Subversion Repositories HelenOS

Rev

Rev 3222 | Rev 3384 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed

  1. /*
  2.  * Copyright (c) 2001-2006 Jakub Jermar
  3.  * All rights reserved.
  4.  *
  5.  * Redistribution and use in source and binary forms, with or without
  6.  * modification, are permitted provided that the following conditions
  7.  * are met:
  8.  *
  9.  * - Redistributions of source code must retain the above copyright
  10.  *   notice, this list of conditions and the following disclaimer.
  11.  * - Redistributions in binary form must reproduce the above copyright
  12.  *   notice, this list of conditions and the following disclaimer in the
  13.  *   documentation and/or other materials provided with the distribution.
  14.  * - The name of the author may not be used to endorse or promote products
  15.  *   derived from this software without specific prior written permission.
  16.  *
  17.  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  18.  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  19.  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  20.  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  21.  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  22.  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  23.  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  24.  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  25.  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  26.  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  27.  */
  28.  
  29. /** @addtogroup genericmm
  30.  * @{
  31.  */
  32.  
  33. /**
  34.  * @file
  35.  * @brief   Address space related functions.
  36.  *
  37.  * This file contains address space manipulation functions.
  38.  * Roughly speaking, this is a higher-level client of
  39.  * Virtual Address Translation (VAT) subsystem.
  40.  *
  41.  * Functionality provided by this file allows one to
  42.  * create address spaces and create, resize and share
  43.  * address space areas.
  44.  *
  45.  * @see page.c
  46.  *
  47.  */
  48.  
  49. #include <mm/as.h>
  50. #include <arch/mm/as.h>
  51. #include <mm/page.h>
  52. #include <mm/frame.h>
  53. #include <mm/slab.h>
  54. #include <mm/tlb.h>
  55. #include <arch/mm/page.h>
  56. #include <genarch/mm/page_pt.h>
  57. #include <genarch/mm/page_ht.h>
  58. #include <mm/asid.h>
  59. #include <arch/mm/asid.h>
  60. #include <preemption.h>
  61. #include <synch/spinlock.h>
  62. #include <synch/mutex.h>
  63. #include <adt/list.h>
  64. #include <adt/btree.h>
  65. #include <proc/task.h>
  66. #include <proc/thread.h>
  67. #include <arch/asm.h>
  68. #include <panic.h>
  69. #include <debug.h>
  70. #include <print.h>
  71. #include <memstr.h>
  72. #include <macros.h>
  73. #include <arch.h>
  74. #include <errno.h>
  75. #include <config.h>
  76. #include <align.h>
  77. #include <arch/types.h>
  78. #include <syscall/copy.h>
  79. #include <arch/interrupt.h>
  80.  
  81. #ifdef CONFIG_VIRT_IDX_DCACHE
  82. #include <arch/mm/cache.h>
  83. #endif /* CONFIG_VIRT_IDX_DCACHE */
  84.  
  85. /**
  86.  * Each architecture decides what functions will be used to carry out
  87.  * address space operations such as creating or locking page tables.
  88.  */
  89. as_operations_t *as_operations = NULL;
  90.  
  91. /**
  92.  * Slab for as_t objects.
  93.  */
  94. static slab_cache_t *as_slab;
  95.  
  96. /**
  97.  * This lock serializes access to the ASID subsystem.
  98.  * It protects:
  99.  * - inactive_as_with_asid_head list
  100.  * - as->asid for each as of the as_t type
  101.  * - asids_allocated counter
  102.  */
  103. SPINLOCK_INITIALIZE(asidlock);
  104.  
  105. /**
  106.  * This list contains address spaces that are not active on any
  107.  * processor and that have valid ASID.
  108.  */
  109. LIST_INITIALIZE(inactive_as_with_asid_head);
  110.  
  111. /** Kernel address space. */
  112. as_t *AS_KERNEL = NULL;
  113.  
  114. static int area_flags_to_page_flags(int aflags);
  115. static as_area_t *find_area_and_lock(as_t *as, uintptr_t va);
  116. static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  117.     as_area_t *avoid_area);
  118. static void sh_info_remove_reference(share_info_t *sh_info);
  119.  
  120. static int as_constructor(void *obj, int flags)
  121. {
  122.     as_t *as = (as_t *) obj;
  123.     int rc;
  124.  
  125.     link_initialize(&as->inactive_as_with_asid_link);
  126.     mutex_initialize(&as->lock, MUTEX_PASSIVE);
  127.    
  128.     rc = as_constructor_arch(as, flags);
  129.    
  130.     return rc;
  131. }
  132.  
  133. static int as_destructor(void *obj)
  134. {
  135.     as_t *as = (as_t *) obj;
  136.  
  137.     return as_destructor_arch(as);
  138. }
  139.  
  140. /** Initialize address space subsystem. */
  141. void as_init(void)
  142. {
  143.     as_arch_init();
  144.  
  145.     as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,
  146.         as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);
  147.    
  148.     AS_KERNEL = as_create(FLAG_AS_KERNEL);
  149.     if (!AS_KERNEL)
  150.         panic("can't create kernel address space\n");
  151.    
  152. }
  153.  
  154. /** Create address space.
  155.  *
  156.  * @param flags Flags that influence way in wich the address space is created.
  157.  */
  158. as_t *as_create(int flags)
  159. {
  160.     as_t *as;
  161.  
  162.     as = (as_t *) slab_alloc(as_slab, 0);
  163.     (void) as_create_arch(as, 0);
  164.    
  165.     btree_create(&as->as_area_btree);
  166.    
  167.     if (flags & FLAG_AS_KERNEL)
  168.         as->asid = ASID_KERNEL;
  169.     else
  170.         as->asid = ASID_INVALID;
  171.    
  172.     atomic_set(&as->refcount, 0);
  173.     as->cpu_refcount = 0;
  174. #ifdef AS_PAGE_TABLE
  175.     as->genarch.page_table = page_table_create(flags);
  176. #else
  177.     page_table_create(flags);
  178. #endif
  179.  
  180.     return as;
  181. }
  182.  
  183. /** Destroy adress space.
  184.  *
  185.  * When there are no tasks referencing this address space (i.e. its refcount is
  186.  * zero), the address space can be destroyed.
  187.  *
  188.  * We know that we don't hold any spinlock.
  189.  */
  190. void as_destroy(as_t *as)
  191. {
  192.     ipl_t ipl;
  193.     bool cond;
  194.     DEADLOCK_PROBE_INIT(p_asidlock);
  195.  
  196.     ASSERT(atomic_get(&as->refcount) == 0);
  197.    
  198.     /*
  199.      * Since there is no reference to this area,
  200.      * it is safe not to lock its mutex.
  201.      */
  202.  
  203.     /*
  204.      * We need to avoid deadlock between TLB shootdown and asidlock.
  205.      * We therefore try to take asid conditionally and if we don't succeed,
  206.      * we enable interrupts and try again. This is done while preemption is
  207.      * disabled to prevent nested context switches. We also depend on the
  208.      * fact that so far no spinlocks are held.
  209.      */
  210.     preemption_disable();
  211.     ipl = interrupts_read();
  212. retry:
  213.     interrupts_disable();
  214.     if (!spinlock_trylock(&asidlock)) {
  215.         interrupts_enable();
  216.         DEADLOCK_PROBE(p_asidlock, DEADLOCK_THRESHOLD);
  217.         goto retry;
  218.     }
  219.     preemption_enable();    /* Interrupts disabled, enable preemption */
  220.     if (as->asid != ASID_INVALID && as != AS_KERNEL) {
  221.         if (as != AS && as->cpu_refcount == 0)
  222.             list_remove(&as->inactive_as_with_asid_link);
  223.         asid_put(as->asid);
  224.     }
  225.     spinlock_unlock(&asidlock);
  226.  
  227.     /*
  228.      * Destroy address space areas of the address space.
  229.      * The B+tree must be walked carefully because it is
  230.      * also being destroyed.
  231.      */
  232.     for (cond = true; cond; ) {
  233.         btree_node_t *node;
  234.  
  235.         ASSERT(!list_empty(&as->as_area_btree.leaf_head));
  236.         node = list_get_instance(as->as_area_btree.leaf_head.next,
  237.             btree_node_t, leaf_link);
  238.  
  239.         if ((cond = node->keys)) {
  240.             as_area_destroy(as, node->key[0]);
  241.         }
  242.     }
  243.  
  244.     btree_destroy(&as->as_area_btree);
  245. #ifdef AS_PAGE_TABLE
  246.     page_table_destroy(as->genarch.page_table);
  247. #else
  248.     page_table_destroy(NULL);
  249. #endif
  250.  
