the left.
  1167.          */
  1168.          
  1169.         if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1170.             /* The interval intersects with the right interval. */
  1171.             return 0;
  1172.         } else if (page + count*PAGE_SIZE == right_pg) {
  1173.             /*
  1174.              * The interval can be added by moving the base of the right interval down
  1175.              * and increasing its size accordingly.
  1176.              */
  1177.             leaf->key[0] = page;
  1178.             leaf->value[0] += count;
  1179.             return 1;
  1180.         } else {
  1181.             /*
  1182.              * The interval doesn't adjoin with the right interval.
  1183.              * It must be added individually.
  1184.              */
  1185.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1186.             return 1;
  1187.         }
  1188.     }
  1189.  
  1190.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1191.     if (node) {
  1192.         __address left_pg = leaf->key[leaf->keys - 1], right_pg = node->key[0];
  1193.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1], right_cnt = (count_t) node->value[0];
  1194.        
  1195.         /*
  1196.          * Examine the possibility that the interval fits
  1197.          * somewhere between the leftmost interval of
  1198.          * the right neigbour and the last interval of the leaf.
  1199.          */
  1200.  
  1201.         if (page < left_pg) {
  1202.             /* Do nothing. */
  1203.         } else if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1204.             /* The interval intersects with the left interval. */
  1205.             return 0;
  1206.         } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1207.             /* The interval intersects with the right interval. */
  1208.             return 0;          
  1209.         } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1210.             /* The interval can be added by merging the two already present intervals. */
  1211.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1212.             btree_remove(&a->used_space, right_pg, node);
  1213.             return 1;
  1214.         } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1215.             /* The interval can be added by simply growing the left interval. */
  1216.             leaf->value[leaf->keys - 1] +=  count;
  1217.             return 1;
  1218.         } else if (page + count*PAGE_SIZE == right_pg) {
  1219.             /*
  1220.              * The interval can be addded by simply moving base of the right
  1221.              * interval down and increasing its size accordingly.
  1222.              */
  1223.             node->value[0] += count;
  1224.             node->key[0] = page;
  1225.             return 1;
  1226.         } else {
  1227.             /*
  1228.              * The interval is between both neigbouring intervals,
  1229.              * but cannot be merged with any of them.
  1230.              */
  1231.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1232.             return 1;
  1233.         }
  1234.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1235.         __address left_pg = leaf->key[leaf->keys - 1];
  1236.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1237.    
  1238.         /*
  1239.          * Investigate the border case in which the right neighbour does not
  1240.          * exist but the interval fits from the right.
  1241.          */
  1242.          
  1243.         if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1244.             /* The interval intersects with the left interval. */
  1245.             return 0;
  1246.         } else if (left_pg + left_cnt*PAGE_SIZE == page) {
  1247.             /* The interval can be added by growing the left interval. */
  1248.             leaf->value[leaf->keys - 1] += count;
  1249.             return 1;
  1250.         } else {
  1251.             /*
  1252.              * The interval doesn't adjoin with the left interval.
  1253.              * It must be added individually.
  1254.              */
  1255.             btree_insert(&a->used_space, page, (void *) count, leaf);
  1256.             return 1;
  1257.         }
  1258.     }
  1259.    
  1260.     /*
  1261.      * Note that if the algorithm made it thus far, the interval can fit only
  1262.      * between two other intervals of the leaf. The two border cases were already
  1263.      * resolved.
  1264.      */
  1265.     for (i = 1; i < leaf->keys; i++) {
  1266.         if (page < leaf->key[i]) {
  1267.             __address left_pg = leaf->key[i - 1], right_pg = leaf->key[i];
  1268.             count_t left_cnt = (count_t) leaf->value[i - 1], right_cnt = (count_t) leaf->value[i];
  1269.  
  1270.             /*
  1271.              * The interval fits between left_pg and right_pg.
  1272.              */
  1273.  
