Subversion Repositories HelenOS

Rev

Rev 4509 | Go to most recent revision | Blame | Compare with Previous | Last modification | View Log | Download | RSS feed

/*
 * 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 <ipc/ipc.h>
#include <async.h>
#include <errno.h>
#include <atomic.h>
#include <stdlib.h>
#include <string.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 fs_node_t *tmpfs_match(fs_node_t *, const char *);
static fs_node_t *tmpfs_node_get(dev_handle_t, fs_index_t);
static void tmpfs_node_put(fs_node_t *);
static fs_node_t *tmpfs_create_node(dev_handle_t, int);
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 *);
static int tmpfs_destroy_node(fs_node_t *);

/* Implementation of helper functions. */
static fs_index_t tmpfs_index_get(fs_node_t *fn)
{
    return TMPFS_NODE(fn)->index;
}

static size_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 bool tmpfs_has_children(fs_node_t *fn)
{
    return !list_empty(&TMPFS_NODE(fn)->cs_head);
}

static fs_node_t *tmpfs_root_get(dev_handle_t dev_handle)
{
    return tmpfs_node_get(dev_handle, TMPFS_SOME_ROOT); 
}

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;
}

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

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

#define NODES_KEY_INDEX 0
#define NODES_KEY_DEV   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);
    return (nodep->index == key[NODES_KEY_INDEX] &&
        nodep->dev_handle == key[NODES_KEY_DEV]);
}

static void nodes_remove_callback(link_t *item)
{
}

/** 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->dev_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(dev_handle_t dev_handle)
{
    fs_node_t *rfn;
    
    rfn = tmpfs_create_node(dev_handle, L_DIRECTORY);
    if (!rfn) 
        return false;
    TMPFS_NODE(rfn)->lnkcnt = 0;    /* FS root is not linked */
    return true;
}

fs_node_t *tmpfs_match(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 = list_get_instance(lnk, tmpfs_dentry_t,
            link);
        if (!str_cmp(dentryp->name, component))
            return FS_NODE(dentryp->node);
    }

    return NULL;
}

fs_node_t *tmpfs_node_get(dev_handle_t dev_handle, fs_index_t index)
{
    unsigned long key[] = {
        [NODES_KEY_INDEX] = index,
        [NODES_KEY_DEV] = dev_handle
    };
    link_t *lnk = hash_table_find(&nodes, key);
    if (!lnk)
        return NULL;
    return FS_NODE(hash_table_get_instance(lnk, tmpfs_node_t, nh_link));
}

void tmpfs_node_put(fs_node_t *fn)
{
    /* nothing to do */
}

fs_node_t *tmpfs_create_node(dev_handle_t dev_handle, int lflag)
{
    assert((lflag & L_FILE) ^ (lflag & L_DIRECTORY));

    tmpfs_node_t *nodep = malloc(sizeof(tmpfs_node_t));
    if (!nodep)
        return NULL;
    tmpfs_node_initialize(nodep);
    nodep->bp = malloc(sizeof(fs_node_t));
    if (!nodep->bp) {
        free(nodep);
        return NULL;
    }
    fs_node_initialize(nodep->bp);
    nodep->bp->data = nodep;    /* link the FS and TMPFS nodes */
    if (!tmpfs_root_get(dev_handle))
        nodep->index = TMPFS_SOME_ROOT;
    else
        nodep->index = tmpfs_next_index++;
    nodep->dev_handle = dev_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_INDEX] = nodep->index,
        [NODES_KEY_DEV] = nodep->dev_handle
    };
    hash_table_insert(&nodes, key, &nodep->nh_link);
    return FS_NODE(nodep);
}

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;
}

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_INDEX] = nodep->index,
        [NODES_KEY_DEV] = nodep->dev_handle
    };
    hash_table_remove(&nodes, key, 2);

    if (nodep->type == TMPFS_FILE)
        free(nodep->data);
    free(nodep->bp);
    free(nodep);
    return EOK;
}

void tmpfs_mounted(ipc_callid_t rid, ipc_call_t *request)
{
    dev_handle_t dev_handle = (dev_handle_t) IPC_GET_ARG1(*request);

    /* accept the mount options */
    ipc_callid_t callid;
    size_t size;
    if (!ipc_data_write_receive(&callid, &size)) {
        ipc_answer_0(callid, EINVAL);
        ipc_answer_0(rid, EINVAL);
        return;
    }
    char *opts = malloc(size + 1);
    if (!opts) {
        ipc_answer_0(callid, ENOMEM);
        ipc_answer_0(rid, ENOMEM);
        return;
    }
    ipcarg_t retval = ipc_data_write_finalize(callid, opts, size);
    if (retval != EOK) {
        ipc_answer_0(rid, retval);
        free(opts);
        return;
    }
    opts[size] = '\0';