  251.     interrupts_restore(ipl);
  252.  
  253.     slab_free(as_slab, as);
  254. }
  255.  
  256. /** Create address space area of common attributes.
  257.  *
  258.  * The created address space area is added to the target address space.
  259.  *
  260.  * @param as Target address space.
  261.  * @param flags Flags of the area memory.
  262.  * @param size Size of area.
  263.  * @param base Base address of area.
  264.  * @param attrs Attributes of the area.
  265.  * @param backend Address space area backend. NULL if no backend is used.
  266.  * @param backend_data NULL or a pointer to an array holding two void *.
  267.  *
  268.  * @return Address space area on success or NULL on failure.
  269.  */
  270. as_area_t *
  271. as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
  272.            mem_backend_t *backend, mem_backend_data_t *backend_data)
  273. {
  274.     ipl_t ipl;
  275.     as_area_t *a;
  276.    
  277.     if (base % PAGE_SIZE)
  278.         return NULL;
  279.  
  280.     if (!size)
  281.         return NULL;
  282.  
  283.     /* Writeable executable areas are not supported. */
  284.     if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
  285.         return NULL;
  286.    
  287.     ipl = interrupts_disable();
  288.     mutex_lock(&as->lock);
  289.    
  290.     if (!check_area_conflicts(as, base, size, NULL)) {
  291.         mutex_unlock(&as->lock);
  292.         interrupts_restore(ipl);
  293.         return NULL;
  294.     }
  295.    
  296.     a = (as_area_t *) malloc(sizeof(as_area_t), 0);
  297.  
  298.     mutex_initialize(&a->lock, MUTEX_PASSIVE);
  299.    
  300.     a->as = as;
  301.     a->flags = flags;
  302.     a->attributes = attrs;
  303.     a->pages = SIZE2FRAMES(size);
  304.     a->base = base;
  305.     a->sh_info = NULL;
  306.     a->backend = backend;
  307.     if (backend_data)
  308.         a->backend_data = *backend_data;
  309.     else
  310.         memsetb(&a->backend_data, sizeof(a->backend_data), 0);
  311.  
  312.     btree_create(&a->used_space);
  313.    
  314.     btree_insert(&as->as_area_btree, base, (void *) a, NULL);
  315.  
  316.     mutex_unlock(&as->lock);
  317.     interrupts_restore(ipl);
  318.  
  319.     return a;
  320. }
  321.  
  322. /** Find address space area and change it.
  323.  *
  324.  * @param as Address space.
  325.  * @param address Virtual address belonging to the area to be changed. Must be
  326.  *     page-aligned.
  327.  * @param size New size of the virtual memory block starting at address.
  328.  * @param flags Flags influencing the remap operation. Currently unused.
  329.  *
  330.  * @return Zero on success or a value from @ref errno.h otherwise.
  331.  */
  332. int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
  333. {
  334.     as_area_t *area;
  335.     ipl_t ipl;
  336.     size_t pages;
  337.    
  338.     ipl = interrupts_disable();
  339.     mutex_lock(&as->lock);
  340.    
  341.     /*
  342.      * Locate the area.
  343.      */
  344.     area = find_area_and_lock(as, address);
  345.     if (!area) {
  346.         mutex_unlock(&as->lock);
  347.         interrupts_restore(ipl);
  348.         return ENOENT;
  349.     }
  350.  
  351.     if (area->backend == &phys_backend) {
  352.         /*
  353.          * Remapping of address space areas associated
  354.          * with memory mapped devices is not supported.
  355.          */
  356.         mutex_unlock(&area->lock);
  357.         mutex_unlock(&as->lock);
  358.         interrupts_restore(ipl);
  359.         return ENOTSUP;
  360.     }
  361.     if (area->sh_info) {
  362.         /*
  363.          * Remapping of shared address space areas
  364.          * is not supported.
  365.          */
  366.         mutex_unlock(&area->lock);
  367.         mutex_unlock(&as->lock);
  368.         interrupts_restore(ipl);
  369.         return ENOTSUP;
  370.     }
  371.  
  372.     pages = SIZE2FRAMES((address - area->base) + size);
  373.     if (!pages) {
  374.         /*
  375.          * Zero size address space areas are not allowed.
  376.          */
  377.         mutex_unlock(&area->lock);
  378.         mutex_unlock(&as->lock);
  379.         interrupts_restore(ipl);
  380.         return EPERM;
  381.     }
  382.    
  383.     if (pages < area->pages) {
  384.         bool cond;
  385.         uintptr_t start_free = area->base + pages*PAGE_SIZE;
  386.  
  387.         /*
  388.          * Shrinking the area.
  389.          * No need to check for overlaps.
  390.          */
  391.  
  392.         /*
  393.          * Start TLB shootdown sequence.
  394.          */
  395.         tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base +
  396.             pages * PAGE_SIZE, area->pages - pages);
  397.  
  398.         /*
  399.          * Remove frames belonging to used space starting from
  400.          * the highest addresses downwards until an overlap with
  401.          * the resized address space area is found. Note that this
  402.          * is also the right way to remove part of the used_space
  403.          * B+tree leaf list.
  404.          */    
  405.         for (cond = true; cond;) {
  406.             btree_node_t *node;
  407.        
  408.             ASSERT(!list_empty(&area->used_space.leaf_head));
  409.             node =
  410.                 list_get_instance(area->used_space.leaf_head.prev,
  411.                 btree_node_t, leaf_link);
  412.             if ((cond = (bool) node->keys)) {
  413.                 uintptr_t b = node->key[node->keys - 1];
  414.                 count_t c =
  415.                     (count_t) node->value[node->keys - 1];
  416.                 unsigned int i = 0;
  417.            
  418.                 if (overlaps(b, c * PAGE_SIZE, area->base,
  419.                     pages * PAGE_SIZE)) {
  420.                    
  421.                     if (b + c * PAGE_SIZE <= start_free) {
  422.                         /*
  423.                          * The whole interval fits
  424.                          * completely in the resized
  425.                          * address space area.
  426.                          */
  427.                         break;
  428.                     }
  429.        
  430.                     /*
  431.                      * Part of the interval corresponding
  432.                      * to b and c overlaps with the resized
  433.                      * address space area.
  434.                      */
  435.        
  436.                     cond = false;   /* we are almost done */
  437.                     i = (start_free - b) >> PAGE_WIDTH;
  438.                     if (!used_space_remove(area, start_free, c - i))
  439.                         panic("Could not remove used space.\n");
  440.                 } else {
  441.                     /*
  442.                      * The interval of used space can be
  443.                      * completely removed.
  444.                      */
  445.                     if (!used_space_remove(area, b, c))
  446.                         panic("Could not remove used space.\n");
  447.                 }
  448.            
  449.                 for (; i < c; i++) {
  450.                     pte_t *pte;
  451.            
  452.                     page_table_lock(as, false);
  453.                     pte = page_mapping_find(as, b +
  454.                         i * PAGE_SIZE);
  455.                     ASSERT(pte && PTE_VALID(pte) &&
  456.                         PTE_PRESENT(pte));
  457.                     if (area->backend &&
  458.                         area->backend->frame_free) {
  459.                         area->backend->frame_free(area,
  460.                             b + i * PAGE_SIZE,
  461.                             PTE_GET_FRAME(pte));
  462.                     }
  463.                     page_mapping_remove(as, b +
  464.                         i * PAGE_SIZE);
  465.                     page_table_unlock(as, false);
  466.                 }
  467.             }
  468.         }
  469.  
  470.         /*
  471.          * Finish TLB shootdown sequence.
  472.          */
  473.  
  474.         tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,
  475.             area->pages - pages);
  476.         /*
  477.          * Invalidate software translation caches (e.g. TSB on sparc64).
  478.          */
  479.         as_invalidate_translation_cache(as, area->base +
  480.             pages * PAGE_SIZE, area->pages - pages);
  481.         tlb_shootdown_finalize();
  482.        
  483.     } else {
  484.         /*
  485.          * Growing the area.
  486.          * Check for overlaps with other address space areas.
  487.          */
  488.         if (!check_area_conflicts(as, address, pages * PAGE_SIZE,
  489.             area)) {
  490.             mutex_unlock(&area->lock);
  491.             mutex_unlock(&as->lock);       
  492.             interrupts_restore(ipl);
  493.             return EADDRNOTAVAIL;
  494.         }
  495.     }
  496.  
  497.     area->pages = pages;
  498.    
  499.     mutex_unlock(&area->lock);
  500.     mutex_unlock(&as->lock);
  501.     interrupts_restore(ipl);
  502.  
