source: mainline/uspace/srv/fs/tmpfs/tmpfs_ops.c@ 46b881c

/*
 * Copyright (c) 2008 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 fs
 * @{
 */

/**
 * @file        tmpfs_ops.c
 * @brief       Implementation of VFS operations for the TMPFS file system
 *              server.
 */

#include "tmpfs.h"
#include "../../vfs/vfs.h"
#include <macros.h>
#include <stdint.h>
#include <async.h>
#include <errno.h>
#include <atomic.h>
#include <stdlib.h>
#include <str.h>
#include <stdio.h>
#include <assert.h>
#include <sys/types.h>
#include <adt/hash_table.h>
#include <as.h>
#include <libfs.h>

#define min(a, b)               ((a) < (b) ? (a) : (b))
#define max(a, b)               ((a) > (b) ? (a) : (b))

#define NODES_BUCKETS   256

/** All root nodes have index 0. */
#define TMPFS_SOME_ROOT         0
/** Global counter for assigning node indices. Shared by all instances. */
fs_index_t tmpfs_next_index = 1;

/*
 * Implementation of the libfs interface.
 */

/* Forward declarations of static functions. */
static int tmpfs_match(fs_node_t **, fs_node_t *, const char *);
static int tmpfs_node_get(fs_node_t **, devmap_handle_t, fs_index_t);
static int tmpfs_node_open(fs_node_t *);
static int tmpfs_node_put(fs_node_t *);
static int tmpfs_create_node(fs_node_t **, devmap_handle_t, int);
static int tmpfs_destroy_node(fs_node_t *);
static int tmpfs_link_node(fs_node_t *, fs_node_t *, const char *);
static int tmpfs_unlink_node(fs_node_t *, fs_node_t *, const char *);

/* Implementation of helper functions. */
static int tmpfs_root_get(fs_node_t **rfn, devmap_handle_t devmap_handle)
{
        return tmpfs_node_get(rfn, devmap_handle, TMPFS_SOME_ROOT); 
}

static int tmpfs_has_children(bool *has_children, fs_node_t *fn)
{
        *has_children = !list_empty(&TMPFS_NODE(fn)->cs_head);
        return EOK;
}

static fs_index_t tmpfs_index_get(fs_node_t *fn)
{
        return TMPFS_NODE(fn)->index;
}

static aoff64_t tmpfs_size_get(fs_node_t *fn)
{
        return TMPFS_NODE(fn)->size;
}

static unsigned tmpfs_lnkcnt_get(fs_node_t *fn)
{
        return TMPFS_NODE(fn)->lnkcnt;
}

static char tmpfs_plb_get_char(unsigned pos)
{
        return tmpfs_reg.plb_ro[pos % PLB_SIZE];
}

static bool tmpfs_is_directory(fs_node_t *fn)
{
        return TMPFS_NODE(fn)->type == TMPFS_DIRECTORY;
}

static bool tmpfs_is_file(fs_node_t *fn)
{
        return TMPFS_NODE(fn)->type == TMPFS_FILE;
}

static devmap_handle_t tmpfs_device_get(fs_node_t *fn)
{
        return 0;
}

/** libfs operations */
libfs_ops_t tmpfs_libfs_ops = {
        .root_get = tmpfs_root_get,
        .match = tmpfs_match,
        .node_get = tmpfs_node_get,
        .node_open = tmpfs_node_open,
        .node_put = tmpfs_node_put,
        .create = tmpfs_create_node,
        .destroy = tmpfs_destroy_node,
        .link = tmpfs_link_node,
        .unlink = tmpfs_unlink_node,
        .has_children = tmpfs_has_children,
        .index_get = tmpfs_index_get,
        .size_get = tmpfs_size_get,
        .lnkcnt_get = tmpfs_lnkcnt_get,
        .plb_get_char = tmpfs_plb_get_char,
        .is_directory = tmpfs_is_directory,
        .is_file = tmpfs_is_file,
        .device_get = tmpfs_device_get
};

