WebSVN – HelenOS – Diff – /trunk/kernel/generic/src/mm/as.c


/*
 * Copyright (c) 2001-2006 Jakub Jermar
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * - Redistributions of source code must retain the above copyright
 *   notice, this list of conditions and the following disclaimer.
 * - Redistributions in binary form must reproduce the above copyright
 *   notice, this list of conditions and the following disclaimer in the
 *   documentation and/or other materials provided with the distribution.
 * - The name of the author may not be used to endorse or promote products
 *   derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */
 
/** @addtogroup genericmm
 * @{
 */
 
/**
 * @file
 * @brief   Address space related functions.
 *
 * This file contains address space manipulation functions.
 * Roughly speaking, this is a higher-level client of
 * Virtual Address Translation (VAT) subsystem.
 *
 * Functionality provided by this file allows one to
 * create address spaces and create, resize and share
 * address space areas.
 *
 * @see page.c
 *
 */
 
#include <mm/as.h>
#include <arch/mm/as.h>
#include <mm/page.h>
#include <mm/frame.h>
#include <mm/slab.h>
#include <mm/tlb.h>
#include <arch/mm/page.h>
#include <genarch/mm/page_pt.h>
#include <genarch/mm/page_ht.h>
#include <mm/asid.h>
#include <arch/mm/asid.h>
#include <synch/spinlock.h>
#include <synch/mutex.h>
#include <adt/list.h>
#include <adt/btree.h>
#include <proc/task.h>
#include <proc/thread.h>
#include <arch/asm.h>
#include <panic.h>
#include <debug.h>
#include <print.h>
#include <memstr.h>
#include <macros.h>
#include <arch.h>
#include <errno.h>
#include <config.h>
#include <align.h>
#include <arch/types.h>
 
#include <syscall/copy.h>
#include <arch/interrupt.h>
 
#ifdef CONFIG_VIRT_IDX_DCACHE
#include <arch/mm/cache.h>
#endif /* CONFIG_VIRT_IDX_DCACHE */
 
/**
 * Each architecture decides what functions will be used to carry out
 * address space operations such as creating or locking page tables.
 */
as_operations_t *as_operations = NULL;
 
/**
 * Slab for as_t objects.
 */
static slab_cache_t *as_slab;
 
/**
 * This lock protects inactive_as_with_asid_head list. It must be acquired
 * before as_t mutex.
 */
SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
 
/**
 * This list contains address spaces that are not active on any
 * processor and that have valid ASID.
 */
LIST_INITIALIZE(inactive_as_with_asid_head);
 
/** Kernel address space. */
as_t *AS_KERNEL = NULL;
 
static int area_flags_to_page_flags(int aflags);
static as_area_t *find_area_and_lock(as_t *as, uintptr_t va);
static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
    as_area_t *avoid_area);
static void sh_info_remove_reference(share_info_t *sh_info);
 
static int as_constructor(void *obj, int flags)
{
    as_t *as = (as_t *) obj;
    int rc;
 
    link_initialize(&as->inactive_as_with_asid_link);
    mutex_initialize(&as->lock);   
   
    rc = as_constructor_arch(as, flags);
   
    return rc;
}
 
static int as_destructor(void *obj)
{
    as_t *as = (as_t *) obj;
 
    return as_destructor_arch(as);
}
 
/** Initialize address space subsystem. */
void as_init(void)
{
    as_arch_init();
   
    as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,
        as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);
   
    AS_KERNEL = as_create(FLAG_AS_KERNEL);
    if (!AS_KERNEL)
        panic("can't create kernel address space\n");
   
}
 
/** Create address space.
 *
 * @param flags Flags that influence way in wich the address space is created.
 */
as_t *as_create(int flags)
{
    as_t *as;
 
    as = (as_t *) slab_alloc(as_slab, 0);
    (void) as_create_arch(as, 0);
   
    btree_create(&as->as_area_btree);
   
    if (flags & FLAG_AS_KERNEL)
        as->asid = ASID_KERNEL;
    else
        as->asid = ASID_INVALID;
   
    as->refcount = 0;
    as->cpu_refcount = 0;
#ifdef AS_PAGE_TABLE
    as->page_table = page_table_create(flags);
#else
    page_table_create(flags);
#endif
 
    return as;
}
 
/** Destroy adress space.
 *
 * When there are no tasks referencing this address space (i.e. its refcount is
 * zero), the address space can be destroyed.
 */
void as_destroy(as_t *as)
{
    ipl_t ipl;
    bool cond;
 
    ASSERT(as->refcount == 0);
   
    /*
     * Since there is no reference to this area,
     * it is safe not to lock its mutex.
     */
    ipl = interrupts_disable();
    spinlock_lock(&inactive_as_with_asid_lock);
    if (as->asid != ASID_INVALID && as != AS_KERNEL) {
        if (as != AS && as->cpu_refcount == 0)
            list_remove(&as->inactive_as_with_asid_link);
        asid_put(as->asid);
    }
    spinlock_unlock(&inactive_as_with_asid_lock);
 
    /*
     * Destroy address space areas of the address space.
     * The B+tree must be walked carefully because it is
     * also being destroyed.
     */
    for (cond = true; cond; ) {
        btree_node_t *node;
 
        ASSERT(!list_empty(&as->as_area_btree.leaf_head));
        node = list_get_instance(as->as_area_btree.leaf_head.next,
            btree_node_t, leaf_link);
 
        if ((cond = node->keys)) {
            as_area_destroy(as, node->key[0]);
        }
    }
 
    btree_destroy(&as->as_area_btree);
#ifdef AS_PAGE_TABLE
    page_table_destroy(as->page_table);
#else
    page_table_destroy(NULL);
#endif
 
    interrupts_restore(ipl);
   
    slab_free(as_slab, as);
}
 
/** Create address space area of common attributes.
 *
 * The created address space area is added to the target address space.
 *
 * @param as Target address space.
 * @param flags Flags of the area memory.
 * @param size Size of area.
 * @param base Base address of area.
 * @param attrs Attributes of the area.
 * @param backend Address space area backend. NULL if no backend is used.
 * @param backend_data NULL or a pointer to an array holding two void *.
 *
 * @return Address space area on success or NULL on failure.
 */
as_area_t *
as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
           mem_backend_t *backend, mem_backend_data_t *backend_data)
{
    ipl_t ipl;
    as_area_t *a;
   
    if (base % PAGE_SIZE)
        return NULL;
 
    if (!size)
        return NULL;
 
    /* Writeable executable areas are not supported. */
    if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
        return NULL;
   
    ipl = interrupts_disable();
    mutex_lock(&as->lock);
   
    if (!check_area_conflicts(as, base, size, NULL)) {
        mutex_unlock(&as->lock);
        interrupts_restore(ipl);
        return NULL;
    }
   
    a = (as_area_t *) malloc(sizeof(as_area_t), 0);
 
    mutex_initialize(&a->lock);
   
    a->as = as;
    a->flags = flags;
    a->attributes = attrs;
    a->pages = SIZE2FRAMES(size);
    a->base = base;
    a->sh_info = NULL;
    a->backend = backend;
    if (backend_data)
        a->backend_data = *backend_data;
    else
        memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data),
            0);
 
    btree_create(&a->used_space);
   
    btree_insert(&as->as_area_btree, base, (void *) a, NULL);
 
    mutex_unlock(&as->lock);
    interrupts_restore(ipl);
 
    return a;
}
 
/** Find address space area and change it.
 *
 * @param as Address space.
 * @param address Virtual address belonging to the area to be changed. Must be
 *     page-aligned.
 * @param size New size of the virtual memory block starting at address.
 * @param flags Flags influencing the remap operation. Currently unused.
 *
 * @return Zero on success or a value from @ref errno.h otherwise.
 */
int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
{
    as_area_t *area;
    ipl_t ipl;
    size_t pages;
   
    ipl = interrupts_disable();
    mutex_lock(&as->lock);
   
    /*
     * Locate the area.
     */
    area = find_area_and_lock(as, address);
    if (!area) {
        mutex_unlock(&as->lock);
        interrupts_restore(ipl);
        return ENOENT;
    }
 
    if (area->backend == &phys_backend) {
        /*
         * Remapping of address space areas associated
         * with memory mapped devices is not supported.
         */
        mutex_unlock(&area->lock);
        mutex_unlock(&as->lock);
        interrupts_restore(ipl);
        return ENOTSUP;
    }
    if (area->sh_info) {
        /*
         * Remapping of shared address space areas
         * is not supported.
         */
        mutex_unlock(&area->lock);
        mutex_unlock(&as->lock);
        interrupts_restore(ipl);
        return ENOTSUP;
    }
 
    pages = SIZE2FRAMES((address - area->base) + size);
    if (!pages) {
        /*
         * Zero size address space areas are not allowed.
         */
        mutex_unlock(&area->lock);
        mutex_unlock(&as->lock);
        interrupts_restore(ipl);
        return EPERM;
    }
   
    if (pages < area->pages) {
        bool cond;
        uintptr_t start_free = area->base + pages*PAGE_SIZE;
 
        /*
         * Shrinking the area.
         * No need to check for overlaps.
         */
 
        /*
         * Start TLB shootdown sequence.
         */
        tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base +
            pages * PAGE_SIZE, area->pages - pages);
 
        /*
         * Remove frames belonging to used space starting from
         * the highest addresses downwards until an overlap with
         * the resized address space area is found. Note that this
         * is also the right way to remove part of the used_space
         * B+tree leaf list.
         */    
        for (cond = true; cond;) {
            btree_node_t *node;
       