  503.     return 0;
  504. }
  505.  
  506. /** Destroy address space area.
  507.  *
  508.  * @param as Address space.
  509.  * @param address Address withing the area to be deleted.
  510.  *
  511.  * @return Zero on success or a value from @ref errno.h on failure.
  512.  */
  513. int as_area_destroy(as_t *as, uintptr_t address)
  514. {
  515.     as_area_t *area;
  516.     uintptr_t base;
  517.     link_t *cur;
  518.     ipl_t ipl;
  519.  
  520.     ipl = interrupts_disable();
  521.     mutex_lock(&as->lock);
  522.  
  523.     area = find_area_and_lock(as, address);
  524.     if (!area) {
  525.         mutex_unlock(&as->lock);
  526.         interrupts_restore(ipl);
  527.         return ENOENT;
  528.     }
  529.  
  530.     base = area->base;
  531.  
  532.     /*
  533.      * Start TLB shootdown sequence.
  534.      */
  535.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  536.  
  537.     /*
  538.      * Visit only the pages mapped by used_space B+tree.
  539.      */
  540.     for (cur = area->used_space.leaf_head.next;
  541.         cur != &area->used_space.leaf_head; cur = cur->next) {
  542.         btree_node_t *node;
  543.         unsigned int i;
  544.        
  545.         node = list_get_instance(cur, btree_node_t, leaf_link);
  546.         for (i = 0; i < node->keys; i++) {
  547.             uintptr_t b = node->key[i];
  548.             count_t j;
  549.             pte_t *pte;
  550.            
  551.             for (j = 0; j < (count_t) node->value[i]; j++) {
  552.                 page_table_lock(as, false);
  553.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  554.                 ASSERT(pte && PTE_VALID(pte) &&
  555.                     PTE_PRESENT(pte));
  556.                 if (area->backend &&
  557.                     area->backend->frame_free) {
  558.                     area->backend->frame_free(area, b +
  559.                         j * PAGE_SIZE, PTE_GET_FRAME(pte));
  560.                 }
  561.                 page_mapping_remove(as, b + j * PAGE_SIZE);            
  562.                 page_table_unlock(as, false);
  563.             }
  564.         }
  565.     }
  566.  
  567.     /*
  568.      * Finish TLB shootdown sequence.
  569.      */
  570.  
  571.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  572.     /*
  573.      * Invalidate potential software translation caches (e.g. TSB on
  574.      * sparc64).
  575.      */
  576.     as_invalidate_translation_cache(as, area->base, area->pages);
  577.     tlb_shootdown_finalize();
  578.    
  579.     btree_destroy(&area->used_space);
  580.  
  581.     area->attributes |= AS_AREA_ATTR_PARTIAL;
  582.    
  583.     if (area->sh_info)
  584.         sh_info_remove_reference(area->sh_info);
  585.        
  586.     mutex_unlock(&area->lock);
  587.  
  588.     /*
  589.      * Remove the empty area from address space.
  590.      */
  591.     btree_remove(&as->as_area_btree, base, NULL);
  592.    
  593.     free(area);
  594.    
  595.     mutex_unlock(&as->lock);
  596.     interrupts_restore(ipl);
  597.     return 0;
  598. }
  599.  
  600. /** Share address space area with another or the same address space.
  601.  *
  602.  * Address space area mapping is shared with a new address space area.
  603.  * If the source address space area has not been shared so far,
  604.  * a new sh_info is created. The new address space area simply gets the
  605.  * sh_info of the source area. The process of duplicating the
  606.  * mapping is done through the backend share function.
  607.  *
  608.  * @param src_as Pointer to source address space.
  609.  * @param src_base Base address of the source address space area.
  610.  * @param acc_size Expected size of the source area.
  611.  * @param dst_as Pointer to destination address space.
  612.  * @param dst_base Target base address.
  613.  * @param dst_flags_mask Destination address space area flags mask.
  614.  *
  615.  * @return Zero on success or ENOENT if there is no such task or if there is no
  616.  * such address space area, EPERM if there was a problem in accepting the area
  617.  * or ENOMEM if there was a problem in allocating destination address space
  618.  * area. ENOTSUP is returned if the address space area backend does not support
  619.  * sharing.
  620.  */
  621. int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
  622.     as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
  623. {
  624.     ipl_t ipl;
  625.     int src_flags;
  626.     size_t src_size;
  627.     as_area_t *src_area, *dst_area;
  628.     share_info_t *sh_info;
  629.     mem_backend_t *src_backend;
  630.     mem_backend_data_t src_backend_data;
  631.    
  632.     ipl = interrupts_disable();
  633.     mutex_lock(&src_as->lock);
  634.     src_area = find_area_and_lock(src_as, src_base);
  635.     if (!src_area) {
  636.         /*
  637.          * Could not find the source address space area.
  638.          */
  639.         mutex_unlock(&src_as->lock);
  640.         interrupts_restore(ipl);
  641.         return ENOENT;
  642.     }
  643.  
  644.     if (!src_area->backend || !src_area->backend->share) {
  645.         /*
  646.          * There is no backend or the backend does not
  647.          * know how to share the area.
  648.          */
  649.         mutex_unlock(&src_area->lock);
  650.         mutex_unlock(&src_as->lock);
  651.         interrupts_restore(ipl);
  652.         return ENOTSUP;
  653.     }
  654.    
  655.     src_size = src_area->pages * PAGE_SIZE;
  656.     src_flags = src_area->flags;
  657.     src_backend = src_area->backend;
  658.     src_backend_data = src_area->backend_data;
  659.  
  660.     /* Share the cacheable flag from the original mapping */
  661.     if (src_flags & AS_AREA_CACHEABLE)
  662.         dst_flags_mask |= AS_AREA_CACHEABLE;
  663.  
  664.     if (src_size != acc_size ||
  665.         (src_flags & dst_flags_mask) != dst_flags_mask) {
  666.         mutex_unlock(&src_area->lock);
  667.         mutex_unlock(&src_as->lock);
  668.         interrupts_restore(ipl);
  669.         return EPERM;
  670.     }
  671.  
  672.     /*
  673.      * Now we are committed to sharing the area.
  674.      * First, prepare the area for sharing.
  675.      * Then it will be safe to unlock it.
  676.      */
  677.     sh_info = src_area->sh_info;
  678.     if (!sh_info) {
  679.         sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
  680.         mutex_initialize(&sh_info->lock, MUTEX_PASSIVE);
  681.         sh_info->refcount = 2;
  682.         btree_create(&sh_info->pagemap);
  683.         src_area->sh_info = sh_info;
  684.         /*
  685.          * Call the backend to setup sharing.
  686.          */
  687.         src_area->backend->share(src_area);
  688.     } else {
  689.         mutex_lock(&sh_info->lock);
  690.         sh_info->refcount++;
  691.         mutex_unlock(&sh_info->lock);
  692.     }
  693.  
  694.     mutex_unlock(&src_area->lock);
  695.     mutex_unlock(&src_as->lock);
  696.  
  697.     /*
  698.      * Create copy of the source address space area.
  699.      * The destination area is created with AS_AREA_ATTR_PARTIAL
  700.      * attribute set which prevents race condition with
  701.      * preliminary as_page_fault() calls.
  702.      * The flags of the source area are masked against dst_flags_mask
  703.      * to support sharing in less privileged mode.
  704.      */
  705.     dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
  706.         AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
  707.     if (!dst_area) {
  708.         /*
  709.          * Destination address space area could not be created.
  710.          */
  711.         sh_info_remove_reference(sh_info);
  712.        
  713.         interrupts_restore(ipl);
  714.         return ENOMEM;
  715.     }
  716.  
  717.     /*
  718.      * Now the destination address space area has been
  719.      * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
  720.      * attribute and set the sh_info.
  721.      */
  722.     mutex_lock(&dst_as->lock); 
  723.     mutex_lock(&dst_area->lock);
  724.     dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
  725.     dst_area->sh_info = sh_info;
  726.     mutex_unlock(&dst_area->lock);
  727.     mutex_unlock(&dst_as->lock);   
  728.  
  729.     interrupts_restore(ipl);
  730.    
  731.     return 0;
  732. }
  733.  
  734. /** Check access mode for address space area.
  735.  *
  736.  * The address space area must be locked prior to this call.
  737.  *
  738.  * @param area Address space area.
  739.  * @param access Access mode.
  740.  *
  741.  * @return False if access violates area's permissions, true otherwise.
  742.  */
  743. bool as_area_check_access(as_area_t *area, pf_access_t access)
  744. {
  745.     int flagmap[] = {
  746.         [PF_ACCESS_READ] = AS_AREA_READ,
  747.         [PF_ACCESS_WRITE] = AS_AREA_WRITE,
  748.         [PF_ACCESS_EXEC] = AS_AREA_EXEC
  749.     };
  750.  