/** Hash table of all TMPFS nodes. */
hash_table_t nodes;

#define NODES_KEY_DEV   0       
#define NODES_KEY_INDEX 1

/* Implementation of hash table interface for the nodes hash table. */
static hash_index_t nodes_hash(unsigned long key[])
{
        return key[NODES_KEY_INDEX] % NODES_BUCKETS;
}

static int nodes_compare(unsigned long key[], hash_count_t keys, link_t *item)
{
        tmpfs_node_t *nodep = hash_table_get_instance(item, tmpfs_node_t,
            nh_link);
        
        switch (keys) {
        case 1:
                return (nodep->devmap_handle == key[NODES_KEY_DEV]);
        case 2: 
                return ((nodep->devmap_handle == key[NODES_KEY_DEV]) &&
                    (nodep->index == key[NODES_KEY_INDEX]));
        default:
                assert((keys == 1) || (keys == 2));
        }

        return 0;
}

static void nodes_remove_callback(link_t *item)
{
        tmpfs_node_t *nodep = hash_table_get_instance(item, tmpfs_node_t,
            nh_link);

        while (!list_empty(&nodep->cs_head)) {
                tmpfs_dentry_t *dentryp = list_get_instance(nodep->cs_head.next,
                    tmpfs_dentry_t, link);

                assert(nodep->type == TMPFS_DIRECTORY);
                list_remove(&dentryp->link);
                free(dentryp);
        }

        if (nodep->data) {
                assert(nodep->type == TMPFS_FILE);
                free(nodep->data);
        }
        free(nodep->bp);
        free(nodep);
}

/** TMPFS nodes hash table operations. */
hash_table_operations_t nodes_ops = {
        .hash = nodes_hash,
        .compare = nodes_compare,
        .remove_callback = nodes_remove_callback
};

static void tmpfs_node_initialize(tmpfs_node_t *nodep)
{
        nodep->bp = NULL;
        nodep->index = 0;
        nodep->devmap_handle = 0;
        nodep->type = TMPFS_NONE;
        nodep->lnkcnt = 0;
        nodep->size = 0;
        nodep->data = NULL;
        link_initialize(&nodep->nh_link);
        list_initialize(&nodep->cs_head);
}

static void tmpfs_dentry_initialize(tmpfs_dentry_t *dentryp)
{
        link_initialize(&dentryp->link);
        dentryp->name = NULL;
        dentryp->node = NULL;
}

bool tmpfs_init(void)
{
        if (!hash_table_create(&nodes, NODES_BUCKETS, 2, &nodes_ops))
                return false;
        
        return true;
}

static bool tmpfs_instance_init(devmap_handle_t devmap_handle)
{
        fs_node_t *rfn;
        int rc;
        
        rc = tmpfs_create_node(&rfn, devmap_handle, L_DIRECTORY);
        if (rc != EOK || !rfn)
                return false;
        TMPFS_NODE(rfn)->lnkcnt = 0;    /* FS root is not linked */
        return true;
}

static void tmpfs_instance_done(devmap_handle_t devmap_handle)
{
        unsigned long key[] = {
                [NODES_KEY_DEV] = devmap_handle
        };
        /*
         * Here we are making use of one special feature of our hash table
         * implementation, which allows to remove more items based on a partial
         * key match. In the following, we are going to remove all nodes
         * matching our device handle. The nodes_remove_callback() function will
         * take care of resource deallocation.
         */
        hash_table_remove(&nodes, key, 1);
}

int tmpfs_match(fs_node_t **rfn, fs_node_t *pfn, const char *component)
{
        tmpfs_node_t *parentp = TMPFS_NODE(pfn);
        link_t *lnk;

        for (lnk = parentp->cs_head.next; lnk != &parentp->cs_head;
            lnk = lnk->next) {
                tmpfs_dentry_t *dentryp;
                dentryp = list_get_instance(lnk, tmpfs_dentry_t, link);
                if (!str_cmp(dentryp->name, component)) {
                        *rfn = FS_NODE(dentryp->node);
                        return EOK;
                }
        }