            ASSERT(!list_empty(&area->used_space.leaf_head));
            node =
                list_get_instance(area->used_space.leaf_head.prev,
                btree_node_t, leaf_link);
            if ((cond = (bool) node->keys)) {
                uintptr_t b = node->key[node->keys - 1];
                count_t c =
                    (count_t) node->value[node->keys - 1];
                int i = 0;
           
                if (overlaps(b, c * PAGE_SIZE, area->base,
                    pages*PAGE_SIZE)) {
                   
                    if (b + c * PAGE_SIZE <= start_free) {
                        /*
                         * The whole interval fits
                         * completely in the resized
                         * address space area.
                         */
                        break;
                    }
       
                    /*
                     * Part of the interval corresponding
                     * to b and c overlaps with the resized
                     * address space area.
                     */
       
                    cond = false;   /* we are almost done */
                    i = (start_free - b) >> PAGE_WIDTH;
                    if (!used_space_remove(area, start_free,
                        c - i))
                        panic("Could not remove used "
                            "space.\n");
                } else {
                    /*
                     * The interval of used space can be
                     * completely removed.
                     */
                    if (!used_space_remove(area, b, c))
                        panic("Could not remove used "
                            "space.\n");
                }
           
                for (; i < c; i++) {
                    pte_t *pte;
           
                    page_table_lock(as, false);
                    pte = page_mapping_find(as, b +
                        i * PAGE_SIZE);
                    ASSERT(pte && PTE_VALID(pte) &&
                        PTE_PRESENT(pte));
                    if (area->backend &&
                        area->backend->frame_free) {
                        area->backend->frame_free(area,
                            b + i * PAGE_SIZE,
                            PTE_GET_FRAME(pte));
                    }
                    page_mapping_remove(as, b +
                        i * PAGE_SIZE);
                    page_table_unlock(as, false);
                }
            }
        }
 
        /*
         * Finish TLB shootdown sequence.
         */
        tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,
            area->pages - pages);
        tlb_shootdown_finalize();
       
        /*
         * Invalidate software translation caches (e.g. TSB on sparc64).
         */
        as_invalidate_translation_cache(as, area->base +
            pages * PAGE_SIZE, area->pages - pages);
    } else {
        /*
         * Growing the area.
         * Check for overlaps with other address space areas.
         */
        if (!check_area_conflicts(as, address, pages * PAGE_SIZE,
            area)) {
            mutex_unlock(&area->lock);
            mutex_unlock(&as->lock);       
            interrupts_restore(ipl);
            return EADDRNOTAVAIL;
        }
    }
 
    area->pages = pages;
   
    mutex_unlock(&area->lock);
    mutex_unlock(&as->lock);
    interrupts_restore(ipl);
 
    return 0;
}
 
/** Destroy address space area.
 *
 * @param as Address space.
 * @param address Address withing the area to be deleted.
 *
 * @return Zero on success or a value from @ref errno.h on failure.
 */
int as_area_destroy(as_t *as, uintptr_t address)
{
    as_area_t *area;
    uintptr_t base;
    link_t *cur;
    ipl_t ipl;
 
    ipl = interrupts_disable();
    mutex_lock(&as->lock);
 
    area = find_area_and_lock(as, address);
    if (!area) {
        mutex_unlock(&as->lock);
        interrupts_restore(ipl);
        return ENOENT;
    }
 
    base = area->base;
 
    /*
     * Start TLB shootdown sequence.
     */
    tlb_shootdown_start(TLB_INVL_PAGES, as->asid, area->base, area->pages);
 
    /*
     * Visit only the pages mapped by used_space B+tree.
     */
    for (cur = area->used_space.leaf_head.next;
        cur != &area->used_space.leaf_head; cur = cur->next) {
        btree_node_t *node;
        int i;
       
        node = list_get_instance(cur, btree_node_t, leaf_link);
        for (i = 0; i < node->keys; i++) {
            uintptr_t b = node->key[i];
            count_t j;
            pte_t *pte;
           
            for (j = 0; j < (count_t) node->value[i]; j++) {
                page_table_lock(as, false);
                pte = page_mapping_find(as, b + j * PAGE_SIZE);
                ASSERT(pte && PTE_VALID(pte) &&
                    PTE_PRESENT(pte));
                if (area->backend &&
                    area->backend->frame_free) {
                    area->backend->frame_free(area, b +
                    j * PAGE_SIZE, PTE_GET_FRAME(pte));
                }
                page_mapping_remove(as, b + j * PAGE_SIZE);            
                page_table_unlock(as, false);
            }
        }
    }
 
    /*
     * Finish TLB shootdown sequence.
     */
    tlb_invalidate_pages(as->asid, area->base, area->pages);
    tlb_shootdown_finalize();
   
    /*
     * Invalidate potential software translation caches (e.g. TSB on
     * sparc64).
     */
    as_invalidate_translation_cache(as, area->base, area->pages);
   
    btree_destroy(&area->used_space);
 
    area->attributes |= AS_AREA_ATTR_PARTIAL;
   
    if (area->sh_info)
        sh_info_remove_reference(area->sh_info);
       
    mutex_unlock(&area->lock);
 
    /*
     * Remove the empty area from address space.
     */
    btree_remove(&as->as_area_btree, base, NULL);
   
    free(area);
   
    mutex_unlock(&as->lock);
    interrupts_restore(ipl);
    return 0;
}
 
/** Share address space area with another or the same address space.
 *
 * Address space area mapping is shared with a new address space area.
 * If the source address space area has not been shared so far,
 * a new sh_info is created. The new address space area simply gets the
 * sh_info of the source area. The process of duplicating the
 * mapping is done through the backend share function.
 *
 * @param src_as Pointer to source address space.
 * @param src_base Base address of the source address space area.
 * @param acc_size Expected size of the source area.
 * @param dst_as Pointer to destination address space.
 * @param dst_base Target base address.
 * @param dst_flags_mask Destination address space area flags mask.
 *
 * @return Zero on success or ENOENT if there is no such task or if there is no
 * such address space area, EPERM if there was a problem in accepting the area
 * or ENOMEM if there was a problem in allocating destination address space
 * area. ENOTSUP is returned if the address space area backend does not support
 * sharing or if the kernel detects an attempt to create an illegal address
 * alias.
 */
int as_area_share(as_t *src_as, uintptr_t src_base, size_t acc_size,
          as_t *dst_as, uintptr_t dst_base, int dst_flags_mask)
{
    ipl_t ipl;
    int src_flags;
    size_t src_size;
    as_area_t *src_area, *dst_area;
    share_info_t *sh_info;
    mem_backend_t *src_backend;
    mem_backend_data_t src_backend_data;
   
    ipl = interrupts_disable();
    mutex_lock(&src_as->lock);
    src_area = find_area_and_lock(src_as, src_base);
    if (!src_area) {
        /*
         * Could not find the source address space area.
         */
        mutex_unlock(&src_as->lock);
        interrupts_restore(ipl);
        return ENOENT;
    }
 
    if (!src_area->backend || !src_area->backend->share) {
        /*
         * There is no backend or the backend does not
         * know how to share the area.
         */
        mutex_unlock(&src_area->lock);
        mutex_unlock(&src_as->lock);
        interrupts_restore(ipl);
        return ENOTSUP;
    }
   
    src_size = src_area->pages * PAGE_SIZE;
    src_flags = src_area->flags;
    src_backend = src_area->backend;
    src_backend_data = src_area->backend_data;
 
    /* Share the cacheable flag from the original mapping */
    if (src_flags & AS_AREA_CACHEABLE)
        dst_flags_mask |= AS_AREA_CACHEABLE;
 
    if (src_size != acc_size ||
        (src_flags & dst_flags_mask) != dst_flags_mask) {
        mutex_unlock(&src_area->lock);
        mutex_unlock(&src_as->lock);
        interrupts_restore(ipl);
        return EPERM;
    }
 
#ifdef CONFIG_VIRT_IDX_DCACHE
    if (!(dst_flags_mask & AS_AREA_EXEC)) {
        if (PAGE_COLOR(src_area->base) != PAGE_COLOR(dst_base)) {
            /*
             * Refuse to create an illegal address alias.
             */
            mutex_unlock(&src_area->lock);
            mutex_unlock(&src_as->lock);
            interrupts_restore(ipl);
            return ENOTSUP;
        }
    }
#endif /* CONFIG_VIRT_IDX_DCACHE */
 
    /*
     * Now we are committed to sharing the area.
     * First, prepare the area for sharing.
     * Then it will be safe to unlock it.
     */
    sh_info = src_area->sh_info;
    if (!sh_info) {
        sh_info = (share_info_t *) malloc(sizeof(share_info_t), 0);
        mutex_initialize(&sh_info->lock);
        sh_info->refcount = 2;
        btree_create(&sh_info->pagemap);
        src_area->sh_info = sh_info;
    } else {
        mutex_lock(&sh_info->lock);
        sh_info->refcount++;
        mutex_unlock(&sh_info->lock);
    }
 
    src_area->backend->share(src_area);
 
    mutex_unlock(&src_area->lock);
    mutex_unlock(&src_as->lock);
 