  751.     if (!(area->flags & flagmap[access]))
  752.         return false;
  753.    
  754.     return true;
  755. }
  756.  
  757. /** Change adress area flags.
  758.  *
  759.  * The idea is to have the same data, but with a different access mode.
  760.  * This is needed e.g. for writing code into memory and then executing it.
  761.  * In order for this to work properly, this may copy the data
  762.  * into private anonymous memory (unless it's already there).
  763.  *
  764.  * @param as Address space.
  765.  * @param flags Flags of the area memory.
  766.  * @param address Address withing the area to be changed.
  767.  *
  768.  * @return Zero on success or a value from @ref errno.h on failure.
  769.  */
  770. int as_area_change_flags(as_t *as, int flags, uintptr_t address)
  771. {
  772.     as_area_t *area;
  773.     uintptr_t base;
  774.     link_t *cur;
  775.     ipl_t ipl;
  776.     int page_flags;
  777.     uintptr_t *old_frame;
  778.     index_t frame_idx;
  779.     count_t used_pages;
  780.  
  781.     /* Flags for the new memory mapping */
  782.     page_flags = area_flags_to_page_flags(flags);
  783.  
  784.     ipl = interrupts_disable();
  785.     mutex_lock(&as->lock);
  786.  
  787.     area = find_area_and_lock(as, address);
  788.     if (!area) {
  789.         mutex_unlock(&as->lock);
  790.         interrupts_restore(ipl);
  791.         return ENOENT;
  792.     }
  793.  
  794.     if (area->sh_info || area->backend != &anon_backend) {
  795.         /* Copying shared areas not supported yet */
  796.         /* Copying non-anonymous memory not supported yet */
  797.         mutex_unlock(&area->lock);
  798.         mutex_unlock(&as->lock);
  799.         interrupts_restore(ipl);
  800.         return ENOTSUP;
  801.     }
  802.  
  803.     base = area->base;
  804.  
  805.     /*
  806.      * Compute total number of used pages in the used_space B+tree
  807.      */
  808.     used_pages = 0;
  809.  
  810.     for (cur = area->used_space.leaf_head.next;
  811.         cur != &area->used_space.leaf_head; cur = cur->next) {
  812.         btree_node_t *node;
  813.         unsigned int i;
  814.        
  815.         node = list_get_instance(cur, btree_node_t, leaf_link);
  816.         for (i = 0; i < node->keys; i++) {
  817.             used_pages += (count_t) node->value[i];
  818.         }
  819.     }
  820.  
  821.     /* An array for storing frame numbers */
  822.     old_frame = malloc(used_pages * sizeof(uintptr_t), 0);
  823.  
  824.     /*
  825.      * Start TLB shootdown sequence.
  826.      */
  827.     tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
  828.  
  829.     /*
  830.      * Remove used pages from page tables and remember their frame
  831.      * numbers.
  832.      */
  833.     frame_idx = 0;
  834.  
  835.     for (cur = area->used_space.leaf_head.next;
  836.         cur != &area->used_space.leaf_head; cur = cur->next) {
  837.         btree_node_t *node;
  838.         unsigned int i;
  839.        
  840.         node = list_get_instance(cur, btree_node_t, leaf_link);
  841.         for (i = 0; i < node->keys; i++) {
  842.             uintptr_t b = node->key[i];
  843.             count_t j;
  844.             pte_t *pte;
  845.            
  846.             for (j = 0; j < (count_t) node->value[i]; j++) {
  847.                 page_table_lock(as, false);
  848.                 pte = page_mapping_find(as, b + j * PAGE_SIZE);
  849.                 ASSERT(pte && PTE_VALID(pte) &&
  850.                     PTE_PRESENT(pte));
  851.                 old_frame[frame_idx++] = PTE_GET_FRAME(pte);
  852.  
  853.                 /* Remove old mapping */
  854.                 page_mapping_remove(as, b + j * PAGE_SIZE);
  855.                 page_table_unlock(as, false);
  856.             }
  857.         }
  858.     }
  859.  
  860.     /*
  861.      * Finish TLB shootdown sequence.
  862.      */
  863.  
  864.     tlb_invalidate_pages(as->asid, area->base, area->pages);
  865.     /*
  866.      * Invalidate potential software translation caches (e.g. TSB on
  867.      * sparc64).
  868.      */
  869.     as_invalidate_translation_cache(as, area->base, area->pages);
  870.     tlb_shootdown_finalize();
  871.  
  872.     /*
  873.      * Map pages back in with new flags. This step is kept separate
  874.      * so that there's no instant when the memory area could be
  875.      * accesed with both the old and the new flags at once.
  876.      */
  877.     frame_idx = 0;
  878.  
  879.     for (cur = area->used_space.leaf_head.next;
  880.         cur != &area->used_space.leaf_head; cur = cur->next) {
  881.         btree_node_t *node;
  882.         unsigned int i;
  883.        
  884.         node = list_get_instance(cur, btree_node_t, leaf_link);
  885.         for (i = 0; i < node->keys; i++) {
  886.             uintptr_t b = node->key[i];
  887.             count_t j;
  888.            
  889.             for (j = 0; j < (count_t) node->value[i]; j++) {
  890.                 page_table_lock(as, false);
  891.  
  892.                 /* Insert the new mapping */
  893.                 page_mapping_insert(as, b + j * PAGE_SIZE,
  894.                     old_frame[frame_idx++], page_flags);
  895.  
  896.                 page_table_unlock(as, false);
  897.             }
  898.         }
  899.     }
  900.  
  901.     free(old_frame);
  902.  
  903.     mutex_unlock(&area->lock);
  904.     mutex_unlock(&as->lock);
  905.     interrupts_restore(ipl);
  906.  
  907.     return 0;
  908. }
  909.  
  910.  
  911. /** Handle page fault within the current address space.
  912.  *
  913.  * This is the high-level page fault handler. It decides
  914.  * whether the page fault can be resolved by any backend
  915.  * and if so, it invokes the backend to resolve the page
  916.  * fault.
  917.  *
  918.  * Interrupts are assumed disabled.
  919.  *
  920.  * @param page Faulting page.
  921.  * @param access Access mode that caused the fault (i.e. read/write/exec).
  922.  * @param istate Pointer to interrupted state.
  923.  *
  924.  * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
  925.  *     fault was caused by copy_to_uspace() or copy_from_uspace().
  926.  */
  927. int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
  928. {
  929.     pte_t *pte;
  930.     as_area_t *area;
  931.    
  932.     if (!THREAD)
  933.         return AS_PF_FAULT;
  934.        
  935.     ASSERT(AS);
  936.  
  937.     mutex_lock(&AS->lock);
  938.     area = find_area_and_lock(AS, page);   
  939.     if (!area) {
  940.         /*
  941.          * No area contained mapping for 'page'.
  942.          * Signal page fault to low-level handler.
  943.          */
  944.         mutex_unlock(&AS->lock);
  945.         goto page_fault;
  946.     }
  947.  
  948.     if (area->attributes & AS_AREA_ATTR_PARTIAL) {
  949.         /*
  950.          * The address space area is not fully initialized.
  951.          * Avoid possible race by returning error.
  952.          */
  953.         mutex_unlock(&area->lock);
  954.         mutex_unlock(&AS->lock);
  955.         goto page_fault;       
  956.     }
  957.  
  958.     if (!area->backend || !area->backend->page_fault) {
  959.         /*
  960.          * The address space area is not backed by any backend
  961.          * or the backend cannot handle page faults.
  962.          */
  963.         mutex_unlock(&area->lock);
  964.         mutex_unlock(&AS->lock);
  965.         goto page_fault;       
  966.     }
  967.  
  968.     page_table_lock(AS, false);
  969.    
  970.     /*
  971.      * To avoid race condition between two page faults
  972.      * on the same address, we need to make sure
  973.      * the mapping has not been already inserted.
  974.      */
  975.     if ((pte = page_mapping_find(AS, page))) {
  976.         if (PTE_PRESENT(pte)) {
  977.             if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
  978.                 (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
  979.                 (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
  980.                 page_table_unlock(AS, false);
  981.                 mutex_unlock(&area->lock);
  982.                 mutex_unlock(&AS->lock);
  983.                 return AS_PF_OK;
  984.             }
  985.         }
  986.     }
  987.    
  988.     /*
  989.      * Resort to the backend page fault handler.