  1274.             if (overlaps(page, count*PAGE_SIZE, left_pg, left_cnt*PAGE_SIZE)) {
  1275.                 /* The interval intersects with the left interval. */
  1276.                 return 0;
  1277.             } else if (overlaps(page, count*PAGE_SIZE, right_pg, right_cnt*PAGE_SIZE)) {
  1278.                 /* The interval intersects with the right interval. */
  1279.                 return 0;          
  1280.             } else if ((page == left_pg + left_cnt*PAGE_SIZE) && (page + count*PAGE_SIZE == right_pg)) {
  1281.                 /* The interval can be added by merging the two already present intervals. */
  1282.                 leaf->value[i - 1] += count + right_cnt;
  1283.                 btree_remove(&a->used_space, right_pg, leaf);
  1284.                 return 1;
  1285.             } else if (page == left_pg + left_cnt*PAGE_SIZE) {
  1286.                 /* The interval can be added by simply growing the left interval. */
  1287.                 leaf->value[i - 1] += count;
  1288.                 return 1;
  1289.             } else if (page + count*PAGE_SIZE == right_pg) {
  1290.                 /*
  1291.                      * The interval can be addded by simply moving base of the right
  1292.                  * interval down and increasing its size accordingly.
  1293.                  */
  1294.                 leaf->value[i] += count;
  1295.                 leaf->key[i] = page;
  1296.                 return 1;
  1297.             } else {
  1298.                 /*
  1299.                  * The interval is between both neigbouring intervals,
  1300.                  * but cannot be merged with any of them.
  1301.                  */
  1302.                 btree_insert(&a->used_space, page, (void *) count, leaf);
  1303.                 return 1;
  1304.             }
  1305.         }
  1306.     }
  1307.  
  1308.     panic("Inconsistency detected while adding %d pages of used space at %P.\n", count, page);
  1309. }
  1310.  
  1311. /** Mark portion of address space area as unused.
  1312.  *
  1313.  * The address space area must be already locked.
  1314.  *
  1315.  * @param a Address space area.
  1316.  * @param page First page to be marked.
  1317.  * @param count Number of page to be marked.
  1318.  *
  1319.  * @return 0 on failure and 1 on success.
  1320.  */
  1321. int used_space_remove(as_area_t *a, __address page, count_t count)
  1322. {
  1323.     btree_node_t *leaf, *node;
  1324.     count_t pages;
  1325.     int i;
  1326.  
  1327.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1328.     ASSERT(count);
  1329.  
  1330.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1331.     if (pages) {
  1332.         /*
  1333.          * We are lucky, page is the beginning of some interval.
  1334.          */
  1335.         if (count > pages) {
  1336.             return 0;
  1337.         } else if (count == pages) {
  1338.             btree_remove(&a->used_space, page, leaf);
  1339.             return 1;
  1340.         } else {
  1341.             /*
  1342.              * Find the respective interval.
  1343.              * Decrease its size and relocate its start address.
  1344.              */
  1345.             for (i = 0; i < leaf->keys; i++) {
  1346.                 if (leaf->key[i] == page) {
  1347.                     leaf->key[i] += count*PAGE_SIZE;
  1348.                     leaf->value[i] -= count;
  1349.                     return 1;
  1350.                 }
  1351.             }
  1352.             goto error;
  1353.         }
  1354.     }
  1355.  
  1356.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1357.     if (node && page < leaf->key[0]) {
  1358.         __address left_pg = node->key[node->keys - 1];
  1359.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1360.  
  1361.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1362.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1363.                 /*
  1364.                  * The interval is contained in the rightmost interval
  1365.                  * of the left neighbour and can be removed by
  1366.                  * updating the size of the bigger interval.
  1367.                  */
  1368.                 node->value[node->keys - 1] -= count;
  1369.                 return 1;
  1370.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1371.                 count_t new_cnt;
  1372.                
  1373.                 /*
  1374.                  * The interval is contained in the rightmost interval
  1375.                  * of the left neighbour but its removal requires
  1376.                  * both updating the size of the original interval and
  1377.                  * also inserting a new interval.
  1378.                  */
  1379.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1380.                 node->value[node->keys - 1] -= count + new_cnt;
  1381.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1382.                 return 1;
  1383.             }
  1384.         }
  1385.         return 0;
  1386.     } else if (page < leaf->key[0]) {
  1387.         return 0;
  1388.     }
  1389.    
  1390.     if (page > leaf->key[leaf->keys - 1]) {
  1391.         __address left_pg = leaf->key[leaf->keys - 1];
  1392.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1393.  
  1394.         if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1395.             if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1396.                 /*
  1397.                  * The interval is contained in the rightmost interval
  1398.                  * of the leaf and can be removed by updating the size
  1399.                  * of the bigger interval.
  1400.                  */
  1401.                 leaf->value[leaf->keys - 1] -= count;
  1402.                 return 1;
  1403.             } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1404.                 count_t new_cnt;
  1405.                