    /*
     * Create copy of the source address space area.
     * The destination area is created with AS_AREA_ATTR_PARTIAL
     * attribute set which prevents race condition with
     * preliminary as_page_fault() calls.
     * The flags of the source area are masked against dst_flags_mask
     * to support sharing in less privileged mode.
     */
    dst_area = as_area_create(dst_as, dst_flags_mask, src_size, dst_base,
        AS_AREA_ATTR_PARTIAL, src_backend, &src_backend_data);
    if (!dst_area) {
        /*
         * Destination address space area could not be created.
         */
        sh_info_remove_reference(sh_info);
       
        interrupts_restore(ipl);
        return ENOMEM;
    }
 
    /*
     * Now the destination address space area has been
     * fully initialized. Clear the AS_AREA_ATTR_PARTIAL
     * attribute and set the sh_info.
     */
    mutex_lock(&dst_as->lock); 
    mutex_lock(&dst_area->lock);
    dst_area->attributes &= ~AS_AREA_ATTR_PARTIAL;
    dst_area->sh_info = sh_info;
    mutex_unlock(&dst_area->lock);
    mutex_unlock(&dst_as->lock);   
 
    interrupts_restore(ipl);
   
    return 0;
}
 
/** Check access mode for address space area.
 *
 * The address space area must be locked prior to this call.
 *
 * @param area Address space area.
 * @param access Access mode.
 *
 * @return False if access violates area's permissions, true otherwise.
 */
bool as_area_check_access(as_area_t *area, pf_access_t access)
{
    int flagmap[] = {
        [PF_ACCESS_READ] = AS_AREA_READ,
        [PF_ACCESS_WRITE] = AS_AREA_WRITE,
        [PF_ACCESS_EXEC] = AS_AREA_EXEC
    };
 
    if (!(area->flags & flagmap[access]))
        return false;
   
    return true;
}
 
/** Handle page fault within the current address space.
 *
 * This is the high-level page fault handler. It decides
 * whether the page fault can be resolved by any backend
 * and if so, it invokes the backend to resolve the page
 * fault.
 *
 * Interrupts are assumed disabled.
 *
 * @param page Faulting page.
 * @param access Access mode that caused the fault (i.e. read/write/exec).
 * @param istate Pointer to interrupted state.
 *
 * @return AS_PF_FAULT on page fault, AS_PF_OK on success or AS_PF_DEFER if the
 *     fault was caused by copy_to_uspace() or copy_from_uspace().
 */
int as_page_fault(uintptr_t page, pf_access_t access, istate_t *istate)
{
    pte_t *pte;
    as_area_t *area;
   
    if (!THREAD)
        return AS_PF_FAULT;
       
    ASSERT(AS);
 
    mutex_lock(&AS->lock);
    area = find_area_and_lock(AS, page);   
    if (!area) {
        /*
         * No area contained mapping for 'page'.
         * Signal page fault to low-level handler.
         */
        mutex_unlock(&AS->lock);
        goto page_fault;
    }
 
    if (area->attributes & AS_AREA_ATTR_PARTIAL) {
        /*
         * The address space area is not fully initialized.
         * Avoid possible race by returning error.
         */
        mutex_unlock(&area->lock);
        mutex_unlock(&AS->lock);
        goto page_fault;       
    }
 
    if (!area->backend || !area->backend->page_fault) {
        /*
         * The address space area is not backed by any backend
         * or the backend cannot handle page faults.
         */
        mutex_unlock(&area->lock);
        mutex_unlock(&AS->lock);
        goto page_fault;       
    }
 
    page_table_lock(AS, false);
   
    /*
     * To avoid race condition between two page faults
     * on the same address, we need to make sure
     * the mapping has not been already inserted.
     */
    if ((pte = page_mapping_find(AS, page))) {
        if (PTE_PRESENT(pte)) {
            if (((access == PF_ACCESS_READ) && PTE_READABLE(pte)) ||
                (access == PF_ACCESS_WRITE && PTE_WRITABLE(pte)) ||
                (access == PF_ACCESS_EXEC && PTE_EXECUTABLE(pte))) {
                page_table_unlock(AS, false);
                mutex_unlock(&area->lock);
                mutex_unlock(&AS->lock);
                return AS_PF_OK;
            }
        }
    }
   
    /*
     * Resort to the backend page fault handler.
     */
    if (area->backend->page_fault(area, page, access) != AS_PF_OK) {
        page_table_unlock(AS, false);
        mutex_unlock(&area->lock);
        mutex_unlock(&AS->lock);
        goto page_fault;
    }
   
    page_table_unlock(AS, false);
    mutex_unlock(&area->lock);
    mutex_unlock(&AS->lock);
    return AS_PF_OK;
 
page_fault:
    if (THREAD->in_copy_from_uspace) {
        THREAD->in_copy_from_uspace = false;
        istate_set_retaddr(istate,
            (uintptr_t) &memcpy_from_uspace_failover_address);
    } else if (THREAD->in_copy_to_uspace) {
        THREAD->in_copy_to_uspace = false;
        istate_set_retaddr(istate,
            (uintptr_t) &memcpy_to_uspace_failover_address);
    } else {
        return AS_PF_FAULT;
    }
 
    return AS_PF_DEFER;
}
 
/** Switch address spaces.
 *
 * Note that this function cannot sleep as it is essentially a part of
 * scheduling. Sleeping here would lead to deadlock on wakeup.
 *
 * @param old Old address space or NULL.
 * @param new New address space.
 */
void as_switch(as_t *old, as_t *new)
{
    ipl_t ipl;
    bool needs_asid = false;
   
    ipl = interrupts_disable();
    spinlock_lock(&inactive_as_with_asid_lock);
 
    /*
     * First, take care of the old address space.
     */
    if (old) {
        mutex_lock_active(&old->lock);
        ASSERT(old->cpu_refcount);
        if((--old->cpu_refcount == 0) && (old != AS_KERNEL)) {
            /*
             * The old address space is no longer active on
             * any processor. It can be appended to the
             * list of inactive address spaces with assigned
             * ASID.
             */
             ASSERT(old->asid != ASID_INVALID);
             list_append(&old->inactive_as_with_asid_link,
                 &inactive_as_with_asid_head);
        }
        mutex_unlock(&old->lock);
 
        /*
         * Perform architecture-specific tasks when the address space
         * is being removed from the CPU.
         */
        as_deinstall_arch(old);
    }
 
    /*
     * Second, prepare the new address space.
     */
    mutex_lock_active(&new->lock);
    if ((new->cpu_refcount++ == 0) && (new != AS_KERNEL)) {
        if (new->asid != ASID_INVALID) {
            list_remove(&new->inactive_as_with_asid_link);
        } else {
            /*
             * Defer call to asid_get() until new->lock is released.
             */
            needs_asid = true;
        }
    }
    SET_PTL0_ADDRESS(new->page_table);
    mutex_unlock(&new->lock);
 
    if (needs_asid) {
        /*
         * Allocation of new ASID was deferred
         * until now in order to avoid deadlock.
         */
        asid_t asid;
       
        asid = asid_get();
        mutex_lock_active(&new->lock);
        new->asid = asid;
        mutex_unlock(&new->lock);
    }
    spinlock_unlock(&inactive_as_with_asid_lock);
    interrupts_restore(ipl);
   
    /*
     * Perform architecture-specific steps.
     * (e.g. write ASID to hardware register etc.)
     */
    as_install_arch(new);
   
    AS = new;
}
 
/** Convert address space area flags to page flags.
 *
 * @param aflags Flags of some address space area.
 *
 * @return Flags to be passed to page_mapping_insert().
 */
int area_flags_to_page_flags(int aflags)
{
    int flags;
 
    flags = PAGE_USER | PAGE_PRESENT;
   
    if (aflags & AS_AREA_READ)
        flags |= PAGE_READ;
       
    if (aflags & AS_AREA_WRITE)
        flags |= PAGE_WRITE;
   
    if (aflags & AS_AREA_EXEC)
        flags |= PAGE_EXEC;
   
    if (aflags & AS_AREA_CACHEABLE)
        flags |= PAGE_CACHEABLE;
       
    return flags;
}
 
/** Compute flags for virtual address translation subsytem.
 *
 * The address space area must be locked.
 * Interrupts must be disabled.
 *
 * @param a Address space area.
 *
 * @return Flags to be used in page_mapping_insert().
 */
int as_area_get_flags(as_area_t *a)
{
    return area_flags_to_page_flags(a->flags);
}
 
/** Create page table.
 *
 * Depending on architecture, create either address space
 * private or global page table.
 *
 * @param flags Flags saying whether the page table is for kernel address space.
 *
 * @return First entry of the page table.
 */
pte_t *page_table_create(int flags)
{
        ASSERT(as_operations);
        ASSERT(as_operations->page_table_create);
 
        return as_operations->page_table_create(flags);
}
 
/** Destroy page table.
 *
 * Destroy page table in architecture specific way.
 *
 * @param page_table Physical address of PTL0.
 */
void page_table_destroy(pte_t *page_table)
{
        ASSERT(as_operations);
        ASSERT(as_operations->page_table_destroy);
 
        as_operations->page_table_destroy(page_table);
}
 
/** Lock page table.
 *
 * This function should be called before any page_mapping_insert(),
 * page_mapping_remove() and page_mapping_find().
 *
 * Locking order is such that address space areas must be locked
 * prior to this call. Address space can be locked prior to this
 * call in which case the lock argument is false.
 *
 * @param as Address space.
 * @param lock If false, do not attempt to lock as->lock.
 */
void page_table_lock(as_t *as, bool lock)
{
    ASSERT(as_operations);
    ASSERT(as_operations->page_table_lock);
 
    as_operations->page_table_lock(as, lock);
}
 
/** Unlock page table.
 *
 * @param as Address space.
 * @param unlock If false, do not attempt to unlock as->lock.
 */
void page_table_unlock(as_t *as, bool unlock)
{
    ASSERT(as_operations);
    ASSERT(as_operations->page_table_unlock);
 
    as_operations->page_table_unlock(as, unlock);
}
 
 
/** Find address space area and lock it.
 *
 * The address space must be locked and interrupts must be disabled.
 *
 * @param as Address space.
 * @param va Virtual address.
 *
 * @return Locked address space area containing va on success or NULL on
 *     failure.
 */
as_area_t *find_area_and_lock(as_t *as, uintptr_t va)
{
    as_area_t *a;
    btree_node_t *leaf, *lnode;
    int i;
   
    a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf);
    if (a) {
        /* va is the base address of an address space area */
        mutex_lock(&a->lock);
        return a;
    }
   
    /*
     * Search the leaf node and the righmost record of its left neighbour
     * to find out whether this is a miss or va belongs to an address
     * space area found there.
     */
   