  990.      */
  991.     if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
  992.         page_table_unlock(AS, false);
  993.         mutex_unlock(&area->lock);
  994.         mutex_unlock(&AS->lock);
  995.         goto page_fault;
  996.     }
  997.    
  998.     page_table_unlock(AS, false);
  999.     mutex_unlock(&area->lock);
  1000.     mutex_unlock(&AS->lock);
  1001.     return AS_PF_OK;
  1002.  
  1003. page_fault:
  1004.     if (THREAD->in_copy_from_uspace) {
  1005.         THREAD->in_copy_from_uspace = false;
  1006.         istate_set_retaddr(istate,
  1007.             (uintptr_t) &memcpy_from_uspace_failover_address);
  1008.     } else if (THREAD->in_copy_to_uspace) {
  1009.         THREAD->in_copy_to_uspace = false;
  1010.         istate_set_retaddr(istate,
  1011.             (uintptr_t) &memcpy_to_uspace_failover_address);
  1012.     } else {
  1013.         return AS_PF_FAULT;
  1014.     }
  1015.  
  1016.     return AS_PF_DEFER;
  1017. }
  1018.  
  1019. /** Switch address spaces.
  1020.  *
  1021.  * Note that this function cannot sleep as it is essentially a part of
  1022.  * scheduling. Sleeping here would lead to deadlock on wakeup. Another
  1023.  * thing which is forbidden in this context is locking the address space.
  1024.  *
  1025.  * When this function is enetered, no spinlocks may be held.
  1026.  *
  1027.  * @param old Old address space or NULL.
  1028.  * @param new New address space.
  1029.  */
  1030. void as_switch(as_t *old_as, as_t *new_as)
  1031. {
  1032.     DEADLOCK_PROBE_INIT(p_asidlock);
  1033.     preemption_disable();
  1034. retry:
  1035.     (void) interrupts_disable();
  1036.     if (!spinlock_trylock(&asidlock)) {
  1037.         /*
  1038.          * Avoid deadlock with TLB shootdown.
  1039.          * We can enable interrupts here because
  1040.          * preemption is disabled. We should not be
  1041.          * holding any other lock.
  1042.          */
  1043.         (void) interrupts_enable();
  1044.         DEADLOCK_PROBE(p_asidlock, DEADLOCK_THRESHOLD);
  1045.         goto retry;
  1046.     }
  1047.     preemption_enable();
  1048.  
  1049.     /*
  1050.      * First, take care of the old address space.
  1051.      */
  1052.     if (old_as) {
  1053.         ASSERT(old_as->cpu_refcount);
  1054.         if((--old_as->cpu_refcount == 0) && (old_as != AS_KERNEL)) {
  1055.             /*
  1056.              * The old address space is no longer active on
  1057.              * any processor. It can be appended to the
  1058.              * list of inactive address spaces with assigned
  1059.              * ASID.
  1060.              */
  1061.             ASSERT(old_as->asid != ASID_INVALID);
  1062.             list_append(&old_as->inactive_as_with_asid_link,
  1063.                 &inactive_as_with_asid_head);
  1064.         }
  1065.  
  1066.         /*
  1067.          * Perform architecture-specific tasks when the address space
  1068.          * is being removed from the CPU.
  1069.          */
  1070.         as_deinstall_arch(old_as);
  1071.     }
  1072.  
  1073.     /*
  1074.      * Second, prepare the new address space.
  1075.      */
  1076.     if ((new_as->cpu_refcount++ == 0) && (new_as != AS_KERNEL)) {
  1077.         if (new_as->asid != ASID_INVALID)
  1078.             list_remove(&new_as->inactive_as_with_asid_link);
  1079.         else
  1080.             new_as->asid = asid_get();
  1081.     }
  1082. #ifdef AS_PAGE_TABLE
  1083.     SET_PTL0_ADDRESS(new_as->genarch.page_table);
  1084. #endif
  1085.    
  1086.     /*
  1087.      * Perform architecture-specific steps.
  1088.      * (e.g. write ASID to hardware register etc.)
  1089.      */
  1090.     as_install_arch(new_as);
  1091.  
  1092.     spinlock_unlock(&asidlock);
  1093.    
  1094.     AS = new_as;
  1095. }
  1096.  
  1097. /** Convert address space area flags to page flags.
  1098.  *
  1099.  * @param aflags Flags of some address space area.
  1100.  *
  1101.  * @return Flags to be passed to page_mapping_insert().
  1102.  */
  1103. int area_flags_to_page_flags(int aflags)
  1104. {
  1105.     int flags;
  1106.  
  1107.     flags = PAGE_USER | PAGE_PRESENT;
  1108.    
  1109.     if (aflags & AS_AREA_READ)
  1110.         flags |= PAGE_READ;
  1111.        
  1112.     if (aflags & AS_AREA_WRITE)
  1113.         flags |= PAGE_WRITE;
  1114.    
  1115.     if (aflags & AS_AREA_EXEC)
  1116.         flags |= PAGE_EXEC;
  1117.    
  1118.     if (aflags & AS_AREA_CACHEABLE)
  1119.         flags |= PAGE_CACHEABLE;
  1120.        
  1121.     return flags;
  1122. }
  1123.  
  1124. /** Compute flags for virtual address translation subsytem.
  1125.  *
  1126.  * The address space area must be locked.
  1127.  * Interrupts must be disabled.
  1128.  *
  1129.  * @param a Address space area.
  1130.  *
  1131.  * @return Flags to be used in page_mapping_insert().
  1132.  */
  1133. int as_area_get_flags(as_area_t *a)
  1134. {
  1135.     return area_flags_to_page_flags(a->flags);
  1136. }
  1137.  
  1138. /** Create page table.
  1139.  *
  1140.  * Depending on architecture, create either address space
  1141.  * private or global page table.
  1142.  *
  1143.  * @param flags Flags saying whether the page table is for kernel address space.
  1144.  *
  1145.  * @return First entry of the page table.
  1146.  */
  1147. pte_t *page_table_create(int flags)
  1148. {
  1149.     ASSERT(as_operations);
  1150.     ASSERT(as_operations->page_table_create);
  1151.    
  1152.     return as_operations->page_table_create(flags);
  1153. }
  1154.  
  1155. /** Destroy page table.
  1156.  *
  1157.  * Destroy page table in architecture specific way.
  1158.  *
  1159.  * @param page_table Physical address of PTL0.
  1160.  */
  1161. void page_table_destroy(pte_t *page_table)
  1162. {
  1163.     ASSERT(as_operations);
  1164.     ASSERT(as_operations->page_table_destroy);
  1165.    
  1166.     as_operations->page_table_destroy(page_table);
  1167. }
  1168.  
  1169. /** Lock page table.
  1170.  *
  1171.  * This function should be called before any page_mapping_insert(),
  1172.  * page_mapping_remove() and page_mapping_find().
  1173.  *
  1174.  * Locking order is such that address space areas must be locked
  1175.  * prior to this call. Address space can be locked prior to this
  1176.  * call in which case the lock argument is false.
  1177.  *
  1178.  * @param as Address space.
  1179.  * @param lock If false, do not attempt to lock as->lock.
  1180.  */
  1181. void page_table_lock(as_t *as, bool lock)
  1182. {
  1183.     ASSERT(as_operations);
  1184.     ASSERT(as_operations->page_table_lock);
  1185.    
  1186.     as_operations->page_table_lock(as, lock);
  1187. }
  1188.  
  1189. /** Unlock page table.
  1190.  *
  1191.  * @param as Address space.
  1192.  * @param unlock If false, do not attempt to unlock as->lock.
  1193.  */
  1194. void page_table_unlock(as_t *as, bool unlock)
  1195. {
  1196.     ASSERT(as_operations);
  1197.     ASSERT(as_operations->page_table_unlock);
  1198.    
  1199.     as_operations->page_table_unlock(as, unlock);
  1200. }
  1201.  
  1202.  
  1203. /** Find address space area and lock it.
  1204.  *
  1205.  * The address space must be locked and interrupts must be disabled.
  1206.  *
  1207.  * @param as Address space.
  1208.  * @param va Virtual address.
  1209.  *
  1210.  * @return Locked address space area containing va on success or NULL on
  1211.  *     failure.
  1212.  */
  1213. as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
  1214. {
  1215.     as_area_t *a;
  1216.     btree_node_t *leaf, *lnode;
  1217.     unsigned int i;
  1218.    
  1219.     a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
  1220.     if (a) {
  1221.         /* va is the base address of an address space area */
  1222.         mutex_lock(&a->lock);
  1223.         return a;
  1224.     }
  1225.    
  1226.     /*
  1227.      * Search the leaf node and the righmost record of its left neighbour
  1228.      * to find out whether this is a miss or va belongs to an address
  1229.      * space area found there.