        *rfn = NULL;
        return EOK;
}

int tmpfs_node_get(fs_node_t **rfn, devmap_handle_t devmap_handle, fs_index_t index)
{
        unsigned long key[] = {
                [NODES_KEY_DEV] = devmap_handle,
                [NODES_KEY_INDEX] = index
        };
        link_t *lnk = hash_table_find(&nodes, key);
        if (lnk) {
                tmpfs_node_t *nodep;
                nodep = hash_table_get_instance(lnk, tmpfs_node_t, nh_link);
                *rfn = FS_NODE(nodep);
        } else {
                *rfn = NULL;
        }
        return EOK;     
}

int tmpfs_node_open(fs_node_t *fn)
{
        /* nothing to do */
        return EOK;
}

int tmpfs_node_put(fs_node_t *fn)
{
        /* nothing to do */
        return EOK;
}

int tmpfs_create_node(fs_node_t **rfn, devmap_handle_t devmap_handle, int lflag)
{
        fs_node_t *rootfn;
        int rc;

        assert((lflag & L_FILE) ^ (lflag & L_DIRECTORY));

        tmpfs_node_t *nodep = malloc(sizeof(tmpfs_node_t));
        if (!nodep)
                return ENOMEM;
        tmpfs_node_initialize(nodep);
        nodep->bp = malloc(sizeof(fs_node_t));
        if (!nodep->bp) {
                free(nodep);
                return ENOMEM;
        }
        fs_node_initialize(nodep->bp);
        nodep->bp->data = nodep;        /* link the FS and TMPFS nodes */

        rc = tmpfs_root_get(&rootfn, devmap_handle);
        assert(rc == EOK);
        if (!rootfn)
                nodep->index = TMPFS_SOME_ROOT;
        else
                nodep->index = tmpfs_next_index++;
        nodep->devmap_handle = devmap_handle;
        if (lflag & L_DIRECTORY) 
                nodep->type = TMPFS_DIRECTORY;
        else 
                nodep->type = TMPFS_FILE;

        /* Insert the new node into the nodes hash table. */
        unsigned long key[] = {
                [NODES_KEY_DEV] = nodep->devmap_handle,
                [NODES_KEY_INDEX] = nodep->index
        };
        hash_table_insert(&nodes, key, &nodep->nh_link);
        *rfn = FS_NODE(nodep);
        return EOK;
}

int tmpfs_destroy_node(fs_node_t *fn)
{
        tmpfs_node_t *nodep = TMPFS_NODE(fn);
        
        assert(!nodep->lnkcnt);
        assert(list_empty(&nodep->cs_head));

        unsigned long key[] = {
                [NODES_KEY_DEV] = nodep->devmap_handle,
                [NODES_KEY_INDEX] = nodep->index
        };
        hash_table_remove(&nodes, key, 2);

        /*
         * The nodes_remove_callback() function takes care of the actual
         * resource deallocation.
         */
        return EOK;
}

int tmpfs_link_node(fs_node_t *pfn, fs_node_t *cfn, const char *nm)
{
        tmpfs_node_t *parentp = TMPFS_NODE(pfn);
        tmpfs_node_t *childp = TMPFS_NODE(cfn);
        tmpfs_dentry_t *dentryp;
        link_t *lnk;

        assert(parentp->type == TMPFS_DIRECTORY);

        /* Check for duplicit entries. */
        for (lnk = parentp->cs_head.next; lnk != &parentp->cs_head;
            lnk = lnk->next) {
                dentryp = list_get_instance(lnk, tmpfs_dentry_t, link); 
                if (!str_cmp(dentryp->name, nm))
                        return EEXIST;
        }

        /* Allocate and initialize the dentry. */
        dentryp = malloc(sizeof(tmpfs_dentry_t));
        if (!dentryp)
                return ENOMEM;
        tmpfs_dentry_initialize(dentryp);