    /* First, search the leaf node itself. */
    for (i = 0; i < leaf->keys; i++) {
        a = (as_area_t *) leaf->value[i];
        mutex_lock(&a->lock);
        if ((a->base <= va) && (va < a->base + a->pages * PAGE_SIZE)) {
            return a;
        }
        mutex_unlock(&a->lock);
    }
 
    /*
     * Second, locate the left neighbour and test its last record.
     * Because of its position in the B+tree, it must have base < va.
     */
    lnode = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf);
    if (lnode) {
        a = (as_area_t *) lnode->value[lnode->keys - 1];
        mutex_lock(&a->lock);
        if (va < a->base + a->pages * PAGE_SIZE) {
            return a;
        }
        mutex_unlock(&a->lock);
    }
 
    return NULL;
}
 
/** Check area conflicts with other areas.
 *
 * The address space must be locked and interrupts must be disabled.
 *
 * @param as Address space.
 * @param va Starting virtual address of the area being tested.
 * @param size Size of the area being tested.
 * @param avoid_area Do not touch this area.
 *
 * @return True if there is no conflict, false otherwise.
 */
bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
              as_area_t *avoid_area)
{
    as_area_t *a;
    btree_node_t *leaf, *node;
    int i;
   
    /*
     * We don't want any area to have conflicts with NULL page.
     */
    if (overlaps(va, size, NULL, PAGE_SIZE))
        return false;
   
    /*
     * The leaf node is found in O(log n), where n is proportional to
     * the number of address space areas belonging to as.
     * The check for conflicts is then attempted on the rightmost
     * record in the left neighbour, the leftmost record in the right
     * neighbour and all records in the leaf node itself.
     */
   
    if ((a = (as_area_t *) btree_search(&as->as_area_btree, va, &leaf))) {
        if (a != avoid_area)
            return false;
    }
   
    /* First, check the two border cases. */
    if ((node = btree_leaf_node_left_neighbour(&as->as_area_btree, leaf))) {
        a = (as_area_t *) node->value[node->keys - 1];
        mutex_lock(&a->lock);
        if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
            mutex_unlock(&a->lock);
            return false;
        }
        mutex_unlock(&a->lock);
    }
    node = btree_leaf_node_right_neighbour(&as->as_area_btree, leaf);
    if (node) {
        a = (as_area_t *) node->value[0];
        mutex_lock(&a->lock);
        if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
            mutex_unlock(&a->lock);
            return false;
        }
        mutex_unlock(&a->lock);
    }
   
    /* Second, check the leaf node. */
    for (i = 0; i < leaf->keys; i++) {
        a = (as_area_t *) leaf->value[i];
   
        if (a == avoid_area)
            continue;
   
        mutex_lock(&a->lock);
        if (overlaps(va, size, a->base, a->pages * PAGE_SIZE)) {
            mutex_unlock(&a->lock);
            return false;
        }
        mutex_unlock(&a->lock);
    }
 
    /*
     * So far, the area does not conflict with other areas.
     * Check if it doesn't conflict with kernel address space.
     */  
    if (!KERNEL_ADDRESS_SPACE_SHADOWED) {
        return !overlaps(va, size,
            KERNEL_ADDRESS_SPACE_START,
            KERNEL_ADDRESS_SPACE_END - KERNEL_ADDRESS_SPACE_START);
    }
 
    return true;
}
 
/** Return size of the address space area with given base.  */
size_t as_get_size(uintptr_t base)
{
    ipl_t ipl;
    as_area_t *src_area;
    size_t size;
 
    ipl = interrupts_disable();
    src_area = find_area_and_lock(AS, base);
    if (src_area){
        size = src_area->pages * PAGE_SIZE;
        mutex_unlock(&src_area->lock);
    } else {
        size = 0;
    }
    interrupts_restore(ipl);
    return size;
}
 
/** Mark portion of address space area as used.
 *
 * The address space area must be already locked.
 *
 * @param a Address space area.
 * @param page First page to be marked.
 * @param count Number of page to be marked.
 *
 * @return 0 on failure and 1 on success.
 */
int used_space_insert(as_area_t *a, uintptr_t page, count_t count)
{
    btree_node_t *leaf, *node;
    count_t pages;
    int i;
 
    ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
    ASSERT(count);
 
    pages = (count_t) btree_search(&a->used_space, page, &leaf);
    if (pages) {
        /*
         * We hit the beginning of some used space.
         */
        return 0;
    }
 
    if (!leaf->keys) {
        btree_insert(&a->used_space, page, (void *) count, leaf);
        return 1;
    }
 
    node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
    if (node) {
        uintptr_t left_pg = node->key[node->keys - 1];
        uintptr_t right_pg = leaf->key[0];
        count_t left_cnt = (count_t) node->value[node->keys - 1];
        count_t right_cnt = (count_t) leaf->value[0];
       
        /*
         * Examine the possibility that the interval fits
         * somewhere between the rightmost interval of
         * the left neigbour and the first interval of the leaf.
         */
         
        if (page >= right_pg) {
            /* Do nothing. */
        } else if (overlaps(page, count * PAGE_SIZE, left_pg,
            left_cnt * PAGE_SIZE)) {
            /* The interval intersects with the left interval. */
            return 0;
        } else if (overlaps(page, count * PAGE_SIZE, right_pg,
            right_cnt * PAGE_SIZE)) {
            /* The interval intersects with the right interval. */
            return 0;          
        } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
            (page + count * PAGE_SIZE == right_pg)) {
            /*
             * The interval can be added by merging the two already
             * present intervals.
             */
            node->value[node->keys - 1] += count + right_cnt;
            btree_remove(&a->used_space, right_pg, leaf);
            return 1;
        } else if (page == left_pg + left_cnt * PAGE_SIZE) {
            /*
             * The interval can be added by simply growing the left
             * interval.
             */
            node->value[node->keys - 1] += count;
            return 1;
        } else if (page + count * PAGE_SIZE == right_pg) {
            /*
             * The interval can be addded by simply moving base of
             * the right interval down and increasing its size
             * accordingly.
             */
            leaf->value[0] += count;
            leaf->key[0] = page;
            return 1;
        } else {
            /*
             * The interval is between both neigbouring intervals,
             * but cannot be merged with any of them.
             */
            btree_insert(&a->used_space, page, (void *) count,
                leaf);
            return 1;
        }
    } else if (page < leaf->key[0]) {
        uintptr_t right_pg = leaf->key[0];
        count_t right_cnt = (count_t) leaf->value[0];
   
        /*
         * Investigate the border case in which the left neighbour does
         * not exist but the interval fits from the left.
         */
         
        if (overlaps(page, count * PAGE_SIZE, right_pg,
            right_cnt * PAGE_SIZE)) {
            /* The interval intersects with the right interval. */
            return 0;
        } else if (page + count * PAGE_SIZE == right_pg) {
            /*
             * The interval can be added by moving the base of the
             * right interval down and increasing its size
             * accordingly.
             */
            leaf->key[0] = page;
            leaf->value[0] += count;
            return 1;
        } else {
            /*
             * The interval doesn't adjoin with the right interval.
             * It must be added individually.
             */
            btree_insert(&a->used_space, page, (void *) count,
                leaf);
            return 1;
        }
    }
 
    node = btree_leaf_node_right_neighbour(&a->used_space, leaf);
    if (node) {
        uintptr_t left_pg = leaf->key[leaf->keys - 1];
        uintptr_t right_pg = node->key[0];
        count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
        count_t right_cnt = (count_t) node->value[0];
       
        /*
         * Examine the possibility that the interval fits
         * somewhere between the leftmost interval of
         * the right neigbour and the last interval of the leaf.
         */
 
        if (page < left_pg) {
            /* Do nothing. */
        } else if (overlaps(page, count * PAGE_SIZE, left_pg,
            left_cnt * PAGE_SIZE)) {
            /* The interval intersects with the left interval. */
            return 0;
        } else if (overlaps(page, count * PAGE_SIZE, right_pg,
            right_cnt * PAGE_SIZE)) {
            /* The interval intersects with the right interval. */
            return 0;          
        } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
            (page + count * PAGE_SIZE == right_pg)) {
            /*
             * The interval can be added by merging the two already
             * present intervals.
             * */
            leaf->value[leaf->keys - 1] += count + right_cnt;
            btree_remove(&a->used_space, right_pg, node);
            return 1;
        } else if (page == left_pg + left_cnt * PAGE_SIZE) {
            /*
             * The interval can be added by simply growing the left
             * interval.
             * */
            leaf->value[leaf->keys - 1] +=  count;
            return 1;
        } else if (page + count * PAGE_SIZE == right_pg) {
            /*
             * The interval can be addded by simply moving base of
             * the right interval down and increasing its size
             * accordingly.
             */
            node->value[0] += count;
            node->key[0] = page;
            return 1;
        } else {
            /*
             * The interval is between both neigbouring intervals,
             * but cannot be merged with any of them.
             */
            btree_insert(&a->used_space, page, (void *) count,
                leaf);
            return 1;
        }
    } else if (page >= leaf->key[leaf->keys - 1]) {
        uintptr_t left_pg = leaf->key[leaf->keys - 1];
        count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
   
        /*
         * Investigate the border case in which the right neighbour
         * does not exist but the interval fits from the right.
         */
         
        if (overlaps(page, count * PAGE_SIZE, left_pg,
            left_cnt * PAGE_SIZE)) {
            /* The interval intersects with the left interval. */
            return 0;
        } else if (left_pg + left_cnt * PAGE_SIZE == page) {
            /*
             * The interval can be added by growing the left
             * interval.
             */
            leaf->value[leaf->keys - 1] += count;
            return 1;
        } else {
            /*
             * The interval doesn't adjoin with the left interval.
             * It must be added individually.
             */
            btree_insert(&a->used_space, page, (void *) count,
                leaf);
            return 1;
        }
    }
   