  1230.      */
  1231.    
  1232.     /* First, search the leaf node itself. */
  1233.     for (i = 0; i < leaf->keys; i++) {
  1234.         a = (as_area_t *) leaf->value[i];
  1235.         mutex_lock(&a->lock);
  1236.         if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
  1237.             return a;
  1238.         }
  1239.         mutex_unlock(&a->lock);
  1240.     }
  1241.  
  1242.     /*
  1243.      * Second, locate the left neighbour and test its last record.
  1244.      * Because of its position in the B+tree, it must have base < va.
  1245.      */
  1246.     lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf);
  1247.     if (lnode) {
  1248.         a = (as_area_t *) lnode->value[lnode->keys - 1];
  1249.         mutex_lock(&a->lock);
  1250.         if (va < a->base + a->pages * PAGE_SIZE) {
  1251.             return a;
  1252.         }
  1253.         mutex_unlock(&a->lock);
  1254.     }
  1255.  
  1256.     return NULL;
  1257. }
  1258.  
  1259. /** Check area conflicts with other areas.
  1260.  *
  1261.  * The address space must be locked and interrupts must be disabled.
  1262.  *
  1263.  * @param as Address space.
  1264.  * @param va Starting virtual address of the area being tested.
  1265.  * @param size Size of the area being tested.
  1266.  * @param avoid_area Do not touch this area.
  1267.  *
  1268.  * @return True if there is no conflict, false otherwise.
  1269.  */
  1270. bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
  1271.               as_area_t *avoid_area)
  1272. {
  1273.     as_area_t *a;
  1274.     btree_node_t *leaf, *node;
  1275.     unsigned int i;
  1276.    
  1277.     /*
  1278.      * We don't want any area to have conflicts with NULL page.
  1279.      */
  1280.     if (overlaps(va, size, NULL, PAGE_SIZE))
  1281.         return false;
  1282.    
  1283.     /*
  1284.      * The leaf node is found in O(log n), where n is proportional to
  1285.      * the number of address space areas belonging to as.
  1286.      * The check for conflicts is then attempted on the rightmost
  1287.      * record in the left neighbour, the leftmost record in the right
  1288.      * neighbour and all records in the leaf node itself.
  1289.      */
  1290.    
  1291.     if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
  1292.         if (a != avoid_area)
  1293.             return false;
  1294.     }
  1295.    
  1296.     /* First, check the two border cases. */
  1297.     if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
  1298.         a = (as_area_t *) node->value[node->keys - 1];
  1299.         mutex_lock(&a->lock);
  1300.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1301.             mutex_unlock(&a->lock);
  1302.             return false;
  1303.         }
  1304.         mutex_unlock(&a->lock);
  1305.     }
  1306.     node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf);
  1307.     if (node) {
  1308.         a = (as_area_t *) node->value[0];
  1309.         mutex_lock(&a->lock);
  1310.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1311.             mutex_unlock(&a->lock);
  1312.             return false;
  1313.         }
  1314.         mutex_unlock(&a->lock);
  1315.     }
  1316.    
  1317.     /* Second, check the leaf node. */
  1318.     for (i = 0; i < leaf->keys; i++) {
  1319.         a = (as_area_t *) leaf->value[i];
  1320.    
  1321.         if (a == avoid_area)
  1322.             continue;
  1323.    
  1324.         mutex_lock(&a->lock);
  1325.         if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
  1326.             mutex_unlock(&a->lock);
  1327.             return false;
  1328.         }
  1329.         mutex_unlock(&a->lock);
  1330.     }
  1331.  
  1332.     /*
  1333.      * So far, the area does not conflict with other areas.
  1334.      * Check if it doesn't conflict with kernel address space.
  1335.      */  
  1336.     if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
  1337.         return !overlaps(va, size,
  1338.             KERNEL_ADDRESS_SPACE_START,
  1339.             KERNEL_ADDRESS_SPACE_END - KERNEL_ADDRESS_SPACE_START);
  1340.     }
  1341.  
  1342.     return true;
  1343. }
  1344.  
  1345. /** Return size of the address space area with given base.
  1346.  *
  1347.  * @param base      Arbitrary address insede the address space area.
  1348.  *
  1349.  * @return      Size of the address space area in bytes or zero if it
  1350.  *          does not exist.
  1351.  */
  1352. size_t as_area_get_size(uintptr_t base)
  1353. {
  1354.     ipl_t ipl;
  1355.     as_area_t *src_area;
  1356.     size_t size;
  1357.  
  1358.     ipl = interrupts_disable();
  1359.     src_area = find_area_and_lock(AS, base);
  1360.     if (src_area){
  1361.         size = src_area->pages * PAGE_SIZE;
  1362.         mutex_unlock(&src_area->lock);
  1363.     } else {
  1364.         size = 0;
  1365.     }
  1366.     interrupts_restore(ipl);
  1367.     return size;
  1368. }
  1369.  
  1370. /** Mark portion of address space area as used.
  1371.  *
  1372.  * The address space area must be already locked.
  1373.  *
  1374.  * @param a Address space area.
  1375.  * @param page First page to be marked.
  1376.  * @param count Number of page to be marked.
  1377.  *
  1378.  * @return 0 on failure and 1 on success.
  1379.  */
  1380. int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
  1381. {
  1382.     btree_node_t *leaf, *node;
  1383.     count_t pages;
  1384.     unsigned int i;
  1385.  
  1386.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1387.     ASSERT(count);
  1388.  
  1389.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1390.     if (pages) {
  1391.         /*
  1392.          * We hit the beginning of some used space.
  1393.          */
  1394.         return 0;
  1395.     }
  1396.  
  1397.     if (!leaf->keys) {
  1398.         btree_insert(&a->used_space, page, (void *) count, leaf);
  1399.         return 1;
  1400.     }
  1401.  
  1402.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1403.     if (node) {
  1404.         uintptr_t left_pg = node->key[node->keys - 1];
  1405.         uintptr_t right_pg = leaf->key[0];
  1406.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1407.         count_t right_cnt = (count_t) leaf->value[0];
  1408.        
  1409.         /*
  1410.          * Examine the possibility that the interval fits
  1411.          * somewhere between the rightmost interval of
  1412.          * the left neigbour and the first interval of the leaf.
  1413.          */
  1414.          
  1415.         if (page >= right_pg) {
  1416.             /* Do nothing. */
  1417.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1418.             left_cnt * PAGE_SIZE)) {
  1419.             /* The interval intersects with the left interval. */
  1420.             return 0;
  1421.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1422.             right_cnt * PAGE_SIZE)) {
  1423.             /* The interval intersects with the right interval. */
  1424.             return 0;          
  1425.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1426.             (page + count * PAGE_SIZE == right_pg)) {
  1427.             /*
  1428.              * The interval can be added by merging the two already
  1429.              * present intervals.
  1430.              */
  1431.             node->value[node->keys - 1] += count + right_cnt;
  1432.             btree_remove(&a->used_space, right_pg, leaf);
  1433.             return 1;
  1434.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1435.             /*
  1436.              * The interval can be added by simply growing the left
  1437.              * interval.
  1438.              */
  1439.             node->value[node->keys - 1] += count;
  1440.             return 1;
  1441.         } else if (page + count * PAGE_SIZE == right_pg) {
  1442.             /*
  1443.              * The interval can be addded by simply moving base of
  1444.              * the right interval down and increasing its size
  1445.              * accordingly.
  1446.              */
  1447.             leaf->value[0] += count;
  1448.             leaf->key[0] = page;
  1449.             return 1;
  1450.         } else {
  1451.             /*
  1452.              * The interval is between both neigbouring intervals,
  1453.              * but cannot be merged with any of them.
  1454.              */
  1455.             btree_insert(&a->used_space, page, (void *) count,
  1456.                 leaf);
  1457.             return 1;
  1458.         }
  1459.     } else if (page < leaf->key[0]) {
  1460.         uintptr_t right_pg = leaf->key[0];
  1461.         count_t right_cnt = (count_t) leaf->value[0];
  1462.    
  1463.         /*
  1464.          * Investigate the border case in which the left neighbour does
  1465.          * not exist but the interval fits from the left.
  1466.          */
  1467.          
  1468.         if (overlaps(page, count * PAGE_SIZE, right_pg,
  1469.             right_cnt * PAGE_SIZE)) {
  1470.             /* The interval intersects with the right interval. */
  1471.             return 0;
  1472.         } else if (page + count * PAGE_SIZE == right_pg) {
  1473.             /*
  1474.              * The interval can be added by moving the base of the
  1475.              * right interval down and increasing its size
  1476.              * accordingly.