        /* Populate and link the new dentry. */
        size_t size = str_size(nm);
        dentryp->name = malloc(size + 1);
        if (!dentryp->name) {
                free(dentryp);
                return ENOMEM;
        }
        str_cpy(dentryp->name, size + 1, nm);
        dentryp->node = childp;
        childp->lnkcnt++;
        list_append(&dentryp->link, &parentp->cs_head);

        return EOK;
}

int tmpfs_unlink_node(fs_node_t *pfn, fs_node_t *cfn, const char *nm)
{
        tmpfs_node_t *parentp = TMPFS_NODE(pfn);
        tmpfs_node_t *childp = NULL;
        tmpfs_dentry_t *dentryp;
        link_t *lnk;

        if (!parentp)
                return EBUSY;
        
        for (lnk = parentp->cs_head.next; lnk != &parentp->cs_head;
            lnk = lnk->next) {
                dentryp = list_get_instance(lnk, tmpfs_dentry_t, link);
                if (!str_cmp(dentryp->name, nm)) {
                        childp = dentryp->node;
                        assert(FS_NODE(childp) == cfn);
                        break;
                }       
        }

        if (!childp)
                return ENOENT;
                
        if ((childp->lnkcnt == 1) && !list_empty(&childp->cs_head))
                return ENOTEMPTY;

        list_remove(&dentryp->link);
        free(dentryp);
        childp->lnkcnt--;

        return EOK;
}

void tmpfs_mounted(ipc_callid_t rid, ipc_call_t *request)
{
        devmap_handle_t devmap_handle = (devmap_handle_t) IPC_GET_ARG1(*request);
        fs_node_t *rootfn;
        int rc;
        
        /* Accept the mount options. */
        char *opts;
        rc = async_data_write_accept((void **) &opts, true, 0, 0, 0, NULL);
        if (rc != EOK) {
                async_answer_0(rid, rc);
                return;
        }

        /* Check if this device is not already mounted. */
        rc = tmpfs_root_get(&rootfn, devmap_handle);
        if ((rc == EOK) && (rootfn)) {
                (void) tmpfs_node_put(rootfn);
                free(opts);
                async_answer_0(rid, EEXIST);
                return;
        }

        /* Initialize TMPFS instance. */
        if (!tmpfs_instance_init(devmap_handle)) {
                free(opts);
                async_answer_0(rid, ENOMEM);
                return;
        }

        rc = tmpfs_root_get(&rootfn, devmap_handle);
        assert(rc == EOK);
        tmpfs_node_t *rootp = TMPFS_NODE(rootfn);
        if (str_cmp(opts, "restore") == 0) {
                if (tmpfs_restore(devmap_handle))
                        async_answer_3(rid, EOK, rootp->index, rootp->size,
                            rootp->lnkcnt);
                else
                        async_answer_0(rid, ELIMIT);
        } else {
                async_answer_3(rid, EOK, rootp->index, rootp->size,
                    rootp->lnkcnt);
        }
        free(opts);
}

void tmpfs_mount(ipc_callid_t rid, ipc_call_t *request)
{
        libfs_mount(&tmpfs_libfs_ops, tmpfs_reg.fs_handle, rid, request);
}

void tmpfs_unmounted(ipc_callid_t rid, ipc_call_t *request)
{
        devmap_handle_t devmap_handle = (devmap_handle_t) IPC_GET_ARG1(*request);

        tmpfs_instance_done(devmap_handle);
        async_answer_0(rid, EOK);
}

void tmpfs_unmount(ipc_callid_t rid, ipc_call_t *request)
{
        libfs_unmount(&tmpfs_libfs_ops, rid, request);
}

void tmpfs_lookup(ipc_callid_t rid, ipc_call_t *request)
{
        libfs_lookup(&tmpfs_libfs_ops, tmpfs_reg.fs_handle, rid, request);
}

void tmpfs_read(ipc_callid_t rid, ipc_call_t *request)
{
        devmap_handle_t devmap_handle = (devmap_handle_t) IPC_GET_ARG1(*request);
        fs_index_t index = (fs_index_t) IPC_GET_ARG2(*request);
        aoff64_t pos =
            (aoff64_t) MERGE_LOUP32(IPC_GET_ARG3(*request), IPC_GET_ARG4(*request));
        