  1406.                 /*
  1407.                  * The interval is contained in the rightmost interval
  1408.                  * of the leaf but its removal requires both updating
  1409.                  * the size of the original interval and
  1410.                  * also inserting a new interval.
  1411.                  */
  1412.                 new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1413.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1414.                 btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1415.                 return 1;
  1416.             }
  1417.         }
  1418.         return 0;
  1419.     }  
  1420.    
  1421.     /*
  1422.      * The border cases have been already resolved.
  1423.      * Now the interval can be only between intervals of the leaf.
  1424.      */
  1425.     for (i = 1; i < leaf->keys - 1; i++) {
  1426.         if (page < leaf->key[i]) {
  1427.             __address left_pg = leaf->key[i - 1];
  1428.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1429.  
  1430.             /*
  1431.              * Now the interval is between intervals corresponding to (i - 1) and i.
  1432.              */
  1433.             if (overlaps(left_pg, left_cnt*PAGE_SIZE, page, count*PAGE_SIZE)) {
  1434.                 if (page + count*PAGE_SIZE == left_pg + left_cnt*PAGE_SIZE) {
  1435.                     /*
  1436.                     * The interval is contained in the interval (i - 1)
  1437.                      * of the leaf and can be removed by updating the size
  1438.                      * of the bigger interval.
  1439.                      */
  1440.                     leaf->value[i - 1] -= count;
  1441.                     return 1;
  1442.                 } else if (page + count*PAGE_SIZE < left_pg + left_cnt*PAGE_SIZE) {
  1443.                     count_t new_cnt;
  1444.                
  1445.                     /*
  1446.                      * The interval is contained in the interval (i - 1)
  1447.                      * of the leaf but its removal requires both updating
  1448.                      * the size of the original interval and
  1449.                      * also inserting a new interval.
  1450.                      */
  1451.                     new_cnt = ((left_pg + left_cnt*PAGE_SIZE) - (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1452.                     leaf->value[i - 1] -= count + new_cnt;
  1453.                     btree_insert(&a->used_space, page + count*PAGE_SIZE, (void *) new_cnt, leaf);
  1454.                     return 1;
  1455.                 }
  1456.             }
  1457.             return 0;
  1458.         }
  1459.     }
  1460.  
  1461. error:
  1462.     panic("Inconsistency detected while removing %d pages of used space from %P.\n", count, page);
  1463. }
  1464.  
  1465. /** Remove reference to address space area share info.
  1466.  *
  1467.  * If the reference count drops to 0, the sh_info is deallocated.
  1468.  *
  1469.  * @param sh_info Pointer to address space area share info.
  1470.  */
  1471. void sh_info_remove_reference(share_info_t *sh_info)
  1472. {
  1473.     bool dealloc = false;
  1474.  
  1475.     mutex_lock(&sh_info->lock);
  1476.     ASSERT(sh_info->refcount);
  1477.     if (--sh_info->refcount == 0) {
  1478.         dealloc = true;
  1479.         link_t *cur;
  1480.        
  1481.         /*
  1482.          * Now walk carefully the pagemap B+tree and free/remove
  1483.          * reference from all frames found there.
  1484.          */
  1485.         for (cur = sh_info->pagemap.leaf_head.next; cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1486.             btree_node_t *node;
  1487.             int i;
  1488.            
  1489.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1490.             for (i = 0; i < node->keys; i++)
  1491.                 frame_free(ADDR2PFN((__address) node->value[i]));
  1492.         }
  1493.        
  1494.     }
  1495.     mutex_unlock(&sh_info->lock);
  1496.    
  1497.     if (dealloc) {
  1498.         btree_destroy(&sh_info->pagemap);
  1499.         free(sh_info);
  1500.     }
  1501. }
  1502.  
  1503. /*
  1504.  * Address space related syscalls.
  1505.  */
  1506.  
  1507. /** Wrapper for as_area_create(). */
  1508. __native sys_as_area_create(__address address, size_t size, int flags)
  1509. {
  1510.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address, AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1511.         return (__native) address;
  1512.     else
  1513.         return (__native) -1;
  1514. }
  1515.  
  1516. /** Wrapper for as_area_resize. */
  1517. __native sys_as_area_resize(__address address, size_t size, int flags)
  1518. {
  1519.     return (__native) as_area_resize(AS, address, size, 0);
  1520. }
  1521.  
  1522. /** Wrapper for as_area_destroy. */
  1523. __native sys_as_area_destroy(__address address)
  1524. {
  1525.     return (__native) as_area_destroy(AS, address);
  1526. }
  1527.