    /*
     * Note that if the algorithm made it thus far, the interval can fit
     * only between two other intervals of the leaf. The two border cases
     * were already resolved.
     */
    for (i = 1; i < leaf->keys; i++) {
        if (page < leaf->key[i]) {
            uintptr_t left_pg = leaf->key[i - 1];
            uintptr_t right_pg = leaf->key[i];
            count_t left_cnt = (count_t) leaf->value[i - 1];
            count_t right_cnt = (count_t) leaf->value[i];
 
            /*
             * The interval fits between left_pg and right_pg.
             */
 
            if (overlaps(page, count * PAGE_SIZE, left_pg,
                left_cnt * PAGE_SIZE)) {
                /*
                 * The interval intersects with the left
                 * interval.
                 */
                return 0;
            } else if (overlaps(page, count * PAGE_SIZE, right_pg,
                right_cnt * PAGE_SIZE)) {
                /*
                 * The interval intersects with the right
                 * interval.
                 */
                return 0;          
            } else if ((page == left_pg + left_cnt * PAGE_SIZE) &&
                (page + count * PAGE_SIZE == right_pg)) {
                /*
                 * The interval can be added by merging the two
                 * already present intervals.
                 */
                leaf->value[i - 1] += count + right_cnt;
                btree_remove(&a->used_space, right_pg, leaf);
                return 1;
            } else if (page == left_pg + left_cnt * PAGE_SIZE) {
                /*
                 * The interval can be added by simply growing
                 * the left interval.
                 */
                leaf->value[i - 1] += count;
                return 1;
            } else if (page + count * PAGE_SIZE == right_pg) {
                /*
                     * The interval can be addded by simply moving
                 * base of the right interval down and
                 * increasing its size accordingly.
                 */
                leaf->value[i] += count;
                leaf->key[i] = page;
                return 1;
            } else {
                /*
                 * The interval is between both neigbouring
                 * intervals, but cannot be merged with any of
                 * them.
                 */
                btree_insert(&a->used_space, page,
                    (void *) count, leaf);
                return 1;
            }
        }
    }
 
    panic("Inconsistency detected while adding %d pages of used space at "
        "%p.\n", count, page);
}
 
/** Mark portion of address space area as unused.
 *
 * The address space area must be already locked.
 *
 * @param a Address space area.
 * @param page First page to be marked.
 * @param count Number of page to be marked.
 *
 * @return 0 on failure and 1 on success.
 */
int used_space_remove(as_area_t *a, uintptr_t page, count_t count)
{
    btree_node_t *leaf, *node;
    count_t pages;
    int i;
 
    ASSERT(page == ALIGN_DOWN(page, PAGE_SIZE));
    ASSERT(count);
 
    pages = (count_t) btree_search(&a->used_space, page, &leaf);
    if (pages) {
        /*
         * We are lucky, page is the beginning of some interval.
         */
        if (count > pages) {
            return 0;
        } else if (count == pages) {
            btree_remove(&a->used_space, page, leaf);
            return 1;
        } else {
            /*
             * Find the respective interval.
             * Decrease its size and relocate its start address.
             */
            for (i = 0; i < leaf->keys; i++) {
                if (leaf->key[i] == page) {
                    leaf->key[i] += count * PAGE_SIZE;
                    leaf->value[i] -= count;
                    return 1;
                }
            }
            goto error;
        }
    }
 
    node = btree_leaf_node_left_neighbour(&a->used_space, leaf);
    if (node && page < leaf->key[0]) {
        uintptr_t left_pg = node->key[node->keys - 1];
        count_t left_cnt = (count_t) node->value[node->keys - 1];
 
        if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
            count * PAGE_SIZE)) {
            if (page + count * PAGE_SIZE ==
                left_pg + left_cnt * PAGE_SIZE) {
                /*
                 * The interval is contained in the rightmost
                 * interval of the left neighbour and can be
                 * removed by updating the size of the bigger
                 * interval.
                 */
                node->value[node->keys - 1] -= count;
                return 1;
            } else if (page + count * PAGE_SIZE <
                left_pg + left_cnt*PAGE_SIZE) {
                count_t new_cnt;
               
                /*
                 * The interval is contained in the rightmost
                 * interval of the left neighbour but its
                 * removal requires both updating the size of
                 * the original interval and also inserting a
                 * new interval.
                 */
                new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
                    (page + count*PAGE_SIZE)) >> PAGE_WIDTH;
                node->value[node->keys - 1] -= count + new_cnt;
                btree_insert(&a->used_space, page +
                    count * PAGE_SIZE, (void *) new_cnt, leaf);
                return 1;
            }
        }
        return 0;
    } else if (page < leaf->key[0]) {
        return 0;
    }
   
    if (page > leaf->key[leaf->keys - 1]) {
        uintptr_t left_pg = leaf->key[leaf->keys - 1];
        count_t left_cnt = (count_t) leaf->value[leaf->keys - 1];
 
        if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
            count * PAGE_SIZE)) {
            if (page + count * PAGE_SIZE ==
                left_pg + left_cnt * PAGE_SIZE) {
                /*
                 * The interval is contained in the rightmost
                 * interval of the leaf and can be removed by
                 * updating the size of the bigger interval.
                 */
                leaf->value[leaf->keys - 1] -= count;
                return 1;
            } else if (page + count * PAGE_SIZE < left_pg +
                left_cnt * PAGE_SIZE) {
                count_t new_cnt;
               
                /*
                 * The interval is contained in the rightmost
                 * interval of the leaf but its removal
                 * requires both updating the size of the
                 * original interval and also inserting a new
                 * interval.
                 */
                new_cnt = ((left_pg + left_cnt * PAGE_SIZE) -
                    (page + count * PAGE_SIZE)) >> PAGE_WIDTH;
                leaf->value[leaf->keys - 1] -= count + new_cnt;
                btree_insert(&a->used_space, page +
                    count * PAGE_SIZE, (void *) new_cnt, leaf);
                return 1;
            }
        }
        return 0;
    }  
   
    /*
     * The border cases have been already resolved.
     * Now the interval can be only between intervals of the leaf.
     */
    for (i = 1; i < leaf->keys - 1; i++) {
        if (page < leaf->key[i]) {
            uintptr_t left_pg = leaf->key[i - 1];
            count_t left_cnt = (count_t) leaf->value[i - 1];
 
            /*
             * Now the interval is between intervals corresponding
             * to (i - 1) and i.
             */
            if (overlaps(left_pg, left_cnt * PAGE_SIZE, page,
                count * PAGE_SIZE)) {
                if (page + count * PAGE_SIZE ==
                    left_pg + left_cnt*PAGE_SIZE) {
                    /*
                     * The interval is contained in the
                     * interval (i - 1) of the leaf and can
                     * be removed by updating the size of
                     * the bigger interval.
                     */
                    leaf->value[i - 1] -= count;
                    return 1;
                } else if (page + count * PAGE_SIZE <
                    left_pg + left_cnt * PAGE_SIZE) {
                    count_t new_cnt;
               
                    /*
                     * The interval is contained in the
                     * interval (i - 1) of the leaf but its
                     * removal requires both updating the
                     * size of the original interval and
                     * also inserting a new interval.
                     */
                    new_cnt = ((left_pg +
                        left_cnt * PAGE_SIZE) -
                        (page + count * PAGE_SIZE)) >>
                        PAGE_WIDTH;
                    leaf->value[i - 1] -= count + new_cnt;
                    btree_insert(&a->used_space, page +
                        count * PAGE_SIZE, (void *) new_cnt,
                        leaf);
                    return 1;
                }
            }
            return 0;
        }
    }
 
error:
    panic("Inconsistency detected while removing %d pages of used space "
        "from %p.\n", count, page);
}
 
/** Remove reference to address space area share info.
 *
 * If the reference count drops to 0, the sh_info is deallocated.
 *
 * @param sh_info Pointer to address space area share info.
 */
void sh_info_remove_reference(share_info_t *sh_info)
{
    bool dealloc = false;
 
    mutex_lock(&sh_info->lock);
    ASSERT(sh_info->refcount);
    if (--sh_info->refcount == 0) {
        dealloc = true;
        link_t *cur;
       
        /*
         * Now walk carefully the pagemap B+tree and free/remove
         * reference from all frames found there.
         */
        for (cur = sh_info->pagemap.leaf_head.next;
            cur != &sh_info->pagemap.leaf_head; cur = cur->next) {
            btree_node_t *node;
            int i;
           
            node = list_get_instance(cur, btree_node_t, leaf_link);
            for (i = 0; i < node->keys; i++)
                frame_free((uintptr_t) node->value[i]);
        }
       
    }
    mutex_unlock(&sh_info->lock);
   
    if (dealloc) {
        btree_destroy(&sh_info->pagemap);
        free(sh_info);
    }
}
 
/*
 * Address space related syscalls.
 */
 
/** Wrapper for as_area_create(). */
unative_t sys_as_area_create(uintptr_t address, size_t size, int flags)
{
    if (as_area_create(AS, flags | AS_AREA_CACHEABLE, size, address,
        AS_AREA_ATTR_NONE, &anon_backend, NULL))
        return (unative_t) address;
    else
        return (unative_t) -1;
}
 
/** Wrapper for as_area_resize(). */
unative_t sys_as_area_resize(uintptr_t address, size_t size, int flags)
{
    return (unative_t) as_area_resize(AS, address, size, 0);
}
 
/** Wrapper for as_area_destroy(). */
unative_t sys_as_area_destroy(uintptr_t address)
{
    return (unative_t) as_area_destroy(AS, address);
}
 