        /*
         * Lookup the respective TMPFS node.
         */
        link_t *hlp;
        unsigned long key[] = {
                [NODES_KEY_DEV] = devmap_handle,
                [NODES_KEY_INDEX] = index
        };
        hlp = hash_table_find(&nodes, key);
        if (!hlp) {
                async_answer_0(rid, ENOENT);
                return;
        }
        tmpfs_node_t *nodep = hash_table_get_instance(hlp, tmpfs_node_t,
            nh_link);
        
        /*
         * Receive the read request.
         */
        ipc_callid_t callid;
        size_t size;
        if (!async_data_read_receive(&callid, &size)) {
                async_answer_0(callid, EINVAL);
                async_answer_0(rid, EINVAL);
                return;
        }

        size_t bytes;
        if (nodep->type == TMPFS_FILE) {
                bytes = min(nodep->size - pos, size);
                (void) async_data_read_finalize(callid, nodep->data + pos,
                    bytes);
        } else {
                tmpfs_dentry_t *dentryp;
                link_t *lnk;
                aoff64_t i;
                
                assert(nodep->type == TMPFS_DIRECTORY);
                
                /*
                 * Yes, we really use O(n) algorithm here.
                 * If it bothers someone, it could be fixed by introducing a
                 * hash table.
                 */
                for (i = 0, lnk = nodep->cs_head.next;
                    (i < pos) && (lnk != &nodep->cs_head);
                    i++, lnk = lnk->next)
                        ;

                if (lnk == &nodep->cs_head) {
                        async_answer_0(callid, ENOENT);
                        async_answer_1(rid, ENOENT, 0);
                        return;
                }

                dentryp = list_get_instance(lnk, tmpfs_dentry_t, link);

                (void) async_data_read_finalize(callid, dentryp->name,
                    str_size(dentryp->name) + 1);
                bytes = 1;
        }

        /*
         * Answer the VFS_READ call.
         */
        async_answer_1(rid, EOK, bytes);
}

void tmpfs_write(ipc_callid_t rid, ipc_call_t *request)
{
        devmap_handle_t devmap_handle = (devmap_handle_t) IPC_GET_ARG1(*request);
        fs_index_t index = (fs_index_t) IPC_GET_ARG2(*request);
        aoff64_t pos =
            (aoff64_t) MERGE_LOUP32(IPC_GET_ARG3(*request), IPC_GET_ARG4(*request));
        
        /*
         * Lookup the respective TMPFS node.
         */
        link_t *hlp;
        unsigned long key[] = {
                [NODES_KEY_DEV] = devmap_handle,
                [NODES_KEY_INDEX] = index
        };
        hlp = hash_table_find(&nodes, key);
        if (!hlp) {
                async_answer_0(rid, ENOENT);
                return;
        }
        tmpfs_node_t *nodep = hash_table_get_instance(hlp, tmpfs_node_t,
            nh_link);

        /*
         * Receive the write request.
         */
        ipc_callid_t callid;
        size_t size;
        if (!async_data_write_receive(&callid, &size)) {
                async_answer_0(callid, EINVAL); 
                async_answer_0(rid, EINVAL);
                return;
        }