    /* Initialize TMPFS instance. */
    if (!tmpfs_instance_init(dev_handle)) {
        ipc_answer_0(rid, ENOMEM);
        return;
    }

    tmpfs_node_t *rootp = TMPFS_NODE(tmpfs_root_get(dev_handle));
    if (str_cmp(opts, "restore") == 0) {
        if (tmpfs_restore(dev_handle))
            ipc_answer_3(rid, EOK, rootp->index, rootp->size,
                rootp->lnkcnt);
        else
            ipc_answer_0(rid, ELIMIT);
    } else {
        ipc_answer_3(rid, EOK, rootp->index, rootp->size,
            rootp->lnkcnt);
    }
}

void tmpfs_mount(ipc_callid_t rid, ipc_call_t *request)
{
    libfs_mount(&tmpfs_libfs_ops, tmpfs_reg.fs_handle, 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)
{
    dev_handle_t dev_handle = (dev_handle_t)IPC_GET_ARG1(*request);
    fs_index_t index = (fs_index_t)IPC_GET_ARG2(*request);
    off_t pos = (off_t)IPC_GET_ARG3(*request);

    /*
     * Lookup the respective TMPFS node.
     */
    link_t *hlp;
    unsigned long key[] = {
        [NODES_KEY_INDEX] = index,
        [NODES_KEY_DEV] = dev_handle,
    };
    hlp = hash_table_find(&nodes, key);
    if (!hlp) {
        ipc_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 (!ipc_data_read_receive(&callid, &size)) {
        ipc_answer_0(callid, EINVAL);   
        ipc_answer_0(rid, EINVAL);
        return;
    }

    size_t bytes;
    if (nodep->type == TMPFS_FILE) {
        bytes = max(0, min(nodep->size - pos, size));
        (void) ipc_data_read_finalize(callid, nodep->data + pos,
            bytes);
    } else {
        tmpfs_dentry_t *dentryp;
        link_t *lnk;
        int 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) {
            ipc_answer_0(callid, ENOENT);
            ipc_answer_1(rid, ENOENT, 0);
            return;
        }

        dentryp = list_get_instance(lnk, tmpfs_dentry_t, link);

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

    /*
     * Answer the VFS_READ call.
     */
    ipc_answer_1(rid, EOK, bytes);
}

void tmpfs_write(ipc_callid_t rid, ipc_call_t *request)
{
    dev_handle_t dev_handle = (dev_handle_t)IPC_GET_ARG1(*request);
    fs_index_t index = (fs_index_t)IPC_GET_ARG2(*request);
    off_t pos = (off_t)IPC_GET_ARG3(*request);

    /*
     * Lookup the respective TMPFS node.
     */
    link_t *hlp;
    unsigned long key[] = {
        [NODES_KEY_INDEX] = index,
        [NODES_KEY_DEV] = dev_handle
    };
    hlp = hash_table_find(&nodes, key);
    if (!hlp) {
        ipc_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 (!ipc_data_write_receive(&callid, &size)) {
        ipc_answer_0(callid, EINVAL);   
        ipc_answer_0(rid, EINVAL);
        return;
    }

    /*
     * Check whether the file needs to grow.
     */
    if (pos + size <= nodep->size) {
        /* The file size is not changing. */
        (void) ipc_data_write_finalize(callid, nodep->data + pos, size);
        ipc_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) {
        ipc_answer_0(callid, ENOMEM);
        ipc_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) ipc_data_write_finalize(callid, nodep->data + pos, size);
    ipc_answer_2(rid, EOK, size, nodep->size);
}

void tmpfs_truncate(ipc_callid_t rid, ipc_call_t *request)
{
    dev_handle_t dev_handle = (dev_handle_t)IPC_GET_ARG1(*request);
    fs_index_t index = (fs_index_t)IPC_GET_ARG2(*request);
    size_t size = (off_t)IPC_GET_ARG3(*request);

    /*
     * Lookup the respective TMPFS node.
     */
    link_t *hlp;
    unsigned long key[] = {
        [NODES_KEY_INDEX] = index,
        [NODES_KEY_DEV] = dev_handle
    };
    hlp = hash_table_find(&nodes, key);
    if (!hlp) {
        ipc_answer_0(rid, ENOENT);
        return;
    }
    tmpfs_node_t *nodep = hash_table_get_instance(hlp, tmpfs_node_t,
        nh_link);

    if (size == nodep->size) {
        ipc_answer_0(rid, EOK);
        return;
    }

    void *newdata = realloc(nodep->data, size);
    if (!newdata) {
        ipc_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;
    ipc_answer_0(rid, EOK);
}

void tmpfs_close(ipc_callid_t rid, ipc_call_t *request)
{
    ipc_answer_0(rid, EOK);
}

void tmpfs_destroy(ipc_callid_t rid, ipc_call_t *request)
{
    dev_handle_t dev_handle = (dev_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_INDEX] = index,
        [NODES_KEY_DEV] = dev_handle
    };
    hlp = hash_table_find(&nodes, key);
    if (!hlp) {
        ipc_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));
    ipc_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)
{
    ipc_answer_0(rid, ENOTSUP);
}

void tmpfs_sync(ipc_callid_t rid, ipc_call_t *request)
{
    /* Dummy implementation */
    ipc_answer_0(rid, EOK);
}

/**
 * @}
 */