/** Print out information about address space.
 *
 * @param as Address space.
 */
void as_print(as_t *as)
{
    ipl_t ipl;
   
    ipl = interrupts_disable();
    mutex_lock(&as->lock);
   
    /* print out info about address space areas */
    link_t *cur;
    for (cur = as->as_area_btree.leaf_head.next;
        cur != &as->as_area_btree.leaf_head; cur = cur->next) {
        btree_node_t *node;
       
        node = list_get_instance(cur, btree_node_t, leaf_link);
       
        int i;
        for (i = 0; i < node->keys; i++) {
            as_area_t *area = node->value[i];
       
            mutex_lock(&area->lock);
            printf("as_area: %p, base=%p, pages=%d (%p - %p)\n",
                area, area->base, area->pages, area->base,
                area->base + area->pages*PAGE_SIZE);
            mutex_unlock(&area->lock);
        }
    }
   
    mutex_unlock(&as->lock);
    interrupts_restore(ipl);
}
 
/** @}
 */
 

Rev 2087	Rev 2089
1	/*	1	/*
2	* Copyright (c) 2001-2006 Jakub Jermar	2	* Copyright (c) 2001-2006 Jakub Jermar
3	* All rights reserved.	3	* All rights reserved.
4	*	4	*
5	* Redistribution and use in source and binary forms, with or without	5	* Redistribution and use in source and binary forms, with or without
6	* modification, are permitted provided that the following conditions	6	* modification, are permitted provided that the following conditions
7	* are met:	7	* are met:
8	*	8	*
9	* - Redistributions of source code must retain the above copyright	9	* - Redistributions of source code must retain the above copyright
10	* notice, this list of conditions and the following disclaimer.	10	* notice, this list of conditions and the following disclaimer.
11	* - Redistributions in binary form must reproduce the above copyright	11	* - Redistributions in binary form must reproduce the above copyright
12	* notice, this list of conditions and the following disclaimer in the	12	* notice, this list of conditions and the following disclaimer in the
13	* documentation and/or other materials provided with the distribution.	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	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.	15	* derived from this software without specific prior written permission.
16	*	16	*
17	* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR	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	18	* IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
19	* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.	19	* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
20	* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,	20	* IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
21	* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT	21	* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
22	* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,	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	23	* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24	* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT	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	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.	26	* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27	*/	27	*/
28		28
29	/** @addtogroup genericmm	29	/** @addtogroup genericmm
30	* @{	30	* @{
31	*/	31	*/
32		32
33	/**	33	/**
34	* @file	34	* @file
35	* @brief Address space related functions.	35	* @brief Address space related functions.
36	*	36	*
37	* This file contains address space manipulation functions.	37	* This file contains address space manipulation functions.
38	* Roughly speaking, this is a higher-level client of	38	* Roughly speaking, this is a higher-level client of
39	* Virtual Address Translation (VAT) subsystem.	39	* Virtual Address Translation (VAT) subsystem.
40	*	40	*
41	* Functionality provided by this file allows one to	41	* Functionality provided by this file allows one to
42	* create address spaces and create, resize and share	42	* create address spaces and create, resize and share
43	* address space areas.	43	* address space areas.
44	*	44	*
45	* @see page.c	45	* @see page.c
46	*	46	*
47	*/	47	*/
48		48
49	#include <mm/as.h>	49	#include <mm/as.h>
50	#include <arch/mm/as.h>	50	#include <arch/mm/as.h>
51	#include <mm/page.h>	51	#include <mm/page.h>
52	#include <mm/frame.h>	52	#include <mm/frame.h>
53	#include <mm/slab.h>	53	#include <mm/slab.h>
54	#include <mm/tlb.h>	54	#include <mm/tlb.h>
55	#include <arch/mm/page.h>	55	#include <arch/mm/page.h>
56	#include <genarch/mm/page_pt.h>	56	#include <genarch/mm/page_pt.h>
57	#include <genarch/mm/page_ht.h>	57	#include <genarch/mm/page_ht.h>
58	#include <mm/asid.h>	58	#include <mm/asid.h>
59	#include <arch/mm/asid.h>	59	#include <arch/mm/asid.h>
60	#include <synch/spinlock.h>	60	#include <synch/spinlock.h>
61	#include <synch/mutex.h>	61	#include <synch/mutex.h>
62	#include <adt/list.h>	62	#include <adt/list.h>
63	#include <adt/btree.h>	63	#include <adt/btree.h>
64	#include <proc/task.h>	64	#include <proc/task.h>
65	#include <proc/thread.h>	65	#include <proc/thread.h>
66	#include <arch/asm.h>	66	#include <arch/asm.h>
67	#include <panic.h>	67	#include <panic.h>
68	#include <debug.h>	68	#include <debug.h>
69	#include <print.h>	69	#include <print.h>
70	#include <memstr.h>	70	#include <memstr.h>
71	#include <macros.h>	71	#include <macros.h>
72	#include <arch.h>	72	#include <arch.h>
73	#include <errno.h>	73	#include <errno.h>
74	#include <config.h>	74	#include <config.h>
75	#include <align.h>	75	#include <align.h>
76	#include <arch/types.h>	76	#include <arch/types.h>
77	#include <typedefs.h>	-
78	#include <syscall/copy.h>	77	#include <syscall/copy.h>
79	#include <arch/interrupt.h>	78	#include <arch/interrupt.h>
80		79
81	#ifdef CONFIG_VIRT_IDX_DCACHE	80	#ifdef CONFIG_VIRT_IDX_DCACHE
82	#include <arch/mm/cache.h>	81	#include <arch/mm/cache.h>
83	#endif /* CONFIG_VIRT_IDX_DCACHE */	82	#endif /* CONFIG_VIRT_IDX_DCACHE */
84		83
85	/**	84	/**
86	* Each architecture decides what functions will be used to carry out	85	* Each architecture decides what functions will be used to carry out
87	* address space operations such as creating or locking page tables.	86	* address space operations such as creating or locking page tables.
88	*/	87	*/
89	as_operations_t *as_operations = NULL;	88	as_operations_t *as_operations = NULL;
90		89
91	/**	90	/**
92	* Slab for as_t objects.	91	* Slab for as_t objects.
93	*/	92	*/
94	static slab_cache_t *as_slab;	93	static slab_cache_t *as_slab;
95		94
96	/**	95	/**
97	* This lock protects inactive_as_with_asid_head list. It must be acquired	96	* This lock protects inactive_as_with_asid_head list. It must be acquired
98	* before as_t mutex.	97	* before as_t mutex.
99	*/	98	*/
100	SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);	99	SPINLOCK_INITIALIZE(inactive_as_with_asid_lock);
101		100
102	/**	101	/**
103	* This list contains address spaces that are not active on any	102	* This list contains address spaces that are not active on any
104	* processor and that have valid ASID.	103	* processor and that have valid ASID.
105	*/	104	*/
106	LIST_INITIALIZE(inactive_as_with_asid_head);	105	LIST_INITIALIZE(inactive_as_with_asid_head);
107		106
108	/** Kernel address space. */	107	/** Kernel address space. */
109	as_t *AS_KERNEL = NULL;	108	as_t *AS_KERNEL = NULL;
110		109
111	static int area_flags_to_page_flags(int aflags);	110	static int area_flags_to_page_flags(int aflags);
112	static as_area_t find_area_and_lock(as_t as, uintptr_t va);	111	static as_area_t find_area_and_lock(as_t as, uintptr_t va);
113	static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,	112	static bool check_area_conflicts(as_t *as, uintptr_t va, size_t size,
114	as_area_t *avoid_area);	113	as_area_t *avoid_area);
115	static void sh_info_remove_reference(share_info_t *sh_info);	114	static void sh_info_remove_reference(share_info_t *sh_info);
116		115
117	static int as_constructor(void *obj, int flags)	116	static int as_constructor(void *obj, int flags)
118	{	117	{
119	as_t as = (as_t ) obj;	118	as_t as = (as_t ) obj;
120	int rc;	119	int rc;
121		120
122	link_initialize(&as->inactive_as_with_asid_link);	121	link_initialize(&as->inactive_as_with_asid_link);
123	mutex_initialize(&as->lock);	122	mutex_initialize(&as->lock);
124		123
125	rc = as_constructor_arch(as, flags);	124	rc = as_constructor_arch(as, flags);
126		125
127	return rc;	126	return rc;
128	}	127	}
129		128
130	static int as_destructor(void *obj)	129	static int as_destructor(void *obj)
131	{	130	{
132	as_t as = (as_t ) obj;	131	as_t as = (as_t ) obj;
133		132
134	return as_destructor_arch(as);	133	return as_destructor_arch(as);
135	}	134	}
136		135
137	/** Initialize address space subsystem. */	136	/** Initialize address space subsystem. */
138	void as_init(void)	137	void as_init(void)
139	{	138	{
140	as_arch_init();	139	as_arch_init();
141		140
142	as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,	141	as_slab = slab_cache_create("as_slab", sizeof(as_t), 0,