        /*
         * Check whether the file needs to grow.
         */
        if (pos + size <= nodep->size) {
                /* The file size is not changing. */
                (void) async_data_write_finalize(callid, nodep->data + pos, size);
                async_answer_2(rid, EOK, size, nodep->size);
                return;
        }
        size_t delta = (pos + size) - nodep->size;
        /*
         * At this point, we are deliberately extremely straightforward and
         * simply realloc the contents of the file on every write that grows the
         * file. In the end, the situation might not be as bad as it may look:
         * our heap allocator can save us and just grow the block whenever
         * possible.
         */
        void *newdata = realloc(nodep->data, nodep->size + delta);
        if (!newdata) {
                async_answer_0(callid, ENOMEM);
                async_answer_2(rid, EOK, 0, nodep->size);
                return;
        }
        /* Clear any newly allocated memory in order to emulate gaps. */
        memset(newdata + nodep->size, 0, delta);
        nodep->size += delta;
        nodep->data = newdata;
        (void) async_data_write_finalize(callid, nodep->data + pos, size);
        async_answer_2(rid, EOK, size, nodep->size);
}

void tmpfs_truncate(ipc_callid_t rid, ipc_call_t *request)
{
        devmap_handle_t devmap_handle = (devmap_handle_t) IPC_GET_ARG1(*request);
        fs_index_t index = (fs_index_t) IPC_GET_ARG2(*request);
        aoff64_t size =
            (aoff64_t) MERGE_LOUP32(IPC_GET_ARG3(*request), IPC_GET_ARG4(*request));
        
        /*
         * Lookup the respective TMPFS node.
         */
        unsigned long key[] = {
                [NODES_KEY_DEV] = devmap_handle,
                [NODES_KEY_INDEX] = index
        };
        link_t *hlp = hash_table_find(&nodes, key);
        if (!hlp) {
                async_answer_0(rid, ENOENT);
                return;
        }
        tmpfs_node_t *nodep = hash_table_get_instance(hlp, tmpfs_node_t,
            nh_link);
        
        if (size == nodep->size) {
                async_answer_0(rid, EOK);
                return;
        }
        
        if (size > SIZE_MAX) {
                async_answer_0(rid, ENOMEM);
                return;
        }
        
        void *newdata = realloc(nodep->data, size);
        if (!newdata) {
                async_answer_0(rid, ENOMEM);
                return;
        }
        
        if (size > nodep->size) {
                size_t delta = size - nodep->size;
                memset(newdata + nodep->size, 0, delta);
        }
        
        nodep->size = size;
        nodep->data = newdata;
        async_answer_0(rid, EOK);
}

void tmpfs_close(ipc_callid_t rid, ipc_call_t *request)
{
        async_answer_0(rid, EOK);
}

void tmpfs_destroy(ipc_callid_t rid, ipc_call_t *request)
{
        devmap_handle_t devmap_handle = (devmap_handle_t)IPC_GET_ARG1(*request);
        fs_index_t index = (fs_index_t)IPC_GET_ARG2(*request);
        int rc;

        link_t *hlp;
        unsigned long key[] = {
                [NODES_KEY_DEV] = devmap_handle,
                [NODES_KEY_INDEX] = index
        };
        hlp = hash_table_find(&nodes, key);
        if (!hlp) {
                async_answer_0(rid, ENOENT);
                return;
        }
        tmpfs_node_t *nodep = hash_table_get_instance(hlp, tmpfs_node_t,
            nh_link);
        rc = tmpfs_destroy_node(FS_NODE(nodep));
        async_answer_0(rid, rc);
}

void tmpfs_open_node(ipc_callid_t rid, ipc_call_t *request)
{
        libfs_open_node(&tmpfs_libfs_ops, tmpfs_reg.fs_handle, rid, request);
}

void tmpfs_stat(ipc_callid_t rid, ipc_call_t *request)
{
        libfs_stat(&tmpfs_libfs_ops, tmpfs_reg.fs_handle, rid, request);
}

void tmpfs_sync(ipc_callid_t rid, ipc_call_t *request)
{
        /*
         * TMPFS keeps its data structures always consistent,
         * thus the sync operation is a no-op.
         */
        async_answer_0(rid, EOK);
}

/**
 * @}
 */