  1477.              */
  1478.             leaf->key[0] = page;
  1479.             leaf->value[0] += count;
  1480.             return 1;
  1481.         } else {
  1482.             /*
  1483.              * The interval doesn't adjoin with the right interval.
  1484.              * It must be added individually.
  1485.              */
  1486.             btree_insert(&a->used_space, page, (void *) count,
  1487.                 leaf);
  1488.             return 1;
  1489.         }
  1490.     }
  1491.  
  1492.     node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
  1493.     if (node) {
  1494.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1495.         uintptr_t right_pg = node->key[0];
  1496.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1497.         count_t right_cnt = (count_t) node->value[0];
  1498.        
  1499.         /*
  1500.          * Examine the possibility that the interval fits
  1501.          * somewhere between the leftmost interval of
  1502.          * the right neigbour and the last interval of the leaf.
  1503.          */
  1504.  
  1505.         if (page < left_pg) {
  1506.             /* Do nothing. */
  1507.         } else if (overlaps(page, count * PAGE_SIZE, left_pg,
  1508.             left_cnt * PAGE_SIZE)) {
  1509.             /* The interval intersects with the left interval. */
  1510.             return 0;
  1511.         } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1512.             right_cnt * PAGE_SIZE)) {
  1513.             /* The interval intersects with the right interval. */
  1514.             return 0;          
  1515.         } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1516.             (page + count * PAGE_SIZE == right_pg)) {
  1517.             /*
  1518.              * The interval can be added by merging the two already
  1519.              * present intervals.
  1520.              * */
  1521.             leaf->value[leaf->keys - 1] += count + right_cnt;
  1522.             btree_remove(&a->used_space, right_pg, node);
  1523.             return 1;
  1524.         } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1525.             /*
  1526.              * The interval can be added by simply growing the left
  1527.              * interval.
  1528.              * */
  1529.             leaf->value[leaf->keys - 1] +=  count;
  1530.             return 1;
  1531.         } else if (page + count * PAGE_SIZE == right_pg) {
  1532.             /*
  1533.              * The interval can be addded by simply moving base of
  1534.              * the right interval down and increasing its size
  1535.              * accordingly.
  1536.              */
  1537.             node->value[0] += count;
  1538.             node->key[0] = page;
  1539.             return 1;
  1540.         } else {
  1541.             /*
  1542.              * The interval is between both neigbouring intervals,
  1543.              * but cannot be merged with any of them.
  1544.              */
  1545.             btree_insert(&a->used_space, page, (void *) count,
  1546.                 leaf);
  1547.             return 1;
  1548.         }
  1549.     } else if (page >= leaf->key[leaf->keys - 1]) {
  1550.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1551.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1552.    
  1553.         /*
  1554.          * Investigate the border case in which the right neighbour
  1555.          * does not exist but the interval fits from the right.
  1556.          */
  1557.          
  1558.         if (overlaps(page, count * PAGE_SIZE, left_pg,
  1559.             left_cnt * PAGE_SIZE)) {
  1560.             /* The interval intersects with the left interval. */
  1561.             return 0;
  1562.         } else if (left_pg + left_cnt * PAGE_SIZE == page) {
  1563.             /*
  1564.              * The interval can be added by growing the left
  1565.              * interval.
  1566.              */
  1567.             leaf->value[leaf->keys - 1] += count;
  1568.             return 1;
  1569.         } else {
  1570.             /*
  1571.              * The interval doesn't adjoin with the left interval.
  1572.              * It must be added individually.
  1573.              */
  1574.             btree_insert(&a->used_space, page, (void *) count,
  1575.                 leaf);
  1576.             return 1;
  1577.         }
  1578.     }
  1579.    
  1580.     /*
  1581.      * Note that if the algorithm made it thus far, the interval can fit
  1582.      * only between two other intervals of the leaf. The two border cases
  1583.      * were already resolved.
  1584.      */
  1585.     for (i = 1; i < leaf->keys; i++) {
  1586.         if (page < leaf->key[i]) {
  1587.             uintptr_t left_pg = leaf->key[i - 1];
  1588.             uintptr_t right_pg = leaf->key[i];
  1589.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1590.             count_t right_cnt = (count_t) leaf->value[i];
  1591.  
  1592.             /*
  1593.              * The interval fits between left_pg and right_pg.
  1594.              */
  1595.  
  1596.             if (overlaps(page, count * PAGE_SIZE, left_pg,
  1597.                 left_cnt * PAGE_SIZE)) {
  1598.                 /*
  1599.                  * The interval intersects with the left
  1600.                  * interval.
  1601.                  */
  1602.                 return 0;
  1603.             } else if (overlaps(page, count * PAGE_SIZE, right_pg,
  1604.                 right_cnt * PAGE_SIZE)) {
  1605.                 /*
  1606.                  * The interval intersects with the right
  1607.                  * interval.
  1608.                  */
  1609.                 return 0;          
  1610.             } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
  1611.                 (page + count * PAGE_SIZE == right_pg)) {
  1612.                 /*
  1613.                  * The interval can be added by merging the two
  1614.                  * already present intervals.
  1615.                  */
  1616.                 leaf->value[i - 1] += count + right_cnt;
  1617.                 btree_remove(&a->used_space, right_pg, leaf);
  1618.                 return 1;
  1619.             } else if (page == left_pg + left_cnt * PAGE_SIZE) {
  1620.                 /*
  1621.                  * The interval can be added by simply growing
  1622.                  * the left interval.
  1623.                  */
  1624.                 leaf->value[i - 1] += count;
  1625.                 return 1;
  1626.             } else if (page + count * PAGE_SIZE == right_pg) {
  1627.                 /*
  1628.                      * The interval can be addded by simply moving
  1629.                  * base of the right interval down and
  1630.                  * increasing its size accordingly.
  1631.                  */
  1632.                 leaf->value[i] += count;
  1633.                 leaf->key[i] = page;
  1634.                 return 1;
  1635.             } else {
  1636.                 /*
  1637.                  * The interval is between both neigbouring
  1638.                  * intervals, but cannot be merged with any of
  1639.                  * them.
  1640.                  */
  1641.                 btree_insert(&a->used_space, page,
  1642.                     (void *) count, leaf);
  1643.                 return 1;
  1644.             }
  1645.         }
  1646.     }
  1647.  
  1648.     panic("Inconsistency detected while adding %" PRIc " pages of used space at "
  1649.         "%p.\n", count, page);
  1650. }
  1651.  
  1652. /** Mark portion of address space area as unused.
  1653.  *
  1654.  * The address space area must be already locked.
  1655.  *
  1656.  * @param a Address space area.
  1657.  * @param page First page to be marked.
  1658.  * @param count Number of page to be marked.
  1659.  *
  1660.  * @return 0 on failure and 1 on success.
  1661.  */
  1662. int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
  1663. {
  1664.     btree_node_t *leaf, *node;
  1665.     count_t pages;
  1666.     unsigned int i;
  1667.  
  1668.     ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
  1669.     ASSERT(count);
  1670.  
  1671.     pages = (count_t) btree_search(&a->used_space, page, &leaf);
  1672.     if (pages) {
  1673.         /*
  1674.          * We are lucky, page is the beginning of some interval.
  1675.          */
  1676.         if (count > pages) {
  1677.             return 0;
  1678.         } else if (count == pages) {
  1679.             btree_remove(&a->used_space, page, leaf);
  1680.             return 1;
  1681.         } else {
  1682.             /*
  1683.              * Find the respective interval.
  1684.              * Decrease its size and relocate its start address.
  1685.              */
  1686.             for (i = 0; i < leaf->keys; i++) {
  1687.                 if (leaf->key[i] == page) {
  1688.                     leaf->key[i] += count * PAGE_SIZE;
  1689.                     leaf->value[i] -= count;
  1690.                     return 1;
  1691.                 }
  1692.             }
  1693.             goto error;
  1694.         }
  1695.     }
  1696.  
  1697.     node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
  1698.     if (node && page < leaf->key[0]) {
  1699.         uintptr_t left_pg = node->key[node->keys - 1];
  1700.         count_t left_cnt = (count_t) node->value[node->keys - 1];
  1701.  
  1702.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1703.             count * PAGE_SIZE)) {
  1704.             if (page + count * PAGE_SIZE ==
  1705.                 left_pg + left_cnt * PAGE_SIZE) {
  1706.                 /*
  1707.                  * The interval is contained in the rightmost
  1708.                  * interval of the left neighbour and can be
  1709.                  * removed by updating the size of the bigger
  1710.                  * interval.