143	as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);	142	as_constructor, as_destructor, SLAB_CACHE_MAGDEFERRED);
144		143
145	AS_KERNEL = as_create(FLAG_AS_KERNEL);	144	AS_KERNEL = as_create(FLAG_AS_KERNEL);
146	if (!AS_KERNEL)	145	if (!AS_KERNEL)
147	panic("can't create kernel address space\n");	146	panic("can't create kernel address space\n");
148		147
149	}	148	}
150		149
151	/** Create address space.	150	/** Create address space.
152	*	151	*
153	* @param flags Flags that influence way in wich the address space is created.	152	* @param flags Flags that influence way in wich the address space is created.
154	*/	153	*/
155	as_t *as_create(int flags)	154	as_t *as_create(int flags)
156	{	155	{
157	as_t *as;	156	as_t *as;
158		157
159	as = (as_t *) slab_alloc(as_slab, 0);	158	as = (as_t *) slab_alloc(as_slab, 0);
160	(void) as_create_arch(as, 0);	159	(void) as_create_arch(as, 0);
161		160
162	btree_create(&as->as_area_btree);	161	btree_create(&as->as_area_btree);
163		162
164	if (flags & FLAG_AS_KERNEL)	163	if (flags & FLAG_AS_KERNEL)
165	as->asid = ASID_KERNEL;	164	as->asid = ASID_KERNEL;
166	else	165	else
167	as->asid = ASID_INVALID;	166	as->asid = ASID_INVALID;
168		167
169	as->refcount = 0;	168	as->refcount = 0;
170	as->cpu_refcount = 0;	169	as->cpu_refcount = 0;
-		170	#ifdef AS_PAGE_TABLE
171	as->page_table = page_table_create(flags);	171	as->page_table = page_table_create(flags);
-		172	#else
-		173	page_table_create(flags);
-		174	#endif
172		175
173	return as;	176	return as;
174	}	177	}
175		178
176	/** Destroy adress space.	179	/** Destroy adress space.
177	*	180	*
178	* When there are no tasks referencing this address space (i.e. its refcount is	181	* When there are no tasks referencing this address space (i.e. its refcount is
179	* zero), the address space can be destroyed.	182	* zero), the address space can be destroyed.
180	*/	183	*/
181	void as_destroy(as_t *as)	184	void as_destroy(as_t *as)
182	{	185	{
183	ipl_t ipl;	186	ipl_t ipl;
184	bool cond;	187	bool cond;
185		188
186	ASSERT(as->refcount == 0);	189	ASSERT(as->refcount == 0);
187		190
188	/*	191	/*
189	* Since there is no reference to this area,	192	* Since there is no reference to this area,
190	* it is safe not to lock its mutex.	193	* it is safe not to lock its mutex.
191	*/	194	*/
192	ipl = interrupts_disable();	195	ipl = interrupts_disable();
193	spinlock_lock(&inactive_as_with_asid_lock);	196	spinlock_lock(&inactive_as_with_asid_lock);
194	if (as->asid != ASID_INVALID && as != AS_KERNEL) {	197	if (as->asid != ASID_INVALID && as != AS_KERNEL) {
195	if (as != AS && as->cpu_refcount == 0)	198	if (as != AS && as->cpu_refcount == 0)
196	list_remove(&as->inactive_as_with_asid_link);	199	list_remove(&as->inactive_as_with_asid_link);
197	asid_put(as->asid);	200	asid_put(as->asid);
198	}	201	}
199	spinlock_unlock(&inactive_as_with_asid_lock);	202	spinlock_unlock(&inactive_as_with_asid_lock);
200		203
201	/*	204	/*
202	* Destroy address space areas of the address space.	205	* Destroy address space areas of the address space.
203	* The B+tree must be walked carefully because it is	206	* The B+tree must be walked carefully because it is
204	* also being destroyed.	207	* also being destroyed.
205	*/	208	*/
206	for (cond = true; cond; ) {	209	for (cond = true; cond; ) {
207	btree_node_t *node;	210	btree_node_t *node;
208		211
209	ASSERT(!list_empty(&as->as_area_btree.leaf_head));	212	ASSERT(!list_empty(&as->as_area_btree.leaf_head));
210	node = list_get_instance(as->as_area_btree.leaf_head.next,	213	node = list_get_instance(as->as_area_btree.leaf_head.next,
211	btree_node_t, leaf_link);	214	btree_node_t, leaf_link);
212		215
213	if ((cond = node->keys)) {	216	if ((cond = node->keys)) {
214	as_area_destroy(as, node->key[0]);	217	as_area_destroy(as, node->key[0]);
215	}	218	}
216	}	219	}
217		220
218	btree_destroy(&as->as_area_btree);	221	btree_destroy(&as->as_area_btree);
-		222	#ifdef AS_PAGE_TABLE
219	page_table_destroy(as->page_table);	223	page_table_destroy(as->page_table);
-		224	#else
-		225	page_table_destroy(NULL);
-		226	#endif
220		227
221	interrupts_restore(ipl);	228	interrupts_restore(ipl);
222		229
223	slab_free(as_slab, as);	230	slab_free(as_slab, as);
224	}	231	}
225		232
226	/** Create address space area of common attributes.	233	/** Create address space area of common attributes.
227	*	234	*
228	* The created address space area is added to the target address space.	235	* The created address space area is added to the target address space.
229	*	236	*
230	* @param as Target address space.	237	* @param as Target address space.
231	* @param flags Flags of the area memory.	238	* @param flags Flags of the area memory.
232	* @param size Size of area.	239	* @param size Size of area.
233	* @param base Base address of area.	240	* @param base Base address of area.
234	* @param attrs Attributes of the area.	241	* @param attrs Attributes of the area.
235	* @param backend Address space area backend. NULL if no backend is used.	242	* @param backend Address space area backend. NULL if no backend is used.
236	* @param backend_data NULL or a pointer to an array holding two void *.	243	* @param backend_data NULL or a pointer to an array holding two void *.
237	*	244	*
238	* @return Address space area on success or NULL on failure.	245	* @return Address space area on success or NULL on failure.
239	*/	246	*/
240	as_area_t *	247	as_area_t *
241	as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,	248	as_area_create(as_t *as, int flags, size_t size, uintptr_t base, int attrs,
242	mem_backend_t backend, mem_backend_data_t backend_data)	249	mem_backend_t backend, mem_backend_data_t backend_data)
243	{	250	{
244	ipl_t ipl;	251	ipl_t ipl;
245	as_area_t *a;	252	as_area_t *a;
246		253
247	if (base % PAGE_SIZE)	254	if (base % PAGE_SIZE)
248	return NULL;	255	return NULL;
249		256
250	if (!size)	257	if (!size)
251	return NULL;	258	return NULL;
252		259
253	/* Writeable executable areas are not supported. */	260	/* Writeable executable areas are not supported. */
254	if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))	261	if ((flags & AS_AREA_EXEC) && (flags & AS_AREA_WRITE))
255	return NULL;	262	return NULL;
256		263
257	ipl = interrupts_disable();	264	ipl = interrupts_disable();
258	mutex_lock(&as->lock);	265	mutex_lock(&as->lock);
259		266
260	if (!check_area_conflicts(as, base, size, NULL)) {	267	if (!check_area_conflicts(as, base, size, NULL)) {
261	mutex_unlock(&as->lock);	268	mutex_unlock(&as->lock);
262	interrupts_restore(ipl);	269	interrupts_restore(ipl);
263	return NULL;	270	return NULL;
264	}	271	}
265		272
266	a = (as_area_t *) malloc(sizeof(as_area_t), 0);	273	a = (as_area_t *) malloc(sizeof(as_area_t), 0);
267		274
268	mutex_initialize(&a->lock);	275	mutex_initialize(&a->lock);
269		276
270	a->as = as;	277	a->as = as;
271	a->flags = flags;	278	a->flags = flags;
272	a->attributes = attrs;	279	a->attributes = attrs;
273	a->pages = SIZE2FRAMES(size);	280	a->pages = SIZE2FRAMES(size);
274	a->base = base;	281	a->base = base;
275	a->sh_info = NULL;	282	a->sh_info = NULL;
276	a->backend = backend;	283	a->backend = backend;
277	if (backend_data)	284	if (backend_data)
278	a->backend_data = *backend_data;	285	a->backend_data = *backend_data;
279	else	286	else
280	memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data),	287	memsetb((uintptr_t) &a->backend_data, sizeof(a->backend_data),
281	0);	288	0);
282		289
283	btree_create(&a->used_space);	290	btree_create(&a->used_space);
284		291
285	btree_insert(&as->as_area_btree, base, (void *) a, NULL);	292	btree_insert(&as->as_area_btree, base, (void *) a, NULL);
286		293
287	mutex_unlock(&as->lock);	294	mutex_unlock(&as->lock);
288	interrupts_restore(ipl);	295	interrupts_restore(ipl);
289		296
290	return a;	297	return a;
291	}	298	}
292		299
293	/** Find address space area and change it.	300	/** Find address space area and change it.
294	*	301	*
295	* @param as Address space.	302	* @param as Address space.
296	* @param address Virtual address belonging to the area to be changed. Must be	303	* @param address Virtual address belonging to the area to be changed. Must be
297	* page-aligned.	304	* page-aligned.
298	* @param size New size of the virtual memory block starting at address.	305	* @param size New size of the virtual memory block starting at address.
299	* @param flags Flags influencing the remap operation. Currently unused.	306	* @param flags Flags influencing the remap operation. Currently unused.
300	*	307	*
301	* @return Zero on success or a value from @ref errno.h otherwise.	308	* @return Zero on success or a value from @ref errno.h otherwise.
302	*/	309	*/
303	int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)	310	int as_area_resize(as_t *as, uintptr_t address, size_t size, int flags)
304	{	311	{
305	as_area_t *area;	312	as_area_t *area;
306	ipl_t ipl;	313	ipl_t ipl;
307	size_t pages;	314	size_t pages;
308		315
309	ipl = interrupts_disable();	316	ipl = interrupts_disable();
310	mutex_lock(&as->lock);	317	mutex_lock(&as->lock);
311		318
312	/*	319	/*
313	* Locate the area.	320	* Locate the area.
314	*/	321	*/
315	area = find_area_and_lock(as, address);	322	area = find_area_and_lock(as, address);
316	if (!area) {	323	if (!area) {
317	mutex_unlock(&as->lock);	324	mutex_unlock(&as->lock);
318	interrupts_restore(ipl);	325	interrupts_restore(ipl);
319	return ENOENT;	326	return ENOENT;
320	}	327	}
321		328
322	if (area->backend == &phys_backend) {	329	if (area->backend == &phys_backend) {
323	/*	330	/*
324	* Remapping of address space areas associated	331	* Remapping of address space areas associated
325	* with memory mapped devices is not supported.	332	* with memory mapped devices is not supported.
326	*/	333	*/
327	mutex_unlock(&area->lock);	334	mutex_unlock(&area->lock);
328	mutex_unlock(&as->lock);	335	mutex_unlock(&as->lock);
329	interrupts_restore(ipl);	336	interrupts_restore(ipl);
330	return ENOTSUP;	337	return ENOTSUP;
331	}	338	}
332	if (area->sh_info) {	339	if (area->sh_info) {
333	/*	340	/*
334	* Remapping of shared address space areas	341	* Remapping of shared address space areas
335	* is not supported.	342	* is not supported.
336	*/	343	*/
337	mutex_unlock(&area->lock);	344	mutex_unlock(&area->lock);
338	mutex_unlock(&as->lock);	345	mutex_unlock(&as->lock);
339	interrupts_restore(ipl);	346	interrupts_restore(ipl);
340	return ENOTSUP;	347	return ENOTSUP;
341	}	348	}
342		349
343	pages = SIZE2FRAMES((address - area->base) + size);	350	pages = SIZE2FRAMES((address - area->base) + size);
344	if (!pages) {	351	if (!pages) {
345	/*	352	/*
346	* Zero size address space areas are not allowed.	353	* Zero size address space areas are not allowed.
347	*/	354	*/
348	mutex_unlock(&area->lock);	355	mutex_unlock(&area->lock);
349	mutex_unlock(&as->lock);	356	mutex_unlock(&as->lock);
350	interrupts_restore(ipl);	357	interrupts_restore(ipl);
351	return EPERM;	358	return EPERM;
352	}	359	}
353		360
354	if (pages < area->pages) {	361	if (pages < area->pages) {
355	bool cond;	362	bool cond;
356	uintptr_t start_free = area->base + pages*PAGE_SIZE;	363	uintptr_t start_free = area->base + pages*PAGE_SIZE;
357		364
358	/*	365	/*
359	* Shrinking the area.	366	* Shrinking the area.
360	* No need to check for overlaps.	367	* No need to check for overlaps.
361	*/	368	*/
362		369
363	/*	370	/*
364	* Start TLB shootdown sequence.	371	* Start TLB shootdown sequence.
365	*/	372	*/
366	tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base +	373	tlb_shootdown_start(TLB_INVL_PAGES, AS->asid, area->base +
367	pages * PAGE_SIZE, area->pages - pages);	374	pages * PAGE_SIZE, area->pages - pages);
368		375
369	/*	376	/*
370	* Remove frames belonging to used space starting from	377	* Remove frames belonging to used space starting from
371	* the highest addresses downwards until an overlap with	378	* the highest addresses downwards until an overlap with
372	* the resized address space area is found. Note that this	379	* the resized address space area is found. Note that this
373	* is also the right way to remove part of the used_space	380	* is also the right way to remove part of the used_space
374	* B+tree leaf list.	381	* B+tree leaf list.
375	*/	382	*/
376	for (cond = true; cond;) {	383	for (cond = true; cond;) {
377	btree_node_t *node;	384	btree_node_t *node;
378		385
379	ASSERT(!list_empty(&area->used_space.leaf_head));	386	ASSERT(!list_empty(&area->used_space.leaf_head));
380	node =	387	node =
381	list_get_instance(area->used_space.leaf_head.prev,	388	list_get_instance(area->used_space.leaf_head.prev,
382	btree_node_t, leaf_link);	389	btree_node_t, leaf_link);
383	if ((cond = (bool) node->keys)) {	390	if ((cond = (bool) node->keys)) {
384	uintptr_t b = node->key[node->keys - 1];	391	uintptr_t b = node->key[node->keys - 1];
385	count_t c =	392	count_t c =
386	(count_t) node->value[node->keys - 1];	393	(count_t) node->value[node->keys - 1];
387	int i = 0;	394	int i = 0;
388		395
389	if (overlaps(b, c * PAGE_SIZE, area->base,	396	if (overlaps(b, c * PAGE_SIZE, area->base,
390	pages*PAGE_SIZE)) {	397	pages*PAGE_SIZE)) {
391		398
392	if (b + c * PAGE_SIZE <= start_free) {	399	if (b + c * PAGE_SIZE <= start_free) {
393	/*	400	/*
394	* The whole interval fits	401	* The whole interval fits
395	* completely in the resized	402	* completely in the resized
396	* address space area.	403	* address space area.
397	*/	404	*/
398	break;	405	break;
399	}	406	}
400		407
401	/*	408	/*
402	* Part of the interval corresponding	409	* Part of the interval corresponding
403	* to b and c overlaps with the resized	410	* to b and c overlaps with the resized
404	* address space area.	411	* address space area.
405	*/	412	*/
406		413
407	cond = false; /* we are almost done */	414	cond = false; /* we are almost done */
408	i = (start_free - b) >> PAGE_WIDTH;	415	i = (start_free - b) >> PAGE_WIDTH;
409	if (!used_space_remove(area, start_free,	416	if (!used_space_remove(area, start_free,
410	c - i))	417	c - i))
411	panic("Could not remove used "	418	panic("Could not remove used "
412	"space.\n");	419	"space.\n");
413	} else {	420	} else {
414	/*	421	/*
415	* The interval of used space can be	422	* The interval of used space can be
416	* completely removed.	423	* completely removed.
417	*/	424	*/
418	if (!used_space_remove(area, b, c))	425	if (!used_space_remove(area, b, c))
419	panic("Could not remove used "	426	panic("Could not remove used "
420	"space.\n");	427	"space.\n");
421	}	428	}
422		429
423	for (; i < c; i++) {	430	for (; i < c; i++) {
424	pte_t *pte;	431	pte_t *pte;
425		432
426	page_table_lock(as, false);	433	page_table_lock(as, false);
427	pte = page_mapping_find(as, b +	434	pte = page_mapping_find(as, b +
428	i * PAGE_SIZE);	435	i * PAGE_SIZE);
429	ASSERT(pte && PTE_VALID(pte) &&	436	ASSERT(pte && PTE_VALID(pte) &&
430	PTE_PRESENT(pte));	437	PTE_PRESENT(pte));
431	if (area->backend &&	438	if (area->backend &&
432	area->backend->frame_free) {	439	area->backend->frame_free) {
433	area->backend->frame_free(area,	440	area->backend->frame_free(area,
434	b + i * PAGE_SIZE,	441	b + i * PAGE_SIZE,
435	PTE_GET_FRAME(pte));	442	PTE_GET_FRAME(pte));
436	}	443	}
437	page_mapping_remove(as, b +	444	page_mapping_remove(as, b +
438	i * PAGE_SIZE);	445	i * PAGE_SIZE);
439	page_table_unlock(as, false);	446	page_table_unlock(as, false);
440	}	447	}
441	}	448	}
442	}	449	}
443		450
444	/*	451	/*
445	* Finish TLB shootdown sequence.	452	* Finish TLB shootdown sequence.
446	*/	453	*/
447	tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,	454	tlb_invalidate_pages(as->asid, area->base + pages * PAGE_SIZE,
448	area->pages - pages);	455	area->pages - pages);
449	tlb_shootdown_finalize();	456	tlb_shootdown_finalize();
450		457
451	/*	458	/*
452	* Invalidate software translation caches (e.g. TSB on sparc64).	459	* Invalidate software translation caches (e.g. TSB on sparc64).
453	*/	460	*/
454	as_invalidate_translation_cache(as, area->base +	461	as_invalidate_translation_cache(as, area->base +
455	pages * PAGE_SIZE, area->pages - pages);	462	pages * PAGE_SIZE, area->pages - pages);
456	} else {	463	} else {
457	/*	464	/*
458	* Growing the area.	465	* Growing the area.
459	* Check for overlaps with other address space areas.	466	* Check for overlaps with other address space areas.
460	*/	467	*/
461	if (!check_area_conflicts(as, address, pages * PAGE_SIZE,	468	if (!check_area_conflicts(as, address, pages * PAGE_SIZE,
462	area)) {	469	area)) {
463	mutex_unlock(&area->lock);	470	mutex_unlock(&area->lock);
464	mutex_unlock(&as->lock);	471	mutex_unlock(&as->lock);
465	interrupts_restore(ipl);	472	interrupts_restore(ipl);
466	return EADDRNOTAVAIL;	473	return EADDRNOTAVAIL;
467	}	474	}
468	}	475	}
469		476
470	area->pages = pages;	477	area->pages = pages;
471		478
472	mutex_unlock(&area->lock);	479	mutex_unlock(&area->lock);
473	mutex_unlock(&as->lock);	480	mutex_unlock(&as->lock);
474	interrupts_restore(ipl);	481	interrupts_restore(ipl);
475		482
476	return 0;	483	return 0;
477	}	484	}
478		485
479	/** Destroy address space area.	486	/** Destroy address space area.
480	*	487	*
481	* @param as Address space.	488	* @param as Address space.
482	* @param address Address withing the area to be deleted.	489	* @param address Address withing the area to be deleted.
483	*	490	*
484	* @return Zero on success or a value from @ref errno.h on failure.	491	* @return Zero on success or a value from @ref errno.h on failure.
485	*/	492	*/
486	int as_area_destroy(as_t *as, uintptr_t address)	493	int as_area_destroy(as_t *as, uintptr_t address)
487	{	494	{
488	as_area_t *area;	495	as_area_t *area;
489	uintptr_t base;	496	uintptr_t base;
490	link_t *cur;	497	link_t *cur;
491	ipl_t ipl;	498	ipl_t ipl;
492		499

Subversion Repositories HelenOS

(root)/trunk/kernel/generic/src/mm/as.c – Rev 2087 → 2089