  1711.                  */
  1712.                 node->value[node->keys - 1] -= count;
  1713.                 return 1;
  1714.             } else if (page + count * PAGE_SIZE <
  1715.                 left_pg + left_cnt*PAGE_SIZE) {
  1716.                 count_t new_cnt;
  1717.                
  1718.                 /*
  1719.                  * The interval is contained in the rightmost
  1720.                  * interval of the left neighbour but its
  1721.                  * removal requires both updating the size of
  1722.                  * the original interval and also inserting a
  1723.                  * new interval.
  1724.                  */
  1725.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1726.                     (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
  1727.                 node->value[node->keys - 1] -= count + new_cnt;
  1728.                 btree_insert(&a->used_space, page +
  1729.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1730.                 return 1;
  1731.             }
  1732.         }
  1733.         return 0;
  1734.     } else if (page < leaf->key[0]) {
  1735.         return 0;
  1736.     }
  1737.    
  1738.     if (page > leaf->key[leaf->keys - 1]) {
  1739.         uintptr_t left_pg = leaf->key[leaf->keys - 1];
  1740.         count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
  1741.  
  1742.         if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1743.             count * PAGE_SIZE)) {
  1744.             if (page + count * PAGE_SIZE ==
  1745.                 left_pg + left_cnt * PAGE_SIZE) {
  1746.                 /*
  1747.                  * The interval is contained in the rightmost
  1748.                  * interval of the leaf and can be removed by
  1749.                  * updating the size of the bigger interval.
  1750.                  */
  1751.                 leaf->value[leaf->keys - 1] -= count;
  1752.                 return 1;
  1753.             } else if (page + count * PAGE_SIZE < left_pg +
  1754.                 left_cnt * PAGE_SIZE) {
  1755.                 count_t new_cnt;
  1756.                
  1757.                 /*
  1758.                  * The interval is contained in the rightmost
  1759.                  * interval of the leaf but its removal
  1760.                  * requires both updating the size of the
  1761.                  * original interval and also inserting a new
  1762.                  * interval.
  1763.                  */
  1764.                 new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
  1765.                     (page + count * PAGE_SIZE)) >> PAGE_WIDTH;
  1766.                 leaf->value[leaf->keys - 1] -= count + new_cnt;
  1767.                 btree_insert(&a->used_space, page +
  1768.                     count * PAGE_SIZE, (void *) new_cnt, leaf);
  1769.                 return 1;
  1770.             }
  1771.         }
  1772.         return 0;
  1773.     }  
  1774.    
  1775.     /*
  1776.      * The border cases have been already resolved.
  1777.      * Now the interval can be only between intervals of the leaf.
  1778.      */
  1779.     for (i = 1; i < leaf->keys - 1; i++) {
  1780.         if (page < leaf->key[i]) {
  1781.             uintptr_t left_pg = leaf->key[i - 1];
  1782.             count_t left_cnt = (count_t) leaf->value[i - 1];
  1783.  
  1784.             /*
  1785.              * Now the interval is between intervals corresponding
  1786.              * to (i - 1) and i.
  1787.              */
  1788.             if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
  1789.                 count * PAGE_SIZE)) {
  1790.                 if (page + count * PAGE_SIZE ==
  1791.                     left_pg + left_cnt*PAGE_SIZE) {
  1792.                     /*
  1793.                      * The interval is contained in the
  1794.                      * interval (i - 1) of the leaf and can
  1795.                      * be removed by updating the size of
  1796.                      * the bigger interval.
  1797.                      */
  1798.                     leaf->value[i - 1] -= count;
  1799.                     return 1;
  1800.                 } else if (page + count * PAGE_SIZE <
  1801.                     left_pg + left_cnt * PAGE_SIZE) {
  1802.                     count_t new_cnt;
  1803.                
  1804.                     /*
  1805.                      * The interval is contained in the
  1806.                      * interval (i - 1) of the leaf but its
  1807.                      * removal requires both updating the
  1808.                      * size of the original interval and
  1809.                      * also inserting a new interval.
  1810.                      */
  1811.                     new_cnt = ((left_pg +
  1812.                         left_cnt * PAGE_SIZE) -
  1813.                         (page + count * PAGE_SIZE)) >>
  1814.                         PAGE_WIDTH;
  1815.                     leaf->value[i - 1] -= count + new_cnt;
  1816.                     btree_insert(&a->used_space, page +
  1817.                         count * PAGE_SIZE, (void *) new_cnt,
  1818.                         leaf);
  1819.                     return 1;
  1820.                 }
  1821.             }
  1822.             return 0;
  1823.         }
  1824.     }
  1825.  
  1826. error:
  1827.     panic("Inconsistency detected while removing %" PRIc " pages of used space "
  1828.         "from %p.\n", count, page);
  1829. }
  1830.  
  1831. /** Remove reference to address space area share info.
  1832.  *
  1833.  * If the reference count drops to 0, the sh_info is deallocated.
  1834.  *
  1835.  * @param sh_info Pointer to address space area share info.
  1836.  */
  1837. void sh_info_remove_reference(share_info_t *sh_info)
  1838. {
  1839.     bool dealloc = false;
  1840.  
  1841.     mutex_lock(&sh_info->lock);
  1842.     ASSERT(sh_info->refcount);
  1843.     if (--sh_info->refcount == 0) {
  1844.         dealloc = true;
  1845.         link_t *cur;
  1846.        
  1847.         /*
  1848.          * Now walk carefully the pagemap B+tree and free/remove
  1849.          * reference from all frames found there.
  1850.          */
  1851.         for (cur = sh_info->pagemap.leaf_head.next;
  1852.             cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
  1853.             btree_node_t *node;
  1854.             unsigned int i;
  1855.            
  1856.             node = list_get_instance(cur, btree_node_t, leaf_link);
  1857.             for (i = 0; i < node->keys; i++)
  1858.                 frame_free((uintptr_t) node->value[i]);
  1859.         }
  1860.        
  1861.     }
  1862.     mutex_unlock(&sh_info->lock);
  1863.    
  1864.     if (dealloc) {
  1865.         btree_destroy(&sh_info->pagemap);
  1866.         free(sh_info);
  1867.     }
  1868. }
  1869.  
  1870. /*
  1871.  * Address space related syscalls.
  1872.  */
  1873.  
  1874. /** Wrapper for as_area_create(). */
  1875. unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
  1876. {
  1877.     if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address,
  1878.         AS_AREA_ATTR_NONE, &anon_backend, NULL))
  1879.         return (unative_t) address;
  1880.     else
  1881.         return (unative_t) -1;
  1882. }
  1883.  
  1884. /** Wrapper for as_area_resize(). */
  1885. unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
  1886. {
  1887.     return (unative_t) as_area_resize(AS, address, size, 0);
  1888. }
  1889.  
  1890. /** Wrapper for as_area_change_flags(). */
  1891. unative_t sys_as_area_change_flags(uintptr_t address, int flags)
  1892. {
  1893.     return (unative_t) as_area_change_flags(AS, flags, address);
  1894. }
  1895.  
  1896. /** Wrapper for as_area_destroy(). */
  1897. unative_t sys_as_area_destroy(uintptr_t address)
  1898. {
  1899.     return (unative_t) as_area_destroy(AS, address);
  1900. }
  1901.  
  1902. /** Print out information about address space.
  1903.  *
  1904.  * @param as Address space.
  1905.  */
  1906. void as_print(as_t *as)
  1907. {
  1908.     ipl_t ipl;
  1909.    
  1910.     ipl = interrupts_disable();
  1911.     mutex_lock(&as->lock);
  1912.    
  1913.     /* print out info about address space areas */
  1914.     link_t *cur;
  1915.     for (cur = as->as_area_btree.leaf_head.next;
  1916.         cur != &as->as_area_btree.leaf_head; cur = cur->next) {
  1917.         btree_node_t *node;
  1918.        
  1919.         node = list_get_instance(cur, btree_node_t, leaf_link);
  1920.        
  1921.         unsigned int i;
  1922.         for (i = 0; i < node->keys; i++) {
  1923.             as_area_t *area = node->value[i];
  1924.        
  1925.             mutex_lock(&area->lock);
  1926.             printf("as_area: %p, base=%p, pages=%" PRIc " (%p - %p)\n",
  1927.                 area, area->base, area->pages, area->base,
  1928.                 area->base + FRAMES2SIZE(area->pages));
  1929.             mutex_unlock(&area->lock);
  1930.         }
  1931.     }
  1932.    
  1933.     mutex_unlock(&as->lock);
  1934.     interrupts_restore(ipl);
  1935. }
  1936.  
  1937. /** @}
  1938.  */
  1939.